A thesis submitted for the degree of Doctor of Philosophy ...689056/s4298612_final_the… · MDevEcon, BCom (Acc) A thesis submitted for the degree of Doctor of Philosophy at The

Modelling the health and economic impact of built environment interventions

María Belén Zapata Diomedi

MDevEcon, BCom (Acc)

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2017

Faculty of Medicine

i

Abstract

Background

In Australia, over half (57%) of the adult population is not meeting national physical activity

guidelines. This is concerning, given that physical inactivity is a well-known preventable risk

factor for several chronic diseases. Physical inactivity was estimated to cost Australians

124,000 disability-adjusted life years annually (2015) and a significant proportion of the

healthcare budget. Furthermore, physically inactive people are more likely to be absent from

work, underperform, or both, impacting negatively on their career prospects and national

output.

In Australia, the majority of the people live in urban areas (90%) and population projections

show significant growth in its main cities. Hence, shaping cities such that they contribute to

healthy behaviours is key to building a healthy and prosperous population. The research

community can contribute to a healthy expansion of Australian cities by producing context

specific evidence for the potential of the built environment to enhance population health.

Aims

The overall aim of this thesis is to contribute to the evidence on the potential health and

economic benefits of investing in built environments that are supportive of physical activity.

Five research questions (RQ) contributed to achieving the aim of this thesis:

RQ 1: What are the attributes of the built environment in Australia that most benefit physical

activity?

RQ 2: What are the physical activity-related health externalities and healthcare costs

associated with changes in the built environment in Australia?

RQ 3: What economic evaluation methods have been used to model future health outcomes

from interventions in active transport?

RQ 4: Can the health impact of changes in physical activity be incorporated more robustly in

cost-benefit analysis of built-environment initiatives in Australia?

RQ 5: What are the potential health and economic impacts of Brisbane meeting its targets for

active travel?

ii

Methods

Two systematic reviews were conducted to address RQ 1 and 3, applying rigorous methods to

ensure objectivity in the selection of literature. For RQ 2, 4 and 5, a multi-state life-table

modelling approach, developed as part of the Assessing Cost Effectiveness in Prevention

study, was updated and expanded. For RQ 2, the model was used to estimate health-adjusted

life years (HALYs) and healthcare costs attributable to changes in exposure to attributes of the

built environment within the neighbourhood area. For RQ 4, the methods applied to answer

RQ 2 were expanded, and describe a method to estimate the health benefits of physical

activity that can be included in cost-benefit analyses of built environment interventions. Lastly,

the model is significantly expanded for RQ 5, with the addition of exposure to air pollution and

road trauma. These are added to physical activity to assess the likely overall health impact of

Brisbane achieving its proposed travel targets (15% of trips walking, 5% cycling and 14%

using public transport).

Main findings

For the Australian setting, recent evidence indicates that features of the built environment in

neighbourhoods, including the availability of destinations and diverse use of lands, can

support physical activity, and especially walking (RQ 1). Improvements in the built

environment within the neighbourhood can therefore significantly contribute to population

health (RQ2). However, it is uncertain whether improved health will translate into healthcare

costs savings, as initial reductions may be offset by healthcare costs in added life years. The

evidence demonstrating the economic merit of active transport interventions has been growing

markedly over the last 10 years (RQ3). Cost-benefit analysis is a widely used method for the

economic appraisal of interventions targeting active transport (RQ 3). However, there are

significant limitations in the literature, mostly in terms of the standards of measurement of the

effect of transport interventions on physical activity (RQ 3). In Australia, per kilometre values

for walking and cycling were proposed in the ‘grey literature’ for inclusion of physical activity-

related health in cost-benefit analysis of transport interventions (RQ 4). In this thesis, an

alternative method based on the multi-state life-table modelling approach is described. In this

method, changes in exposure to features of the built environment are directly linked to

monetised health outcomes of physical activity. Lastly, achieving the proposed active-travel

targets would accrue significant health and economic benefits for the city of Brisbane (RQ 5).

Increased walking and cycling could add approximately 33,000 HALYs gained over the life-

course of the Brisbane adult population.

iii

Implications

This thesis demonstrates that investing in built environments that support physical activity can

accrue significant health and economic gains. Hence, Australian governments should support

built environments that facilitate active lifestyles. Currently, decisions about the built

environment are usually made without a full consideration of health outcomes. Typically,

consideration of health effects is limited to road trauma and exposure to poor air quality.

Therefore, decisions are made with an incomplete picture of the social consequences of

investments. In the future, researchers and practitioners should aim to work in collaboration to

ensure that the research community is producing timely, policy-relevant evidence of high

scientific rigour on the potential health outcomes of built environment initiatives. In addition,

equal efforts should be placed to ensure that the evidence is translated into policy and

practice aiming at creating healthier built environments.

iv

Declaration by author

This thesis is composed of my original work, and contains no material previously published or

written by another person except where due reference has been made in the text. I have

clearly stated the contribution by others to jointly-authored works that I have included in my

thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical

assistance, survey design, data analysis, significant technical procedures, professional

editorial advice, and any other original research work used or reported in my thesis. The

content of my thesis is the result of work I have carried out since the commencement of my

research higher degree candidature and does not include a substantial part of work that has

been submitted to qualify for the award of any other degree or diploma in any university or

other tertiary institution. I have clearly stated which parts of my thesis, if any, have been

submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library

and, subject to the policy and procedures of The University of Queensland, the thesis be

made available for research and study in accordance with the Copyright Act 1968 unless a

period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright

holder(s) of that material. Where appropriate I have obtained copyright permission from the

copyright holder to reproduce material in this thesis.

v

Publications during candidature

Peer reviewed papers

1. Nomaguchi, T, Cunich M, Zapata-Diomedi B & Veerman JL, ‘The impact on

productivity of a hypothetical tax on sugar-sweetened beverages’. Health Policy, 2017.

doi: http://doi.org/10.1016/j.healthpol.2017.04.001 (in press)

2. Zapata-Diomedi, B & Veerman, JL 2016, 'The association between built environment

features and physical activity in the Australian context: a synthesis of the literature',

BMC Public Health, vol. 16, no. 1, pp. 1-10.

doi: 10.1186/s12889-016-3154-2

3. Zapata-Diomedi, B, Herrera, AMM & Veerman, JL 2016, 'The effects of built

environment attributes on physical activity-related health and health care costs

outcomes in Australia', Health & Place, vol. 42, pp. 19-29.

doi: 10.1016/j.healthplace.2016.08.010

4. Zapata-Diomedi, B, Barendregt, JJ & Veerman, JL 2016, 'Population attributable

fraction: names, types and issues with incorrect interpretation of relative risks', British

Journal of Sports Medicine. Published Online First: 08 March 2016.

doi: 10.1136/bjsports-2015-095531

5. Veerman, JL, Zapata-Diomedi, B, Gunn, L, McCormack, GR, Cobiac, LJ, Mantilla

Herrera, AM, Giles-Corti, B & Shiell, A 2016, 'Cost-effectiveness of investing in

sidewalks as a means of increasing physical activity: a RESIDE modelling study', BMJ

Open, vol. 6, no. 9.

doi: http://dx.doi.org/10.1136/bmjopen-2016-011617

6. Brown, V, Zapata-Diomedi, B1, Moodie, M, Veerman, JL & Carter, R 2016, 'A

systematic review of economic analyses of active transport interventions that include

physical activity benefits', Transport Policy, vol. 45, pp. 190-208.

doi: http://dx.doi.org/10.1016/j.tranpol.2015.10.003

1 Shared first co-authorship

http://doi.org/10.1016/j.healthpol.2017.04.001

http://dx.doi.org/10.1016/j.tranpol.2015.10.003

vi

Conference abstracts

International

1. Oral presentation, International Conference in Transport and Health-June 2017,

Barcelona, Spain: Zapata-Diomedi B, Knibbs LD, Ware RS, Tainio M, Woodcock J,

Heesch KC, Veerman LJ. A shift from motorised travel to walking, cycling and public

transport: what are the potential health gains for an Australian city? (forthcoming)

2. Oral presentation, International Society of Physical Activity and Health (ISPAH)

conference-November 2016, Bangkok, Thailand: Zapata-Diomedi B, Brown V,

Moodie M, Veerman JL, Carter R. A systematic review of economic analyses of active

transport interventions that include physical activity benefits.

3. Oral presentation, International Conference in Transport and Health-June 2016, San

Jose, California, United States: Zapata-Diomedi B, Veerman JL, Gunn L, McCormack

G, Giles-Corti B, Cobiac L, Mantilla-Herrera AM, Shiell A. Cost effectiveness of

investing in footpaths as a means of increasing physical activity in an Australian city: a

modelling study.

4. Poster presentation, International Conference in Diet and Activity Methods 9-

September 2015, Brisbane, Australia: Zapata-Diomedi B, Brown V, Moodie M,

Veerman JL, Carter R. A systematic review of economic analyses of active transport

interventions that include physical activity benefits.

National

1. Poster presentation, Australian and New Zealand Obesity Society Annual Scientific

Meeting-October 2015, Melbourne, Australia: Zapata-Diomedi B, Veerman JL. The

effects of built environment attributes on physical activity: a systematic review of the

Australian literature.

2. Poster presentation, Charles Perkins Centre symposium on Promoting lifestyle

physical activity and -tackling sedentary behaviour: the need for multidisciplinary

approaches-November 2014, Sydney, Australia: Zapata-Diomedi B, Mantilla Herrera

AM, Barendregt JJ, Veerman JL. A model to estimate physical activity effects on

health.

vii

Local

1. Oral presentation, University of Queensland School of Public Health Research Higher

Degree Conference-November 2016, Brisbane, Australia: Zapata-Diomedi B, Knibbs

LD, Ware RS, Heesch KC, Veerman LJ. A shift from motorised travel to walking,

cycling and public transport: what are the potential health gains for an Australian city?

2. Oral presentation, University of Queensland School of Public Health Research Higher

Degree Conference-November 2015, Brisbane, Australia: Zapata Diomedi B,

Veerman JL. The effects of built environment attributes on physical activity: a

systematic review of the Australian literature.

Reports

1. Zapata-Diomedi B, Brown V, Veerman L. An evidence review and modelling exercise:

the effects of urban form on health: costs and benefits. An evidence review

commissioned by the Centre for Population Health, NSW Ministry of Health, and

brokered by the Sax Institute for The Australian Prevention Partnership Centre; 2015.

http://preventioncentre.org.au/wp-content/uploads/2016/04/1604-urban-form-rapid-

review_to-authors_1-April.pdf

http://preventioncentre.org.au/wp-content/uploads/2016/04/1604-urban-form-rapid-review_to-authors_1-April.pdf

http://preventioncentre.org.au/wp-content/uploads/2016/04/1604-urban-form-rapid-review_to-authors_1-April.pdf

viii

Publications included in this thesis

1. Zapata-Diomedi, B & Veerman, JL 2016, 'The association between built environment



doi: 10.1186/s12889-016-3154-2 - incorporated as Chapter 4.

Contributor Statement of contribution

ZAPATA-DIOMEDI (Candidate) Adapted the research question from the research report prepared for the New South Wales Department of Health: 100%

Designed the methods: 100%

Conducted the review: 100%

Wrote the manuscript: 100%

VEERMAN Advised on the research question: 100%

Advised on the design of methods: 100%

Critically reviewed the manuscript: 100%

Edited the manuscript: 100%

ix

2. Zapata-Diomedi, B, Herrera, AMM & Veerman, JL 2016, 'The effects of built


outcomes in Australia', Health & Place, vol. 42, pp. 19-29.

doi: 10.1016/j.healthplace.2016.08.010 - incorporated as Chapter 5.


ZAPATA-DIOMEDI (Candidate) Adapted the research question from research report prepared for the New South Wales Department of Health: 100%

Updated the methods and prepared the model input data: 90%

Conducted the analyses: 100%

Wrote the manuscript: 100%

HERRERA Provided the epidemiological estimates: 10%

VEERMAN Advised on the research question: 100%

Advised on the design of methods: 100%

Critically reviewed the manuscript: 100%

Edited the manuscript: 100%

x

3. Brown, V, Zapata-Diomedi, B2, Moodie, M, Veerman, JL & Carter, R 2016, 'A



doi: http://dx.doi.org/10.1016/j.tranpol.2015.10.003 - incorporated as Chapter 6.


ZAPATA-DIOMEDI (Candidate) Conceived the research question: 100%

Formulated the methods: 50%

Wrote the methods: 100%

Wrote the results: 100%

Co-edited the abstract: 50%

Co-edited the introduction: 50%

Co-edited the discussion: 50%

Co-edited the conclusion: 50%

BROWN Formulated the methods: 50%

Wrote the abstract: 100%

Wrote the introduction: 100%

Wrote the discussion: 100%

Co-edited the abstract: 50%

Co-edited the introduction: 50%

Co-edited the discussion: 50%

Co-edited the conclusion: 50%

MOODIE Critically reviewed the paper: 100%

VEERMAN Critically reviewed the paper: 100%

CARTER Critically reviewed the paper: 100%



xi

Contributions by others to the thesis

I acknowledge my co-authors who provided invaluable advice for my unpublished

manuscripts: Dr Kristi Heesch, Dr Luke Knibbs, Dr James Woodcock, Dr Marko Tainio, Dr

Lucy Gunn, Professor Rob Ware, Professor Alan Shiell and Professor Billie Giles-Corti.

Statement of parts of the thesis submitted to qualify for the award of

another degree

None

xii

Acknowledgements

This thesis would have not been possible without the financial support from the Australian

Postgraduate Award and the Alan Lopez Public Health Award top-up scholarship. I would also

like to thank to the Centre of Research Excellence in Healthy, Liveable Communities for the

financial support to attend research-related events.

I would like to express my deepest gratitude to my principal advisor Dr Lennert Veerman for

his invaluable support throughout my PhD candidature. I am thankful for his patience, support

and continuous guidance during the process of completing this thesis and related research

activities. I would also like to extend my gratitude to my associate advisers Professor Alan

Shiell (La Trobe University), Associate Professor Jan Barendregt (The University of

Queensland) and Professor Rob Ware (Griffith University). I am grateful to my co-authors who

provided invaluable advice for my manuscripts: Dr Kristi Heesch, Dr Luke Knibbs, Dr James

Woodcock, Dr Marko Tainio, Dr Lucy Gunn and Professor Billie Giles-Corti.

Very importantly, this research would have not been possible without the leadership of

Professor Billie Giles-Corti (RMIT) who is the director of the Centre of Research Excellence in

Healthy, Liveable Communities. I am grateful to Billie for the opportunity to be a contributor to

the Centre’s important work in creating liveable built environments. I also extend my

gratefulness to all member of the Centre for their enthusiasms and good will in progressing

science, and most importantly, making a positive difference to society.

My thanks to all my PhD colleagues at the School of Public Health who enriched my student

experience and provided continuous support, especially to Sarah Blondell, Jessica Bogard,

Preetha Thomas, Caroline Salom, Adnan Choudhury, Shamsir Ahmed and Leopold Aminde.

I am grateful to my parents Alicia Diomedi and Nestor Zapata for their encouragement during

the candidature process and for their lifelong support. Special thanks to my partner Tom

Patterson for his continuous encouragement, understanding and support.

xiii

Keywords

Built environment, transport, land use, physical activity, air quality, road trauma, economic

evaluation, health impact assessment, public health

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 140208 Health economics (50%)

ANZSRC code: 111706 Epidemiology (50%)

Fields of Research (FoR) Classification

FoR code: 1402 Applied Economics (80%)

FoR code: 1117 Public Health and Health Services (20%)

xiv

Table of contents

Chapter 1 Introduction ............................................................................................................. 1

1.1 Background .................................................................................................................... 1

1.2 Aim and research questions .......................................................................................... 3

1.3 Significance .................................................................................................................... 4

1.4 Thesis structure .............................................................................................................. 4

Chapter 2 Literature review ..................................................................................................... 6

2.1 Introduction ..................................................................................................................... 6

2.2 PA and health ................................................................................................................. 8

PA and all-cause mortality ..................................................................................... 9

PA and chronic disease ....................................................................................... 11

2.3 PA in Australia .............................................................................................................. 14

2.4 The BE and PA ............................................................................................................ 17

2.5 The BE and road trauma ............................................................................................. 19

2.6 The BE and ambient air pollution ................................................................................ 20

2.7 Summary ...................................................................................................................... 21

Chapter 3 Overview of health prediction approach used in this thesis ................................ 23

3.1 General methodology .................................................................................................. 24

3.2 Differences with ACE Prevention ................................................................................ 26

Terminology .......................................................................................................... 26

Healthcare costs ................................................................................................... 27

3.3 Comparison with other prediction modelling approaches .......................................... 31

Chapter 4 The association between built environment features and physical activity in the

Australian context: a synthesis of the literature .......................................................................... 32

xv

Chapter 5 The effects of built environment attributes on physical activity-related health and

healthcare costs outcomes in Australia ...................................................................................... 50

Chapter 6 A systematic review of economic analyses of active transport interventions that

include physical activity benefits ................................................................................................. 76

Chapter 7 A method for the inclusion of physical activity-related health benefits in cost-

benefit analysis of built environment initiatives ........................................................................ 107

Chapter 8 A shift from motorised travel to active transport: what are the potential health

gains for an Australian city? ...................................................................................................... 123

Chapter 9 General discussion ............................................................................................. 148

9.1 Introduction ................................................................................................................. 148

9.2 Summary of findings .................................................................................................. 149

9.3 Implications ................................................................................................................ 152

9.4 Directions for future research .................................................................................... 161

9.5 Conclusions ................................................................................................................ 164

Chapter 10 References ......................................................................................................... 166

Chapter 11 Appendix 1: Supplementary material in support of paper in Chapter 4 ........... 201





xvi

List of Figures

Figure 1-1 Conceptual framework CRE in Healthy, Liveable Communities ............................... 3

Figure 2-1 Ecological approach to active living ............................................................................ 7

Figure 2-2 Theoretical framework model ...................................................................................... 8

Figure 2-3 Sufficient PAa age and sex groups, 2011–12 .......................................................... 14

Figure 2-4 Sufficient PA by disadvantage index quintilea, 2011–12 .......................................... 15

Figure 2-5 Contribution of top five risk factors to the total burden of disease and injuries, 2011

.............................................................................................................................................. 16

Figure 3-1 Schematic description of a proportional MSLT ........................................................ 25

Figure 3-2 Disease process model ............................................................................................. 26

Figure 4-1 Summary of included studies .................................................................................... 40

Figure 4-2 Proportion of tested associations for built environment features with sufficient

evidence in the expected direction ...................................................................................... 43

Figure 5-1 Analytical framework of the process of quantifying HALYs and healthcare costs of

changes in exposure to selected built environment attributes. .......................................... 55

Figure 6-1 PRISMA table ............................................................................................................ 83

Figure 6-2 Selected cost benefit ratios by interventiona ............................................................. 98

Figure 7-1 Study framework ...................................................................................................... 113

Figure 7-2 Monetised PA-related health benefits per year per adult living in a neighbourhood

where built environment changes are made (A$ 2016)a .................................................. 119

Figure 8-1 Analytical framework ............................................................................................... 133

Figure 8-2 HALYs by risk factor over the life course of the Brisbane adult population (95%

uncertainty interval) ........................................................................................................... 142

Figure 8-3 Change % on disease prevalence and mortality over the life course of the Brisbane

adult population (error bars indicate the 95% uncertainty interval) .................................. 143

xvii

List of Tables

Table 2-1 Summary of the literature for PA and all-cause mortality .......................................... 10

Table 2-2 Relative risks for PA and chronic diseases................................................................ 13

Table 2-3 Sufficient PA measure by health characteristics (% persons over 18 years old) .... 16

Table 3-1 Differences in prediction modelling studies ............................................................... 23

Table 3-2 Comparison healthcare cost of physical inactivity diseases (A$ 2013) .................... 28

Table 4-1 Inclusion criteria .......................................................................................................... 35

Table 4-2 Categorisation of built environment attributes ........................................................... 37

Table 4-3 Summary of associations between built environment attributes and physical activity

.............................................................................................................................................. 44

Table 5-1 Proportional multi-state life table inputs ..................................................................... 59

Table 5-2 Uncertainty parameters for evaluation health effects ................................................ 60

Table 5-3 Univariate sensitivity analysis ..................................................................................... 61

Table 5-4 Built environment attributes modelled ........................................................................ 62

Table 5-5 HALYs, healthcare costs savings and all other healthcare costs in added life years

per 100,000 people per year for built environment scenarios ............................................ 70

Table 6-1 Methods for full economic evaluation ......................................................................... 80

Table 6-2 An overview of included studies as per the CHEERS guidelines for quality of

reporting ............................................................................................................................... 85

Table 6-3 Interventions included in the review ........................................................................... 88

Table 6-4 Other non-PA benefits/disbenefits included in the cost-benefit analyses ................. 96

Table 7-1 Comparison of included outcomes in per kilometre estimates ............................... 120

Table 8-1 Mode-specific mean (95% Uncertainty Interval (UI)) trips per weekday in 2009, by

age and sex ....................................................................................................................... 129

Table 8-2 Mode share travel targets ......................................................................................... 131

Table 8-3 Proportional multi-state life table Markov model input parameters ......................... 134

Table 8-4 Percentage of trips made by distance travelled and transport mode, for baseline and

travel target scenarios ....................................................................................................... 139

Table 8-5 Mean trips per week (weekdays only) for baseline and travel targets scenarios, by

age and sex ....................................................................................................................... 139

Table 8-6 Additional mean minutes per week of transport physical activity undertaken in the

travel targets scenario compared to the baseline scenario (statu-quo), by age and sex 140

Table 8-7 Road trauma rates per 100 million kilometres travelled by transport mode ........... 140

xviii

Table 8-8 PM2.5 values baseline and sensitivity scenarios ...................................................... 141

Table 8-9 Healthcare costs and health outcomes for base case by sex over the life course of

the Brisbane adult population (95% uncertainty interval) ................................................. 142

Table 8-10 Change in prevalent cases and mortality over the life course of the Brisbane adult

population (95% uncertainty interval) ................................................................................ 143

xix

List of abbreviations

ABS Australian Bureau of Statistics

ACE Prevention Assessing Cost Effectiveness in Prevention

AIHW Australian Institute of Health and Welfare

AP Air pollution

BCC Brisbane City Council

BCR Benefit cost ratio

BE Built environment

BoD Burden of Disease

CBA Cost Benefit Analysis

CCD Census Collection District

CEA Cost Effectiveness Analysis

CEDAR Centre for Diet and Physical Activity

CHEERS Consolidated Health Economic Evaluation Reporting Standards

CI Confidence interval

CO2 Carbon dioxide

CRA Comparative Risk Assessment

CRE CRE Healthy, Liveable Communities

CUA Cost Utility Analysis

CURF Confidentialised Unit Record File

DALY Disability-adjusted life year

mailto:CO@

xx

GBD Global Burden of Disease study

HALY Health-adjusted life year

HEAT Health Economic Assessment Tool

IHME Institute of Health Metrics

ITHIM Integrated Transport and Health Impact Model

LTPA Leisure time physical activity

LUM Land use mix

MET Metabolic equivalent of task

MSLT Multi-state life table

NICE National Institute for Health and Care Excellence

NHRMC National Health and Medical Research Council

NNPA Nutrition and Obesity Policy and Evaluation Network

NOPREN Prevention Research Centre on Nutrition and Physical Activity

NZTA New Zealand Transport Agency

OLS Ordinary least squares

PA Physical activity

PIF Potential Impact Fraction

PM2.5 Particulate matter (PM) with an aerodynamic diameter smaller than 2.5 microns

PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses

QALY Quality-adjusted life year

xxi

QLD Queensland

RESIDE RESIDential Environment Study

RR Relative risk

RT Road trauma

SA South Australia

SIN Safety in numbers

THE PEP The Transport, Health and Environmental Pan-European Programme

TMR Department of Transport and Main Roads

UI Uncertainty Interval

UK United Kingdom

US United States

VSL Values of Statistical Life

VSLY Value of Statistical Life Year

WA Western Australia

WHO World Health Organization

YLD Years lived with disability

YLL Years of life lost

xxii

Glossary

6 D's Attributes of the built environment that influence physical

activity behaviours and start with the letter D (density,

diversity, design, destination accessibility, distance to

transit, demand management)

ACE Prevention ACE Prevention was a 5-year study funded by the

National Health and Medical Research Council (NHMRC)

investigating the cost-effectiveness of 123 disease

prevention and 23 treatment interventions.

Active travel Form of transport that requires physical activity of the

human being (e.g. walking, cycling)

Built environment Physical features of urban spaces, including

transportation systems and urban planning

Burden of disease Measure of the impact of disease and injury on population

health and the economy. Burden of disease studies

quantify outcomes in terms of disability-adjusted life

years, years lived with disability, years of life lost, deaths,

prevalence, incident cases (e.g. Global Burden of Disease

study).

Chronic disease Long-lasting human health condition

Comparative risk

assessment

Methods developed by the World Health Organization to

systematically assess changes in population health of

changes in exposure to risk factors

Cost-benefit analysis Type of economic evaluation where the expected benefits

of an intervention are measured in monetary terms and

compared to the costs of the intervention. Results are

reported as cost per unit of benefit.

xxiii

Cost-effectiveness analysis Type of economic evaluation where health outcomes are

expressed as a unit of effect, for example life years saved

or prevalent cases averted with an associated cost.

Results can be presented as cost per life year saved or

prevalent cases averted.

Cost-utility analysis Type of economic evaluation where the expected health

outcomes of an intervention are measured in terms of the

quality and quantity of life attributable to the intervention.

Health outcomes can be expressed as disability adjusted

life years (DALYs) or quality adjusted life years (QALYs).

Results can be presented as cost per averted DALY or

gained QALY.

Disability-adjusted life

years

Population health measured created by the World Health

Organization for burden of disease studies. Disability-

adjusted life years as the sum of years lived with disability

and years of life lost

Economic evaluation Systematic study for the optimal allocation of scarce

resources for the maximisation of benefit to society

Health-adjusted life years Population health measure that includes measures of

mortality and morbidity. One health-adjusted life year is a

year of life adjusted for quality of life or disability due to

morbid states.

Health impact assessment Systematic approach of the potential health impacts of

public or private initiatives using quantitative or qualitative

techniques

MET One MET is equivalent to the metabolic rate of an

individual at rest and is equal to 3.5 ml O2/kg per minute

(approximately 1 kcal/kl per hour).

xxiv

Multi-state multi-cohort life

table

Life table that includes multi-disease states and age and

sex cohorts

Physical activity Any bodily movement produced by skeletal muscles that

results in energy expenditure

Physical inactivity Typically used to refer to the level of physical activity

below national or international guidelines for physical

activity

Population impact fraction Measure of the change in disease risk attributable to a

change in exposure to a related risk factor

Quality-adjusted life year Measure of disease burden accounting for the quantity

and quality of life. QALYs are measured using states of

health (utility) on a scale from 1 (perfect health) to 0

(death).

Years lived with disability Years lived with disability are estimated as the number of

prevalent cases by the corresponding disability weight.

Past Burden of Disease studies used to calculate years

lived with disability as incident cases multiplied by

average duration of the condition by the associated

disability weight of the condition.

Years of life lost Years of life lost due to premature death

1

Chapter 1 Introduction

1.1 Background

The World Health Organization (WHO) estimates that approximately 23% of the world’s adult

population qualified as physically inactive in 2010 (less than 150 minutes of moderate to

vigorous physical activity per week) (1). This presents a public health challenge, as physical

inactivity has been recognised as an important risk factor for chronic conditions and all-cause

mortality (2, 3). Physical inactivity contributes to 6%, 7%, 10% and 10% respectively of the

burden of coronary heart disease, type 2 diabetes, breast cancer and colon cancer (4). In

addition, it has also been related to mental health conditions (5), weight gain (6) and falls in

older people (7). A comprehensive public health approach that goes beyond the healthcare

sector is needed to tackle the increasing burden of physical inactivity (8). Calls for a multi-

sectoral approach have been increasing with the realisation that individual-based interventions

within the health sector have several limitations. Hence, ecological solutions should be sought

to provide effective answers to reduce the prevalence of risk factors, such as physical

inactivity (9).

The built environment (BE), defined as those elements of the environment that are man-

made, including transportation systems, urban planning and individual buildings (10 p28), has

been drawing increasing attention as to its effect on health-risk factors, including physical

inactivity. The BE includes what many see as two different fields, planning and transport.

There is a growing body of evidence that indicates that certain features of the BE may

facilitate or impede physical activity (PA) (11-14). Creating PA-friendly environments would

also produce co-benefits such as environmentally sustainable transportation options that

could lead to reductions in air pollution and road trauma (15, 16). Urban designs that promote

walking and cycling may also encourage social capital by enabling incidental interactions

among people (17). Furthermore, it has been suggested that increased walking and cycling

generates natural surveillance and therefore safety (18). However, in highly motorised cities,

health trade-offs—ranging from greater exposure to motor-vehicle emissions to road trauma—

can arise from urban designs that promote active travel (19).

There is consensus internationally that the BE has a large impact on health and wellbeing.

One of the most important and promising acknowledgements of the importance of suitable

environments for health is the Ottawa Charter for Health Promotion (20). The creation of

supporting environments for health was one of the five actions specified to build healthy public

2

policy. Further development occurred thereafter: the Jakarta Declaration (1997), Health for All

Targets (1997), Health 21 (1999), the Bangkok Charter (2005), the Nairobi Call for Action

(2009), the Helsinki Statement of Health in all Policies (2014) and the Bangkok Declaration on

Physical Activity for Global Health and Sustainable Development (2016). All of them included

recommendations to government leaders for the inclusion of health in all policies, including the

transport and planning sectors, with the aim of achieving a multi-sectoral approach to public

health. The formation of the Alliance for Healthy Cities initiated by the WHO is an informative

case study that highlights the increasing commitment by governments, non-government

organisations, and the private sector to create and maintain BEs that are conducive to healthy

living (21). The role of the BE for health and wellbeing is also part of the United Nations 2016

Sustainable Development Goals (Goal 11) (22). For example, expanding public transport, with

special attention to accessibility, safety and sustainability are emphasised with the aim of

improving inclusion of vulnerable groups, women and children. Sustainable urban designs that

facilitate walking and cycling for improved health and wellbeing are also explicitly mentioned in

the outcome documents of the most recent (2016) United Nations HABITAT meeting (23).

In keeping with these international developments, in Australia, governments at all levels

(national, state and local) and non-government bodies are paying increasing attention to the

important role of transport and planning for PA (24-29). However, decisions within the BE

sector are usually made without a full appraisal of health impacts, including those from PA

(30). An important advancement towards a comprehensive assessment of BE interventions is

the creation of the Centre of Research Excellence in Healthy, Liveable Communities (CRE)

funded by the Australian National Health and Medical Research Council (NHRMC) (31). The

aim of the CRE is ‘to be a source of high-quality and policy-relevant research that informs

healthy urban designs and planning’ (31). The CRE research program is divided into five

themes (Figure 1-1), working over a period of five years (2013–18, A$2.5 million). The

structure of the CRE was set up to investigate the association of BE features with health and

well-being outcomes (Themes 1–3), the economic merit of BE interventions (Theme 4) and

the translation into policy and practice (Theme 5).

The research conducted for this thesis contributes to Theme 4 ‘Economic evaluations of BE

interventions for health, liveable and equitable communities’. Given that all themes started

working simultaneously, the research conducted in this body of work is independent of other

themes presented in the CRE conceptual framework. However, it serves to inform Theme 5.

Another important contribution led by the author of this thesis to the field in the Australian

3

context was a project funded by the Centre of Population Health of the New South Wales

Department of Health (A$36,000, 2015) to summarise and quantify the evidence on cost and

benefits of urban form (built environment) (32). The centre aimed to have readily available

evidence to support their work with the departments of transport and planning for the

incorporation of PA-related health in cost-benefit analysis.

Figure 1-1 Conceptual framework CRE in Healthy, Liveable Communities

Source: University of Melbourne. NHMRC Centre of Research Excellence in Healthy, Liveable Communities (31). Available from: http://mccaughey.unimelb.edu.au/programs/cre. The graph is from the grant proposal.

HWB: Health and wellbeing

1.2 Aim and research questions

The overall aim of this thesis is to contribute to the evidence, in the Australian context, of

potential benefits and harms to health, and the economic merit, of a more active adult

population achieved by BE interventions and policies. The aim is to be accomplished by

addressing five research questions:

RQ 1: What are the attributes of the built environment in Australia that most benefit physical

activity?

http://mccaughey.unimelb.edu.au/programs/cre

4

RQ 2: What are the physical activity-related health externalities and healthcare costs


RQ 3: What economic evaluation methods have been used to model future health outcomes

from interventions in active transport?

RQ 4: Can the health impact of changes in physical activity be incorporated more robustly in

cost-benefit analysis of built-environment initiatives in Australia?

RQ 5: What are the potential health and economic impacts of Brisbane meeting its targets for

active travel?

1.3 Significance

Health benefits and harms and economic outcomes from changes in PA attributable to the BE

are not part of the routine evaluation of BE interventions. As a result, decisions are made with

an incomplete picture of the potential of the BE for population health and the economy. This

may result in suboptimal decisions from a societal perspective because health is undervalued.

Generating evidence for Australian-specific BE correlates of PA, as well as the potential health

outcomes of PA-friendly environments, is fundamental to supporting informed decision-

making on BE initiatives. Moreover, quantifying the potential societal benefits of government

policies for greater levels of walking, cycling, and use and of public transport serves to support

continuity of investment. Prior to this thesis, this was a significant knowledge gap in Australia.

There are a high number of individual studies for BE correlates of PA; however, there has

been no attempt to systematically summarise the evidence. Quantification of health and

economic outcomes of BE policies is also scarce for the Australian setting. The studies in this

thesis addressed these knowledge gaps and produced policy-relevant research, enabled

through collaborative work with government and academic partnerships.

1.4 Thesis structure

This thesis consists of fifteen chapters, starting with an introductory chapter (Chapter 1).

Chapter 2 consists of a literature review that served to identify the significance of this thesis

and address knowledge gaps. The following broad topics were reviewed: PA and health and

PA in the Australian setting, evidence of the association of the BE with PA and road trauma

and air pollution. Chapter 3 is an overview of the prediction modelling methods for the

assessment of BE initiatives used in this thesis. Chapter 4 to 8 are based on journal articles

5

that have been published, are under review or have been submitted. Each chapter based on a

journal article is an individual study with an introduction that includes a brief review of the

literature, methods, results and discussion. Papers based on the work in Chapter 4 to 6 have

been published in peer-reviewed journals and the formatting in the corresponding chapters is

as per the general formatting of this thesis. References for chapters 1 to 9 are placed at the

end of the document before the appendices. Chapters 4 to 8 have extensive supplementary

materials. For published research, these are available as links to the corresponding journal

webpages (Chapters 11 to 13). For unpublished research, these are available in the

Appendices of this thesis (Chapters 14 to 15). Chapter 4, 5 and 7 are related. Chapter 5 builds

on the work presented in Chapter 4, and Chapter 7 on the work in Chapter 5. Chapter 5, 7

and 8 have modelling methods in common; however, the model used in Chapter 8 is more

comprehensive as it adds road trauma and air pollution to the base version, which only

includes PA. Some repetition in the introductions of the manuscripts included in the chapters

(4 to 8) was inevitable as they share a common theme. Furthermore, due to the commonality

of methods for Chapter 5, 7 and 8, there is repetition in the supplementary materials and

methods sections of these manuscripts. Chapters consisting of a journal article (Chapter 4 to

8) start with a brief introduction explaining the contribution the study makes to achieving the

overall aim of this thesis and a corresponding research question, declaration of copyrights,

authors’ contribution statement and a link to publication (only for published work). Chapter 9

contains a summary of the findings in this thesis, a discussion of the implications of the

findings, and recommendations for future research.

6

Chapter 2 Literature review

2.1 Introduction

The publication of the special issue Built environment and health in the American Journal of

Public Health in 2003 marked a renewed interest in collaborative work between the health and

urban planning sectors to tackle common risk factors for chronic diseases (33). In Australia,

the Preventive Health Taskforce specifically addressed the BE by recommending

‘environmental changes throughout the community to increase levels of physical activity (PA)

and reduce sedentary behaviour’ (34 p13). BE interventions with an impact on population

levels of PA such as building sidewalks, bikeways and providing convenient public transport

are long lasting and appealing compared with the short-term impacts of individual-level

initiatives (35).

In 2006, Sallis and colleagues proposed an ecological framework to emphasise the interacting

factors determining population PA levels, including the BE (36). Sallis et al. proposed a

multilevel approach to achieve a positive change in PA (Figure 2-1). For instance, the policy

environment determines active travel behaviours through policies such as zoning and land

use and transport investment. Those policy decisions have an impact on the layout of

neighbourhoods that may or may not facilitate transport walking and cycling3 and use of public

transport. Additionally, available information has a considerable impact on perceptions of the

level of safety, and thus influences individuals’ decisions on active commuting. Furthermore,

natural characteristics of the environment such as topography, weather and air quality need

consideration as they affect active-travel behaviour. Lastly, all the above-mentioned

environmental factors impact on the personal perceptions of the environment and, together

with personal characteristics, define PA behaviours.

3 Walking and cycling with the purpose of getting from one place to another.

7

Figure 2-1 Ecological approach to active living

Source: Sallis et al. (36)

This thesis addressed PA in the ‘active transport’ domain and to a lesser extent in the ‘active

recreation’ domain of active living (Figure 2-1), and contributes to the development of methods

to predict the potential health and economic outcomes of changes in BE attributes that have

an impact on population levels of PA and scenarios in which travel patterns change. In this

thesis, the WHO definition of the BE is adopted, which includes transportation systems and

urban planning but excludes individual buildings (10 p28). The BE, therefore encompasses

land use managed by urban planners and transport infrastructure including roads, local

streets, bike lanes, sidewalks, etc. managed by transport engineers and planners. The

following theoretical framework is the base for the developments in this body of work (Figure

2-2). It resembles the framework proposed by Theme 4 of the CRE in Healthy and Liveable

Communities (please refer to Figure 1-1). The framework adds two risk factors to PA:

exposure to ambient air pollution and road trauma. Air pollution and road trauma may partially

offset the potential health benefits of BE interventions with the aim of improving population

levels of PA (37, 38). This thesis focuses on the adult population (18+ years).

8

Figure 2-2 Theoretical framework model

The remainder of this chapter is structured as follows. Section 2.2 is a summary of research

for the association of PA with health outcomes. Section 2.3 presents information regarding PA

and the burden from physical inactivity in the Australian context. Section 2.4 presents studies

of the association of the BE with PA. Section 2.5 and 2.6 are dedicated to discussing the role

of the BE in road trauma and air pollution, and how these exposures impact on health.

Section 2.7 is a brief overview of prediction studies of BE interventions. The last section is a

summary of identified gaps in the literature.

2.2 PA and health

PA has been recognised as an important protective factor for a number of chronic diseases as

well as premature mortality. The United States Surgeon General’s report released in 1996

was a landmark review of epidemiological research of PA and health (39). The findings from

this report indicated that moderate levels of PA (e.g. 30 minutes of brisk walking on most

Δ Built Environment/Transport targets

Δ Exposure risk factors

Physical activity

Road trauma

Air pollution (particulate matter)

Road traffic injuries/ fatalities

Ischemic heart disease

Breast cancer

Diabetes

Ischemic stroke

Colon cancer

Chronic obstructive pulmonary disease

Lung cancer

Summary measures of population health

Economic outcomes

Other health conditions

9

week days) reduces the risk of all cause-mortality and the risk of developing cardiovascular

diseases, type 2 diabetes, and colon cancer. Furthermore, being physically active was found

to improve mental health and the health of muscles, bones and joints. Higher levels of PA

were associated with greater health gains.

In general, government health departments and international organisations recommend 150

minutes per week of moderate PA or equivalent, distributed among most weekdays for adults

(40-42). In Australia, at least 150 minutes of moderate intensity PA, 75 minutes of vigorous PA

or a combination of both, distributed among most weekdays is recommended for health and

wellbeing (41). In addition, the guidelines state that doing any PA is better than none. Muscle

strengthening activity at least twice per week is also recommended.

National statistical offices often present population PA-prevalence estimates in categories

such as sedentary, inactive, sufficiently active and highly active—depending on the level of PA

achieved (43). Such categorisation has been used in the majority of epidemiological research

to assess the risk of mortality and morbidity associated with each category. I will now present

a summary of evidence from meta-analysis studies, associating PA with all-cause mortality

and disease-specific mortality, and risk of chronic diseases.

PA and all-cause mortality

Table 2-1 is a summary of relative-risk estimates for the relationship of PA with all-cause

mortality from recent meta-analyses. It is noticeable that tests of associations for walking,

cycling or domain-specific PA (transport and occupational) indicate lower risk reductions

compared with those including more comprehensive exposures (3, 44, 45). Also, the studies

by Woodcock et al. and Kelly et al. indicate that the greatest health gains are achieved at low

levels of PA, with decreasing reductions in the risk of all-cause mortality for higher levels of PA

(3, 44). Lastly, the results by Samitz et al. suggest that perhaps, greater benefits from

reduction in mortality are achieved by doing higher intensity PA (exercise and sport vs PA for

transport and occupation ) (45).

10

Table 2-1 Summary of the literature for PA and all-cause mortality

Study Exposure Outcomes Resultsa

Woodcock et al. (2011) (3)

Non-vigorous (moderate, light for older adults) PA, walking, cycling and active commuting

All-cause mortality

Reduction in all-cause mortality for: Moderate PA

• 2.5 h/wk compared with no activity: 19% (95%

CI 15 to 24).

• 7 h/wk of moderate and light PA compared

with no activity: 24% (95% CI 19 to 29) and 22% (95% CI 17 to 26) Walking

• 2.5 h/wk compared with no walking: 11%

(95% CI 4 to 12) Active commuting (excludes walking and cycling for other purposes)

• Only assessed in two studies and these did

not indicate an association with the probability of dying from all causes Cycling

• Two studies were found, with one indicating a

significant association and the other no association with all-cause mortality.

Samitz et al. (2011) (45)

Total PA: leisure-time PA, routine activities of daily living and occupational PA Leisure time with routine activities of daily living: leisure-time PA and one or more components of PA of daily living Leisure time PA: recreational activities including callisthenics, dancing, walking, hiking, golf, bicycling, swimming, games, exercise and sports Exercise and sports: structured aerobic and muscle-strengthening exercise and sports PA of daily living: non-exercise activities including housework, gardening, stair climbing, walking and cycling as part of daily life PA for transport: walking/cycling to and from work Occupational PA: PA as part of work

All-cause mortality

Reduction in all-cause mortality forb: Total PA

• Comparing the last and most active groups:

35% (95% CI 29 to 40) Leisure time with routine activities of daily living


36% (95% CI 29 to 39) Leisure time PA


26% (95% CI 23 to 30) Exercise and sports


34% (95% CI 29 to 39) PA of daily living


36% (95% CI 25 to 45) PA for transport


12% (95% CI 2 to 21) Occupational PA


17% (95% CI 3 to 29)

Kelly et al. (2014) (44)

Walking and cycling All-cause mortality

Reduction in all cause motility for: Walking:

• For about 168 min/wk of walking compared

with no walking: 11% (95% CI 4 to 17) Cycling

• For about 100 min/wk of cycling compared

with no cycling: 10% (95% CI 6 to13) Sensitivity analyses indicated that the greatest benefits are achieved for the first increases in walking and cycling, with decreasing benefits for further walking and cycling

CI = confidence interval; wk = week; h = hours; min = minutes

11

a Results are adjusted for common confounders (e.g. age, cigarette smoking, body mass index, diabetes mellitus, lipid factors, alcohol consumption, socioeconomic status, marital status) and PA in other domains. b Source study also presents results for one hour increases in PA.

PA and chronic disease

Table 2-2 consists of a summary of the associations of PA with chronic diseases. Studies

have mostly focused on ischemic heart disease, ischemic stroke, type 2 diabetes, colon

cancer and breast cancer (46-48). These diseases have been chosen because of the

biologically plausible mechanisms to explain the associations and well-stablished evidence of

causality (46 p798), plus their importance for population health. Most of the evidence is

assessed for categorical exposures to PA, as opposed to continuous exposures.

Bull and colleagues conducted one of the first comprehensive systematic reviews and meta-

analyses of studies assessing PA with the risk of chronic diseases in 2004 (46) (Table 2-2).

Exposure to PA was treated as a categorical variable consisting of inactive (none or very little

PA), insufficiently active (up to 150 minutes of moderate-intensity PA per week or 60 minutes

of vigorous-intensity) and unexposed (greater than 150 minutes of moderate-intensity PA per

week or 60 minutes of vigorous-intensity). Overall PA was defined as PA in all four domains

(transport, recreation, occupation and domestic). However, domestic PA was excluded due to

lack of data. Generally, the results indicated reductions in the probability of developing chronic

diseases for those shifting from inactive to unexposed. Associations were weaker when

comparing the reduction in probability of developing chronic diseases from insufficiently active

to unexposed. In a later study in 2009, Danaei and colleagues rescaled Bull et al.’s results to

four categories: inactive, low active, moderately active and highly active (47) (Table 2-2).

Results remain consistent to those in Bull et al., with Danaei et al.’s results indicating greater

gains for a shift from inactive to highly active.

There are many challenges in summarising the evidence for the association of PA with health

outcomes (46, 48). Instruments used to measure and classify PA are heterogeneous. For

example, some studies use continuous measures (e.g. hours per week), while others use

categorical variables (active vs inactive). Categorisation of exposure also varies greatly

between studies, mostly ranging from two to four categories. Furthermore, a wide range of

instruments are available for measuring PA, with some only capturing PA for specific domains

(transport, recreational, domestic or occupational). Kyu and colleagues addressed these

limitations in a recent study of systematic reviews and meta-analyses for the dose-response of

PA and the risk of ischemic heart disease, ischemic stroke, colon cancer, breast cancer, and

12

type 2 diabetes (48) (Table 2-2). Compared with past research, three to five times more

prospective cohort studies were included. Total PA was the exposure variable, which was

homogenised across studies applying complex mapping methods and translated to MET4

minutes of activity per week. Kyu et al. findings are in agreement with past studies, indicating

greater gains at lower levels of PA. However, cut-offs are much higher compared with past

studies due to the inclusion of PA in all domains.

PA has also been associated with mental health outcomes. Cross-sectional evidence to date

indicates that PA may have a protective effect for the development of depressive disorders

and anxiety (49, 50). However, relying on cross-sectional studies would imply making an

assumption about the direction of the association (51). Longitudinal studies show a bi-

directional association or no associations (52, 53). While PA may not be causally associated

with depression, evidence from a meta-analysis indicates that it may prevent the development

of cognitive decline and dementia (54).

4 One MET is equivalent is the metabolic rate of an individual at rest and is equal to 3.5 ml O2/kg per minute

(approximately 1 kcal/kl per hour) [50]

13

Table 2-2 Relative risks for PA and chronic diseases

Study Outcome Exposure Age and sex group

Relative risks (95% Confidence Interval)


Ischemic stroke Type 2 diabetes Colon cancer Breast cancer

Bull et al. (46)a Incidence and mortality

Inactiveb

30–44, femalesd - - - - 1.25 (1.20 to 1.30)

45–69, females - - - - 1.34 (1.29 to 1.39)

30–69, males & females 1.71 (1.58 to 1.85) 1.53 (1.31 to 1.79) 1.45 (1.37 to 1.54) 1.68 (1.55 to 1.82) -

70–79, males & females 1.50 (1.38 to 1.61) 1.38 (1.18 to 1.60) 1.32 (1.25 to 1.40) 1.48 (1.36 to 1.60) 1.25 (1.21 to 1.30)

≥ 80, males & females 1.30 (1.21 to 1.41) 1.24 (1.06 to 1.45) 1.20 (1.14 to 1.28) 1.30 (1.20 to 1.40) 1.16 (1.11 to 1.20)

Insufficiently active

30–44, femalesd - - - - 1.13 (1.04 to 1.22)

45–69, females - - - 1.13 (1.04 to 1.22)



≥ 80, males & females 1.20 (1.07 to 1.35) 1.05 (0.85 to 1.30) 1.11 (0.99 to 1.25) 1.08 (0.97 to 1.22)

Danaei et al. (47)c

Mortality

Inactive

30–44, femalesd - - - - 1.56 (1.30 to 1.87)

45–69, females - - - - 1.67 (1.44 to 1.94)




Insufficiently active

30–44, femalesd - - - - 1.41 (0.84 to 2.36)

45–69, females - - - - 1.41 (0.84 to 2.36)

30–69, males & females 1.66 (1.14 to 2.42) 1.23 (0.41 to 3.67) 1.50 (0.90 to 2.50) 1.27 (0.86 to 1.87)



Recommended level active

30–44, femalesd - - - - 1.25 (0.90 to 1.74)

45–69, females - - - - 1.25 (0.90 to 1.74)




Kyu et al. (48)f Incidence

< 600g All ages, males & females 1.25 (1.19 to 1.30) 1.26 (1.19 to 1.34) 1.28 (1.23 to 1.32) 1.21 (1.15 to 1.27) 1.14 (1.10 to 1.17)

600–3999 All ages, males & females 1.23 (1.16 to 1.30) 1.19 (1.06 to 1.31) 1.25 (1.20 to1.30) 1.17 (1.10 to 1.23) 1.06 (1.02 to 1.10)

4000–7999 All ages, males & females 1.16 (1.11 to 1.21) 1.16 (1.08 to 1.22) 1.14 (1.10 to 1.18) 1.10 (1.05 to 1.15) 1.03 (1.00 to 1.06) a Relative risks adjusted for confounding variables (e.g. age, sex) and measurement error but not for intermediate variables (blood pressure and cholesterol) b Compared with reference group unexposed (< 150 min of moderate-intensity PA/wk or 60 min of vigorous-intensity) c Confidence intervals were not reported in the source document, authors provided standard errors and confidence intervals were estimates as mean±1.96*SE. d For all diseases except breast cancer, > 69 was categorised as 30–69, for breast cancer in women the category was split into 30–44 and 45–69 e Compared with reference group highly active = > 1 h/wk of vigorous activity and => 1,600 MET-min/wk f Reporting for categorical dose-response. Change reference category from < 600 to ≥ 8000 (unexposed) for consistency of reporting with other RRs g MET-min/wk

14

2.3 PA in Australia

In Australia, like in other high income countries, the low level of PA observed in the adult

population is a public health concern (55, 56). National health statistics based on self-

report measures showed that less than 50% of adults (≥18 years old) achieved the

recommended PA level to maintain good health and wellbeing in 2011–12 (57). When

stratified by age groups, young males (18–24 years old) showed a higher proportion

achieving recommended levels compared with females of the same age and older age

groups; however, meeting guidelines is still low across the whole age spectrum in both

sexes (Figure 2-3).

Figure 2-3 Sufficient PAa age and sex groups, 2011–12

a Sufficient PA: at least 150/75 minutes of moderate/vigorous physical activity on most days Source: Australian Bureau of Statistics (57)

National estimates also indicate that in 2011–12, there was a significant difference in the

proportion of people reporting sufficient PA when stratified by a national index that

measures disadvantage (Figure 2-4). While half of the least disadvantaged (fifth quintile)

reported doing at least 150/75 minutes of moderate/vigorous PA or a combination of

both per week, only one-third in the most disadvantaged group met the PA threshold. As

15

well, those living in urban areas were more likely to be PA compared with their

counterparts in regional and remote Australia (45% vs 36% and 39%).

Figure 2-4 Sufficient PA by disadvantage index quintilea, 2011–12

a Index of Relative Socio-Economic Disadvantage: relative index that summarises information about the economic and social conditions of people and household within an area. The first quintile represents the greatest relative disadvantage, and the fifth quintile the greatest lack of relative disadvantage [54]. Source: Australian Bureau of Statistics (57)

There is a robust body of evidence that supports that PA people are at lower risk of

premature mortality and risk from chronic diseases (see Section 2.2). In Australia in

2011, physical inactivity was ranked among the top five risk factors responsible for the

total burden of diseases (58) (Figure 2-5). Furthermore, a physically inactive population

represents an economic burden. For instance, chronic diseases, including those

associated with low levels of PA, consumed a significant proportion of the health budget

(e.g. cardiovascular disease) (59). Also, Australians suffering from chronic conditions

are more likely to be absent from their jobs or not participate in the workforce compared

with their healthy counterparts (60). Another important implication of individual levels of

PA is a person’s rating of overall health and wellbeing (57). In 2011–12, 57% of adults

who rated their health as ‘excellent’ also reported doing sufficient PA, compared with

27% of adults reporting ‘fair’ and 26% indicating ‘poor’ health (Table 2-3).

16

Figure 2-5 Contribution of top five risk factors to the total burden of disease and injuries, 2011

Source: Author compilation from online dataset (58)

Table 2-3 Sufficient PA measure by health characteristics (% persons over 18 years old)

Self-reported health status Inactivea Insufficiently activeb

Sufficiently active for healthc

Excellent 12.4 30.5 56.6

Very good 17.0 33.8 47.8

Excellent/Very good 15.5 32.7 50.6

Good 20.7 40.4 38.0

Fair 32.4 39.3 27.3

Poor 46.4 27.5 25.8

Fair/Poor 36.0 36.2 26.9

Total 20.3 35.7 43.0 a No walking, moderate or vigorous intensity PA b Some activity* but not enough to reach the levels required for 'sufficiently active' c 150/75 minutes of moderate/vigorous PA on most days Source: Table 6.3 Australian Bureau of Statistics (57)

In Australia, there are opportunities to increase population levels of PA by creating BEs

that support active travel (walking, cycling and public transport) and recreation. Nearly

80% of trips to work in Australia are in motor vehicles (61). Therefore, building

supportive BEs for active transport is essential to facilitate PA and change the current

trend of high use of motor vehicles. There is also considerable scope to increase PA for

leisure purposes. In 2009–10, only 30% of Australian adults reported participating in

sports and leisure time PA at least twice per week in the last year (62). Of those who

17

reported participation, the most commonly used facilities were parks, walking trails, and

beaches to do unstructured exercise.

In sum, a high proportion of Australian adults do too little PA to accrue health benefits.

Despite some sociodemographic and geographical differences, overall the trend is

consistent across the country. Improving the BE to facilitate PA is a feasible path to

contribute to a more active Australia. International evidence from health prediction

models suggest that BE initiatives that support PA can accrue important health benefits

(37, 63). However, the evidence also suggests negative health externalities may arise

from interventions that increase the uptake of active transport. Hence, for Australia, low

participation in active travel and leisure time PA in public spaces indicates that

investments in infrastructure and programs to support the use of facilities may contribute

to increased PA participation. The main pathways from the BE to health (exposure to

PA, road trauma and air pollution) are investigated below.

2.4 The BE and PA

A large number of studies have investigated cross-sectional and longitudinal

associations of a range of BE features and PA outcomes, including reviews (11) and

meta reviews (64).

A recent example of research for the association BE-PA is based on the International

Physical Activity and the Environment Study, which is one of the largest research efforts

investigating BE and PA across a range of countries (12). For this study, Sallis et al

combined objectively measured BE and PA data for 14 cities across 10 countries

(including high-, middle- and low-income countries) to assess cross-sectional

associations within 0.5 and 1.0 km buffer areas from residents’ homes. High residential,

intersection and public transport density, and number of parks were determinants of

overall PA. Diversity in the uses of land (e.g. commercial, residential, and industrial) and

proximity to the closest public transit stop were not significantly related to PA. An earlier

study in the United States by Frank and colleagues estimated objectively measured BE

and PA associations (within the neighbourhood area) in Atlanta (65). Similar to Sallis et

al., they found that residential density and intersection density were determinants of

overall PA. Contrary to Sallis et al.’s results, land-use mix was not associated with PA.

In spite of the accuracy of measurement for the exposure and outcome variables in the

18

studies by Sallis et al. and Frank et al., they rely on cross-sectional designs which, on

their own, do not allow establishing causation.

Quasi-experiments and natural experiments that control for potential confounders are

suggested as rigorous approaches to establishing causal attribution (11, 66).

McCormack and Shiell conducted a review of the literature, including quasi-experiments

and cross-sectional studies that control for self-selection and measured the BE

objectively (11). Self-selection refers to the bias introduced by residents who prefer to be

physically active, choosing to live in places that facilitate PA rather than doing more PA

because of attributes of the BE. In most of the studies in the McCormack and Shiell

review, PA was measured with self-reports. The diversity of land use, population

density, and street connectivity were determinants of overall PA. McCormack and Shiell

presented their findings per a PA domain, with most of the positive evidence related to

walking for transport purposes. In Australia, a quasi-experiment of people moving into

new homes found improvements in transport and recreational walking among those

participants with increased access to domain-specific destinations after relocation (67).

In this study, a geographic information system (GIS) and self-reports were used to

measure the BE, and PA was measured using a questionnaire. Results for

improvements in domain-specific walking were attenuated by BE perceptions. In the

United Kingdom, a quasi-experiment involving new active transport (walking, cycling,

and using public transport) infrastructure showed improvements in transport cycling and

overall active commuting for the participants who were least active at baseline (68). In

this study, PA was measured with a self-report questionnaire.

In conclusion, the literature presented in this section indicates that there is evidence

supporting a positive association between BE attributes (density, land-use mix, street

connectivity, access to public transport, park density) and overall PA. There is a lack of

agreement for some of the BE attributes such as land-use mix and park density.

Transport PA seems to be determined by a number of BE attributes including land-use

mix, residential density, street connectivity, availability of transport destinations and

transport infrastructure. However, the evidence is less supportive of the link of

recreational PA with BE attributes. Furthermore, the evidence is mostly confined to high-

income countries.

19

2.5 The BE and road trauma

Features related to the design of cities and transportation systems can greatly contribute

to, or reduce, the burden of road trauma (19, 69). For example, a study in Germany

found that improving the rail service frequency by 10% was associated with a 3%

reduction in use of motorised private vehicles and a 5% reduction in road accidents (70).

Chen and Zou conducted a study for the city of Seattle in the United States and found

that BE attributes that facilitate walking (e.g. high transit-stop density) were positively

associated with pedestrian crashes (71). This study highlights the importance of

providing infrastructure that is safe for all road users (72). A review of the literature for

the relationship of transport infrastructure and cyclist road trauma suggested that off-

road cycling facilities are associated with lower risk for cyclists (73). Multi-purpose trails

and major roads presented a higher risk. Findings from an empirical analysis for the city

of San Antonio in the United States suggested that higher population density and more

pedestrian destinations were correlated with fewer crashes compared with modern

urban designs of disconnected streets and destinations in non-residential areas (74).

Studies consistently show that pedestrians and cyclists suffer a higher rate of road

accidents than car occupants. For example, a study in the United States found that

compared with car occupants, pedestrians and cyclists were 23 and 12 times more

likely to be killed per kilometre travelled (75). A study in the Rhode County in France

found that compared with car occupants, cyclists were three times more likely to be

killed per time travelled, eight times more likely to be injured and 16 times more likely to

be seriously injured (76). For pedestrians, this study indicated that despite being half as

likely to be injured in a road accident, they were twice as likely to be seriously injured

compared with car occupants. There are large discrepancies in the incidence rates of

road injuries and fatalities, with studies indicating that compared with the United States,

European countries such as Sweden, Denmark and Germany are considerably safer for

pedestrians and cyclists (75, 77, 78). For cycling, Pucher and colleagues concluded that

a range of policies to improve safety, including off-road bikeways and traffic calming on

busy roads, are important determining factors of the high level of cycling observed in

Denmark, the Netherlands, and Germany (79).

20

Even though pedestrians and cyclists are exposed to higher levels of road injuries and

fatalities, increases in walking and cycling volumes do not necessarily translate into a

proportional growth in road trauma. Ecological studies (80-82) show a non-linear

relationship between exposure to traffic and injuries as the proportion of walkers and

cyclists increases, termed ‘safety in numbers’ (SIN) effect (80). However, a critical

review of the literature suggested that there is likely to be a threshold mode share for the

SIN effect to take place (83). A recent meta-analysis confirms the existence of a SIN

effect, however, the mechanisms are still unclear, which precludes drawing conclusions

about causality (84). The effect may be due to better infrastructure, which leads to more

safety and therefore more cyclists and pedestrians. Bhatia and Wier also point out that

the majority of the studies assessing SIN are based on cross-sectional evidence, which

precludes drawing strong conclusions regarding causality.

The evidence presented in this section indicates that pedestrians and cyclists are more

vulnerable to road trauma than car occupants. However, policies aiming at

improvements in walking and cycling may not result in proportional increases in road

trauma due to the SIN effect. Nevertheless, safety infrastruture as well as consirable

shifts from car travel to walking and cycling seem to be necessary for the SIN effect to

take place.

2.6 The BE and ambient air pollution

A number of air pollutants negatively impact on people’s health, including airborne

particulate matter (PM), ozone, nitrogen dioxide, and sulphur dioxide (85, 86). Of these

pollutants, airborne PM was identified by the WHO as causing the greatest burden of

disease (86). PM with an aerodynamic diameter smaller than 2.5 microns or less is

more health damaging than larger particles as it can penetrate deeper and stay in the

lungs (85). In studies estimating the potential health outcomes of transport and urban

design policies, PM2.5 has been a widely-used proxy for exposure to traffic-related air

pollution (37, 38, 87).

Several epidemiological studies indicated a strong dose-response relationship between

long- and short-term exposure to PM2.5 and cardiopulmonary diseases and lung cancer

(85, 88). For instance, Hoek and colleagues meta-analysed the results of

epidemiological studies assessing the impact of ambient air pollution on mortality from

21

all causes, cardiovascular disease, and respiratory disease (89). They found a

significant association of all-cause and cardiovascular mortality with long-term exposure

to PM2.5, but no significant association was found for non-malignant respiratory disease

mortality. In a systematic review and meta-analysis, Hamra and colleagues found an

increase in risk of lung cancer of 9% (95% CI 4 to 14) per 10ug/m3 increase in ambient

PM2.5 (90). The WHO recently updated the meta-analyses conducted by Hoek et al.,

indicating an increase of 7% (95% CI 4 to 9), 10% (95 CI% 5 to 15 and 10% (95% CI -2

to 24) per 10ug/m3 increase in ambient PM2.5 for all-cause, cardiovascular disease and

respiratory disease mortality, respectively (91). Furthermore, an integrated exposure-

response model was developed to be applied in Global Burden of Disease studies (88).

This model integrates relative risk information for exposure to ambient air pollution,

second-hand tobacco smoke, household solid cooking fuel, and active smoking, with the

aim of deriving a dose-response function that is inclusive of a wide range of exposures.

In summary, it is now well established that air pollution is a health risk factor, and that

PM2.5 affects a large number of people worldwide. The evidence summarised in this

section indicates that PM2.5 is associated with the risk of all-cause and cardiopulmonary

mortality, and the risk of lung cancer. For the BE in particular, PM2.5 has been

consistently used as a marker of transport-related air pollution. Using PM2.5 as a proxy

for transport air pollution may be due to the small number of studies that investigated

transport specific air pollution and health (92, 93).

2.7 Summary

The health and economic burden of chronic diseases in Australia is high, and it

consumes a significant amount of resources. A proportion of the burden from chronic

diseases can be prevented by targeting modifiable risk factors such as physical

inactivity. PA within the transport and recreational domains can be encouraged by BE

initiatives, and there is robust evidence indicating that a set of environmental factors are

conducive to PA. Hence, solutions to target low levels of PA should be sought from a

holistic perspective, involving non-health sectors, such as transport and planning.

Low levels of PA have been associated with all-cause mortality and several chronic

diseases (e.g. cardiovascular disease, cancers, type 2 diabetes, and mental health).

22

However, to date the evidence for the protective effect of PA on the development of

mental health conditions is not sufficient to conclude causality.

BE interventions such as active transport infrastructure and community programs to

encourage walking and cycling have great potential to improve population health.

However, trade-offs may arise. Those who take up PA in response to a BE intervention

would potentially experience greater exposure to air pollution and heat. Furthermore, the

risk of road trauma is higher among pedestrians and cyclists compared with car

occupants (drivers and passengers).

Chapters 4 to 8 aim to provide evidence of the potential of the BE for health and

methods for the incorporation of physical activity related health in the economic

appraisal of BE interventions in Australia. First, while there is a large body of evidence

from reviews of the literature for the association of BE attributes with PA, none of the

published reviews are specific to Australia. Chapter 4 is a systematic review of evidence

for the association BE-PA specific to the Australian setting. Second, the literature

estimating health and economic impacts of BE interventions mostly addressed

hypothetical scenarios (e.g. shifts from motorised travelled to active transport) without a

direct link to specific changes in features of the BE, and Australian evidence is limited.

The developments in Chapter 5 estimate the potential health and healthcare cost

savings of changes in exposure to 28 BE attributes based on evidence from real

observed difference in PA from alternative BE exposures. Third, findings from Chapter 6

serve to provide up-to-date evidence on the state of economic evaluations of transport

interventions with an active travel component. Fourth, the evidence from Chapter 6

indicates that cost-benefit analysis is a widely-used method in the field of transport.

Hence, Chapter 7 describes a method for the incorporation of PA-health benefits in

cost-benefit analysis of a broad range of BE interventions (e.g. improvements in

connectivity, land-use mix, etc.). Lastly, Chapter 8 is a quantitative health-impact

assessment of increased active transport based on a scenario developed to reflect

specific travel targets for a major Australian city.

23

Chapter 3 Overview of health prediction approach used in this thesis

This thesis includes three studies (Chapters 5, 7 and 8) that predict the potential health

and economic outcomes of changes in exposure to health risk factors. While these

studies share the same baseline methodology described in Section 3.1 and in the

methods sections of the corresponding studies, there are some differences among

them. Table 3-1 summarises the main differences between these three studies and how

they are related to other research chapters of this thesis.

Table 3-1 Differences in prediction modelling studies

Study Chapter 5 Study Chapter 7 Study Chapter 8

Aim Demonstrate the potential health and healthcare cost savings from improvements in PA attributable to the BE in Australia.

Develop a method for the robust incorporation of PA-related health in costs-benefit analysis (CBA) in Australia.

Demonstrate the potential of achieving travel targets of increased active travel for population health.

Scenarios 28 scenarios of changes in attributes of the BE based on the literature search in Chapter 4.

28 scenarios of changes in attributes of the BE based on the literature search in Chapter 4.

Increased active travel from a hypothetical scenario of achieving travel targets.

Relationship with other research chapters

Uses effect sizes reported in the systematic review conducted in Chapter 4.

Builds upon Chapter 5 by translating HALYs into monetised values. Based on findings from systematic review in Chapter 6 indicating that CBA is the most widely used method for the economic evaluation of BE interventions. An extensive discussion and justification for the monetisation of HALYs is provided. Adds an additional component by estimating monetised values per kilometre walked and cycled.

Methods

Included health risk factors

Physical activity Physical activity Physical activity, air pollution and road trauma

Dose-response physical activity and health baseline

Linear (non-linear in sensitivity analysis)

Non-linear Non-linear

Healthcare costs ACE Prevention inflated with healthcare cost indexed

ACE Prevention inflated with healthcare cost indexed

Modified version of ACE Prevention costs. (see Section 3.2.2)

24

3.1 General methodology

Health and economic predictions in this thesis were made using an updated and

adapted version of the PA model developed by Cobiac et al to assess PA interventions

for Assessing Cost-Effectiveness in Prevention (ACE Prevention) (94, 95). The original

model was designed to do full economic evaluations; however, in this research its use

was limited to projecting health outcomes and healthcare costs. Furthermore, for the

study described in Chapter 8, air pollution and road trauma were added as risk factors.

Only a brief description of the model is presented here, to avoid repetition with the

corresponding thesis chapters and supplementary materials, which fully develop the

methods and input parameters.

In ACE Prevention, a macro simulation approach was developed to calculate changes

in HALYs from changes in PA levels, applying a proportional multi-state multi-cohort life

table (MSLT) Markov model (96) (Figure 3-1). Also, healthcare costs for modelled

diseases (colon cancer, breast cancer, ischemic heart disease, ischemic stroke and

type 2 diabetes) and all other healthcare costs in added life years were estimated. The

model can be adapted to produce other measures of population health (e.g. life years,

health-adjusted life expectancy) and to estimate changes in the burden from specific

diseases (e.g. prevalence, mortality). ‘Macro’ or ‘cohort’ models are population-level

approaches that have entire populations as the unit of modelling (97). Markov models

are based on a set of transition probabilities (e.g. incidence and mortality hazards) that

determine how cohorts move between health states at discrete time intervals (98). The

Markov model is a suitable approach for modelling processes that progress over time,

such as chronic diseases (98, 99). DisMod II, which was designed to estimate internally

consistent epidemiological estimates (e.g. incidence, prevalence and mortality hazards)

and used in this thesis, is also based on a Markov model (Figure 3-2) (98, 100).

Two populations are simulated in the MSLT until they die or reach 100 years of age: the

population of interest as it is, and an identical population that is exposed to changes in

PA. The potential impact fraction (PIF) is used to link changes in exposure to the

determinant of health (PA) and incidence of PA related diseases (breast cancer, colon

cancer, ischaemic heart disease, ischaemic stroke, and type 2 diabetes) (101). The PIF

calculates the proportional change in disease risk attributable to a change in exposure.

The ‘relative risk (RR) shift’ method is used to calculate PIFs, translating changes in PA

25

into changes in disease incidence. Incidence has an impact on prevalence, and from

there on mortality, for each modelled disease. HALYs are calculated as life years lived

by the modelled cohorts in the MSLT, adjusted for disability (which is derived from

changes in prevalence). Two MSLT are modelled, one for the population as is and

another for an intervention population. HALYs are calculated as the difference in HALYs

between the two populations (Lwx in Figure 3-1). The MSLT allows for a time

component, hence, when modelling interventions, decays in effect as well as build-up of

effect can be incorporated. Also, the interaction between multiple diseases is accounted

for, with proportions of the population being able to be in more than one disease state.

Figure 3-1 Schematic description of a proportional MSLT

i = incidence; p = prevalence; m = mortality; w = disability adjustment; q = probability of dying; l = number of survivors; L = life years; Lw = disability adjusted life years

Interaction between life-table parameters and disease parameters. All the parameters are age specific, denoted with x, ‘-’ denotes parameter related to diseases or causes not included in the models and ‘+’ relates to all modelled diseases included in the model.

Note: A change in the determinant of health (PA) translates into changes in incidence (ix), which changes disease specific prevalence (px) and mortality (mx). The ‘relative risk shift’ method for the calculation of the PIF was used to estimate new levels of incidence due to changes in PA, where pi is PA prevalence at level I, RRi is the relative risk of PA for each of the diseases associated with PA level i and RRi' is the relative risk of PA for each associated disease after the intervention.

Disease process 2

m x - m

x + q x l

x L x e

x w x - w

x + Lw x HALE

x

i x p x m x

w

Life table

i x p x m x

w

3

m x - m

x + q x l

x L x e

x w x - w

x + Lw x

i x p x m x

w

Life table

i x p x m x

w

4 5

Risk factor

Disease process 1

I’=I (1-PIF)

PIF=∑ p

iRRi

ni=1 - ∑ p

iRRi

'ni=1

∑ piRRi

ni=1

26

Figure 3-2 Disease process model

Note: The disease conceptual model for each disease process, applied to each disease process separately (see Figure 3-1), has four health states (healthy, diseased, dead from the disease, and dead from other causes) and transition hazards between health states.

Source: Barendregt, Oortmarssen (100), Barendregt (102)

3.2 Differences with ACE Prevention

Terminology

In this thesis the term health-adjusted life years (HALYs) was used instead of DALYs

(used in ACE Prevention) to differentiate them from the DALYs produced in burden of

disease studies (103). Both are population health measures that incorporate quantity

and quality of life. The DALY was developed for use in burden of disease studies as a

measure of health loss due to disease (103). DALYs estimated with the ACE Prevention

method represent a health gain, hence are more like the quality-adjusted life years

(QALYs) that are widely used in economic evaluation of healthcare interventions (97).

However, QALYs use measures of utility weights that traditionally represent individual

experiences of health, whereas HALYs estimated with the ACE Prevention approach in

this thesis use disability weights linked to specific diseases, which were developed for

the Global Burden of Disease study (104). In addition, HALYs estimated in this thesis

are not the addition of years of life lost (YLL) and years lived with disability (YLD), but

instead calculated as the total number of life years lived, adjusted for the average

health-related quality of life in those years (by age and sex). The main difference is

related to the way the YLL component is calculated in burden of disease studies relative

to our use of HALYs. First, the mortality rates we used are those expected for the

Healthy

Diseased

Dead

(disease)

Dead

(other)

Mortality (other)

Remission

Incidence Mortality (other)

Case fatality

27

population in question, not those for a hypothetical population that has the lowest

observed mortality at every age, as is the case with the YLL component of DALYs.

Second, in contrast to our HALYs, YLLs used in the burden of disease studies are not

adjusted for quality of life, and implicitly assume all added life years are in full health.

This leads to an over-estimation in health gain when examining interventions that

postpone death.

Healthcare costs

In this thesis, prediction modelling studies in Chapters 5, 7 and 8 include estimates of

healthcare costs attributable to the modelled diseases and injuries (i.e. related

healthcare costs) and all other healthcare costs in added life years. A brief explanation

for each of these categories is presented below, with an emphasis in differences with

the original ACE Prevention model.

3.2.2.1 Healthcare costs of modelled diseases and injuries

The average healthcare cost for diseases and injuries are based on a rather crude

macro-level methodology developed in the ACE Prevention study. Healthcare costs

include hospital services, out-of-hospital medical services, pharmaceuticals, and health

professionals. In ACE Prevention, average cost per disease case by age and sex were

estimated as total disease costs (public and private expenditure), divided by prevalent or

incident cases (105). Healthcare costs were from a 2001 disease-cost study by the

Australian Institute of Health and Welfare (AIHW) (106). The denominators used to

estimate per case costs in ACE Prevention were based on the 2001 predictions and

2003 data from the 2003 Australian Burden of Disease (BoD) study (107). Original per

case estimates from ACE Prevention, indexed with healthcare sector inflation indices,

were used in the manuscripts in Chapters 5 and 7. During the writing of Chapter 8,

updated epidemiological data from the Global Burden of Diseases (GBD) study became

available that included estimates of prevalent and incident cases for Australia for 2000

(104). Given the lack of updated healthcare costs data at the level required for this

thesis, it was deemed appropriate to at least update the denominators used to calculate

the healthcare costs per case, with 2000 GBD data for Australia as a proxy for 2001.

For the study in Chapter 8, healthcare costs per case were then inflated using

healthcare sector price indices. The most salient difference was observed for healthcare

28

costs per case of ischemic heart disease and ischemic stroke. As depicted in Table 3-2,

estimates are considerably lower when using the Australian 2003 BoD estimates of

prevalent cases, compared with estimates from the GBD study for 2000. Healthcare

costs for the other physical inactivity–related diseases (type 2 diabetes, colon cancer,

and breast cancer) do not differ greatly. Crude estimates of healthcare costs were also

observed in the literature included in Chapter 6 and in a recent study of trials in the

United States (US) (108). For instance, in the US study, total costs of treatment were

translated into cost per case, in combination with assumptions regarding diseases’

duration.

Table 3-2 Comparison healthcare cost of physical inactivity diseases (A$ 2013)

Age and sex

group

Cost prevalent case of Ischemic heart

disease (A$) (prevalent cases)

Cost prevalent case of Ischemic

stroke (A$) (prevalent cases)

GBD 2000 Australian BoD

2001/2003 GBD 2000

Australian BoD

2001/2003

< 55, males 9,007 (25,528) 3,829 (60,052) 10,056 (5,382) 2,880 (18,793)

55–64, males 8,309 (35,936) 2,570 (116,194) 13,459 (5,650) 6,387 (11,905)

65–74, males 7,243 (53,907) 2,151 (181,522) 18,270 (10,052) 12,316 (14,912)

75+, males 8,384 (44,970) 1,922 (196,149) 24,047 (10,801) 17,640 (14,724)

< 55, females 6,021 (12,648) 2,367 (32,166) 6,848 (5,996) 1,501 (27,201)

55–64, females 7,209 (13,855) 1,965 (50,833) 9,225 (4,250) 2,702 (14,511)

65–74, females 8,041 (26,998) 2,061 (105,321) 11,664 (8,280) 6,600 (14,633)

75 +, females 8,052 (52,378) 2,066 (204,124) 33,981 (13,640) 20,078 (23,085)

3.2.2.2 Healthcare costs in added life years

All other healthcare costs in added life years were also estimated adopting the macro

approach developed in ACE Prevention using the 2001 healthcare cost data above

mentioned. All other healthcare costs per person per year, by sex and 5-year age group,

were calculated as all healthcare costs minus those for the diseases explicitly modelled,

divided by the number of person-years. For example, in the prediction models used in

Chapters 5 and 7, all other healthcare costs in added life years exclude costs of physical

inactivity-related diseases, to avoid double counting. For the studies in Chapters 5, 7

and 8, all other health care costs in added life years were indexed with healthcare sector

inflation indices. The methodology applied assumes that the current costs structure

29

(2001) is constant over time. In ACE Prevention, all other healthcare costs in life years

gained were explicitly left out of the cost-effectiveness studies as the lead authors

deemed that there was no agreement in the literature on this topic (105). As pointed out

by Van Baal et al, the definition of related and unrelated made in healthcare economic

evaluations refers to the diseases affected by the intervention (109). Therefore, in ACE

Prevention, indirectly, the authors did include part of the healthcare costs in added life

years that are considered unrelated the interventions of interest. This is because added

life years do incur costs for the diseases that are explicitly modelled (i.e., related to the

risk factors included). The structure of the multi-state life table makes it impossible to

exclude these. Since ACE Prevention, the incorporation of all other healthcare costs in

added life years in economic evaluations of health interventions has been gaining

support (110-115). In what follows a brief discussion of the literature that opposes and

supports the inclusion of all other healthcare costs in added life years is presented.

Russel argued that healthcare interventions should only be assessed on their own merit,

hence, costs attributable to extending life should not be included (Russel 1986 cited in

114). Nyman in 2004 also suggested the exclusion of such costs since the utility derived

in the added life years is not directly linked with all other healthcare costs in added life

years (116). In addition, to date, most of the guidelines for economic evaluations in

healthcare recommend the exclusion of all other healthcare costs in added life years

due to the high level of uncertainty about the future (117).

Rappange et al (114) in 2008 discussed in detail the arguments for inclusion of all other

healthcare costs in added life years in economic appraisal of healthcare interventions.

The authors argued that such costs should be included regardless of the study

perspective adopted, otherwise, results would not be optimal and consistent. In

principle, either societal or healthcare sector perspectives warrant their inclusion. From

a societal perspective, all costs should be addressed, regardless of who incurs them

and who pays for them (118). From a healthcare sector perspective, prolonging life has

important costs implication as if people live longer, they are also likely to need care that

without an intervention would not have been used. Hence, in terms of optimal allocation

of resources, ignoring all other healthcare costs in added life years would have

important implications, especially for interventions with significant gains in life years.

Notably, funding would be biased towards interventions that extend life among the older

age cohort, over interventions that improve quality of life (119). In addition, it is

30

inconsistent to only include healthcare costs related to the illnesses/injuries related to

the intervention but account for all future health gains. Van Baal et al (112)

demonstrated the implication of including/excluding all other healthcare costs in added

life years in cost utility ratios for a smoking cessation intervention (112). The analysis

consisted of estimating four cost-utility ratios as follows: (1) Intervention costs over total

quality-adjusted life years (QALYs); (2) Intervention costs plus related healthcare costs

(savings) over total QALYs gained; (3) Intervention costs plus related healthcare costs

over QALYs attributable to the related costs; and (4) Intervention costs, related

healthcare costs, and all other healthcare cost in added life years over all QALY gains.

The authors argued that only approaches 3 and 4 are consistent, since there is a clear

link between costs and health gains. However, the third approach presents calculation

difficulties as a breakdown of QALYs to reflect gains due to the intervention and those

that are not directly linked to the intervention are required. In addition, Van Baal and

colleagues argue that the third approach does not make sense from a decision makers

perspective, as all gains in health are included in the objective function in health

economic evaluations (112). Hence the fourth approach is appropriate. It also gives the

least favourable cost-utility ratio. As explained by Rappange et al (114 p826):

It is not surprising that including all costs and effects in the cost-utility ratios as compared with only

some costs and all effects results in higher estimate of the costs per QALY gained. The reason for

this is simple. Living longer brings about competing diseases that result in healthcare expenditures

(which in turn yield health).

In a recent editorial Morton et al (115) likewise support the inclusion of all other

healthcare costs in added life years in health technology appraisal. The authors stated

that while they appreciate that there are uncertainties as to what future health

treatments may cost, this is not a valid reason to exclude them. Morton et al recommend

that an assessment of the effects of including all other healthcare costs in added life

years in the funding structure of health technologies should be performed (for the United

Kingdom). In addition, the authors urge for the development and validation of models to

estimate lifetime healthcare costs.

For this thesis, all other healthcare costs in added life years were estimated and

presented in Chapters 5, 7 and 8 (see Table 3-1). The aim of this thesis was to predict

health and healthcare cost outcomes of BE scenarios, and propose methods for the

incorporation of PA-related health in cost-benefit analysis of BE interventions. Hence, it

31

was deemed appropriate to estimate a complete picture of the cost implications to the

health system. In addition, economic evaluations within the transport and planning

sectors tend to adopt a societal approach and use cost-benefit analysis to assess the

merit of their proposed projects (120, 121). Therefore, from a societal perspective, all

benefits and costs should be included regardless of who incurs them.

3.3 Comparison with other prediction modelling approaches

Similar macro simulation methods based on Markov models have been used in the

past. For example, in the United Kingdom, the economic merit of building pathways to

increase population levels of PA was assessed applying the Prevent model (102, 122).

There are two main differences between ACE Prevention and Prevent. First, ACE

Prevention allows assessing uncertainty, while Prevent does not. Second, Prevent

incorporates a time lag for the time that the risk of disease remains unchanged following

a change in the risk factor. In ACE Prevention it is assumed that incidence changes

immediately after a change in exposure to risk factors. However, changes in prevalence

and case fatality are based on a set of transition probabilities, and therefore eventuate

over the years in the remaining lifetime of the cohorts. Another example is the study, by

Mansfield and MacDonald, assessing three BE interventions in three locations in three

communities, in which the DYNAMO-HIA model was applied (123). DYNAMO-HIA

combines a partial stochastic micro-simulation approach to project risk factors and a

deterministic macro approach based on a MSLT Markov model to project health

outcomes (124). This split into two models was mainly done for computational reasons;

the MSLT becomes time and memory demanding when the number of health states

increases. ACE Prevention does not face this issue to the same extent, because the

MSLT approach is based on the ‘proportional’ method which avoids the problem of

exponential growth of health states (96).

32

Chapter 4 The association between built environment features and

physical activity in the Australian context: a synthesis of the literature

Introduction to manuscript

The body of evidence of BE correlates of PA for the Australian setting has been growing

exponentially over recent years. The field has been progressing from cross-sectional

designs to longitudinal studies and quasi-experiments with greater sample sizes.

However, to date, no attempt has been made to summarise Australian-based evidence,

as has been done in other regions (e.g. European countries) (14). A systematic

summary of Australian studies would help local policymakers to recognise what

environmental features most benefit PA. Furthermore, understanding the local setting

enables international comparison and assessment of discrepancies in what works in

other places and not in Australia. This chapter addresses Research Question 1: What

are the attributes of the built environment in Australia that most benefit physical activity?

The magnitude of effects reported in the studies presented in this chapter were used in

Chapter 5 to estimate the potential health and economic benefits of changes in the BE

attributes.

Citation

Zapata-Diomedi, B & Veerman, JL 2016, 'The association between built environment


BMC Public Health, vol. 16, no. 1, pp. 1-10. doi: 10.1186/s12889-016-3154-25

Authors’ contribution: Zapata-Diomedi B adapted the research question from the

research report prepared for the New South Wales Department of Health, designed

the methods, conducted the review and wrote the manuscript. Veerman JL advised

on the research question and methods, critically reviewed and co-edited the paper.

5 Appendices are available as online material accessible via this link: https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-016-3154-2

https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-016-3154-2

33

Abstract

Background: There is growing evidence indicating that the built environment is a

determinant of physical activity. However, despite the well-established health benefits of

physical activity this is rarely considered in urban planning. We summarised recent

Australian evidence for the association built environment-physical activity among adults.

This summary aims to inform policy makers who advocate for the consideration of

health in urban planning.

Methods: A combination of built environment and physical activity terms were used to

systematically identify relevant peer reviewed and grey literature.

Results: A total of 23 studies were included, providing 139 tests of associations

between specific built environment features and physical activity. Of the total, 84

relationships using objective measures of built environment attributes were evaluated,

whereas 55 relationships using self-reported measures were evaluated. Our results

indicate that walkable neighbourhoods with a wide range of local destinations to go to,

as well as a diverse use of land, encourage physical activity among their residents.

Conclusions: This research provides a summary of recent Australian evidence on built

environments that are most favourable for physical activity. Features of walkability and

availability of destinations within walking distance should be accounted for in the

development or redevelopment of urban areas. Our findings emphasise the importance

of urban planning for health via its impact on population levels of physical activity.

Key words

Built environment, physical activity, Australia, association, review, health, policy

34

Introduction

Physical inactivity is a significant public health concern given the detrimental impact on

population health. For example, low levels of physical activity have been associated with

higher all-cause mortality (3) and mortality and morbidity of chronic diseases (4, 46). In

Australia, less than 50 per cent of the adult population meets the recommended

physical activity guidelines (57). Physical inactivity across the Australian adult population

is responsible for 6%, 8%, 10%, 11% and 10% respectively of the burden of coronary

heart disease, type 2 diabetes, breast cancer, colon cancer and all-cause mortality (4).

It has also been suggested that inactive lifestyles are related to poorer mental health

outcomes (5, 54), falls in older people (7), and higher risk of overweight (6). Given the

detrimental impact of physical inactivity on population health, much emphasis is placed

on ways to improve physical activity behaviours.

Researchers acknowledge that to positively change physical activity behaviours across

populations, holistic approaches that consider individual as well as environmental

interventions are needed. For example, Sallis and colleagues (36) proposed an

ecological framework for ‘active living’ which identifies a number of environmental

features and their influence on physical activity behaviours. The built environment is the

overarching term used in the literature to describe those objective and subjective

features in the physical setting in which people spend their time (125). According to the

World Health Organization (WHO), the built environment incorporates the building and

transportation design of a city, including factors such as open green spaces, bike

ways/sidewalks, shopping centres, business complexes, and residential

accommodation (10). In recent years, the literature assessing the association between

the built environment and physical activity outcomes has grown, mainly in the developed

world. This includes a number of survey studies assessing correlations between built

environment features and physical activity and obesity (11, 13, 14, 126-130) as well as

reviews of reviews (64, 131, 132).

In this research we reviewed evidence for the association between built environments

and physical activity in the Australian context, with the aim of giving an indication of

which environmental factors stand out as being related to physical activity. The review

was prepared for the Centre of Population Health (CPH) of the Government of New

South Wales (Australia) to assist in decision-making regarding the inclusion of physical

activity in urban planning. The objective was to summarise the evidence in Australia

http://www.aihw.gov.au/risk-factors-overweight-obesity/

35

from 2009 to date for the association between built environment attributes and adult

(>18 years) physical activity.

Methods

Search strategy, data sources and inclusion criteria

One author (BZD) systematically searched peer-reviewed and ‘grey’ literature, in

English, restricted to human subjects from 2009 onwards limited to Australia. The

scope of the review was defined by the CPH in collaboration with the authors to reflect

recent context specific evidence for the associations between built environment

characteristics and physical activity. Search strategies were defined in collaboration with

members of the CPH and applied to both academic datasets and the grey literature

(Appendix A). The following academic databases were systematically searched: Web of

Science, Scopus, EBSCOHost (which includes Business Source Complete, CINAHL,

MEDLINE, SportDiscus and Econlit), GeoRef and Leisure Tourism. Google was used to

search for Government reports and experts in the field were consulted to ensure that all

relevant literature was included. Reporting was based on PRISMA guidelines (133)

(Appendix B). Studies included in selected reviews were assessed against the inclusion

criteria (see Table 4-1).

Studies that compared physical activity behaviours before and after relocation into a

different neighbourhood without direct association to a particular built environment

attribute were excluded. Studies assessing mediating variables in the association

between built environment attributes and physical activity were excluded when a direct

association was not provided. Only studies targeting the adult general population were

included (see Table 4-1).

Table 4-1 Inclusion criteria

Criteria

1. Published in English from 1 January 2009 to 15 March 2015

2. Study conducted in the Australian context 3. Primary study or review 4. Presented evidence on the direct association between built environment features and physical

activity 5. Adult population (>18 years)

Built environment attributes

We grouped built environment features into one of seven categories, including five of

the “6 Ds” proposed by Ewing and Cervero (134), plus safety, and aggregated

36

neighbourhood measures (see Table 4-2). We subdivided broad categories (e.g. design

and destinations) given the heterogeneity of measures included in them (135). Features

included in each category are presented in Table 4-2 with the expected direction of the

association based on past literature (136-138).

37

Table 4-2 Categorisation of built environment attributes

Categorya Built environment attributes Expected direction of association

Density Population density/jobs density Positive

Diversity Land use mix/non-residential zone Positive

Design Street Network: street connectivity/few cul de sacs/space syntax measures (e.g. local and control integration)/traffic slowing devices/ pedestrian crossing/ active transport route options/3/4 or more ways intersections

Positive

Road traffic volume / busy roads Negative

Transport infrastructure: sidewalks/bikeways/street lights/aesthetics and attractiveness

Positive

Green and recreational space: area/presence /number/distance(shorter)/quality/attractiveness/

maintenance/aesthetics

Positive

Destination Transport related: shorter distance (or access within walking distance) to: neighbourhood destinations, retail, school/better job accessibility by public transport

Positive

Job accessibility by car Negative

Recreation related: shorter distance (or access within walking distance) to recreational destinations

Positive

Distance to transit Shorter distance (or access within walking distance) to bus stops/ train stations

Positive

Safety Neighbourhood lighting Positive

Crime/Traffic Negative

Aggregated neighbourhood characteristics

Walkability index/environmental score Positive

a Note: Ewing and Cervero have a 6th D relating to the Demand for Parking. It has been excluded in this list as no relevant research was found.

Coding of evidence

Most of the studies tested multiple associations as a result of different domains of

physical activity assessed, outcomes evaluated, neighbourhood definitions, and spatial

area evaluated. Similar approaches were taken in past studies (13, 136). Results were

coded in terms of whether the associations between built environment attributes and

physical activity behaviours were in the expected direction (+), in the opposite direction

(-), or not statistically significant (0) according to the level of significance stated in the

study. We present results for studies in which built environment attributes were both

objectively measured and subjectively measured. We also report physical activity based

on perceptions of the purpose (e.g. transport and recreational) and total physical

38

activity). Total physical activity was explicitly derived in most of the studies (139-144) as

any physical activity in the transport and recreational domains. However, in two studies

total physical activity included households and gardening chores (143, 144) (detail can

be found in Appendix D).

We considered objective measures of built environment attributes as showing sufficient

evidence if they were assessed in at least three independent studies (14). Of the built

environment categories with sufficient evidence, it was deemed convincing if at least

50% of all associations were in the expected direction (14, 136). Self-reported built

environment attributes showing convincing evidence (≥50% associations in the

expected direction from at least 3 independent studies) are presented to assess

whether they support objective findings.

Quality assessment

We assessed the quality of studies (see Appendix F, Tables 1 and 2) using tools from a

similar review (126). The quality assessment focused on the representativeness of the

sample, measurement of outcome variables, and control for confounding variables.

Longitudinal (n=2) and quasi-experimental designs (n=2) were assessed separately

from cross-sectional designs (n=19). Studies were classified as being of poor, fair, or

good quality according to the number of criteria met. We assessed the strength of the

associations with and without quality assessment, following recommendations in the

literature to not rely on ‘vote counting’ techniques (145).

Results

A total of 22 studies from the database search and one additional study recommended

by experts in the field provided 139 associations of built environment attributes and

physical activity (Figure 4-1). Of the total, 84 associations were evaluated against

objective measures of built environment attributes and 55 associations were evaluated

against subjective measures (Table 4-3). A list of excluded papers and reasons for

exclusion is provided in Appendix C.

Study characteristics

The largest proportion of studies was conducted in Western Australia (n=8), followed by

South Australia (n=7), New South Wales (n=3), Victoria (n=3) and Queensland (n=2)

(Table 1, Appendix D). Most of the studies were cross-sectional in design (n=19), with

39

two longitudinal studies (146, 147) and two quasi-experiments (67, 148). The median

response rate across studies reporting it was 31%, ranging from 11.5% (147, 149-152)

to 68.5% (139). The majority of included studies randomly selected participants, with the

exception of five studies from the RESIDE project which selected participants according

to their intention to relocate to new developments (140, 141, 148, 153, 154). The

median sample size for studies reporting it was 2,194 with a range from 320 individuals

(155, 156) to 203,883 individuals (143, 144). All included studies were from urban areas

with one exception for rural zones (157). For the studies that reported participants’

ages, averages across studies ranged from 35 (157) to 61 (143, 144) years with a mean

of 45 years. The older participants were selected only if they were 45 years old or

above. The majority of the studies included both genders, with one exception that only

included women (157). For the studies that reported gender distribution, women

represented on average 55% of the samples across the included studies, with the

highest proportion at 62% (142). Only one study sampled individuals from a specific

income group (low socio-economic status) (157).

Physical activity measures

All included studies used self-reported measures of physical activity for a usual week

(i.e. previous seven days/week), or the past month. Walking was the most commonly

assessed physical activity outcome (n=16), followed by cycling (n=3) (151, 153, 158),

moderate to vigorous physical activity (n=3) (143, 144, 152), leisure time physical

activity (n=1) (157), and use of active travel modes (n=1) (159). One study assessed

both walking and moderate to vigorous physical activity (144). In less than half of the

included studies (n=9) (67, 140, 141, 148, 152-156) physical activity was measured

using questionnaires that specified the location (e.g. neighbourhood) in which activities

took place.

Built environment measures

All studies, except one (159), assessed built environment attributes in the

neighbourhood area, commonly defined as the 1.6 kilometre street network service

area, or 1 kilometre radius from a participant’s residence, or walking area within 10 to 15

minutes. Ten studies used objective measures of built environment attributes, eight

studies used both objective and subjective measures, and five studies used only

subjective measures.

40

Figure 4-1 Summary of included studies

Summary of findings

The greatest number of associations that were evaluated against objective measures of

the built environment addressed total physical activity (n=32), physical activity related to

transport (n=28), and leisure specific physical activity (n=24). In the following section

and in Figure 4-2 and Table 4-3, we present a summary of the evidence with complete

results for each physical activity domain and subcategories of the built environment

attributes available in Appendix E. Density

41

After adjustment for other explanatory variables, the evidence for the effect of density on

physical activity outcomes is not convincing. For the Australian context, only 33% of

cases indicated a positive association of density with physical activity.

Diversity

The findings indicate convincing evidence of a positive relationship between built

environment diversity measures and physical activity, with four out of six studied

associations in the expected direction (67%). This indicates that greater diversity in the

built environment is associated with greater physical activity. All four positive

associations for measures of land use mix were related to physical activity in the

transport domain.

Design

The evidence for the relationship between design features and physical activity

outcomes is not convincing, with only 28% of associations in the expected direction.

When assessing the subcategories of design (e.g. street network, transport

infrastructure, and green and recreational spaces), the evidence remains unconvincing,

or not sufficient to draw conclusions.

Destinations

The evidence of a relationship between availability of destinations and physical activity

outcomes is convincing with 70% of the associations showing an effect in the expected

direction. The majority (6/8) of the evidence for destination measures relates to transport

destinations such as retail zones, services, post offices, food outlets, transit stops, job

locations, and open public spaces such as parks. All but one positive association related

to transport or total physical activity.

Distance to transit

The current evidence provides convincing evidence for the association between shorter

distance to transit and physical activity, with 80% of associations in a positive direction.

It should be noted that many of the studies include transit as a measure under

’destinations’. Most of the associations for physical activity measures relate to transport

physical activity, with one exception relating to total physical activity.

Safety

42

Two studies indicated that safety is associated with total physical activity. However, in

one of the studies the direction of the associations were not in the expected direction,

indicating that less safe places were associated with positive physical activity outcomes.

The evidence remains inconclusive since we found only two studies and there is no

consensus in the potential effect of safer places.

Aggregated neighbourhood measures

Aggregated neighbourhood measures, such as walkability, are composite indices that

include a number of built environment features such as density, connectivity, and land

use mix (160). Convincing evidence was found in the Australian context for aggregated

neighbourhood measures with 74% of the associations indicating a positive impact on

physical activity. Walkability measures indicated a stronger association for transport

related physical activity (7/7) and total physical activity (5/6) in comparison to physical

activity for recreational purposes (2/6).

Evidence for self-reported built environment attributes

For perceived measures of the built environment, most studies evaluated leisure

physical activity outcomes (n=41) and physical activity for transport purposes (n=14)

(Appendix E).

Associations between physical activity and perceived built environment attributes were

not always in line with similar relationships that used objective measures. As shown in

Figure 4-2 and Table 4-3, there is sufficient evidence to draw conclusions for

destinations, safety, and design. For destinations, the evidence was similar to objective

measures. On the other hand, the evidence from self-reported measures indicated that

built environment attributes related to design are positively associated with physical

activity outcomes.

43

Figure 4-2 Proportion of tested associations for built environment features with sufficient evidence in the expected direction

Sensitivity of results to study quality

If we consider only studies judged to be of good quality, there is not enough evidence to

draw conclusions for any of the associations between objectively measured built

environment attributes and physical activity outcomes (Table 4-3). The only exception is

measures of design, for which the evidence is sufficient in quantity, however this

evidence does not convincingly show an association with physical activity outcomes

(25%). When the components of design are individually analysed, there is convincing

evidence for a positive association between the street network subcategory (street

connectivity) and physical activity outcomes (50%). For the case of evidence using self-

reported built environment measures, conclusions remained unchanged for all except

for measures of design (from 50% to 40%) after taking the quality of the studies into

account.

33%

67%

28%

70%

80%

74%

50%

71%

33%

0% 20% 40% 60% 80% 100%

Density

Diversity

Design

Destinations

Distance to transit

Safety

Aggregated Neighbourhood measures

Self-reported measures Objective measures

44

Table 4-3 Summary of associations between built environment attributes and physical activity

Built environment attributes

Objective Built Environment Self-reported built environment

All studies Good and fair

qualityb All studies

Good and fair quality

Density 3/9 (33%) [4]a 1/5 (20%) [2]

Diversity 4/6 (67%) [3] 2/4 (50%) [1]

Design 8/29 (28%) [6] 6/24 (25%) [4] 16/32 (50%) [3] 11/27 (40%) [3]

Destinations 7/10 (70%) [4] 3/6 (50%) [2] 10/14 (71%) [3] 10/14 (71%) [3]

Distance to transit 4/5 (80%) [3] 3/4 (75%) [2] 1/2 (50%) [1] 1/2 (50%) [1]

Safety 2/6 (33%) [2] 0/0 (N/A) 3/9 (33%) [3] 2/9 (22%) [3]

Aggregated neighbourhood measures

14/19 (74%) [3] 8/15 53% [2] 1/1 (100%) [1] 1/1 (100%) [1]

Note: Results represent the proportion (%) of tested associations with results in the expected direction a Number of independent studies b Only two cross-sectional design studies rated as good quality, ten qualified as fair quality and seven as poor. All studies with longitudinal and quasi-experiment designs rated as good quality (Appendix F Tables 1 and 2)

Discussion

In this review we summarise the recent Australian literature measuring the association

between physical activity and built environment attributes. A total of 23 quantitative

studies that focused on adults’ physical activity were reviewed for both objective and

self-reported measures of the built environment. As a whole, evidence indicates a

positive relationship between built environment attributes and physical activity for adults.

Objective measures of built environment attributes that were positively associated with

physical activity included destinations within walking or cycling distance from the

residence, shorter distance to transportation, such as bus stops, train stations, and ferry

terminals, walkability, and higher diversity of land uses. Findings were similar for both

objective and self-reported measures of availability of destinations. Although self-

reported measures of design indicated convincing evidence of an association with

physical activity, this was not the case for objective measures. Both objective and

perceived measures of the built environment are considered important as they provide

insight into different relationships with physical activity outcomes (145). For example, a

range of social, economic and demographic factors are likely to influence individuals’

perceptions of the built environment, which not necessarily correspond to objective

measures (161, 162). We did not differentiate results in regards to self-reported and

objective measures of physical activity as all included studies relied on self-reported

measures. However, in our sensitivity analysis we did consider the quality of reporting

(i.e. the use of a validated questionnaire). We could not identify a consistent pattern for

45

results when comparing studies using validated questionnaires against those that did

not. Nevertheless, four out of six studies not using validated questionaries were

classified as being of poor quality according to the criteria used in this study.

For objective measures of design and density, there was not convincing evidence to

indicate that these variables are associated with physical activity from recent studies in

the Australian context. However, measures of density, connectivity (design feature), and

open public spaces (design and destinations feature), are commonly included in

aggregated neighbourhood measures, which shows convincing evidence of having a

positive relationship with physical activity. Additionally, having more places to visit

implies various components of design such as parks and green open spaces. While

density itself is unlikely to stimulate physical activity, higher density allows for mass

transit and commercial and non-commercial destinations and therefore tends to

increase the number of potential destinations within walking or cycling distance (67,

163). In accounting for these mediating variables, there is a risk of over-adjustment and

‘explaining away’ real associations. Hence, it may be that a mix of built environment

attributes is needed to have a positive impact on physical activity. The overall evidence

summarised in this review suggests that having access to a wide variety of destinations

within walking distance supports higher levels of physical activity.

We reported physical activity outcomes in two domains (recreational and transport) as

well as total physical activity following recommendations from the literature regarding

the different uses of built environment features for physical activity (35, 145). However,

results should be interpreted with caution. For example, we cannot conclude that

diversity is more important for transport physical activity than for recreational physical

activity as we do not have the same number of associations across all domains of

physical activity.

After excluding studies that did not meet quality criteria, there was insufficient evidence

to draw conclusions for any of the objective measures of the built environment except

street connectivity (design measures within street network category). The results for

self-reported measures remained unchanged, expect for the case of measures of

design. Our results highlight the importance of taking the quality of the studies into

account in the process of summarising the literature addressing association between

the built environment and physical activity. For cross- sectional studies the most

problematic quality criteria were the low response rate (less than 40% in 16 out of 19)

46

and the lack of reporting on whether the samples were representative of the population

of interest. Of the longitudinal and quasi-experiment studies, none reported information

about the comparability of the exposed and non-exposed groups. Compared to

longitudinal and quasi-experimental designs (n=4), cross-sectional evidence was rated

as being of lower quality with only 10% of the studies scoring as good quality,

compared to 100% for longitudinal and quasi-experimental studies. As highlighted in the

past, quality assessment of primary studies is rarely carried out in despite of being one

of the main criteria of a systematic review (145).The traditional hierarchical classification

of evidence recommended by the National Health and Medical Research Council of

Australia (NHMRC) (164) is ill-suited for the quality evaluation of studies on the

relationship between the built environment and physical activity. This is because it does

not distinguish between different observational designs. It focuses on experiments,

which are seldom feasible in this field. We decided to use a tool based on quality criteria

for the evaluation of observational studies proposed by Petticrew and Roberts (164) as

adapted by Grasser and colleagues (126). However, we added a criterion to assess

whether cross-sectional studies included a measure to control for self-selection, given

its attenuating effect for the association built environment and physical activity (11).

Our findings are specific to the Australian context. Nevertheless, they are in line with

internationally conducted literature surveys. Recent reviews found that availability of

destinations (overlapping with land use mix) and walkability are facilitators of physical

activity. McCormack and Shiell (11) conducted a systematic review of the international

literature on the association of objectively measured built environment features and

physical activity, including only studies that controlled for self-selection (cross sectional

controlling for self-selection and quasi-experimental designs). They found consistent

associations between physical activity and land use mix, composite walkability indices,

and neighbourhood type (i.e. neo-traditional versus conventional). A study focusing only

on European countries found convincing evidence for an association between physical

activity and walkability, access to shops, services, and work, and environmental quality

(14). Grasser and colleagues (126) found consistent associations between physical

activity and density (i.e. population, housing, and intersections) and walkability indices.

Strengths of this study include the systematic review of evidence that is recent and

directly applicable to the Australian context, and the ascertainment of study quality. It is

worth noting that the inclusion of quality criteria for studies assessing the association

47

between the built environment and physical activity is uncommon in the literature (145).

Furthermore, the search strategy was defined in collaboration with a group of experts in

the field and policy makers, as this review is part of a broader review for a government

body. Limitations of this study should be mentioned. While a comprehensive search

strategy was followed, only one reviewer was in charge of systematically reviewing the

literature. However, given that the process was overseen by a group of experts, the

potential of missing relevant studies was small. Besides, it can be argued that in the aim

of showing recent Australian evidence and limiting the review to 2009 onwards,

important literature may have been missed. Furthermore validated physical activity

questionnaires were not used in six out of 23 studies which may had resulted in biased

results for the assessed associations (165). We attempted to pool results in a meta-

analysis, however, given the diversity in the ways in which different studies report their

findings this was not possible. Besides, the greatest majority of the evidence relies on

cross-sectional designs, which does not allow for causal inference.

Recommendations for future research

We observed a number of limitations in the literature that should be addressed in future

research assessing the relationship between built environment attributes and physical

activity.

1. Use of standard methods for reporting the association between the built

environment and physical activity allowing for the statistical combination of

results. This may include moving from categorical exposures to continuous

measures. As recently suggested by Lamb and White (166), using continuous

exposure measures would also avoid the loss of exposure information which

occurs in the categorisation process. Pooling results from studies has numerous

advantages, including a higher number of observations for a given association

and hence greater statistical power and improved estimates of effect size (167).

2. Provide sufficient information on the exposure variable to enable a direct

interpretation of results. For example, presenting results in terms of associations

of physical activity outcomes with z-scores is meaningless without descriptive

information about the distribution of the exposure variable (i.e. mean and

standard deviation). Furthermore, the categorisation of exposure variables in

quantiles, or ordinal data without an indication of the mean value of each

category makes it impossible to know what level of change in the exposure

48

variable is needed to achieve a certain outcome. In plain summary, all we know

from the literature is that more is better than less (or vice versa, depending on the

exposure), however, we are unsure about how much change is needed. Hence,

researchers investigating the potential effect of physical activity of changes in the

built environment should be specific in the level of change in the exposure

variable (e.g. increase in 8 dwelling per hectare). This is of particular relevance

for policy makers who need robust information on what environmental factors are

associated with physical activity behaviours and how much of each is needed to

achieve meaningful health benefits.

3. Researchers should take into account mechanisms to diminish the potential bias

introduced by self-selection such as longitudinal and quasi-experimental designs,

or inclusion of question to assess potential self-selection in cross sectional

studies. The majority of studies are cross sectional in design, which does not

allow for a direct causal interpretation. The association may be due to the built

environment influencing physical activity; this is the hypothesis underlying this

research. Alternatively, it may be due to physically active people choosing to live

in neighbourhoods that facilitate that behaviour. By adjusting for self-selection,

some studies try to avoid this ‘reverse causal’ interpretation. McCormack and

Shiell (11) systematically reviewed the international literature and found that

adjusting for self-selection tended to diminish the strength of the associations, but

only to a small extent. Finally, the associations could be due to other (observed

or unobserved) factors causing both (i.e., confounders). Most studies use

statistical adjustment to minimise the impact of measured factors. However, it is

unclear what unobserved factors could explain the associations.

Conclusions

This is the first review for built environment correlates of physical activity among adults

specific to the Australian context. We found convincing evidence that people who live in

neighbourhoods with a large availability of destinations within walking/cycling distance

are more likely to engage in physical activity. Objectively measured distance to transit,

destinations and land use measures supported this conclusion. Likewise, self-reported

measures of destinations and design indicated a positive relationship with physical

activity. On the other hand, for objectively measured density and design, the evidence of

association with physical activity was insufficient or inconclusive. This finding should be

49

interpreted cautiously, as without density there would not be people to use and support

the destinations, and design features such as connectivity enable people to reach

destinations. In this review we found that commonly cited correlates of physical activity

in international literature also apply to Australia (destinations, diversity and measures of

walkability).

This review has also demonstrated that results for objectively measured built

environment features differ from those for self-reported measures. Investigating

objective and self-reported measures of built environment attributes has been

recommended in the literature, as these are likely to relate differently to physical activity

outcomes. For example, even though a neighbourhood could be unsafe in terms of

objective measures of crime, people may not perceive this and rate crime as a non-

issue for being physically active (or vice versa).

Results presented in this review are of use to policy makers in the health sector who

advocate for the inclusion of physical activity in urban and transport planning.

50

Chapter 5 The effects of built environment attributes on physical

activity-related health and healthcare costs outcomes in Australia


Chapter 4 indicates that for the Australian context, people living in neighbourhoods with

a wide range of destinations, diverse use of land, and high walkability are more likely to

be physically active compared with those living in areas where such attributes are

lacking. Showing the potential health and healthcare cost outcomes associated with

changes in specific features of the BE would assist to build-up the case for a broader

consideration of outcomes in the assessment of BE interventions. In this chapter, we

used evidence from the correlation of BE attributes with PA from studies included in

Chapter 4 to predict HALYs and healthcare costs of changes in exposure to the BE.

Specifically, Research Question 2 is addressed: What are the physical activity-related

health externalities and healthcare costs associated with changes in the built

environment in Australia?

Citation

Zapata-Diomedi, B, Herrera, AMM & Veerman, JL 2016, 'The effects of built


outcomes in Australia', Health & Place, vol. 42, pp. 19-29. doi:

10.1016/j.healthplace.2016.08.01067

Order Detail ID: 70181942

Health & place by PERGAMON. Reproduced with permission of PERGAMON in the

format Thesis/Dissertation via Copyright Clearance Center.

Authors’ contribution: Zapata-Diomedi B adapted the research question from the

research report prepared for the New South Wales Department of Health, updated

the methods and prepared the input data, conducted the analysis and wrote the

6 A corrigendum was published online 2 May 2017. The version presented here includes corrigendum. 7 Appendices are available as online material accessible via this link: http://www.sciencedirect.com/science/article/pii/S1353829216301368?via%3Dihub

http://www.sciencedirect.com/science/article/pii/S1353829216301368?via%3Dihub

51

manuscript. Herrera AMM provided epidemiological estimates. Veerman JL advised

on the research question and methods, critically reviewed and co-edited the paper.

52

Abstract

Attributes of the built environment can positively influence physical activity of urban

populations, which results in health and economic benefits. In this study, we derived

scenarios from the literature for the association built environment-physical activity and

used a mathematical model to translate improvements in physical activity to health-

adjusted life years and healthcare costs. We modelled 28 scenarios representing a

diverse range of built environment attributes including density, diversity of land use,

availability of destinations, distance to transit, design and neighbourhood walkability.

Our results indicated potential health gains in 20 of the 28 modelled built environment

attributes. Healthcare cost savings due to prevented physical activity-related diseases

ranged between A$2,800to A$99,600 per 100,000 adults per year. On the other hand,

additional healthcare costs of prolonged life years attributable to improvements in

physical activity were nearly 50 percent higher than the estimated healthcare costs

savings. Our results give an indication of the potential health benefits of investing in

physical activity-friendly built environments.

Key Words

Built environment, physical activity, health, economic evaluation, health impact

assessment

53

Introduction

In Australia, just over half of the adult population meets the recommended physical

activity (PA) guidelines (168). This is a public health concern, given the strong evidence

of a causal association between low levels of physical activity and ischemic heart

disease, stroke, colon cancer, breast cancer in women, and type 2 diabetes (46). The

high prevalence of physical inactivity in Australia is taking its toll with nearly 10,000

premature deaths and 31,000 years lived with disability annually (169). A physically

inactive population also represents an economic burden for the society by means of

high healthcare costs and loss of productivity (170).

Population levels of physical activity could be increased via multilevel approaches that

include the individual, institutional, community, and built and policy environments (171).

The built environment (BE), defined as those elements of the environment that are man-

made, including transportation systems, urban planning, and individual buildings (10 p.

28), has drawn increasing attention to its effect on health. This is reflected in the

exponential growth over recent years of studies investigating the links between physical

activity and built environment attributes (11, 14, 126, 172, 173). These studies have

shed light on the effect of the built environment on levels of physical activity. However,

demonstrating the potential health value of built environments that facilitate physical

activity may help to convince policy makers to consider health impacts in project

appraisals.

In recent years, a number of quantitative studies have been conducted to predict health

and economic outcomes of built environment interventions. Health impact assessment

(HIA) studies mostly investigated hypothetical or policy scenarios, including health

impacts via physical activity, air pollution, and road injuries. For example, Woodcock

and colleagues developed the Integrated Transport and Health Impact Modelling

(ITHIM) tool and applied it to assess transport and urban form scenarios in the United

Kingdom (UK), Europe, India and the United States (174). In one of the applications of

ITHIM, three alternative urban land transport scenarios (low-carbon emission motor

vehicles, increased active travel and a combination of both) were assessed for London,

UK and Delhi, India (175). The findings from this study indicated that decreased use of

motor vehicles and more active travel produced the highest health benefits with 7,332

averted disability-adjusted life years in London and 12,516 in Delhi on average per year

54

per million population. A recent systematic review of HIAs and economic evaluations

assessing mode shifts towards active transport found that in most of the included

studies, health benefits from physical activity outweighed other potential health harms of

active transport (e.g. road injuries and greater exposure to air pollution) (37). The

literature in the field is now advancing towards more specific scenarios linking built

environment to physical activity, followed by health impact assessments and economic

evaluations as opposed to basing prediction on hypothetical scenarios. For instance, a

recent study conducted cost-benefit analyses (CBAs) of proposed built environment

changes designed to improve walkability in three different communities: one urban, one

suburban, and one rural (123). In this study estimates for the association between a

walkability score and sidewalk density were used to predict changes in walking for

transport. The study found that the health benefits of the built environment projects

exceeded the project costs in the urban area and the rural town, with benefit-cost ratios

of 20.2 (95% CI: 8.7-30.6) and 4.7 (95% CI: 2.1-7.1). The suburban project's costs

exceeded benefits by 40% (benefit-cost ratio= 0.6, 95% CI 0.3-0.9). Unlike the urban

and rural projects, the suburban project involved only the installation of sidewalks,

without other improvements such as addition of walking destinations, in an area that

was lacking in destinations. Gibson and colleagues recently developed a simulation

model linking changes in the built environment to time spent walking which was

translated into health and economic outcomes (176). The study results indicated

potential economic benefits of US$ 234 million (95% CI: US$53-US$393 million)

attributable to decreased mortality and diseases prevalence. A benefit-cost ratio of 29

(95% CI: 6.5-48) was estimated including only the cost of sidewalk infrastructure.

In Australia, building and maintaining healthy places has become a priority given the

rising levels of chronic diseases (177). Creating healthy built environments is already on

the agenda of health professionals, who are working closely with urban planners to

influence city designs that support healthy lifestyles (178). However, for the inclusion of

physical activity in urban and transport projects, context specific estimates for the

association built environment-physical activity, in combination with agreed methods to

determine the health benefits of physical activity are required.

In this study, we quantified physical activity-related improvements in mortality and

morbidity measured in health-adjusted life years (HALYs) associated with specific built

environment attributes along with potential savings/increases in healthcare costs for the

55

Australian context. The results can serve as a reference for the inclusion of physical

activity-related health outcomes in the appraisal of built environment projects. This

research originated as an initiative from the Centre for Population Health, Government

of New South Wales (NSW), to demonstrate the potential costs and benefits of changes

in urban form (built environment).

Methods

We reviewed the Australian literature assessing the association BE-PA for the adult

population and used reported effect estimates to quantify the potential health benefits

and healthcare costs associated with improving population levels of PA attributable to

the BE. There are three sections to our analysis: (1) selection of BE attributes; (2)

estimation of change in PA attributable to the BE expressed as average minutes of PA

per week across the population; and (3) translation of changes in population levels of

PA into HALYs gained and healthcare costs, using a mathematical model. We explain

each step in turn (Figure 5-1).

Figure 5-1 Analytical framework of the process of quantifying HALYs and healthcare costs of changes in exposure to selected built environment attributes.

Selection of built environment attributes

We reviewed the current Australian literature for the association BE-PA for the adult

population (18 years +) (For complete review see Zapata-Diomedi and Veerman (179)).

Given the wide diversity of BE attributes reported, we grouped them in seven

categories, including five of the six “D’s” from Ewing and Cervero (134) (density,

diversity of land use, availability of destinations, distance to transit, and design) plus

measures of safety and neighbourhood walkability. We assessed studies for the quality

56

of their design, representativeness of the data, and control for confounding variables

using tools applied for similar purposes (126). We only modelled attributes from studies

of good and fair quality that measured the BE objectively and were based on samples of

over 1,000 individuals.

Estimation of changes in physical activity

Three types of measures for the association BE-PA were used in the source literature:

(1) odds ratios for the likelihood of doing PA for a given BE exposure (139, 141, 148,

151, 154); (2) beta coefficients for the additional time or sessions of PA for a given BE

exposure (67, 140, 150) and (3) marginal probabilities of doing PA for those exposed

compared to non-exposed to a given BE attribute (140). Given the diversity of reporting

styles we applied different methods to translate effect estimates into average population

change in minutes of PA per week.

Two steps were required to translate OR into average additional minutes of PA across

the population. Firstly, we converted OR into relative risks (RR) to estimate the

additional proportion doing PA if exposed to an alternative BE. We used the formula

proposed by Grant (180) which was developed by Zhang and Yu (181) to convert OR to

RR (Formula 1).

(1) 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑅𝑖𝑠𝑘 =𝑂𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜

(1−𝑝0+(𝑝0∗𝑂𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜))

Here, 𝑝0is the incidence of the outcome of interest in the non-exposed group (physical

activity among those not exposed to the built environment of interest). None of the

source studies provided information for 𝑝0, hence we assumed that this was equivalent

to the prevalence of PA for the sample under consideration (sample prevalence physical

activity in Table 5-4 in Results section). Our assumption is likely to be an over estimation

of 𝑝0 (we would expect that those not exposed would be less physically active),

therefore we conducted a sensitivity analysis to explore the impact of alternative

assumptions (see univariate sensitivity analysis). Secondly, we assumed that those

taking up PA would increase the weekly dose to reach the level equivalent to the

sample mean PA (sample weekly dose of physical activity in Table 5-4 in Results

section). We conducted a sensitivity analysis to test our results to the assumption made

on additional minutes (see univariate sensitivity analysis). RR and sample mean

minutes of PA per week were then applied to calculate the change in average minutes

57

of PA across the population (Formula 2). The first component of the left hand side of the

formula indicates the additional proportion doing PA if exposed to an alternative BE

which is then multiplied by the sample baseline minutes of PA to obtain the average

change in minutes of PA across the population.

(2) ∆ 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 𝑜𝑓 𝑃𝐴 𝑎𝑐𝑟𝑜𝑠𝑠 𝑡ℎ𝑒 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 = (𝑅𝑅 ∗ 𝑆𝑎𝑚𝑝𝑙𝑒 𝑝𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 𝑃𝐴 −

𝑆𝑎𝑚𝑝𝑙𝑒 𝑝𝑟𝑒𝑣𝑎𝑙𝑒𝑛𝑐𝑒 𝑃𝐴) ∗ 𝐵𝑎𝑠𝑒𝑙𝑖𝑛𝑒 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 𝑝𝑒𝑟 𝑤𝑒𝑒𝑘

Beta coefficients were reported for three scenarios from two studies (67, 150) and we

interpreted them as the average increase in time/sessions of PA per week across the

population for a given change in exposure. For instance, a study reported that every

additional transport destination (67) within 800-1600 metres of a person’s residence was

associated with an average increase in 5.8 minutes of walking per week. In our

scenario, we assumed that the whole population would have one such additional

destination.

Thirdly, McCormack and colleagues (140) reported the marginal effect of changes in the

BE on the proportion of the population that did any walking, as well as the change in the

average minutes per week walked among the walkers. Where this was the case, we

incorporated both effects in our calculation of the change in minutes walked across the

population.

A number of included studies presented effect sizes for more than one PA threshold

(e.g. walk>30/60/90 mins/wk.), domain (PA for transport, recreation or both) or more

than one model were used. All decisions regarding chosen effect sizes are presented in

Table S1 of the supplementary material.

Mathematical model

We translated changes in average minutes of PA across the adult population into

undiscounted HALYs using an updated version of the mathematical model developed

for the Assessing Cost Effectiveness in Prevention (ACE Prevention) project (94). Using

2010 as the base year, we discounted 3% per annum to healthcare costs (182).The

supplementary material (Section 2.2) gives a detailed description of the model and input

parameters.

The mathematical model uses a macro simulation approach based on the proportional

multi-state life table. It calculates changes in the occurrence of PA related diseases and

58

‘health-adjusted life years’ (HALYs) (96). One HALY is the equivalent of one year in full

health that is gained due to avoidance of disease (adjusted for severity) and

postponement of death. The analysis is conducted by comparing health outcomes

associated with a ‘status quo’ scenario against those in an alternative scenario in which

PA levels are changed. Health outcomes were calculated from changes in the

occurrence of diseases causally related to PA (ischemic heart disease, stroke, type 2

diabetes, breast cancer and colon cancer) (47). Incidence rates for each disease are

modified via potential impact fraction (PIF) calculations, which gives the proportional

change in incidence as a function of a change in exposure, using the “relative risk shift”

method (101) (See Figure S2 in the supplementary material). That is, rather than

proportions moving to higher PA categories (e.g. from inactive to insufficiently active),

the population remains in the same category (inactive, insufficiently active, etc.) but the

risk of disease is reduced for that category. Changing incidence rates has an impact on

the number of prevalent cases in the future and consequently mortality and years lived

with disability (compared to the base case scenario). To calculate the PIF we used

information for PA prevalence and RRs before and after an improvement in PA. Danaei,

Ding (47) proposed a four-tier dose-response for PA and health outcomes: highly active

(≥1,600 Metabolic Equivalents of Task minutes (MET-minutes)/wk. and 1h/wk of

vigorous PA), recommended level active (600≤MET-minutes/wk. ≤1,600 and 1 h of

vigorous PA/wk. or 2.5 h of moderate PA/wk.), insufficiently active (0<MET-minutes/wk.

≤600 or <2.5 h/wk. of moderate PA) and inactive (no moderate or vigorous PA). PA

prevalence was derived from mean minutes spent in a usual week doing moderate PA,

vigorous PA, walking for transport and walking for recreation, with the included

categories being mutually exclusive (57). We translated population mean minutes of PA

per week into MET-minutes, applying intensity values from the physical activity

compendium (183). We fitted linear functions to reported RRs with decreasing levels of

risk associated with increasing levels of weekly energy expenditure (mean METs per

week) (94). We tested the sensitivity of our results by assuming an alternative non-

linear dose-response function. The source studies lacked of information regarding

likelihood of doing PA according to PA membership; hence we assumed that all groups

(inactive, insufficiently active, etc.) increased PA equally. Furthermore, given the nature

of our macro approach for modelling health outcomes we modelled the “average

change in PA” rather than individual change. Those in the highly active group (≥1,600

59

MET-min/wk.) had a relative risk of one in the source literature, implying no additional

benefit from extra physical activity (47).

The model requires baseline age and sex specific epidemiological and demographic

data, prevalence of the risk factor (PA), relative risks for PA related diseases, MET-

minutes values and healthcare costs (Table 5-1). Given that type 2 diabetes is a risk

factor for cardiovascular disease, relative risks were applied to incorporate the

increased risk of ischemic heart disease and stroke among those with type 2 diabetes.

To avoid double counting, we reduced the direct effect of PA on ischemic heart disease

and stroke commensurately, using correction factors from the Global Burden of Disease

(GBD) study (184). Healthcare costs for PA-related diseases were calculated by dividing

total cost related to a disease by the number of incident cases (breast cancer and colon

cancer) or prevalent cases (ischaemic heart disease, stroke and type 2 diabetes).

Healthcare costs for the modelled diseases are from the original ACE-prevention study,

which used data from the Disease Costs and Impact Study 2001 prepared by the

Australian Institute of Health and Welfare, inflated with the Health Price Index (185).

Healthcare costs due to any other diseases that occur across the life course are

estimated in the same fashion (if an intervention prolongs people’s lives, they spend

more in healthcare). We refer to this costs as all other healthcare costs in added life

years.

Table 5-1 Proportional multi-state life table inputs

Input parameter Source

2010 mortality rates and population numbers Australian Bureau of Statistics (186), Australian Bureau of Statistics (187)

2010 epidemiological data (prevalence, incidence, case fatality and mortality)a

Institute for Health Metrics and Evaluation (188)

Prevalence of physical activity (supplementary material Figure S3)

National Nutrition and Physical Activity Survey Basic Confidentialised Unit Record File (CURF) (57)

Physical activity related diseases relative risk (supplementary material Table S2)

Danaei et al. (47)

Relative risks of ischaemic heart disease and ischaemic stroke due to diabetes (supplementary material Table S2)

Asia Pacific Cohort Studies Collaboration (189)

Mediating effect factors for diabetes in the association physical activity-ischemic heart disease/ischemic stroke

GBD 2013 study (184p. 711 Supplementary Material)

MET-minutes (walking 3.5 and cycling 5) Ainsworth, Haskell (183)

Healthcare costs (supplementary material Tables S3 and S4)

ACE Prevention study

a Epidemiological data for the five physical activity related diseases (ischemic heart disease, stroke, type 2 diabetes, colon cancer and breast cancer in women) were derived with the help of DISMOD II

60

(available free of charge at http://www.epigear.com/index_files/dismod_ii.html) to obtain data in metrics not explicitly reported (incidence and case fatality from prevalence and mortality).

Uncertainty and sensitivity analyses

Ninety-five percent uncertainty intervals were determined for all outcome measures by

Monte Carlo simulation (2,000 iterations), using the Excel add-in tool Ersatz (Epigear,

Version 1.33). Uncertainty parameters are presented in Table 5-2.

Table 5-2 Uncertainty parameters for evaluation health effects

Parameter Mean (SE) Distribution Source

Relative Risks of diseases

See Table S2 in supplementary material

Normal (Ln RR)

Physical activity: (47) Diabetes: (189)

Healthcare costs

See Table S3 and S4 in supplementary material

Uniform

Australian Institute of Health and Welfare Impacts Study 2001. Maximum/minimum assumed at ±25% of mean value

Mediating effect diabetes on ischemic heart disease (IHD) and stroke

IHD: 0.14 (0.02) Stroke: 0.08 (0.03)

Normal GBD 2013 Risk Factors Collaborators (184p. 711 Supplementary Material)

Minutes per week See Table 5-4 Lognormal McCormack, Shiell (140), Koohsari, Sugiyama (150)

Odds ratios See Table 5-4 Lognormal See Table 5-4 studies reporting Odds ratios

We tested our results to the sensitivity of a number of assumptions we had to make

given the lack of information provided in the studies reporting the modelled scenarios as

well as decisions inherent to our mathematical model. To translate OR into additional

minutes per week we made two assumptions, one on the 𝑝0 value used in Formula 1

and the other on the additional minutes per week for those increasing PA. We tested the

sensitivity of results of varying both parameters upwards and downwards. We also

tested the sensitivity of our results to discounting HALYs and using a higher rate for

healthcare costs. Given the increasing literature suggesting a curvilinear association for

PA with specific diseases (190) we modelled an alternative scenario assuming that PA

is log linearly associated with a power transformation in MET minutes per week (0.75)

(See Figures S4 and S5 in the Supplementary Material). Lastly, we produced estimates

without taking into account the mediating effect of diabetes in the association PA-

cardiovascular disease. A summary of sensitivity analyses performed is presented in

Table 5-3.

61

Table 5-3 Univariate sensitivity analysis

Parameter Base case Sensitivity

Physical activity estimates

Sample weekly dose of physical activity/Effect estimatea

See Table 4

±50%

𝑝0 (see formula 1) Sample prevalence physical activity (see Table 4)

-20%

Mathematical model

Discount rate health outcomes and healthcare costs

0% health and 3% healthcare costs per annum

3% health effects and 6% healthcare costs (191)

PA RR Linear Log-linear with power transformation of MET-mins/wk.

Potential impact factor

N/A Exclude mediation effect of diabetes in the association physical activity-cardiovascular disease

N.B 1000 iterations for Monte Carlo simulation a In the study by Giles-Corti et al. 2013 only p-values were reported from which we could not derive uncertainty parameters, hence we applied sensitivity analysis to the additional minutes per week as a result of increases in the number of destinations.

Results

Scenarios

We modelled a total of 28 scenarios from eight studies (67, 139-141, 148, 150, 151,

154) in density (n=3), diversity (n=2), design (n=7), destinations (n=6), distance to transit

(n=4) and walkability indices (n=6). No studies for the safety category met the inclusion

criteria. We present evaluated scenarios in Table 5-4 (e.g., density, diversity, etc.),

detailing the change in the built environment assessed (see supplementary material

Table S5 for studies’ details). Besides, we provide information on the outcome

measured in the scenarios (e.g. walking, cycling) and measures of effect (odds ratios,

beta coefficients and marginal effects + beta coefficients). Reported baseline data for

the sample prevalence of PA and sample weekly dose of PA served to translate OR to

additional minutes of PA per week as explained in the methods section.

62

Table 5-4 Built environment attributes modelled

Category Scenario/Study/

Location Change in built environment attribute Outcome

Effect estimate

(SE)

Baseline data

Measure of effect

Sample prevalence

physical activity

Sample weekly dose of physical activity

DE

NS

ITY

Density/ Christian et al. (2011)/Perth (WA)

Density standardised to z-scores. One unit increase in density (1 SD) represents an increase of 8 dwellings per ha.d within a participant’s 1.6 km network service area

Any walking Odds Ratio

1.04 (0.06)a 62% 93.5

Density/Knuiman et al. (2014)/Perth (WA)

Density standardised to z-scores. One unit increase in density (1 SD) represents an increase of 8 dwellings per ha.e within a participant’s 1.6 km network service area

Walking for transport

Odds Ratio

0.96 (0.09) a 33% 18.75f

Density/Wilson et al (2011)/Brisbane (QLD)

Decrease from 9205 (mean lowest quintile) to 650 (mean highest quintile) average size of residential zone lande within a one-kilometre radius of participant’s residence


1.37 (0.12) a 23% 30g

DIV

ER

SIT

Y

Land use mix (LUM)i/Christian et al. (2011)/Perth (WA)

LUM standardised to z-scores. One unit increase in the LUM represents an increase in 0.15 units in diversityd within a participant’s 1.6 km network service area


Odds Ratio

1.15 (0.05) a 26 % 26

LUMi/Knuiman et al. (2014)/Perth (WA)

LUM standardised to z-scores. One unit increase in the LUM represents an increase in 0.15 units in diversitye within a participant’s 1.6 km network service area


Odds Ratio

1.33 (0.07) a 33% 18.75f

DE

SIG

N

Connectivity/ Christian et al. (2011)/Perth (WA)

Connectivity standardized to z-scores. One unit increase represents an increase of 18 three or more ways intersections per km2 d within a participant’s 1.6 km network service area


Odds Ratio

1.15 (0.06) a 26 % 26

Connectivity/ Koohsari et al. (2014)/Adelaide (SA)

Increase from 1 to 10 intersections (3-way or more) per km2. Mean 245 (range 12 to 901) within a participant’s Census Collection Districts (CCD) area.


Beta coefficient

0.27 (0.06)b

63



Effect estimate

(SE)

Baseline data

Measure of effect

Sample prevalence

physical activity


Connectivity/ Knuiman et al. (2014)/Perth (WA)

Connectivity standardized to z-scores. One unit increase represents an increase of 18 three or more ways intersections per km2 e within a participant’s 1.6 km network service area


Odds Ratio

1.13 (0.06) a 33% 18.75 f

Connectivity/ Wilson et al (2011)/Brisbane (QLD)

Increase from 4 (mean lowest quintile) to 51 (mean highest quintile) 4-way intersectionse within a one-kilometre radius of a participant’s residence


1.44 (0.13) a 23% (43%) 30g

Sidewalks/ McCormack et al. (2012)/Perth (WA)

10 km. increase in sidewalk availability within a participant’s 1.6 km network service area

Transport walking

Marginal effect +Beta

coefficient

2.97%c, 5.38 (3.01)c

Off road bikeways/Wilson et al (2011)/Brisbane (QLD)

Increase from 0 km. (mean lowest quintile) to 7 km. (mean highest quintile) of off road bikewayse within a one-kilometre radius of a participant’s residence


1.34 (0.11) a 23% 30g

Street lights/Wilson et al (2011)/Brisbane (QLD)

Increase from 315 (mean lowest quintile) to 783 (mean highest quintile) of street lights within a one-kilometre radius of a participant’s residence


1.25 (0.12) a 23% 30g

DE

ST

INA

TIO

NS

Transport destinations/ Giles-Corti et al. (2013)/Perth (WA)

Per increase in one transport destination (after

relocation)/ Post office, bus stops, delicatessens,

supermarkets within 800 m of a participant’s

residence and train stations, shopping centres or

CD and DVD stores within 1.6 km

Transport walking

Beta coefficient

5.8h

Recreation destinations/ Giles-Corti et al. (2013)/Perth (WA)

Per increase in one recreational destination (after relocation)/ Beaches within 800 m of a

Recreational walking

Beta coefficient

17.6h

64



Effect estimate

(SE)

Baseline data

Measure of effect

Sample prevalence

physical activity


participant’s residence and parks and sport fields within 1.6 km

Distance to retail/Wilson et al. (2011)/Brisbane (QLD)

From a retail zone within >1 km to one within >0.2 km within the street network distance in kilometres from a participant’s residence


1.46 (0.13) a 23% 30g

Distance to parks/Wilson et al. (2011)/Brisbane (QLD)

From a park zone land within >1 km to one within >0.2 km within the street network distance in kilometres from a participant’s residence


1.08 (0.13) a

23% 30g

Destinations/ Knuiman et al. (2014)/Perth (WA)

From =<3 to 4-7 general destinations (services, convenience stores and public open spaces) accessible along the street network within 1.6 km from a participant’s residence

Transport walking

Odds Ratio

1.08 (0.15) a 33% 18.75 f

Destinations/ Knuiman et al. (2014)/Perth (WA)

From =<3 to 5-15 general destinations (services, convenience stores and public open spaces) accessible along the street network within 1.6 km from a participant’s residence

Transport walking

Odds Ratio

1.40 (0.21) a 33% 18.75 f

DIS

TA

NC

E T

O T

RA

NS

IT

Bus stops/Knuiman et al. (2014)/Perth (WA)

From 0-14 to 5-19 general destinations bus stops accessible along the street network within 1.6 km from a participant’s residence

Transport walking

Odds Ratio

1.99 (0.16) a 33% 18.75 f

Bus stops/ Knuiman et al. (2014)/Perth (WA)

From 0-14 to =>30 general destinations bus stops accessible along the street network within 1.6 km from a participant’s residence

Transport walking

Odds Ratio

2.33 (0.20) a 33% 18.75 f

Train station/ Knuiman et al. (2014)/Perth (WA)

Train station accessible along the street network within 1.6 km from a participant’s residence

Transport walking

Odds Ratio

1.79 (0.29) a 33% 18.75 f

Transit stops/ Access to the nearest transit stop within >0.2 km compared to >1 km within the street network


1.34 (0.16) a 23% 30g

65



Effect estimate

(SE)

Baseline data

Measure of effect

Sample prevalence

physical activity


Wilson et al. (2011)/Brisbane (QLD)

distance in kilometres from a participant’s residence

WA

LK

AB

ILIT

Y I

ND

EX

Walkability indexj/ Christian et al. (2011)/Perth (WA)

Increase in one unit in the index (z-score) within a participant’s 1.6 km network service area


1.06 (0.02) a 62% 93.5

Walkability indexk-Suburb/Learnihan (2011)/Perth (WA)

Highly walkable compared to low within a participant’s suburb area

Transport walking

Odds Ratio

1.63 (0.15) a 36% 26

Walkability indexk- Census Collection District (CCD) scale/Learnihan (2011)/Perth (WA)

Highly walkable compared to low within a participant’s CCD area

Transport walking

Odds Ratio

2.07 (0.13) a 36% 26

Walkability indexk-15 mins walk scale/ Learnihan (2011)/Perth (WA)

Highly walkable compared to low within a participant’s 15 minutes walking area

Transport walking

Odds Ratio

2.79 (0.15) a 36% 26

Walkability indexj/ McCormack et al. (2012)/Perth (WA)

Increase in one unit in the index (z-score) within a participant’s 1.6 km network service area

Transport walking

Marginal effect +Beta

coefficient

2.16%c , 3.32 (6.21)c

Walkability indexk/Owen et al. (2010)/Adelaide (SA)

High compared to low within a participant’s CCD area

Any cycling Odds Ratio

1.82 (0.19) a 14% 10

a Odds ratio (OR) (SE(lnOR)). Standard errors were estimated from the confidence interval applying the formula proposed on page 33 of Ersatz user guide (102). b β coefficient from negative binomial regression converted into additional minutes per week by multiplying by 10 which represented the minimum walking time for a trip (outcome assessed). Similar procedure was followed to estimate the standard error (150). c Two stage modelling approach: Probit regression to estimate marginal probabilities followed by OLS to estimate β for additional walking minutes (140).

66

d We assumed that the information provided by Knuiman et al. for the value of 1SD applies here as both studies are based on the same data set (RESIDential Environment Study (RESIDE)). e Study authors provided information for the value of the mean and SD of the built environment attributes assessed. f Average trips over 4 data collections by trip time of 15 minutes (148). g Lower bound walking range assessed (see Table S1 supplementary material). h β coefficient from Generalize Linear Mixed Models representing the effect of one unit change in the continuous independent variable on the continuous outcome (walking). i LUM includes the following land uses: ‘Residential’, ‘Retail’, ‘Office’, ‘Health, welfare and community’ and ‘Entertainment, culture and recreation’ land use classes (141). j Walkability index based on three built environment characteristics: residential density, street connectivity and land use mix. k Walkability index based on four built environment characteristics: residential density, street connectivity, land use mix and retail floor area. WA: Western Australia; QLD: Queensland; SA: South Australia

67

Health outcomes

In the following paragraphs we present findings per 100,000 adults per year for HALYs

gained for the 28 evaluated scenarios. There was large variability in the results, most of

which can be attributed to the different reporting methods in the source literature for the

association BE-PA. All results are presented in Table 5-5 and discussed in the following

paragraphs.

Density

Only one of the three density scenarios indicated statistically significant results for health

outcomes, with estimated HALYs gained of 1.98 for a decrease from 9,205m2 to 650m2

of average residential zone land per hectare (10,000 m2) (Wilson et al.’s scenario).

Wilson and colleagues’ scenario represents approximately an increase from 1 to 15

dwellings/ha.8, which is considerably higher than the increase in 8 dwelling/ha. for the

scenarios from Knuiman et al. and Christian and colleagues. Despite the scenarios

derived from Knuiman and co-authors and Christian et al. being based on the same

study (RESIDential Environment Study (RESIDE)), their results differed. One possible

explanation is that Knuiman et al. evaluated walking for transport whereas for Christian

and co-authors we used estimates for walking for any purpose (see Table S1 of the

supplementary material). Further, the estimate from Knuiman et al. was based on

longitudinal data collected over four waves whereas Christian et al. used baseline data.

Diversity

On average, an improvement in diversity represented by one unit increase in the

composite measure of LUM, within the area of 1.6 km street network from a participant’s

residence, could potentially accrue 0.94 HALYS gained (scenario derived from Christian

et al.) to 1.37 (scenario derived from Knuiman et al.). The interpretation of improvement

in LUM is rather difficult. However, the source information did not allow us to translate

such change into an explicit scenario (see explanation i from Table 5-4). While both

estimates of effect of LUM on PA are based on the same study, the same conceptual

definition of LUM, and the same physical activity outcome (walking for transport), the

8 If the average residential land size is 650m2, there would be approximately 15 houses in a hectare (10,000/650=15.38), whereas only one house fits in a hectare for an average land size of 9,205m2 (10,000/9250=1.08),

68

results are different. The odds ratio from the longitudinal analysis by Knuiman and

colleagues (see Table 5-4) and prevalence of walking for transport at the baseline are

greater to those in the cross-sectional study by Christian et al. This implies a greater

proportion taking up walking for transport in the scenarios based on the analysis by

Knuiman et al. However, the additional weekly dose of transport in the scenario derived

from Knuiman and co-authors is smaller than that in the scenario resulting from

Christian et al.

Design

Seven scenarios for measures of design were evaluated, including connectivity (the

number of intersections within an area), availability of sidewalks or bikeways, and

number of street lights. The average HALYs gained from improvements in connectivity

ranged from 0.56 for an increase of 18 three- or more- way intersections per km2

(scenario derived from Knuiman et al.) to 3.03 for an increase from 1 to 10 three- or

more- ways intersections per km2 (scenario derived from Koohsari et al.). In Wilson et al.

walking for any purpose was evaluated, whereas in the rest of the scenarios the

outcome was walking for transport purposes. The mean HALYs gained for increases in

the availability of sidewalks and off-road bikeways ranged from 1.85 for a change in the

availability of bikeways from none to 7 km (scenario derived from Wilson et al.’s study)

to 4.82 for an additional 10 km of sidewalk (scenario resulting from McCormack et al.’s

analysis) within the neighbourhood area defined in the source studies. In the scenario

by McCormack and colleagues, walking for transport was evaluated, whereas any

walking was the outcome in Wilson et al.’s analysis. Lastly, an improvement in street

lights from 315 to 783 within 1 km from a participant’s residence accrues on average

1.36 HALYs gained (Wilston et al.) as a result of improvements in walking for any

purpose. However, the estimate for Wilson et al.’s scenario includes 0 in the uncertainty

interval.

Destinations

Improvements in walking for transport, in the scenario derived from Giles-Corti, resulted

in HALYs gains of 6.53 for an increase in one transport destination within the area of 1.6

km street network from a participant’s residence. Increasing general destinations was

not associated with statistically significant changes in walking based on Knuiman et al.’s

scenarios. Providing an additional recreational destination within 1.6 km street network

69

from residence accrues 19.81 HALYs in the scenario based on Giles Corti and

colleagues’ study. In Wilson et al., having a retail zone within less than 0.2 km

compared to less than 1 km results in potential HALYs gained of 2.45. Also, Wilson et

al. provided a scenario for an improvement in access to park land, from one within 1 km

to one within less than 0.2 km, however, they did not find a statistically significant

association.

Distance to transit

Two of the three scenarios derived from Knuiman et al. for improvements in the

availability of transit stops indicated health benefits due to increased walking for

transport. Increasing the availability of bus stops, from less than 14 to 15-19 within 1.6

km street network from residence, translates into HALYs gained of 3.39. Slightly higher

HALY gains of 4.14 can be achieved if the improvement is up to more than 30 bus

stops.

Walkability

All six evaluated scenarios indicated health benefits in terms of HALYS gained for

improvements in measures of walkability within the studies’ defined neighbourhood

areas. Average values from improvements in walking for transport, ranged from 3.23

HALYs for an increase in one unit in the standardised walkability index for the scenario

derived from McCormack et al.’s, to 7.2 for an improvement from low to high walkability

in the scenario resulting from Learnihan and colleagues’ analysis (15 minutes area

scale). In Christian et al., an increase in one unit in the standardised walkability index

would potentially accrue on average 1.44 HALYs due to improvements in walking for

any purpose. Lastly, in Owen and co-authors benefits from improvements in cycling

were modelled, with results indicating HALY gains of 1.56 for a change from low to high

walkability.

Healthcare costs

Savings in healthcare costs per year for PA-related diseases ranged from A$1,558 to

A$99,568 per 100,000 adults. On the other hand, healthcare costs in added life years

were approximately 50% higher than the savings obtained by having to treat fewer

cases of PA related disease in earlier years, even after discounting at 3% (Table 5-5). It

is important to note that there is great uncertainty in the healthcare costs estimates.

70

Table 5-5 HALYs, healthcare costs savings and all other healthcare costs in added life years per 100,000 people per year for built environment scenarios

Changes in built environment attribute HALYs Healthcare costs

(A$ 2010)a All other healthcare costs in added

life years (A$ 2010)a

Density. + 1 SD (Christian et al. 2011) 0.95 (-1.75 to 3.7) -$4,713 (-$21,286 to $9,380) $6,836 (-$12,808 to $28,544)

Density. + 1 SD (Knuiman et al. 2014) -0.17 (-1.01 to 0.72) $813 (-$3,784 to $5,754) -$1,182 (-$7,490 to $5,173)

Density. From 650 sqm2 to 9205 sqm2 average size of residential zone land within 1 km radius of residence (Wilson et al. 2011) 1.98 (0.33 to 3.75) -$9,837 (-$22,844 to $875) $14,200 ($1,481 to $28,946)

Land use mix. + 1 SD (Christian et al. 2011) 0.94 (0.24 to 1.71) -$4,634 (-$10,158 to $420) $6,703 ($870 to $12,986)

Land use mix. + 1 SD (Knuiman et al. 2014) 1.37 (0.65 to 2.11) -$6,813 (-$13,186 to $434) $9,851 ($2,651 to $16,316)

Connectivity. + 1 SD (Christian et al. 2011) 0.94 (0.19 to 1.69) -$4,642 (-$10,306 to $487) $6,695 ($741 to $13,121)

Connectivity. Increase from 1 to 10 intersections (3-way or more) (Koohsari et al. 2014) 3.03 (2.29 to 3.43) -$15,028 (-$23,449 to $1093) $21,729 ($7,362 to $27,232)

Connectivity. + 1 SD (Knuiman et al. 2014) 0.56 (0.02 to 1.09) -$2,790 (-$6,478 to $449) $4,036 (-$113 to $8,378)

Connectivity. From 4 to 51 four-way intersections (Wilson et al 2011) 2.34 (0.64 to 4.22) -$11,662 (-$25,706 to $692) $16,836 ($2,463 to $32,579)

Sidewalk. 10 km increase in sidewalk. (McComack et al. 2012)

4.82 (2.91 to 8.65) -$24,224 (-$52,545 to $32) $34,653 ($11,337 to $65,230)

Bikeways. From 0 km to 7km (mean highest quintile) of off road bikeways (Wilson et al 2011) 1.85 (0.41 to 3.47) -$9,150 (-$20,974 to $708) $13,215 ($1,674 to $26,576)

Street lights. From 315 to 783 street lights (Wilson et al 2011) 1.36 (-0.13 to 3.05) -$6,732 (-$17,833 to $1699) $9,722 (-$1,251 to $23,179)

Destinations. + 1 transport destination (Giles-Corti et al. 2013)

6.53 (5.02 to 7.25) -$32,812 (-$50,726 to $45) $46,971 ($16,988 to $57,660)

Destinations. + 1 recreational destination (Giles-Corti et al. 2013)

19.81 (15.22 to

22.01) -$99,568 (-$153,929 to $130) $142,537 ($51,560 to $174,973)

Destinations. From retail zone land within >1km to >0.2km (Wilson et al. 2011) 2.45 (0.74 to 4.4) -$12,189 (-$26,972 to $857) $17,609 ($3,411 to $34,273)

Destinations. From park zone land within >1km to >0.2km (Wilson et al. 2011) 0.44 (-1.04 to 2.13) -$2,228 (-$12,429 to $5458) $3,175 (-$7,444 to $15,693)

71

Changes in built environment attribute HALYs Healthcare costs

(A$ 2010)a All other healthcare costs in added

life years (A$ 2010)a

Destinations. From =<3 to 4-7 (Knuiman et al. 2014) 0.32 (-1.09 to 1.82) -$1,558 (-$10,026 to $5851) $2,282 (-$8,013 to $13,904)

Destinations. From =<3 to 8-15 (Knuiman et al. 2014) 1.56 (-0.52 to 3.73) -$7,735 (-$21,746 to $3514) $11,200 (-$3,971 to $28,523)

Bus stops. From 0-14 to 15-19 (Knuiman et al. 2014) 3.39 (1.67 to 5.22) -$16,852 (-$31,583 to $1168) $24,342 ($6,754 to $39,916)

Bus stops. From 0-14 to =>30 (Knuiman et al. 2014) 4.14 (2.02 to 6.33) -$20,546 (-$39,917 to $1686) $29,656 ($8,697 to $48,995)

Train station. Railway station present within 1.6km compared to no rail way station (Knuiman et al. 2014) 2.74 (-0.12 to 5.83) -$13,630 (-$34,381 to $3616) $19,691 (-$1,404 to $44,171)

Transit stop. From one within >1km to one within >0.2km (Wilson et al. 2011) 1.87 (-0.25 to 4.47) -$9,320 (-$25,001 to $2598) $13,473 (-$1,972 to $33,758)

Walkability index. + 1SD (Christian at al. 2011) 1.44 (0.28 to 2.67) -$7,119 (-$15,704 to $755) $10,312 ($1,270 to $20,446)

Walkability index neighbourhood scale. High walkable compared to low (Learnihan et al. 2011) 3.36 (1.2 to 5.55) -$16,704 (-$33,948 to $800) $24,086 ($4,054 to $42,223)

Walkability index CCD scale. High walkable compared to low (Learnihan et al. 2011) 5.12 (2.94 to 7.24) -$25,446 (-$45,885 to $1787) $36,754 ($11,128 to $55,864)

Walkability index 15 minutes area scale. High walkable compared to low (Learnihan et al. 2011) 7.2 (4.78 to 9.58) -$35,737 (-$61,919 to $2692) $51,646 ($16,012 to $74,379)

Walkability index. + 1SD (McCormack et al. 2012) 3.23 (1.63 to 8.89) -$16,156 (-$46,915 to $20) $23,170 ($6,143 to $65,448)

Walkability index. High walkable compared to low (Owen et al. 2010) 1.56 (0.45 to 2.91) -$7,744 (-$17,630 to $622) $11,180 ($1,834 to $22,081)

a Negative figures represent costs saving

72

Results from sensitivity analyses

The results are sensitive to some of the assumptions made in this study. Firstly, results

are sensitive to the assumption around the number of additional minutes per week for

scenarios derived from studies reporting odds ratios. Increasing or decreasing the dose

of physical activity for those taking up walking or cycling as a result of a change in the

built environment by 50%, translates to proportional changes in the estimated health

and healthcare costs values (Tables S6 and S7 of the supplementary material). Similar

sensitivity results were obtained from scenarios derived from studies reporting beta

coefficients for which we did not have uncertainty parameters (destinations scenarios

from Giles-Corti et al.). A lower value for p0 (20% lower) (incidence physical activity in

the non-exposed) used in the formula to translate odds ratios into relative risks resulted

in an upper variation in the estimated values ranging from 5% to 33% (supplementary

material Table S8). Notably, the scenarios for density and walkability derived from

Christian et al. were the most sensitive, which can be attributed to the high level of p0

(refer to Table 4 Sample prevalence physical activity). Excluding the mediating effect of

diabetes on cardiovascular diseases results in slightly higher estimates, 9% for HALYs

gained and PA-healthcare costs savings and 10% for healthcare costs in added life

years (supplementary material Table S9). Applying a 3% per annum discount rate for

health outcomes results into a decrease of 40% in potential HALYs gained

(supplementary material Table S10). Discounting healthcare costs at a higher rate (6%)

results in lower estimates ranging from 20% to 35% (Table S11 supplementary

material). Lastly, changing the shape of the dose response function for PA with health

outcomes to a curvilinear dose-response has a major impact with results doubling in

some cases (supplementary material Table S12).

Discussion

To our knowledge, this is the first study that attempts to estimate the potential health

gains and healthcare cost savings associated with specific attributes of the built

environment for the Australian context. Past studies specific to Australia provided

general estimates of economic value per kilometre walked or cycled that included both

mortality and morbidity measures of improvements in PA (30, 121). However, these

general estimates do not specify what built environment attributes need to be targeted to

achieve these benefits. The results from our research add to the existing literature by

73

producing a series of health and economic values for specific changes to the built

environment based on well-established methods of the proportional multi-state life table

(96). Our estimates could be used to incorporate the value of physical activity-related

health outcomes in HIAs and economic evaluations of interventions to the built

environment.

Overall, 20 of the 28 modelled scenarios indicated potential annual health benefits

represented by HALYs gained per 100,000 adults per year. Most of the health benefits

in terms of HALYs gained presented in Table 5-5 ranged from 1 to 7 per 100,000 adults

exposed to an improvement in the built environment per year. The greatest majority of

results for savings in healthcare costs of improvements in PA related disease ranged

between A$4,634 and A$35,737 per 100,000 adults per year. All other healthcare costs

in added life years ranged mostly between A$9,851 and A$51,646 per 100,000 people

(+18) per year. Our estimates are specific to the data collection areas in three main

Australian cities (Brisbane, Perth and Adelaide). However, in the absence of locally

derived alternatives, they could be used as a reference for other metropolitan areas with

similar characteristics.

To our knowledge, no other studies have evaluated the potential health outcomes in

terms of health-adjusted life years of improvements in the BE. Boarnet, Greenwald and

McMillan (192) did perform an analysis that had mortality as outcome measure. They

used regression analysis on travel survey data from Portland, Oregon, to quantify the

impact of built environment attributes (population/jobs density, number of intersections

and distance to business centre) on distance walked and translated improvements in

walking to lives saved. Their results suggested that at a minimum 0.0031 to 0.0912 lives

per 1000 people per year would be saved from improvements in the BE towards more

walkable places. These figures translate into 0.31 to 9.12 lives saved per year per

100,000 people. Even though our estimates are not directly comparable as we adjust

life years gained for disability, these include the range estimated by Boarnet and

colleagues.

Quantifying the potential health and healthcare costs attributable to improvements in the

BE involved a number of challenging assumptions. To assess the potential impact on

results of these assumptions, we conducted an extensive sensitivity analysis. The

greatest majority of studies reported results in terms of the odds of doing physical

74

activity for those exposed to the assessed BE feature, compared to those not exposed,

without indicating the dose. The only exception was the study by McCormack, Shiell

(140), which assessed not only the marginal probability for an individual walking if

exposed to an environmental attribute, but also the change in the average weekly dose

among those walking. As presented in our sensitivity results, our estimates are highly

sensitive to the assumption of the dose of PA for scenarios derived from studies

reporting odds ratios. Our estimates are also sensitive to discounting HALYs and

variations in the discounting rate for healthcare costs. However, the choice of discount

rate is dependent on the agency carrying out the evaluations, hence; it is not an issue of

empirical uncertainty but of choice. Whether health should be discounted has been

debated in the past, with some literature suggesting applying the same rate as cost as

well as conducting sensitivity analyses (182) while others recommend not discounting

health outcomes (193). Discounting the future is a common practice for monetary costs

to account for people’s time preference (individuals would rather have something good

today than something good in the future, and the reverse for something bad) (97, 194),

but is controversial when applied to the health of others, or of future generations.

Applying alternative dose-response function for the effect of physical activity on health

outcomes has a great impact on results. However, past studies also indicated major

variations in results depending on the dose-response function used (195).

Some further limitations related of this study should be discussed. Firstly, the diversity in

the ways in which different studies report their findings for the relationship of built

environment with physical activity outcomes hinders direct comparison and pooling, and

in some cases insufficient information is provided to enable meaningful interpretation.

The use of more uniform measurement methods for both exposure (instruments used

and domains measured) and physical activity would facilitate pooling and comparability

of results. Furthermore, the great variability of measurements methods and results of PA

exposure has a large impact in our estimated results. Secondly, there is potentially

some imprecision in the measurement of exposure in the source studies, which leads to

‘regression dilution bias’, that is, improved measurement of relevant exposures (i.e., BE

attributes) may lead to larger, more precise effect estimates. A further limitation is that

the greatest majority of scenarios are based on cross-sectional studies, which does not

allow for a direct causal interpretation. The association can be due to the built

environment influencing physical activity; this is the hypothesis underlying this research.

75

Alternatively, it could be due to physically active people choosing to live in

neighbourhoods that facilitate that behaviour. By adjusting for self-selection, some

studies try to avoid this ‘reverse causal’ interpretation. McCormack and Shiell (11)

systematically reviewed the international literature for the relationship BE-PA and found

that adjusting for self-selection tended to diminish the strength of the associations, but

only to a small extent. Nonetheless, the associations could be due to other (observed or

unobserved) factors causing both (confounding). Most studies use statistical adjustment

to minimise the impact of measured factors. From the literature we do not know whether

those taking up physical activity due to an intervention may respond by simultaneously

reducing other forms of physical activity. Along this analysis, we made the assumption

that there was no substitution effect, as has been done in the past (192). In our model,

the proportion of the population that is sufficiently active (~25%) receives no benefit from

additional physical activity, which may led to underestimation of health impact. Also,

there is growing evidence suggesting a causal association between PA and dementia

which were not included in our estimates resulting into a potential under estimation of

outcomes (54, 196).

Conclusion

In this research we produced estimates for the physical activity-related health benefits of

specific built environment attributes, and the economic value in terms of healthcare

costs these represent. To our knowledge, there has been no study in the past that has

attempted to demonstrate the potential health and economic value of such a broad

range of specific built environment attributes. The results of this study can be

incorporated into health impact assessments and cost-benefit analyses conducted to

inform infrastructural developments.

76

Chapter 6 A systematic review of economic analyses of active

transport interventions that include physical activity benefits


Attributes of the design of cities, including transportation systems can greatly improve

population levels of PA. Findings from Chapter 4 indicate that people living in walking

distance of transit destinations are more likely to walk compared with those not living in

walking distance of transit destinations. Chapter 5 shows that between three to four

HALYs could be gained annually per 100,000 adults living in a neighbourhood if

availability of transit stops improves. These studies provide important information as to

the potential health gains of specific changes in features of the BE related to the

transport sector. The next step is to investigate methods used in the field of transport to

conduct economic evaluations of interventions with a PA component. This chapter is a

systematic review of the literature of transport intervention evaluations that include PA

health-related benefits. The findings from this review serve to inform the developments

of Chapter 7 in terms of what types of economic evaluations are predominantly used in

transport. A similar literature review was conducted in the past (2008) (197). However,

since then a number of studies have been published in peer-reviewed and grey

literature, hence, this paper serves as an update. The quality of reporting was evaluated

in detail, which provides important information as to the state of the field compared with

international reporting standards for economic evaluations in the healthcare sector.

Specifically, in this chapter I and my co-authors address Research Question 3: What

economic evaluation methods have been used to model future health outcomes from

interventions in active transport?

77

Citation

Brown, V, Zapata-Diomedi, B9, Moodie, M, Veerman, JL & Carter, R 2016, 'A systematic

review of economic analyses of active transport interventions that include physical

activity benefits', Transport Policy, vol. 45, pp. 190-208. doi:


Order Detail ID: 70181927

Transport policy by WCTR Society Reproduced with permission of PERGAMON in the

format Thesis/Dissertation via Copyright Clearance Center.

Authors’ contribution: Zapata-Diomedi B conceived the research question,

formulated the methods with Brown V, wrote the methods and results and co-edited

the manuscript. Brown V wrote the abstract, introduction and discussion and co-

edited the manuscript. Moodie M, JL Veerman and Carter R critically reviewed the

paper.


10 Appendices are available as online material accessible via this link: http://www.sciencedirect.com/science/article/pii/S0967070X15300639


http://www.sciencedirect.com/science/article/pii/S0967070X15300639

78

Abstract

Physical inactivity is one of the leading causes for the growing prevalence of non-

communicable diseases worldwide and there is a need for more evidence on the

effectiveness and cost-effectiveness of interventions that aim to increase physical

activity at the population level. This study aimed to update a systematic review

published in 2008 by searching peer-reviewed and unpublished literature of economic

evaluations of transport interventions that incorporate the health related effects of

physical activity. Our analysis of methods for the inclusion of physical activity related

health effects into transport appraisal over time demonstrates that methodological

progress has been made. Thirty-six studies were included, reflecting an increasing

recognition of the importance of incorporating these health effects into transport

appraisal. However, significant methodological challenges in the incorporation of wider

health benefits into transport appraisal still exist. The inclusion of physical activity related

health effects is currently limited by paucity of evidence on morbidity effects and of more

rigorous evidence on the effectiveness of interventions. Significant scope exists for

better quality and more transparent reporting. A more consistent approach to the

inclusion of benefits and disbenefits would reinforce the synergies between the health,

environmental, transport and other sectors. From a transport sector perspective the

inclusion of physical activity related health benefits positively impacts cost effectiveness,

with the potential to contribute to a more efficient allocation of scarce resources based

on a more comprehensive range of merits. From a public health perspective the

inclusion of physical activity related health benefits may result in the funding of more

interventions that promote active transport, with the potential to improve population

levels of physical activity and to reduce prevalence of physical activity related diseases.

Keywords

Active transport, economic evaluation, physical activity

79

Introduction

Physical inactivity is the fourth leading risk factor for mortality worldwide (198) and is

one of the main contributors to the global burden of non-communicable diseases.

Physical inactivity increases the risk of many adverse health conditions, including

obesity, coronary heart disease, stroke, breast and colon cancer, diabetes, dementia

and depression (54, 170). Rates of physical inactivity are high worldwide, with

technological progress meaning that we now spend less energy in our everyday lives

than our predecessors (4, 199, 200). Coupled with the fact that we also have more

access to energy dense foods, this constitutes increasingly obesogenic environments

requiring ecological solutions (201-203). In order to address the observed low levels of

physical activity across populations, it is widely recognised that the incorporation of

more incidental physical activity into everyday life is required through environmental,

social, cultural and behavioural approaches (204).

Active forms of transport, such as walking, cycling and use of public transport, have

been recognised as possible avenues to increase the daily physical activity levels of

populations through incidental exercise, providing an alternative to more traditional

physical activity domains such as sport and exercise (205-207). Active transport is often

referred to as utilitarian physical activity, as it involves walking, cycling or use of public

transport for functional purposes. It is increasingly recognised that synergistic policies in

sectors outside of health, including that of transportation, may have significant potential

to improve physical activity rates and hence the health status of populations (208).

Ecological evidence suggests that countries with higher rates of active transport have

lower rates of obesity (209) and that a positive association may exist between motor

vehicle usage and body weight (210-213). Although establishing the health effects of

active transport policies and interventions is challenging, a recent systematic review of

trials and cohort studies found consistent support for the health benefits of active

transport over longer periods and distances (214).

This has led to increasing recognition of the importance of using a broad definition of

benefits in the economic evaluation of transportation policies and infrastructure (30, 87,

215). Table 6-1 lists the most common methods for economic evaluation, with a brief

definition given for each method. The transport sector traditionally uses cost benefit

analysis (CBA) for project appraisal, where costs and benefits are expressed in

80

monetary terms and health effects are most commonly limited to the effects of injuries

and exposure to environmental effects such as air pollution. This narrow incorporation of

health potentially undervalues active transport projects, especially in light of the

emerging evidence on the potential health benefits of walking and cycling for transport

and the well-recognised health benefits of physical activity (216).

Table 6-1 Methods for full economic evaluation

Economic evaluation method Definition

Cost Benefit Analysis (CBA) The expected benefits of an intervention are measured in monetary terms and compared to the costs of the intervention. Results are reported as cost per unit of benefit.

Cost Utility Analysis (CUA) The expected health outcomes of an intervention are measured in terms of the quality and quantity of life attributable to the intervention. Health outcomes can be expressed as disability adjusted life years (DALYs) or quality adjusted life years (QALYs). Results can be presented as cost per averted DALY or gained QALY.

Cost Effectiveness Analysis (CEA) Health outcomes are expressed as a unit of effect, for example life years saved or prevalent cases averted with an associated cost. Results can be presented as cost per life year saved or prevalent cases averted.

Following a number of early, pioneering studies (217-219), recent methodological

advances have been made in the inclusion of physical activity related health effects in

transport appraisal. A systematic review conducted in 2008 by Cavill et al. found 16

economic evaluations of transport infrastructure and policies incorporating physical

activity related health effects (197). At that time the approaches to the inclusion of

physical activity related health outcomes differed considerably among studies, as did

study quality and transparency. The review by Cavill et al. called for a more harmonised

approach and identified the method taken in the study by Rutter (217) as having the

greatest potential for inclusion of physical activity related health effects into transport

appraisal.

This knowledge was used in the development of the World Health Organisation (WHO)

Health Economic Assessment Tools (HEAT) for walking and cycling, with the aim of

devising a more consistent approach to monetising the physical activity related health

impacts of active transport for inclusion into CBA of transport projects (220). The HEAT

tool estimates the mean and maximum annual reduction in mortality attributable to an

increase in walking or cycling. The assessment of mortality benefits relies on a number

of assumptions which are clearly stated in the HEAT user guide (220). The economic

81

value of decreased mortality is estimated by applying the value of a statistical life (VSL).

The main justification for using the VSL lies on planners who are accustomed to this

valuation technique as the end users of HEAT. Due to a lack of evidence for the effect

of walking and cycling on morbidity HEAT currently however only incorporates mortality

effects, although the inclusion of morbidity effects has been identified as important in

future refinements of the tool.

It has now been several years since the original systematic review by Cavill et al. (197)

and the availability of the WHO HEAT tools. Whilst methodological advances in the

incorporation of physical activity related health effects into transport appraisal have been

made, it is uncertain whether this has translated into more routine incorporation of these

effects. In this paper we aim to provide an up-to-date overview of the literature through

the conduct of a systematic review of economic evaluations of transport interventions

and policies that include health effects of physical activity.

Methods

Inclusion criteria

To be considered for inclusion, studies had to meet the following criteria:

1. Be published in English between 1 January 1990 and 3 July 2014.

2. Be in the public domain, either as academic papers in peer reviewed journals or

studies from the ‘grey’ literature such as government reports and commissioned

documents.

3. Be a primary study. Reviews and commentaries were excluded.

4. Present a full economic evaluation (including CBA, cost utility analysis (CUA) or cost

effectiveness analysis (CEA)) of a real or hypothetical transport intervention or policy

in an urban setting that included health effects related to a change in physical

activity. Full economic evaluations consider both costs and consequences of all

alternatives examined and methods are listed in Table 6-1 (97).

5. Interventions must have resulted in changes to predominantly utilitarian physical

activity (i.e. strictly leisure time physical activity (LTPA) interventions were excluded).

6. All age groups were considered.

7. Interventions and/or policies targeting special groups, such as patients with a

disability or any other health condition, were excluded.

82

Search strategy and data sources

A comprehensive search of the literature was conducted independently by two

researchers (VB and BZ) based on Cochrane’s guidelines for systematically reviewing

public health interventions (221) and Preferred Reporting Items for Systematic Reviews

and Meta-Analyses (PRISMA) guidelines (133). The following academic databases

were searched: Web of Science, Scopus, EBSCOHost (including: Business Source

Complete, CINAHL Complete, Health Economic Evaluation Database, MedLine

Complete, PsycInfo, SportDiscus), PubMed, EMBASE, GeoBase, Compendex, Inspec,

NTIS and GeoRef. Search strategies were developed for each of the databases in

conjunction with two subject-specific librarians. The reference lists of included papers

and the index of the Journal of Transport and Health were also searched.

Specific strategies were used to search the ‘grey’ literature in well-known organisational

websites including: WHO-Cost effectiveness and strategic planning (WHO-CHOICE),

the National Institute for Health and Care Excellence (NICE), the Transport, Health and

Environment Pan-European Programme (THE PEP), the Centre for Diet and Physical

Activity (CEDAR), the Nutrition and Obesity Policy and Evaluation Network (NOPREN)

and Active Living Research. A strategy was also designed for the search engine Google

and experts in the field were consulted to ensure that all relevant literature was included.

All search strategies are given at Appendix A.

Results

Search results

The database search resulted in 7,475 papers, the titles of which were assessed for

relevance independently by each reviewer. Title and abstracts of 162 studies were

examined for relevance, with the full text of 34 studies then retrieved and reviewed. After

further exclusions, 13 studies from the database searches were included in the final

review. A list of excluded papers and reasons for exclusion is given in Appendix B. Only

one paper met the inclusion criteria from the hand search of the index of the Journal of

Transport and Health and an additional 7 papers were included from the reference list

search.

A further 15 papers were located from the grey literature. Overall 36 papers were

assessed for quality and relevant data was extracted from them (Figure 6-1).

83

Figure 6-1 PRISMA table

Data extraction and review

Included studies were assessed by two reviewers (VB and BZ) and data were extracted

with the aim of providing an overview of the main aspects, including study type, whether

the economic evaluation was of a real or hypothetical intervention, methodological

approach, targeted population, measurement of health benefits and disbenefits and

costs. These data are available on request from the corresponding author. Main results

of the analyses were also extracted, but variations in assumptions between studies

precluded the summarising of results in a single measure.

Potentially relevant publications identified through

search strategy (n=9,170)

Abstracts examined for potential relevance

(n=162)

Full text for potentially relevant publications

located and reviewed (n=34)

Total studies included from the academic

databases search (n=13)

Synthesis of results (n=36)

Duplicates removed (n=1,695)

Excluded publications, not relevant (n=7,313)

Excluded publications, not relevant (n=128)

Partly reviewed and excluded (n=21)

Additional papers from hand search (n= 1)

Titles screened for potential relevance (n=7,475)

Additional papers located from the grey literature

and reviewed (n=15)

Additional papers located from reference list of

relevant publications and reviewed (n=7)

84

In this review, the specific grading of studies according to their quality has been avoided

on the basis that such a method may unfairly judge studies where economic evaluation

was not the primary purpose or where the assigning of a grading may be difficult to

undertake in an objective manner. The use of scales for assessing quality or risk of bias

is challenging as it invariably involves assigning weights to different items on the scale

to reflect proportional value. Whilst this approach offers simplicity, its use has been

discouraged because of the potential for unreliability of results (221).

The 36 included studies were instead assessed independently by each reviewer using

the Consolidated Health Economic Evaluation Reporting Standards (CHEERS)

checklist (222). The CHEERS checklist was formulated to improve the quality and

transparency of the reporting of economic evaluations with the overarching goal of

supporting and facilitating interpretation and comparability of results. The approach

taken in this paper was to organise the quality assessment by CHEERS items. Table 6-

2 gives an overview of the quality of studies as per the CHEERS guidelines.

85

Table 6-2 An overview of included studies as per the CHEERS guidelines for quality of reporting

CHEERS no. 1 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 14 15 16 17 18 19 20 20 21 22 22 23 24

Checklist item Titl

e ex

plic

itly

iden

tifie

s st

ud

y as

an

eco

no

mic

ap

pra

isal

or

eval

uati

on

Titl

e u

ses

bro

ader

eco

no

mic

ter

ms

to id

enti

fy a

s an

eco

nom

ic a

pp

rais

al

Titl

e ex

plic

itly

incl

ud

es in

terv

enti

on

Pro

vid

es s

tru

ctu

red

su

mm

ary

or

abst

ract

Clea

r st

atem

ent

on

th

e b

road

er c

on

text

of t

he

stu

dy

Des

crib

es c

har

acte

rist

ics

of t

he

targ

et p

op

ula

tio

n

Des

crib

es s

etti

ng

and

loca

tio

n

Des

crib

es p

ersp

ecti

ve o

f th

e st

ud

y

Com

para

tor

exp

licit

ly s

tate

d

Tim

e ho

rizo

n g

iven

Rep

orts

ch

oic

e o

f dis

cou

nt

rate

Clea

rly

def

ines

hea

lth

ou

tco

mes

of

inte

rest

Clea

rly

det

ails

evi

den

ce fo

r ef

fect

iven

ess

Clea

rly

det

ails

met

ho

ds

for

mea

sure

men

t o

f p

refe

ren

ce

base

d o

utc

om

es (i

f ap

plic

able

)

Clea

rly

lists

all

rele

van

t co

sts

for

the

inte

rven

tio

n, g

ives

sour

ces,

un

it p

rice

s an

d q

uan

titi

es

Bas

e ye

ar r

epo

rted

Curr

ency

an

d c

on

vers

ion

met

ho

ds

(if a

pp

licab

le) r

epo

rted

Des

crib

es a

nd

just

ifie

s u

se o

f dec

isio

n a

nal

ytic

al m

od

el

Des

crib

es a

ll as

sum

pti

on

s re

qu

ired

Des

crib

es a

nal

ytic

met

ho

ds

Rep

orts

stu

dy

par

amet

ers

Rep

orts

incr

emen

tal c

ost

s an

d o

utc

om

es

Un

dert

akes

sen

siti

vity

an

alys

is

Rep

orts

un

cert

ain

ty

Dis

cuss

es h

eter

oge

nei

ty

Dis

cuss

es li

mit

atio

ns

Dis

cuss

es g

ener

alis

abili

ty

Rep

orts

on

so

urc

e o

f fu

nd

ing

Rep

orts

on

po

ten

tial

co

nfl

icts

of i

nte

rest

Academic literature

Beale et al. (2012)

Cobiac et al. (2009)

Dallat et.al. (2014)

Deenihan & Caulfield (2014) NA Gotschi (2011)

NA

Guo, & Gandavarapu, (2010) NA Macmillan et al. (2014) NA

Moodie et al. (2009)

Moodie et al. (2011)

Rabl & de Nazelle (2012) NA

Saelensminde (2004) NA

Schweizer & Rupi (2014) NA

Stokes et al. (2008) NA

Wang et al. (2005) NA

86

“Grey” literature

AECOM (2010) NA

Buis & Wittink (2000) NA

Cope et al. (2010) NA

Co & Vautin (2014) NA

COWI (n.d.) NA

Department for Transport (2014) NA

Fishman et al. (2011) NA

Foltynova & Kohlova 2002 NA

Krag (2007) NA

Li & Faghri (2014) NA

Lind et al. (2005) NA

Meggs & Schweizer (n.d.) NA

PricewaterhouseCoopers (2009) NA

Saari et al. (2007) NA

SinclairKnight and PWC (2011) NA

Sinnett & Powell (2012) NA

SQW (2007) NA

SQW (2008) NA

Sustrans Scotland (2013) NA

Transport for Greater Manchester (2011) NA

Transport for London (2004) NA

Wilson & Cope (2011) NA

Total (n=36) 17 8 14 19 29 12 25 8 10 30 25 29 18 2 20 22 31 5 17 20 26 8 17 6 1 21 7 10 11

87

Study descriptors (CHEERS items 1-3, 10)

Cost benefit analysis (CBA) was the dominant method of economic appraisal

undertaken, with 32 of the 36 included studies reporting results as cost per unit of

benefit or as cost benefit ratios (CHEERS item 10) (218, 219, 223-252). Cost

effectiveness analysis (CEA) was performed in five papers, reporting results as cost per

disability adjusted life-year (DALY) averted (94, 122, 253, 254) or quality adjusted life-

year (QALY) gained (224) (CHEERS item 10). It should be noted however that the

terms CEA and cost utility analysis (CUA) are used interchangeably in the literature (97)

and that one study undertook both CBA and CUA (224). Less than half of all included

studies clearly identified the study as an economic evaluation as per the CHEERS

guidelines (94, 218, 219, 223, 224, 226-228, 232, 234, 235, 237, 239, 244, 251, 253,

254) (CHEERS item 1).

Only 14 papers (122, 219, 223, 227, 232, 235-237, 244, 247, 249, 251, 253, 254)

reported the intervention being evaluated in the title of the study as recommended by

the CHEERS guidelines (CHEERS item 1). Seventeen studies (47% of included

studies) undertook analyses of hypothetical interventions or scenarios (Table 6-3).

Seven studies (19% of included studies) evaluated proposed interventions and thirteen

studies (36% of included studies) examined implemented interventions (Table 6-3). The

majority of studies (n=29) assessed the economic credentials of hypothetical, proposed

or implemented cycling and walking infrastructure or facilities (122, 218, 219, 223-226,

228-230, 232-239, 241, 242, 244-246, 248-252, 255) (Table 6-3).

88

Table 6-3 Interventions included in the review

Type of intervention evaluated Studies included Intervention

Hypothetical interventions

AECOM 2010 Cycling infrastructure

Beale et al. 2012 Multi-use trail, Cycling/walking infrastructure

Buis & Wittink 2000 Cycling infrastructure

Co & Vautin 2014 Congestion charging, Cycling/walking infrastructure

Department for Transport 2014 Cycling/walking infrastructure

Fishman et al. 2011 Active transport to school program

Foltynova & Kohlova 2002 Cycling infrastructure

Gotschi 2011 Cycling infrastructure

Guo & Gandavarapu 2010 Cycling/walking infrastructure

Krag 2007 Cycling infrastructure

Lind et al. 2005 Cycling infrastructure

Macmillan et al. 2014 Cycling infrastructure

PricewaterhouseCoopers 2009 Cycling infrastructure

Saari et al. 2007 Cycling infrastructure

Schweizer & Rupi 2014 Cycling infrastructure

Sinclair Knight & PricewaterhouseCoopers 2011

Cycling/walking infrastructure

Transport for London 2004 Cycling infrastructure, Cycle education programs

Proposed interventions

Dallat et al. 2014 Urban greenway incorporating active transport infrastructure

Deenihan & Caulfield 2014 Cycling infrastructure

Li & Faghri 2014 Cycling infrastructure

Meggs & Schweizer n.d. Cycling infrastructure

Stokes et al 2008 Light rail infrastructure

Transport for Greater Manchester 2011

Cycling infrastructure

Macmillan et al. 2014 Cycling infrastructure

Implemented interventions

Cobiac et al. 2009 TravelSmart program

Cope et al. 2010 English Cycling Town investment program

COWI and the City of Copenhagen n.d.

Cycling infrastructure

Moodie et al. 2009 Walking School Bus program

Moodie et al. 2011 TravelSmart Schools program

Rabl & de Nazelle 2012 Bicycle share scheme Saelensminde 2004 Cycling infrastructure

Sinnett & Powell 2012 Living Streets program

SQW 2007 Cycling/walking infrastructure, Cycle education programs

SQW 2008 Cycling infrastructure

Sustrans Scotland 2013 Cycling/walking infrastructure

Wang et al. 2005 Cycling/walking infrastructure

Wilson & Cope 2011 Cycling/walking infrastructure

The abstracts of academic papers were generally more succinct and targeted than the

abstracts of studies found in the grey literature (CHEERS item 2). The context and

relevance of the studies also differed between peer-reviewed and grey literature

(CHEERS item 3). Generally, peer-reviewed studies presented a case for the inclusion

of health outcomes of transport interventions or assessed changes in population health

89

attributable to active transport and were undertaken to build the evidence for the

inclusion of physical activity related health effects into transport appraisal (94, 122, 219,

224, 229, 233, 234, 238, 240, 242, 247, 251, 253, 254). Studies from the grey literature

and reference list searches were mostly reports developed by government or non-

government organisations, with several of the economic evaluations being undertaken

as supporting case-studies or as part of broader guiding documents (223, 226-228, 231,

235, 237, 239, 244-246, 248, 249, 252, 255).

Methods (CHEERS items 4-17)

Target population and subgroups (CHEERS item 4)

Health consequences of physical inactivity vary for adults and children and therefore

clear reporting of an interventions target population is required to assess whether

appropriate health outcomes are being evaluated and whether an intervention is cost-

effective. Only eleven of the 36 included studies explicitly described age ranges or gave

some clear indication of the intervention target population (for example, the adult

population) (94, 223, 224, 226, 231, 232, 238, 252-255).Three interventions targeted

children exclusively (231, 253, 254). Sub-group analyses, for example by age cohort or

by socioeconomic position (SEP), were not undertaken in any of the included studies.

Setting and location (CHEERS item 5)

Studies were undertaken in France (240), Norway (219), the Czech Republic (232),

Denmark (218, 228), Sweden (236), Finland (241) and the Netherlands (225). Two

studies looked at interventions in a number of European cities (237, 242). Nine studies

were undertaken in England (224, 227, 230, 244-246, 249, 250, 252), one in Scotland

(248), one in Ireland (229) and one in Northern Ireland (122). Six studies were

undertaken in the United States (226, 233-235, 247, 251), one in New Zealand (238)

and seven studies were undertaken in Australia (94, 223, 231, 239, 243, 253, 254).

Due to the nature of the interventions examined, the majority of the studies were

conducted in community settings amongst the general population.

Study perspective and comparators (CHEERS items 6-7)

Determining the appropriate health outcomes and resources and methods for

quantifying and valuing them is dependent on the study perspective (222) .Only nine

studies reported their perspectives explicitly. Four applied a health sector perspective

90

(94, 122, 247, 251), one a public payer perspective (218), two a societal perspective

(253, 254), one used both a health sector and a societal perspective (224) and one

used an ecosystem health perspective (238). Economic evaluation entails the

incremental assessment of both the costs and benefits of an intervention against an

alternative option. Shortcomings in reporting comparison scenarios were observed with

less than one third indicating them explicitly (94, 223, 224, 226, 232, 238, 243, 252-

254), although a “do-nothing” comparator may have been implied particularly for the

relatively large number of studies evaluating new cyclists and infrastructure.

Time horizon, discounting and base year (CHEERS items 8, 9 and 14)

Reporting of time horizons and discount rates in the included studies was variable. Time

horizons were reported in 30 of the included studies (94, 122, 218, 219, 223-227, 229-

235, 237-240, 242-245, 249-254), ranging from one year to lifetime horizons. Discount

rates were explicitly reported in 25 of the included studies (94, 122, 218, 219, 223-227,

229-237, 239, 243, 244, 249, 250, 253, 254). Choice of discount rate ranged from 2.5%

(231) to 7% (223, 231, 232, 239, 243). The base year of the study was clearly reported

in 22 studies (94, 122, 219, 223-228, 230, 231, 233-235, 238, 239, 243, 250, 251, 253,

254) and the majority of studies reported the currency for costs and benefits (94, 122,

218, 219, 223-225, 227-232, 236-248, 250, 252-254).

Measurement of effectiveness (CHEERS item 11)

The quality of evidence for all included studies in our review can only be considered as

weak by traditional epidemiological standards. The studies evaluating hypothetical or

proposed interventions (Table 6-3) used differing methods for estimating effect. Three

studies applied stated willingness to change transport behaviours to walking or cycling,

collected through surveys (229, 232, 237). Two studies estimated indicative diversion

rates from intercept surveys or user counts of similar active transport infrastructure

(250). Four studies based estimates of effect on values from the literature (224, 230,

238, 243) and four studies assumed estimates of effect (122, 218, 231, 233, 249). Five

studies used demand forecasting or simulation modelling (223, 225, 226, 238, 247) and

two studies applied regression analysis based on built environment attributes to

estimate demand for active travel (234, 242). One study used a combination of

approaches, including using an assumed estimate of effect based on an aspirational

target, the use of survey data and estimates of effect from the literature (239). It was not

91

clear how the estimate of effect was derived for three hypothetical intervention studies

(235, 236, 241).

Methods for estimating effect sizes for implemented interventions included in our review

also differed. Eleven studies examining implemented interventions (Table 6-3) based

effectiveness on observed effects derived from survey or count data (94, 227, 228, 244-

246, 248, 251-254). Due to limitations of the data collected most of these studies relied

on a number of assumptions in order to derive these effects. Rabl and de Nazelle (240)

included a case study of the Velib bicycle share scheme in Paris to illustrate the

potential health benefits of a shift from car to active transport however only incorporated

an assumed effect estimate in their calculations. Two studies based estimate of effect

on assumptions and evidence from the literature (219, 238). It should also be noted that

the effectiveness data of three implemented interventions (94, 253, 254) was then

extrapolated to apply to the Australian population to estimate cost-effectiveness.

Limited detail on methods for inclusion of cross-sectional study data (from survey or

counts) was given in all relevant studies, making it difficult to comment on the overall

quality of the data and factors such as bias or seasonality. None of the studies

controlled for any possible substitution effect of a potential uptake in utilitarian physical

activity on leisure time physical activity, probably due to a lack of rigorous evidence of

any potential effect (233).

The health benefits of physical activity may accrue differently in persons who are

sedentary as compared to those who are already physically active (256), however data

were not available at the required level for the impact of these effects to be

comprehensively considered in any of the included studies. A variety of methods were

used to account for a lack of rigorous evidence on health benefit accrual in different

groups. In some studies, the effect of an increase in physical activity as a result of an

intervention only accrued in persons who were previously inactive (226, 232) and in one

study only in obese people (243, 247). Effects were included only for new users in two

studies of cycling interventions (239, 245), whilst another study (253) assumed that half

of the participants in the intervention program were new to active transport. Sinnett et al.

(244) attributed 50% of the uptake of active transport to the intervention. Only one study

controlled for “non-traders” (i.e those who would not take up active transport despite the

intervention) (223).

92

Timing to intervention uptake was considered in four studies included in our review.

Deenihan et al. (229) assumed two years of build up to reach full use of the cycleway.

Cope et al. (227) considered three years until the intervention achieved the level of

cycling applied as a measure of effectiveness. In the study by Schweizer and Rupi (242)

it was assumed that it would take 10 years to reach the bicycle mode share full

potential. In one study different timing scenarios were assessed for the intervention to

take effect and health benefits to be fully realised (239). Only one study was specific in

terms of the level of usage of the intervention, with new cycling facilities assumed to be

used at 75% of full capacity (249).

Methods to account for the sustainability of intervention effect also varied between

studies. Cobiac et al. (94) assumed a level of effectiveness decay of 50% after the first

year and Macmillan et al. (238) considered two years. In the hypothetical Department

for Transport intervention (230) it was assumed that the effect of the intervention would

decay at an annual rate of 10%. The studies by Moodie et al. (253, 254) assumed 100%

maintenance of effect. There is a risk of overestimating the benefits of an intervention if

sustainability of effects over time is not taken into consideration. This may be the case

with the remaining studies in this review.

Evaluation of benefits/disbenefits and costs (CHEERS items 13 and 14)

Our analysis highlights that a variety of potential benefits/disbenefits and cost categories

have been included into the economic evaluation of active transport interventions, with

limited uniformity in terms of type or methodology of inclusions between studies. These

inclusions incorporate a multitude of health, social, economic and environmental

considerations. As the focus of this review is on physical activity related health benefits

we present our findings on these first, with discussion around the inclusion of other

benefits/disbenefits and costs following.

Physical activity related health benefits

Different methodological approaches to the evaluation of health benefits of

increased physical activity were identified in the relevant studies, including the

incorporation of mortality outcomes, morbidity outcomes or a combination of both

(Appendix C).

Mortality outcomes

93

Sixteen studies included only mortality related outcomes associated with

an increase in physical activity (223, 227, 229, 230, 235, 237-240, 242,

244, 248-250, 252) (Appendix C). Eleven studies applied the WHO

HEAT tool for walking and cycling (220) to estimate changes in all-cause

mortality attributable to increases in physical activity levels (223, 227, 229,

235, 237, 239, 240, 242, 244, 248, 252). Six studies allowed for a period

of 5 years to fully achieve health benefits as a result of the intervention as

per HEAT recommendations (220). Given the methodological limitations

of the WHO HEAT tool for use in those aged under 20 years, Cope et al.

(227) omitted any physical activity related health benefits as a result of the

intervention in children or young people despite the potential of the

intervention to change active transport behaviours in this group (220).

Conversely, Sinnett & Powell (244) assumed that all those affected by the

intervention were aged between 20 and 74 years so that the WHO HEAT

tool could be used.

Two studies applied the HEAT all-cause mortality relative risks estimates

indirectly, following the UK Department for Transport WebTAG guidance

(230, 249). In one study avoidable deaths from cardiovascular heart

diseases, stroke and colon cancer were estimated for those moving from

physically inactive to active (250). The study by PWC (239) included

mortality outcomes for cardiovascular diseases assessed as per

published values by the Road and Traffic Authority of New South Wales

for the main analysis and the HEAT tool for sensitivity testing. In the study

by PricewaterhouseCoopers different scenarios for the full realisation of

health effects were assessed (239). Macmillan and colleagues (238)

applied relative risks for all-cause mortality from the literature to estimate

impacts of increased cycling levels assuming a two-year build up for

achieving full health effects.

Morbidity outcomes

Five studies included only morbidity related outcomes associated with an

increase in physical activity (219, 226, 234, 247, 251), with different

approaches taken between studies. Four studies included health effects

94

related to a potential change in physical activity through cost savings of

diseases averted (219, 226, 234, 247) although the specific diseases

included varied (Appendix C). Of these four studies, two included the

healthcare cost savings specifically related to obesity prevention (234,

247). In one case (251) physical activity related health effects were

incorporated through healthcare cost savings incurred from moving from

physical inactivity to physical activity.

Mortality and morbidity effects

Eleven studies included both mortality and morbidity related outcomes

associated with an increase in physical activity (94, 122, 224, 231-233,

243, 245, 246, 253, 254)(Appendix C). The evaluations by SQW

Consulting in 2008 (246) and Gotschi (233) included morbidity effects by

incorporating healthcare costs saved as a result of moving from physical

inactivity to activity alongside mortality outcomes assessed with the HEAT

tool. The evaluations by SQW Consulting in 2007 (245) and Foltynova &

Kohlova (232) incorporated both mortality using the value of statistical life

and morbidity effects but did not use the HEAT tool. Foltynova & Kohlova

(232) used a cost of illness approach for morbidity effects and assumed a

9% decrease in mortality from cardiovascular diseases to estimate the

mortality value. Another study by SQW Consulting (245) used estimates

from the literature to estimate the value of loss of life and savings to the

healthcare system. Fishman et al. (231) assessed an intervention

targeting children accounting for mortality and morbidity applying values

from the New Zealand Transport Agency (NZTA) for adults, supporting

this decision based on the argument of applicability posited by Genter et

al (257).

In the cost utility studies QALYs gained (224) or DALYs averted (94, 122,

253, 254) are both measures that include mortality and morbidity

outcomes. Two methodologies for inclusion of health outcomes were

identified: the Assessing Cost Effectiveness (ACE) approach (94, 253,

254) and the PREVENT model (122). Both methods apply the concept of

population impact fraction (PIF) to estimate the change in future incidence

95

of diseases. However, PREVENT is a full dynamic population model and

incorporates only sensitivity analysis, whereas ACE models per cohort

and considers both sensitivity and uncertainty around the input

parameters. The study by Beale et al. (224) used both regression analysis

and cost savings through diseases averted to estimate QALY gains from

an increase in physical activity. Cobiac et al. (94) was the only study to

clearly justify the use of DALYs as a measure of health over QALYs.

Unspecified outcomes

Six studies lacked specificity of health outcomes and it was unclear exactly

what physical activity related health benefits had been included (218, 224,

225, 228, 236, 241). In one case internal costs for the user and external

costs for society were given, however from the text it was not possible to

identify whether these refer to mortality, morbidity or other measures of

health (228). Buis & Wittink (225) only considered health attributable to an

increase in physical activity for one of four case studies undertaken and

values were taken from the literature. The studies by Krag (218), Lind et al.

(236) and Saari et al. (241) applied values from the literature (258) without

specifying end health outcomes accounted for. Krag (218) assumed that it

would take 12 years after the intervention for the full health benefits from

the intervention to be achieved.

Other benefits/disbenefits

Cost benefit studies varied widely in terms of the other health and non-health

benefits and disbenefits that were included (Table 6-4). Whilst influenced by the

study perspective chosen, it is clear that little consensus exists around what

impacts should be included and how to include them. Several studies were quite

comprehensive in their inclusion of a range of potential benefits and disbenefits

(219, 223, 239, 243, 250), whilst others were not (228, 229, 237, 242, 247, 251).

Environmental effects were the most included category (62.5% of studies),

followed by the inclusion of the effects of accidents and injuries (50% of studies).

The cost utility analyses undertaken using the ACE approach incorporated other

96

factors for consideration in the decision-making process such as equity and

feasibility qualitatively (94, 253, 254).

Table 6-4 Other non-PA benefits/disbenefits included in the cost-benefit analyses

STUDY

BENFEFITS/DISBENEFITS INCLUDED

CO

MF

OR

T/

SE

CU

RIT

Y

PR

OD

UC

TIV

ITY

TR

AV

EL

TIM

E

AC

CID

EN

TS

/ IN

JU

RIE

S

EN

VIR

ON

ME

NT

CO

NG

ES

TIO

N

PA

RK

ING

VE

HIC

LE

OP

ER

AT

ING

OT

HE

R

AECOM 2010 Journey ambience, public transport

cost savings

Beale et al 2012

Buis & Wittink 2000 Bike theft

Co & Vautin 2014

Cope et al 2010 Road infrastructure

COWI n.d.

Deenihan & Caulfield 2014

Department for Transport 2014 Journey quality, indirect taxes, road

infrastructure

Fishman et al 2011 Public transport cost savings.

Foltynova & Kohlova 2002

Gotschi 2011

Guo & Gandavarapu 2010

Krag 2007 Reduced income from reduced

public transport demand

Li & Faghri 2014

Lind et al 2005

Macmillan et al 2014

Meggs & Schweizer n.d.

PWC 2009 Road infrastructure

Rabl & de Nazelle 2012

Saari et al 2007 Road infrastructure

Saelensminde 2004 Public transport provision

Schweizer & Rupi 2014

Sinclair Knight and PWC 2011 Road infrastructure

Sinnett & Powell 2012

SQW Consulting 2007 Journey ambience

SQW Consulting 2008 Agglomeration

Stokes et al 2008

Sustrans Scotland 2013 Road infrastructure

Transport for Greater Manchester 2011

Cyclist user charges

Transport for London 2004

Wang et al 2005

Wilson & Cope 2011

Total (n=32) 9 13 10 16 20 12 7 8

Costs

Costs included for infrastructure interventions were mostly construction and

maintenance costs. For policies or programs, the included costs were mostly

related to the delivery of the program, with four including costs to the individual

and the family (235, 239, 253, 254). The effect on physical activity of

complementary interventions was considered in two studies (219, 223), however

97

no costs were attributed to such interventions. The results of such scenarios are

therefore likely to overestimate cost effectiveness.

The quality of cost data varied, with some studies reporting data sources and unit

costs clearly and transparently (94, 122, 219, 230, 233, 234, 237, 238, 243, 244,

250, 251, 253, 254) whilst other studies gave limited detail (218, 225-228, 232,

236, 240-242, 246-249, 252). Five studies relied on estimates of costs from the

literature, which may be very specific to a geographical location and therefore not

necessarily generalisable to other settings (219, 223, 224, 229, 239). Since a

large proportion of the included studies assessed hypothetical or modelled

interventions, there is potentially a large margin of error in the cost estimation.

Results (CHEERS items 18-21)

Results for the included cost benefit analyses were reported as ratios of benefits to

costs, ranging from -31.9:1 (244) to 59:1 (226). Results cannot be combined due to the

high level of heterogeneity in study design, quality, evidence of effectiveness, outcomes

considered and costs and benefits included. Figure 6-2 shows the cost benefit ratios

from selected studies. Twenty-six of the 32 cost benefit studies reported benefits greater

than costs thus indicating good value for money based on their underlying assumptions

(218, 219, 223-227, 229-231, 233-239, 241, 242, 245, 246, 248-252). One study

evaluating an implemented intervention reported results as net present value and

internal rate of return estimates (228). Two studies did not explicitly state cost benefit

ratios but gave inputs for their calculation, one examined an implemented intervention

(240) and one examined a proposed intervention (247).

Eight studies reporting cost benefit ratios of implemented interventions were included in

our review (219, 227, 244-246, 248, 251, 252). Six of these were considered cost-

effective (219, 227, 245, 248, 251, 252). The study by Sinnett & Powell (244) evaluated

Fitter for Walking projects in a number of locations and applied several assumptions. It

should be noted that the results of this study varied widely in terms of its cost-

effectiveness according to location and estimate of effect used. Cost effectiveness of

interventions examined by SQW Consulting (246) also varied dependent on location

examined, with 60% (3/5) of the cycling infrastructure projects considered cost-effective.

Of the seventeen cost benefit studies reporting cost benefit ratios for hypothetical

interventions (218, 223-226, 230-234, 236, 238, 239, 241-243, 250), all except one

98

(232) indicated benefits greater than costs. Four of the proposed interventions also

reported benefits greater than costs (229, 235, 237, 249).

Figure 6-2 Selected cost benefit ratios by interventiona

a Where included studies reported more than one benefit cost ratio (BCR) the smallest value was used, except in the case of Schweizer & Rupi (2014) where only average BCR value was clearly presented. It should also be noted that the direct comparison of results between studies is not recommended due to differences in methodologies between studies.

For the cost utility studies conducted in Australia examining implemented interventions

(94, 253, 254), only one study result (94) was under the commonly used threshold of

AUD50,000 per DALY averted (95). The studies utilising the ACE approach presented

cost effectiveness planes and results in terms of costs per averted DALY (94, 253, 254).

In the study by Cobiac et al. , an intervention pathway for the base case scenario and

sensitivity analyses were presented, indicating how much health is gained by

cumulatively adding each intervention from the most to the least efficient (94).

Beale et al. (224) reported incremental cost effectiveness ratios (ICERs) and a

comparative analysis indicating the conditions required under each approach for the

results to be most similar for two hypothetical scenarios. In the UK a threshold of

£20,000 to £30,000 per QALY is the standard applied (259), in which case estimates of

£94 per QALY to £9439 per QALY are considered cost effective. In the study by Dallat

et al. (122), results were presented for each of the three evaluated scenarios in terms of

-40 -30 -20 -10 0 10 20

Infrastructure (AECOM 2010)Infrastructure (Beale et al. 2012)Infrastructure (Buis & Wittink 2010)Infrastructure (Co & Vautin 2014)Infrastructure (Deenihan & Caulfield 2014)Infrastructure (Department for Transport 2014)Infrastructure (Foltynova & Kohlova 2002)Infrastructure (Gotschi 2011)Infrastructure (Guo & Gandavarapu 2010)Infrastructure (Krag 2007)Infrastructure (Li & Faghri 2014)Infrastructure (Lind et al. 2005)Infrastructure (MacMillan et al. 2014)Infrastructure (Meggs & Schweizer n.d.)Infrastructure (PriceWaterhouseCoopers 2009)Infrastructure (Saari et al. 2007)Infrastructure (Saelensminde 2004)Infrastructure (Schweizer & Rupi 2014)Infrastructure (Sinclair Knight & PWC 2011)Infrastructure (Sinnett & Powell 2012)Infrastructure (SQW Consulting 2007)Infrastructure (SQW Consulting 2008)Infrastructure (Sustrans 2013)Infrastructure (Transport for Greater…Infrastructure (Transport for London 2004)Infrastructure (Wang et al. 2005)Infrastructure (Wilson & Cope 2011)Community programs (Cope et al.2010)Community programs (Fishman et al. 2011)

99

costs per averted DALY, ranging from approximately £4470 per DALY to just over

£18,400 per DALY.

Our analysis indicates some confusion in the literature on the different meanings of

sensitivity and uncertainty analysis. In sensitivity analysis (or deterministic sensitivity

analysis) input parameters are changed manually to evaluate the sensitivity of the

model’s outputs to specific input parameters (260). Model outputs can be tested by

changing one input parameter at a time (one-way sensitivity analysis) or a group of

them simultaneously (multi-way sensitivity analysis). Sensitivity analyses were

performed in 22 of the 36 included studies (94, 122, 219, 223, 224, 227-229, 231, 232,

235, 237-239, 242-244, 247, 251, 253, 254), although only four studies explicitly

reported it (122, 224, 253, 254). The study by Macmillan et al (238) was the only study

to perform multi-way sensitivity analysis, with the others performing one-way analysis.

The input parameters most commonly tested for sensitivity included discount rates,

intervention effects, intervention costs, intervention time decay and lag time for disease.

In one case (94) the intervention became cost ineffective when the effect decay rate

was varied from 75% to 100% in the first year. The intervention assessed by Dallat et al.

(122) became cost ineffective when the discount rate was changed to 5% for one of the

assessed scenarios (scenario A 2% shift from inactive to active). In the study by

Macmillan et al. (238) results were sensitive to assumptions regarding safety in

numbers, which relates to the non-linear relationship between the number of road

injuries and number of people engaging in active transport (whereby more people

walking and cycling may result in fewer accidents) (80).

There are different types of uncertainty: parameter uncertainty and structural uncertainty

(99). Parameter uncertainty is also commonly tested in probabilistic sensitivity analysis

(260) and refers to the uncertainty introduced into the model by uncertainty in the input

variables. Structural uncertainty refers to uncertainty due to assumptions made in the

model, and model structure. Uncertainty around selected input parameters was

performed in seven of the included studies (94, 238, 240, 247, 251, 253, 254). Only four

studies provided detailed information in terms of input parameter distributions and the

assumptions made to account for uncertainty (94, 238, 253, 254). The study by

Macmillan et al was the only study to report performing structural uncertainty analysis

(238).

100

Assumptions, limitations and generalisablity of studies (CHEERS item 22)

Transport interventions by their very nature can be extremely context specific and

therefore inputs to the analyses and results are difficult to generalise between studies.

Transport behavioural change is complex and modal choice is influenced by a number

of factors, such as individual preference, the built environment, topography and climate,

culture and perceptions of safety (261). Context specific interventions require context

specific input parameters however our analysis has shown that many studies rely on

generalised input parameters (for example, for effectiveness, cost estimates, health

benefits), which may potentially limit the reliability of results.

All of the included studies relied on a number of assumptions, most of which have been

highlighted in the previous sections. Assumptions made most commonly related to the

lack of effectiveness data, with other commonly cited limitations including a reliance on

self-reported data and the potential for bias (224, 229), low response rate (254), the

attenuation of intervention effect over time (94, 230, 238) and limited evidence on the

time lag between intervention and health effect. In those studies that considered health

benefits of active and inactive people, an assumption had to be made regarding the

threshold level of physical activity above which people were deemed to be active. For

instance, in the research by Gotschi (233, 238) a 30 minute per day cut off was

assumed. Saelensminde (219) assumed that health benefits only accrued to 50% of

new pedestrians and cyclists, arguing that otherwise health benefits would be

overestimated.

Two studies explicitly stated linearity in health effects (224). Despite this being implicit in

the majority of studies, reporting of this assumption was not the norm. Only one study

explicitly reported that individuals were 100% compliant with the extra physical activity

induced by the intervention (224, 238). An increase in walking as a result of the

intervention was assumed to grow in line with the population in the study by Macmillan

et al. (238). An increase in cycling was assumed to grow at a rate of 5% in the

evaluation by Sustrans Scotland (248).

The WHO HEAT tool uses estimates for health from the Danish population (220).

Studies applying the WHO HEAT tool therefore are based on the underlying assumption

that the subject population is similar to that of the Danish population, which is unlikely to

be the case for some of the included interventions.

101

Source of funding and conflicts of interest (CHEERS items 23-24)

Only 16 of the 36 included studies were from peer-reviewed sources and therefore more

likely to have been through a rigorous evaluation process (94, 122, 219, 224, 226, 229,

233-235, 238, 240, 242, 247, 251, 253, 254). This is an indication that special care should

be taken in the interpretation of results of some of the analyses, as well as potential

funding sources for conducting the studies.

Discussion

The aim of this review was to provide a current overview of the state of the literature

regarding the inclusion of physical activity related health effects into transport appraisal.

Our analysis gives an overview of the methodological challenges in the incorporation of

broader health effects into transport appraisal, and highlights the lack of an agreed

approach to the inclusion of physical activity effects into transport economic evaluation.

A comprehensive search strategy was developed so as to avoid missing relevant

studies. Despite our best efforts, the wide range of terminologies used in the active

transport area means that some studies may have been missed.This study did not

consider comparative risk assessments or health impact assessments as they did not fit

the study inclusion criteria of having undertaken a CBA, CUA or CEA. In addition, this

review may be susceptible to publication bias as it is possible that only the most cost-

effective interventions have been reported.

Heterogeneity of study methods and approaches made a meta-analysis unfeasible.

Studies included in our review varied greatly in terms of the active transport

interventions that they evaluated and other relevant contextual factors.

It is clear that the advent of the WHO HEAT tool for walking and cycling (220) has led to

more interest in the inclusion of physical activity related health effects into transport

appraisal. The review by Cavill et al (197) identified only 16 studies, whereas our study

included 36 studies. This is despite the fact that Cavill’s review used wider inclusion

criteria by including economic valuations of any kind whereas our review examined only

full economic evaluations, or more specifically CBAs, CEAs and CUAs (Table 6-1). For

example, Cavill et al included the study by Rutter (217) whereas our review excluded

this study as it did not consider costs.

102

Whilst there have been notable improvements since the original publication by Cavill

and colleagues (197) in terms of harmonisation of estimation techniques applied for

mortality related physical activity outcomes, our analysis suggests that many of the

issues highlighted in the Cavill et al. review remain. Slightly over 50 per cent of studies

included in our review and published after HEAT inception have applied the tool.

However, the current version of HEAT only incorporates mortality effects of an uptake in

walking or cycling. Therefore those studies seeking to incorporate morbidity as well as

mortality effects are still using differing methods. A novel approach developed in recent

years is the Integrated Transport and Health Impact Modelling (ITHIM) tool developed

by Woodcock et al. (195), which serves to measure the impact of transport policies on

health outcomes related to changes in physical activity including mortality, morbidity and

exposure to road injuries and air pollution. The ITHIM has however only been applied to

conduct health impact assessments, and therefore is not included in this review.

Our analysis of the literature using the CHEERS checklist (222) has highlighted that

significant scope exists to improve the rigour of effectiveness analyses being used. The

majority of studies included in our review examined the economic credentials of

hypothetical or proposed active transport interventions. This is expected given the

relative importance of economic evaluation in the decision-making process in both the

health and more specifically the transportation sectors. However the level of uncertainty

of an economic evaluation relies partially on the sum of its inputs and this highlights one

of the complexities of establishing rigorous estimates of impact of active transport

interventions on which to base analyses.

Our review of the literature suggests that the quality of effectiveness data used for

evaluating implemented interventions is only marginally better than that used to

evaluate hypothetical interventions. All evaluations required a number of assumptions in

terms of effectiveness, including those evaluating implemented interventions. Whilst it is

recognised that the collection of high quality evidence of effectiveness in this area is

challenging (214, 238, 262, 263), this highlights the importance of incorporating rigorous

and comprehensive evaluation programs into interventions prior to implementation.

There is enormous variety in the structure, form and purpose of transport related

interventions. Often health is a secondary consideration to the primary purpose of a

transport intervention, which may be to ease road congestion or to address

environmental concerns. Whatever the primary purpose of the intervention, a more

103

thorough and considered approach to the measurement of impact on rates of walking

and cycling is required.

Whilst it has been suggested that more appropriate and feasible levels of evidence be

used in the evaluation of effectiveness of transport and built environment interventions

(208, 264), it is important that these more feasible levels of evidence retain enough

rigour to be able to draw conclusions. For instance, much of the research treats walking

and cycling as a single behaviour, although they may have different correlates (261,

265, 266) and the potential health benefits between them may differ (267, 268). Data

also rarely exists on pace, intensity and magnitude of active transport, precluding more

rigorous analysis. None of the studies included in our review adequately dealt with the

residual confounding that may exist, for example due to the effect of active commuters

having higher rates of physical activity but also potentially being more health conscious

and living a more healthful life through diet and other health-related behaviours. The

current evidence base is limited, and it is clear that more and better quality evaluation of

implemented interventions is required to provide better data on transport behaviours.

This is particularly important then given the proportion of studies that are reliant on

evidence from the literature on which to base their analyses.

The generalisability of study findings should however also be approached with caution.

Transport interventions can be highly situation specific and the potential impact of a

range of factors that may influence modal choice should be considered (214). Many of

the included studies in our review relied on estimates from the literature, with no

guarantees that such estimates would prove reliable in different contexts. The

assumptions made about transferability of data from one setting to another is a concern,

as noted by Cavill et al. in 2008 – and our analysis suggests these assumptions remain

a concern several years after the issue was first highlighted.

Difficulties also exist in terms of defining and measuring target populations of

environmental interventions, with included studies again limited by data. For instance,

the WHO HEAT tool was primarily designed for use in the adult population (aged 20-64

years for cycling and aged 20-74 years for walking) due to the fact that evidence for

calculating relative risks in children and teens is not currently deemed sufficient. The

application of values based on adult relative risks in studies such as Fishman et al.

(231) and as recommended by the New Zealand Transport Agency (NZTA) (257)

104

highlights the need for more evidence to be generated to better inform results across

the spectrum of target populations. Active transport interventions may have an impact

on the travel behaviours of children and young people, and a more appropriate

representation of these potential benefits would be preferable to using adult values or

simply omitting any possible effect (227). More robust evidence is required on the

potential health benefits of walking and cycling for transport in children and youth,

despite the inherent challenges presented by the fact that many of these potential health

benefits may be realised over long time horizons.

Approaches to the measurement of physical activity varied widely between studies,

which was another issue highlighted by Cavill et al. several years ago. Recent studies

have used a range of measures, including the number of new users, the percentage of

all trips shifted to active transport, number of trips, MET minutes per week spent in

active transport, time spent in active transport, the proportion of physically inactive that

became active, vehicle miles saved and distance walked or cycled. The WHO HEAT

tools require data on the number of people walking or cycling as a result of an

intervention and the average time spent (which can be calculated by using duration,

distance, trips or steps). A more consistent approach to measuring physical activity as a

result of active transport interventions may prove more useful, could facilitate

comparison and may minimise the number of assumptions required to estimate a

change in travel behaviours.

Scope also exists for a more standardised approach to the inclusion of benefits and

disbenefits into the economic evaluation of transport projects. It is interesting to note that

those studies that sought to include a more comprehensive range of possible benefits

and disbenefits into their analyses were mostly from the grey literature (223, 239, 243,

250), with one exception (219). Studies found within the academic literature tended to

focus on the inclusion of health benefits related to physical activity, with little regard to

other possible impacts. This suggests that despite growing awareness of the need for a

more multi-sectoral approach to increasing physical activity incorporating health,

environmental, transport and other sectors (269-271), more work is required to put this

theory into practice. At present there still seems to be a focus on single sector

consequences of public policies and program, within the academic literature at least,

where more of a systems approach may prove more useful (272).

105

Two of the studies included in our review discussed the proportion of overall benefit

attributable to physical activity related health benefits as part of their analyses (224,

227). Whilst this highlights the importance of the inclusion of physical activity related

health effects into transport evaluation it is important that studies do not overstate

relative importance, especially given the wide variation in benefits and disbenefits

included between studies. Such statements are more valid in studies that incorporate a

wider range of benefits and disbenefits (227) than those that only include a limited range

in their analysis (224).

Our analysis highlights that more consistency and transparency in reporting economic

evaluations of transport interventions incorporating health outcomes is needed, and

tools such as the CHEERS guidelines (222) should be used more widely and

consistently. There is great scope for improvement in the reporting of study

perspectives, comparators, time horizons, evidence for effectiveness, choice of discount

rates, assumptions and the costs and benefits included in the analyses. A lack of

transparency limits both the application of study results and potential advances in

methodologies for the incorporation of physical activity related health effects into

transport economic appraisals.

Finally, our analysis suggests that active transport projects should be considered based

on a wide range of their potential merits, such as the ability to reduce traffic congestion,

but also on their health and environmental benefits. This will result in the more efficient

allocation of scarce transport resources, with more informed transport decision making

leading to transport systems that encourage a variety of modes of transport based on

their relative value. From a public health perspective, this may result in an increase in

incidental physical activity across populations as the incorporation of physical activity

related health benefits contribute to the cost effectiveness of active transport policies

and programs.

Conclusion

Our review demonstrates that whilst important progress has been made towards more

routine recognition of active transport health benefits in transport planning, there is still

more work to be done. Increasing evidence suggests that the health effects of active

transport behaviours may be more far-reaching than the effect of injuries and emissions,

106

to include physical activity related health benefits and even possible benefits related to

mental health and quality of life.

Better understanding is required of the effect of transport interventions on transport

behaviours and the ways that both mortality and morbidity related health effects can be

taken into account. Research time and effort should be placed on understanding and

incorporating the broad range of health benefits into transport appraisal, so that better

informed decision-making can ensure the most efficient allocation of society’s scarce

resources. At present, a significant degree of uncertainty exists on the effectiveness

and impact of interventions (214, 273, 274) and this uncertainty is reflected in

subsequent economic evaluations. A more uniform and comprehensive approach to

measurement of physical activity behaviours across populations would assist, as would

more attention to clear and transparent reporting of economic evaluations.

Positive steps are being taken and it is very encouraging that more studies are being

generated into the important links between transport, health and the environment. This

growing body of evidence has the potential for future positive public health ramifications,

through more transparent, comprehensive and fair appraisal of active versus motorised

transport policies and programs.

107

Chapter 7 A method for the inclusion of physical activity-related

health benefits in cost-benefit analysis of built environment initiatives


Cost-benefit analysis is a widely used method for the economic evaluation of

transport projects in high-income countries. However, PA-related health benefits are

not included in the economic appraisal of transportation initiatives on a routine basis.

The same can be said for the planning sector. Ignoring the potential of the BE to

contribute to population PA and therefore health may result in low levels of

investment in infrastructure that supports active life styles. The lack of inclusion may

be due to the absence of agreed and robust methods, or scarce evidence of the

potential of the sector for public health, or both. This chapter explores a method for

the incorporation of PA-related health benefits in the appraisal of BE initiatives based

on the research conducted in Chapter 4 to 6. Specifically, this chapter addresses

Research Question 4: Can the health impact of changes in physical activity be

incorporated more robustly in cost-benefit analysis of built-environment initiatives in

Australia?

Zapata-Diomedi B, Gunn L, Shiell A, Giles-Corti B and Veerman JL, A method for

the inclusion of physical activity health outcomes in cost-benefit analysis of built

environment initiatives, under review Preventive Medicine (submitted 14/07/2017).

Authors’ contribution: Zapata-Diomedi B conceived the research question,

designed the methods, conducted the analyses, wrote the manuscript and co-edited

the paper. Gunn L, Shiell A, Giles-Corti B and Veerman JL critically reviewed and co-

edited the paper.

108

Abstract

The built environment has a significant influence on population levels of physical activity

(PA) and therefore health. However, PA-related health benefits are seldom considered

in transport and urban planning (i.e. built environment) economic evaluations. The usual

method used in economic evaluations of built environment initiatives is cost-benefit

analysis, which requires that the benefits of any intervention or policy are valued in

monetary terms to make them commensurable with costs. This leads to the need for

monetised values of the health benefits of PA. In the Australian setting, a range of such

values has been published. These are expressed in terms of dollars per kilometre

walked or cycled, and are specific to the transport sector. The aim of this study was to

explore a method for the incorporation of monetised PA-related health benefits in cost-

benefit analysis of a broader variety of urban planning interventions. We also identified a

range of values for these benefits in the Australian context. Our approach is based on

the proportional multi-state multi-cohort life table Markov model, which has been widely

used in peer-reviewed health literature. These methods could be adapted to assess the

health and economic impacts of specific urban development scenarios by working in

collaboration with urban planners.

109

List of abbreviations

BE Built environment

CBA Cost-benefit analysis

PA Physical activity

SP Stated preferences

VSL Value of statistical life

VSLY Value of statistical life year

WTP Willingness to pay

110

Introduction

It is now well established that city and transport planning influence the health and

wellbeing of urban populations (16). Worldwide, transport and land-use policies have

contributed to the increasing burden of non-communicable diseases and injuries, mainly

via physical inactivity, air pollution and road trauma (15). The transport and land

planning sectors are part of the broader “built environment” (BE) concept, defined by the

World Health Organization as ’Elements of the physical environment that are man-

made, in contrast to the natural environment. The BE includes everything from

metropolitan land-use patterns to urban transportation systems to individual buildings

and the spaces around them‘ (10 p28). There is a growing body of evidence on the

influence of the BE on health, specifically by either facilitating or hindering physical

activity (PA) (11, 179). However BE initiatives that improve population levels of PA may

expose individuals to increased risk from road trauma and air pollution (e.g. active travel

programs). There is evidence to suggest that PA benefits outweigh these other health

harms (37, 63).

For public initiatives in the transport and land use sectors, cost-benefit analysis (CBA) is

the recommended method for the ex-ante appraisal of public policies in Australia and

elsewhere (120, 121). As per Prest and Turvey, CBAs aim to ’maximise the present

value of all benefits less that of all costs, subject to specified constraints‘ (275 p686 ).

CBAs of public sector initiatives are commonly referred to as Social Cost-Benefit

Analysis (30, 276, 277). The philosophical underpinnings of CBA are in welfare

economics (97). Welfare economics is a branch of economics that aims to maximise

societal welfare, interpreted as the sum of individuals’ welfare (or utility) (278). CBA

aims to support the decision-making process by providing relevant information, and is

used together with other information to make decisions (277).

There are increasing calls for PA-related health benefits to be included in the appraisal

of BE initiatives (30, 120, 279). Lack of full consideration of health outcomes may

misdirect allocation of resources towards developments that do not facilitate PA (e.g.

motorways, disconnected streets, isolated housing developments, etc.), which may lead

to societally sub-optimal outcomes. The United States, Australia, New Zealand and

countries in Europe already produce guidance for the incorporation of PA-related health

benefits in CBAs of initiatives within the transport sector (280). For Australia, a range of

111

values have been suggested for incorporation of PA-related health benefits in CBAs of

transport initiatives. Such values are expressed on per kilometre basis and range from

less than A$0.02 to A$1.46 (2016) (281, 282) per kilometre cycled (Figure S1

supplementary material). For walking, values ranged from A$1.79 to A$2.92 (2016)

(121, 281) (Figure S2 supplementary material). The range of values identified can be

explained by both differences in the benefits included in the evaluation (e.g. mortality

and morbidity, healthcare costs and productivity gains) and differences in the methods

used to quantify those benefits. In Table S1 of the supplementary material we present a

summary of the methods used to estimate per kilometre monetised values. These

values are easily applicable to transport interventions with an active travel component

where estimates of additional kilometres walked or cycled are available or can be made

(see Table S2 in the supplementary material). However, these estimates were produced

in the grey literature, and have not been scrutinized in a formal peer-review process. In

addition, per kilometre estimates from the literature are applicable specifically in the

transport sector, rather than the BE more broadly. It would therefore be useful to link the

health benefits directly to characteristics of the BE (such as density, design

characteristics and diversity of land use), where estimates of the effects are not readily

translated into increases in kilometres travelled.

The aim of this study was to explore a method and a range of values that could

incorporate monetised PA-related health benefits in CBAs assessing a broad range of

BE initiatives. In addition, we used our methods to produce monetised values of the PA

health gains per kilometre walked and cycled for comparison with existing estimates

from the Australian ‘grey literature’.

Methods

The work presented here is based on the framework depicted in Figure 7-1 (179, 283).

First, we systematically reviewed Australian contemporary literature for measures of the

association between BE attributes and PA outcomes (179). Secondly, we translated

effect sizes (e.g. odd ratios, beta coefficients) reported in our systematic review into

average minutes of PA per week per adult living in a neighbourhood where an attribute

of the BE changes (283). Minutes of PA per week (walking and cycling) as well as times

per week are the most commonly collected type of data for studies of PA within the

neighbourhood area (284). We grouped effect sizes reported in the literature according

112

to five of the six ‘D’s proposed by Ewing and Cervero (134) (density, diversity of land

uses, destinations, distance to transit and design) plus aggregated neighbourhood

measures (i.e. walkability index). Third, we predicted annual average health-adjusted life

years (HALYs) and healthcare costs per adult residing in a neighbourhood where there

is a change in a feature of the BE using a mathematical model (283). HALYs are

population health measures that incorporate mortality and morbidity (285). Healthcare

costs include both savings from a reduction in physical inactivity related diseases and

increases in costs due to health needs in the prolonged life years attributable to

improvements in PA. Lastly, we multiplied predicted HALYs by the value of statistical life

year (VSLY) and added net healthcare costs to produce overall monetised values linked

to changes in features of the BE.

In the next section we explain the mathematical model and briefly discuss the

monetisation of HALYs and chosen VSLY. Detailed information about the literature

review and calculation of average minutes of PA per week per adult living in a

neighbourhood with a change in a BE feature can be found in Zapata-Diomedi and

Veerman and Zapata-Diomedi et al (179, 283).

113

Figure 7-1 Study framework

Mathematical model

Our mathematical model is based on the proportional multi-cohort multi-state life table

Markov model (MSLT) developed for the Assessing Cost-Effectiveness in Prevention

project (ACE Prevention) (94, 96, 100, 286). The MSLT model consists of a general life

table and a life table for each of the modelled diseases. We included five diseases

related to low levels of PA (ischemic heart disease, ischemic stroke, type 2 diabetes,

colon cancer and breast cancer in women (46-48)). We modelled 5-year age groups by

sex in the MSLT until everyone reaches the age of 100 or dies. The population impact

fraction (PIF) ‘relative risk’ shift method was used to estimate the effect of changes in

PA on disease incidence rates (101). Changes in incidence impact on prevalent

numbers of cases in later years, and consequently on years lived with disability and

mortality. HALYs represent life years adjusted for disability attributable to disease and

injury. The ‘relative risk’ method for calculating the PIF requires estimates of PA

prevalence at baseline, and corresponding relative risks (RR) of physical inactivity-

related diseases before and after a change in PA levels. The Australian Bureau of

Statistics collects data on four types of PA undertaken in the previous week by age and

$HALY

+

Net healthcare costs

=

Monetised PA-health related

benefit

Change in HALY and

Net healthcare costs

Average change in physical

activity minutes per week

Change in built environment

attributes

Review Australian literature

Mathematical model

Monetised PA-related health

benefits

114

sex: walking for transport, walking for recreation, moderate PA (excluding walking) and

vigorous PA (57). From this we derived PA scores by multiplying mean time spent in

each of the aforementioned PA categories by an assigned metabolic equivalent value11

(MET) (183) and then summing the results. These scores were used to categorise

people into those who were highly active (≥1,600 MET-minutes)/wk.), those who met the

recommended level of activity (600 to <1,600 MET-minutes/wk.), those who were

insufficiently active <600 MET-minutes/wk.) and those who were inactive (0 MET-

minutes/wk.) (47)12. We used Excel slope and intercept functions to fit RRs reported by

Danaei et al (2009) with RRs as the dependent variable and mean MET-minutes/wk.

per PA category at baseline as the independent variable. We used the slope and

intercept parameters to estimate RRs for the modelled scenarios with baseline MET-

minutes per PA category changing accordingly for each of the modelled scenarios.

Healthcare costs in the source study were expressed in 2010 values (283); we indexed

these to 2016 using the Health Price Index (287). Results are based on the results in

Zapata-Diomedi, Mantilla Herrera (283) for a non-linear dose-response function for PA

with health outcomes (See Table 3 Zapata-Diomedi, Mantilla Herrera (283)). Healthcare

costs and monetised HALYs were discounted at an annual rate of 3% (182). The MSLT

model was set up in Excel and an uncertainty analysis was conducted with the Ersatz

add-in tool (102). Uncertainty parameters are described in Table 2 from Zapata-

Diomedi, Mantilla Herrera (283).

Monetisation of HALYs

Although a full discussion on the monetisation of health outcomes is beyond the scope

of this paper, assigning a dollar value to health outcomes for their inclusion in CBAs is a

controversial topic (97). For example, Ackerman and Heinzerling (288) have argued that

11 “The ratio of the work metabolic rate to the resting metabolic rate. One MET is defined as 1 kcal/kg/hour and is roughly equivalent to the energy cost of sitting quietly. A MET also is defined as oxygen uptake in ml/kg/min with one MET equal to the oxygen cost of sitting quietly, equivalent to 3.5 ml/kg/min.” 183. Ainsworth BE, Haskell WL, Herrmann SD, Meckes N, Bassett DR, Jr., Tudor-Locke C, et al. 2011 compendium of physical activities: a second update of codes and MET values. Med Sci Sports Exerc. 2011;43(8):1575-81.

12 Note that Danaei et al.’s categorisation is as follows: highly active (≥1,600 MET-minutes)/wk. and ≥1h/wk. of vigorous PA), recommended level active (600 to <1,600 MET-minutes/wk. and either ≥1 h of vigorous PA/wk. or ≥2.5 h of moderate PA/wk.), insufficiently active <600 MET-minutes/wk. or <2.5 h/wk. of moderate PA) and inactive (0 MET-minutes/wk. of moderate or vigorous PA). We only based the categories on MET-minutes per week otherwise some participants would have not fitted in any category.

115

health and life cannot be valued and that any attempt to do so is fundamentally flawed.

Conversely, others (Johannesson and Jönsson (289)), supported the use of CBA for

health economic evaluations. In the following paragraph the focus is on methods for the

monetisation of HALYs and the approach used in this research.

As highlighted in the introduction to this paper, CBAs aim to maximise societal welfare,

understood as the sum of individuals’ welfare (utility, preferences) (97). Consistent with

welfare economics, HALYs should represent the collective utility value of an intervention

(97, 290). In welfare economics such utility value represents the societal willingness to

pay to benefit from an intervention’s outcome (97). Hence, for CBAs of health related

initiatives, methods are needed to elicit a monetary value as a measure of what society

is willing to sacrifice in order to gain an improvement in health.

To date, the literature presents the monetary value of a quality-adjusted life year

(QALY), which like our estimated HALYs represents the length and quality of life (97,

290-292). Another population health measure that represents length and quality of life is

the disability-adjusted life year (DALY). The DALY is a measure of health loss that was

developed for use in Burden of Disease studies (285). Like QALYs, our HALYs

represent health gains, but we use disability weights from the Global Burden of Disease

study to measure the quality component. We expand on the differences between these

measures in the supplementary material.

As per Mason, Jones‐Lee (292) two methods may be applied to elicit the willingness to

pay to derive the monetary value of a QALY. One approach is to elicit its monetary

value directly using survey methods. Another approach is to derive it from the monetary

value of a prevented fatality, also referred to as the value of statistical life (VSL). It is

important to note that the VSL is also derived using survey methods.

Revealed preference or contingent valuation (stated preferences) studies (97, 290-292)

have been used to directly elicit the monetised value of a QALY. Revealed preference

studies are based on real world market transactions that represent actual risk-taking

behaviours (97, 293). The most common type of revealed preference studies are wage-

risk trade-off studies, which measure the compensating wage differential that individuals

are willing to accept for jobs that represent health risks (97). In contingent valuation

studies, individuals are presented with hypothetical scenarios and are asked to reveal

the maximum amount that they are willing to pay for the potential health benefits. The

116

most common approach to elicit the willingness to pay for a QALY is to ask a sample

from the general population about improvements in general health, rather than targeted

health conditions (290). Both revealed preferences and contingent valuation have some

potential drawbacks, mostly related to the bias introduced by the high context specificity

of the jobs (revealed preferences) and framing of the questions (contingent valuation)

(97, 293). A common disadvantage to both is that eliciting the monetary value that the

society places on a QALY for each individual intervention is time and resource

consuming. A commonly used alternative is therefore to derive the monetary value of a

QALY from an established VSL (291).

Generally, the VSL is understood as the marginal dollar value placed on a small

reduction in the risk of death (293, 294). Estimates for the VSL have long been

produced using contingent valuation studies of life saving policies in the transport and

environment sectors (97). For example, in the transport context, contingent valuation

studies are conducted to elicit the willingness to pay for a reduction in the risk of road

fatality. To derive the monetised value of a QALY from the VSL, a present value

discount formula assuming an average age and life expectancy based on the population

of interest is used. As an example, Ryen and Svensson (290) proposed solving x in

equation 1 to derive the willingness to pay for a QALY from a VSL estimated at $3

million for a population with an average age of 38 and a remaining life expectancy of 40

years.

3 𝑚𝑖𝑙𝑙𝑖𝑜𝑛 = ∑𝑞𝑡+38∗𝑥

(1+0.03)𝑡𝑡=39𝑡=0

Equation 1

Where q represents the quality of life adjustment for an average 38-year old and t

represents the remaining life expectancy at age x.

In Equation 1, each future life year is adjusted for age specific quality of life weights. For

example, in Hirth, Chernew (295) an adjustment for quality-of-life of 0.9 was applied for

a life year of an average 45-year old. If the quality of life component is removed from

equation 1, the value of statistical life year (VSLY) is calculated. The VSLY is broadly

interpreted as the marginal dollar value of a healthy human life year (293).The VSLY

has been widely used to monetise DALYs attributable to low level of PA for their

incorporation in CBAs of active transport interventions (see table S1 in the

supplementary material). However, this is likely to underestimate the value of the DALY

117

since a DALY represents a year in full health lost, and the VSLY represents a year at

the average quality at middle and old age.

Nevertheless, using the VSLY is a convenient method that is widely used to monetise

PA-related health benefits for incorporation in CBAs of active transport interventions.

This method assumes that individuals’ preferences from a reduction in the risk of

mortality and morbidity attributable to low levels of PA are the same as those from

lifesaving interventions.

The Australian context

In this research we used the VSLY to translate HALYs into monetised values. There is

not a single, universally accepted VSLY in use for Australia. At the federal level, the

Office of Best Practice Regulation (OBPR) recommends using a VSLY of A$182,000 in

2014 dollars in the preparation of CBAs for interventions that aim to reduce physical

harm (OBPR, 296)13.The OBPR estimate was recommended by Abelson (297, 298)

after reviewing literature from Australia, the United States and European countries

(VSLY 2003 value A$108,000). The VSLY in Abelson is based on European values that

used wage-risk and contingent valuation methods (297, 298). The VSLY was derived

from a VSL of A$2.5 million (2003) assuming a remaining life expectancy of 40 years

with a 3% discount rate applied (297, 298). A life expectancy of 40 years is based on the

average life expectancy and average age of the population (298).

Using the multi-state life table approach to estimate per kilometre values

To compare with estimates produced in the literature, and so provide a degree of

“comparative validation” of our method, we also used the MSLT model to estimate

monetised values per kilometre walked and cycled. We produced annual values

assuming that walking one kilometre takes 12 minutes (at a speed of 5 km per hour)

and cycling one kilometre takes 3 minutes (20 km per hour). We used MET rates from

the PA compendium for walking and cycling (3.5 and 6.8) (183). We projected the

13 The original document says: “Many regulations have the benefit of reducing the risk of injury, diseases or disability. One method to value these benefits is to adjust the value of statistical life year (which could be interpreted as the value of a year of life free of injury, disease and disability) by a factor that accounts for the type of injury, disease or disability” 296. Office of Best Practice Regulation (OBPR). Best Practice Regulation Guidance Note: Value of statistical life: Department of Prime Minister and Cabinet. Australian Goverment; 2014 [5 May 2015]. Available from: https://www.dpmc.gov.au/sites/default/files/publications/Value_of_Statistical_Life_guidance_note.pdf.

118

monetised value for walking and cycling 1 kilometre per week divided by 52 (weeks in a

year).

Results

Monetised PA-health related values linked to changes to the built

environment

The range of estimated values is depicted in Figure 7-2. The interpretation is in terms of

the annual average monetised PA-related health benefit for a person living in a

neighbourhood where BE changes are made, which could serve as reference values in

CBAs of BE interventions. The monetised values presented here are specific to the

changes in exposures to attributes of the BE in the source studies. For example, while

some estimates may be directly applicable to CBAs of BE initiatives such as an

improvement in one transport destination. Others are rather specific, for example, an

increase in street lights in a neighbourhood from 315 to 780.

Overall, the results suggest that the greatest gains would be accrued from increasing

destinations, walkability and attributes of design within the neighbourhood area. Annual

PA-related health benefits worth up to A$70 per resident were estimated for

improvements in neighbourhood destinations. Improving neighbourhood walkability was

estimated to be worth up to A$30 and improvements in sidewalks availability (measure

of design) up to A$22 per adult resident. There is great variability in our estimated

monetised values per BE category attributable to the estimates of effect from our source

studies.

As highlighted in the introduction, we aimed to explore a methodology, and our methods

can be adapted to specific evaluations needs. For further details on results, including

uncertainty ranges and the magnitude of change in BE attributes and in PA, see Table

S3 of the supplementary material. Our results indicated that the monetised value of a

HALY determines the overall value of PA. The savings in treatment cost of PA-related

diseases are much smaller, and are negated by all other healthcare costs in added

years of life.

119

Figure 7-2 Monetised PA-related health benefits per year per adult living in a neighbourhood where built environment changes are made (A$ 2016)a

a Uncertainty intervals can be found in Table S3 of the supplementary material

120

Monetised PA-related health values per kilometre walked and cycled

Our results indicate that the value of PA-related health benefits associated with walking

is A$0.98 (95% Uncertainty Interval (UI) 0.73 to 1.24) per kilometre. For cycling the

benefits are worth A$0.62 (95% UI 0.76 to 0.79) per kilometre.

These estimates are in the lower end of the range compared with those proposed in the

‘grey literature’. In Table 7-1 we present the values and summarise the outcomes in

each of the estimates from the literature and our estimates. We cannot conclude that

our values are lower due to a narrower inclusion of outcomes. Instead, this must be

attributed to differences in methods or data.

Table 7-1 Comparison of included outcomes in per kilometre estimates

Study

Values per km (A$ 2016)a

Mortality Morbidity Healthcare

costs

Healthcare costs in

added life years

Productivity

ATAP (299) Walking: $2.92 Cycling: $1.46

Transport for New South Wales (121)

Walking: $1.79 Cycling: $1.26

SKM & PWC (243)


Mulley, Tyson (30)


PWC (300) Walking: $2.36

AECOM (2010)

Cycling: $0.26

PWC (282)b Cycling: (1) $0.01 and (2) $0.71

This study Walking: $0.98 Cycling: $0.62

a For further information for values from the literature please refer to Table S1 in the supplementary

material b Value (1) only includes healthcare costs of mortality from heart attack. Value (2) is from a sensitivity analysis using the HEAT tool.

Discussion

Summary

We described a method and proposed a range of values that could be used for the

incorporation of monetised PA-related health benefits in CBAs of BE initiatives. Our

values directly link monetised health outcomes to the D categories (density, diversity of

land uses, destinations, distance to transit and design) of the BE, plus aggregated

121

neighbourhood measures (i.e. walkability index). Our proposed values can be used as

ball-park estimates to assess specific urban planning scenarios where the planner aims

to include wider societal benefits, such as health effects. We offer a flexible approach

applicable to the Australian setting, however, given the availability of required data, it

can be adapted to any context.

In addition, we estimated monetised PA-related health benefits per kilometre walked or

cycled. To the authors’ knowledge such values have not previously been published in

the peer-reviewed literature.

Strengths and limitations

The main strength of this study is the robust prediction modelling approach that is based

on the established and well-validated methods of the MSLT. It can be implemented in

spreadsheets, which makes it transparent. The MSLT has been widely used for the

assessment of health risk factors including high body mass index, tobacco smoking and

physical inactivity (286, 301). Our modelling framework is more complex and demanding

in terms of data inputs than those used in in the literature presented in the

supplementary material for monetised values per kilometre walked and cycled.

However, for Australia, the required data were available from the national statistics office

and burden of disease studies. In this paper we focused on HALYs, but the model can

be customised to include other outcomes such as disease incidence and prevalence,

health-adjusted life expectancy and life years.

A number of limitations should be highlighted. Most of the literature assessing the

association BE-PA indicates the odds of doing PA for a given exposure rather than an

indication of magnitude that can be used to estimate health outcomes (192). Thus,

assumptions were needed to translate effect sizes reported in the literature into minutes

of PA per week, implying that our results should be only used as ballpark estimates (see

Zapata-Diomedi et al. (2016)). In addition, most of the evidence for the BE-PA

association is from cross-sectional studies, which by themselves do not prove causality

(11). Furthermore, we assumed that a change in the BE translates into a change in

overall PA, but it could be that the additional activity partly or fully replaces PA

previously performed for other purposes, which would imply our results are biased

upwards. Quasi-experimental studies of real interventions that measure total PA at

122

multiple time points with case and control groups would provide a better indication of the

effect of an intervention on total PA. Cluster randomised control trials would provide the

highest quality of evidence for the effect of a BE intervention. Such trials are, however,

not practical in this field. A further advancement would be to have complete information

for the participants as to their initial levels of PA (e.g. inactive, insufficiently active and

sufficiently active) so that an assessment of who benefits from the intervention could be

made, and hence more precise estimates produced. In this study we assumed that

everyone changes PA, with those already achieving 1,600 MET-minutes per week or

more at baseline not receiving health benefits. This is supported by a recent quasi-

experiment investigating the effect of transport infrastructure on walking, which found

that the interventions encouraged walking in both active and inactive participants at

baseline (302).

Conclusions

Traditionally, CBAs of BE initiatives do not include potential PA-related health benefits,

resulting in suboptimal welfare outcomes for society. This study provides a range of

estimates for the value of PA-related health outcomes that can be used as reference

values in CBAs of BE interventions for the Australian setting. Incorporating PA-related

health in the appraisal of BE interventions may lead to healthier urban designs.

123

Chapter 8 A shift from motorised travel to active transport: what are

the potential health gains for an Australian city?


In Chapters 5 and 7 it was demonstrated that the BE can contribute to population health

and the economy by improving levels of PA. This is now acknowledged by Australian

governments, with many examples of transport departments working towards travel

mode–share targets aimed at reducing private car travel and increasing active travel—

walking, cycling, and use of public transport. However, for change to take place,

investments for the provision of safe and convenient alternative options to private cars

are needed. In this chapter, potential health and healthcare costs associated with the

Australian city of Brisbane achieving travel targets (1.1 million residents in 2013) are

estimated. The findings from this study are supportive of the need to continue investing

in active-travel initiatives as the potential gains in terms of health would contribute to

lessen the disease and economic burden from an unhealthy population. This chapter

shares common prediction modelling methods with Chapters 5 and 7, however, adds

two additional exposures to risk factors (road trauma and air pollution). In addition, while

Chapters 5 and 7 estimates are based on estimates of effects from the literature, here

prediction are for a hypothetical scenario of travel targets aspirations of greater walking,

cycling and public transport by local and state governments. This chapter addresses

Research Question 5: What are the potential health and economic impacts of Brisbane

meeting its targets for active travel?

Zapata-Diomedi B, Knibbs LD, Ware RS, Heesch KC, Tainio M, Woodcock J &

Veerman JL, ‘A shift from motorised travel to active transport: What are the potential

health gains for an Australian city?’, under review PLoS One (submitted 25/03/2017).

Authors’ contribution: Zapata-Diomedi B conceived the research question, designed

the methods, conducted the analyses, wrote the manuscript and co-edited. Knibbs LD

contributed to design of the air- quality component; and critically reviewed and co-edited

the paper. Ware RS contributed to the analysis of the household travel data; and

critically reviewed and co-edited the manuscript. Heesch KC contributed to the analysis

124

of the household travel data; and critically reviewed and co-edited manuscript and co-

edited. Tainio M contributed to the design of the air-quality component; and critically

reviewed and co-edited the manuscript. Woodcock J critically reviewed and co-edited

the manuscript. Veerman JL contributed to the design of the study; and critically

reviewed and co-edited the manuscript.

125

Abstract

Introduction

An alarmingly high proportion of the Australian adult population does not meet national

physical activity guidelines (57%). This is concerning because physical inactivity is a risk

factor for several chronic diseases. In recent years, an increasing emphasis has been

placed on the potential for transport and urban planning to contribute to increased

physical activity via greater uptake of active transport (walking, cycling and public

transport). In this study, we aimed to estimate the potential health gains and savings in

healthcare costs of an Australian city achieving its stated travel targets for the use of

active transport.

Methods

Additional active transport time was estimated for the hypothetical scenario of Brisbane

(1.1 million population 2013) in Australia achieving specified travel targets. A multi-state

life table model was used to estimate the number of health-adjusted life years, life-

years, changes in the burden of diseases and injuries, and the healthcare costs

associated with changes in physical activity, fine particle (<2.5 µm; PM2.5) exposure, and

road trauma attributable to a shift from motorised travel to active transport. Sensitivity

analyses were conducted to test alternative modelling assumptions.

Results

Over the life course of the Brisbane adult population in 2013 (860,000 persons), 33,000

health-adjusted life years could be gained if the travel targets were achieved by 2026.

This was mainly due to lower risks of physical activity related diseases, with life course

reductions in prevalence and mortality risk in the range of 1.5%-6%. Prevalence and

mortality of respiratory diseases increased slightly (≥0.27%) due to increased exposure

of larger numbers of cyclists and pedestrians. The burden of road trauma increased by

30% for mortality and 7% for years lived with disability. We calculated substantial net

savings ($AU183 million, 2013 values) in healthcare costs.

Conclusion

In cities, such as Brisbane, where over 80% of trips are made by private cars, shifts

towards walking, cycling and public transport would cause substantial net health

126

benefits and savings in healthcare costs. However, for such shifts to occur, investments

are needed to ensure safe and convenient travel.

Key words

Active transport, health, physical activity, air pollution, road trauma, policy

127

Introduction

The built environment, largely determined by policies in the planning and transport

sectors, contributes greatly to health risk factors including physical inactivity, traffic-

related air pollution and road trauma globally (15, 303).

An emerging body of literature is examining the impacts of built environment initiatives

on health and economic outcomes. Two recent reviews of economic evaluation and

health impact assessments of active transport interventions found that the greatest

individual health gains of active travel policies are achieved by increasing physical

activity (PA) levels (37, 304). For those who undertake active travel, trade-offs can arise

in terms of higher exposure to environmental hazards, including pollution (air and noise),

heat and road injuries (37, 87), but the evidence suggests that PA benefits outweigh

these other risks (19, 37, 305).

In Australia, 57% of adults do not meet the national PA guidelines (57), and inactivity

has been estimated to result in the annual loss of nearly 124,000 disability-adjusted life

years in 2015 (DALYs) (306). With nearly 80% of adults’ travel for work or education by

private cars (307), encouraging active travel is a feasible avenue to improve population

health. Government and non-government agencies are working towards a shift from

private cars towards active transport (24-26, 177, 308). For Brisbane, the capital of the

State of Queensland and Australia’s third most populous city, strategic planning at the

city and state level aims to achieve a mode share of 15% for walking, 5% for cycling and

14% for public transport (25, 308). Quantifying health outcomes of Brisbane’s transport

strategy could support the case for the required investment.

In this study, we quantified health outcomes and healthcare costs of replacing private

car trips with active transport in Brisbane. A shift from car travel to active transport was

based on the transport targets for South East Queensland (25) and Brisbane (308)

(aiming for 15% for walking, 5% for cycling and 14% for public transport). We projected

the potential health benefits and healthcare cost savings of a linear annual increase in

active travel at the expense of private car travel from 2013 up to fully achieving the

targets in 2026.

128

Methods

Study area

The Brisbane Local Government Area is located in South East Queensland, Australia.

Brisbane consists of 188 mainland suburbs as well as additional islands and localities in

Moreton Bay (309). A total of 1,131,191 people lived in Brisbane in 2013, with over 80%

of the population aged over 15 years (median 34.5 years) (310, 311).

Survey data

We used daily travel information collected from the South East Queensland Household

Travel survey (312). Data were collected from April to September 2009 (313). Data

from the 2012 version of the survey were available, but were not used because the

survey was incomplete (314).

For the survey the sampling unit was at the household level, and data was collected on

all individuals (≥5 years of age) living in selected households. Information was self-

reported via paper-based questionnaire about all trips taken in a one week period.

These were selected using a three-stage, variable proportion, clustered sampling of

household addresses within Census Collection Districts (CCDs). CCDs are the second

smallest geographical collection unit for census and population data collection and

processing (315). Sampling regions were divided into those participating and not

participating in the Travel Smart program, a behaviour change program that aimed to

reduce private car travel by increasing active transport and share rides, and that was

being implemented at the time of data collection (25). The sampling process consisted

of randomly selecting CCDs in each sampling region, followed by a random selection of

56 dwellings within each CCD, with 42 of them kept in the primary sample and 14 as

potential replacements ( 313). Field checks took place for the identification of sample

loss and ensure that at least 42 dwellings per CCD would be included.

In total 4,240 households in Brisbane responded to the survey, representing a response

rate of 52%, 11,191 persons and 32,536 trips (313). As the sample of selected

households may not have been representative of Brisbane households, weights at the

trip and person level were applied (313). Trips that were made by persons aged <17

years; were taken on weekends; or did not include walking, cycling, public transport or

private car were excluded, leaving 19,385 trips available for analysis. Analyses were

129

conducted using Stata v.13 (StataCorp, College Station, TX: StataCorp LP). A trip was

defined as a one-way movement from one place to another with a single purpose

(transport or recreational). One trip could include multiple modes (313). Two variables

were provided in the data set to represent the main mode for a multi-modal trip: (1)

longest-time mode; and (2) priority mode based on hierarchical order for public transport

(313). The longest-time mode assigned the main mode for a multi-modes trip to the

mode used with the longest time. These variables were highly correlated (r=0.99). We

used the longest-time mode variable given that no justification is given in the source

document for the chosen hierarchy.

Travel patterns

Of all weekday trips made by Brisbane adults in 2009, 24% were <2km, 24% 2-5km,

33% 6-16 km and 19% >17 km. Our estimates for trips >5 km are higher than those

reported for Australia as a whole (25% (307)). However, national census data refer to

trips for work or full time study only, whereas we included all transport trips (made to get

from one place to another). A Government report based on census data stated that for

Brisbane, 24% of the trips for work commutes were > 5 km and 22% were 5-10 km (61).

Similarly, in our dataset 21% of work commute trips were >5 km and 23% were 5-10

km. Table 8-1 depicts 2009 average weekday trips by mode and overall by age and sex.

Table 8-1 Mode-specific mean (95% Uncertainty Interval (UI)) trips per weekday in 2009, by age and sex

Age (years) and sex Car

occupant Walk Bicycle

Public Transport

Total

17-49, male 2.07

(1.94 to 2.19) 0.22

(0.18 to 0.27) 0.063

(0.038 to 0.087) 0.28

(0.23 to 0.33) 2.71

( 2.59 to 2.83)

17-49, female 2.80

(2.66 to 2.95) 0.33

(0.28 to 0.38) 0.018

(0.009 to 0.026) 0.29

(0.25 to 0.33) 3.46

(3.32 to 3.60 )

50-74, male 2.51

(2.37 to 2.68) 0.25

(0.19 to 0.31) 0.031

(0.012 to 0.05) 0.13

(0.09 to 0.16) 2.96

(2.79 to 3.13)

50-74, female 2.38

(2.22 to 2.54) 0.27

(0.31 to 0.33) 0.004

(-0.002 to 0.010) 0.18

(0.14 to 0.23) 2.85

(2.70 to 3.02)

75 plus, male 1.81

(1.45 to 2.18) 0.14

(0.06 to 0.22) -

0.09 (0.015 to 0.16)

2.07 (1.72 to 2.42)

75 plus, female 1.17

(0.87-1.45) 0.13

(0.05-0.20) -

0.15 (0.06-0.24)

1.48 (1.18-1.77 )

Travel targets

Brisbane aims for a travel mode share of 15% for walking and 5% for cycling (308).

Because no targets were proposed at the city level for public transport, we applied the

regional (South East Queensland) target of a 14% share (25). At the regional level,

130

similar targets were proposed for walking and cycling, with an overall share of 20% for

active transport. We interpreted the targets to apply only to adults, given that separate

targets were proposed at the regional level for trips to school (made by children) (25).

We assumed that the targets apply to weekday trips, given that this was specified at the

regional level (25). We followed the South East Queensland report that outlined the

travel targets (25), and which recommends that trips <5 km can be made by bicycle.

The TMR strategy suggests that <1.2 km is a walkable distance, whereas national l

active transport statements suggest <2 km (316). In this study, we transferred trips

made by car of <2 km to walking and trips between 2-5 km to cycling. We assumed that

trips of 5-16 km made by car can be replaced by public transport. Our assumption of

replaceable distances is aligned with past data on average distance travelled reported

by mode of 1 km for walking, 4 km for cycling and 15 km for public transport (317).

Transfers from car occupant (driver and passenger) to active transport trips per week

reflected the age and sex distributions of transport mode share at baseline. To estimate

the distance of walking and cycling trips that replaced car occupant trips, the mean

kilometres travelled was calculated for each of the three distance categories (<2; 2-<5

km; 5-16 km) separately for each age and sex group (Table 8-2). These estimates were

used to estimate the health impact of increasing PA levels. To estimate the health

impact of increasing PA levels by shifting car occupant travel to public transport walking,

the mean minutes of public transport walking was calculated for each of the three

distance categories separately for each age and sex group. Decreases in car occupant

kilometres travelled per year were used to estimate the health impact of exposure to

PM2.5 and road trauma.

131

Table 8-2 Mode share travel targets

Baseline Travel targets Change in number of weekday trips (% mode)

9 % walking 1 % cycling 8 % public transport 82 % car occupants

15% walking 5% cycling 14% public transport 66% car occupants

291,834 (65%) 196,864 (390%) 291,834 (73%) -780,531 (-19%)

Baseline mode distribution by age and sex

17-49, male 17-49, female

50-74, male

50-74, female

75 +, male 75 +, female

Walking 26% 41% 14% 15% 1% 2% Cycling 61% 20% 17% 2% 0% 0% Public transport

37% 40% 8% 12% 1% 2%

Travel target scenario mean increase in average weekday trips by age and sexa

Persons 475,481 486,670 217,226 226,625 35,807 49,980

Walking 0.8 1.24 0.95 0.99 0.56 0.49 Cycling 1.27 0.4 0.77 0.10 0 0 Public transport 1.15 1.19 0.53 0.76 0.46 0.63

Baseline average car trip length by distance category by age and sex

> 2km 1.27 1.27 1.17 1.18 1.19 1.19 2-5 km 3.37 3.32 3.34 3.31 3.18 3.04 5-16 km 9.47 8.91 9.57 9.44 8.74 7.31

Baseline mean minutes walking public transport trips by age and sex

13.39 13.15 12.58 12.77 10.03 13.31

Decrease in car occupant kilometres per year by age and sexb

Walking 24,880,419 39,893,226 12,584,681 13,781,943 1,240,537 1,531,142 Cycling 105,524,358 33,231,580 29,150,158 3,732,041 - - Public transport

269,397,592 267,323,368 57,491,462 84,207,133 7,424,688 11,975,262

a Equals: Change in number of weekday trips*Baseline mode distribution by age and sex/Persons in age and sex group*5 (weekdays). b Equals: Travel target scenario mean increase in daily trips by age and sex * Baseline car trip length by distance

category by age and sex*Persons in age and sex group*260 (weekdays in a calendar year).

Quantification of health outcomes and healthcare costs

Health outcomes and healthcare costs were estimated over the lifetime of the Brisbane

adult population with 2013 serving as the baseline year. Health outcomes were derived

from changes in average population PA, exposure to annual mean ambient fine

particulate matter (PM2.5), exposure to on-road PM2.5, and road trauma (injuries and

fatalities). PM2.5 is a widely-used proxy for exposure to air pollution during travel (37).

We used a mathematical model based on the proportional multi-cohort multi-state life

table Markov model (MSLT) developed for the Assessing Cost-Effectiveness in

Prevention project (ACE-prevention) (94, 96, 100, 286).

We compared health outcomes and costs for a scenario in which age- and sex-specific

travel patterns persist from 2013 to 2026, with a scenario in which proposed travel

targets would be achieved by 2026. Outcomes and costs associated with the travel

132

targets scenario were assumed to be the result of a gradual increase in active travel.

Outcomes were estimated by dividing status-quo and scenario populations into 5-year

age groups (20-24 to 95 plus) by sex and simulating each cohort in the MSLT until

everyone dies or reaches the age of 100. Health outcomes included: health-adjusted life

years (HALYs), life years, prevalent cases, deaths and years lived with disability (YLDs).

HALYs are estimated as years of life lived adjusted for disease-related quality of life.

Healthcare costs of included diseases were calculated by dividing total cost of a disease

by the number of incident or prevalent cases for each 5-year age-sex group. For road

injuries we estimated the healthcare costs per year lived with disability. We present

undiscounted health outcomes (193) and used a 3% annual discount rate for healthcare

costs (182). We tested the sensitivity of our results to discounting health at 3% and

healthcare costs at 5% (Table S14 supplementary material). Ninety-five percent

uncertainty intervals were determined for all outcome measures by Monte Carlo

simulation (2,000 iterations), using the Excel add-in tool Ersatz (Epigear, Version 1.34)

(102). Figure 8-1 depicts the study’s analytical framework. Input parameters and

uncertainty distributions are presented in Table 8-3 and the supplementary material.

133

Figure 8-1 Analytical framework

Achieving the travel targets translates into an increase in cycling, walking and use of public transport at the expense of

private car travel (thick solid lines), which was translated into gains in HALYs, gained life years, reduced healthcare

costs, prevented/increased prevalent cases (diseases) and changes in death rates (thick lines at the bottom of the

graph). Averted years lived with disability were estimated for road trauma. We modelled the effect of PA and PM2.5 via

their impact on incidence of diseases (thin lines) and road trauma via its impacts on disability and mortality (captured by

HALYs and YLDs) (interrupted thick lines). The effect of less road trauma was quantified as improvements in ambient

PM2.5, which benefits the population as a whole (interrupted thin lines).

Travel targets

Decreased private car use

Increased cycling

Increased walking

Increased public transport

Physical activity Ambient PM2.5 On-road PM2.5 Road trauma

Diabetes

Breast cancer

Colon cancer

Ischemic stroke

Healthcare costs


HALYs

Chronic Obstructive Pulmonary Disease

Tracheal, bronchus and lung cancers

Life years Prevalent cases Deaths YLDs

134

Table 8-3 Proportional multi-state life table Markov model input parameters

Input parameter Uncertainty/Parametersa

Source

Baseline and travel target scenario

2013 mortality rates and population numbers

N/A Australian Bureau of Statistics (318, 319)

Years live with disability (YLD) (all causes and road trauma)

N/A Institute of Health Metrics and Evaluation (169)

Incidence and case fatality modelled diseasesb

N/A

Derived with DisMod II (102) from Global Burden of Disease (GBD) 2013 data (188) and Australian Institute of Health and Welfare (AIHW) data (320) (see supplementary material ‘Notes on DisMod II modelling’)

Disability weights modelled diseases

N/A Derived from prevalence and years lived with disability from GBD 2013 (see supplementary material ‘Disability weights’)

Relative risks physical activity Normal (Ln RR)c

Danaei et al. (47)

Relative risks of ischaemic heart disease and ischaemic stroke due to diabetes

Normal (Ln RR)c

Asia Pacific Cohort Studies Collaboration (189)

Relative risks PM2.5 Normal (Ln RR)c

World Health Organization (91), Hamra et al. (90)

Mediating effect factors for diabetes in the association physical activity-ischemic heart disease/ischemic stroke

Normal GBDd 2013 study (184p. 711 supplementary material)

Categories of physical activity Dirichlet

MET-minutes PA categories Lognormal National Nutrition and Physical Activity Survey Basic Confidentialised Unit Record File (CURF) (57, 321)

MET-minutes (walking 3.5, cycling 6.8, moderate PA 5 and vigorous PA 7.5)

N/A Ainsworth et al. (183) (walking and cycling) (57) (walking, moderate and vigorous PA)

Healthcare costs N/A

AIHW (106) all diseases except COPD (322) indexed to 2013 using AIHW reported health sector indices (323, 324). Denominators for per case costs (incidence, prevalence and years lived with disability) (IHME 325)

Discount rate for healthcare costs

N/A Health: Murray et al. (193) , healthcare costs: Gold et al. (182)

Travel targets scenario

Contribution to transport mode by age and sex

Dirichlet South East Queensland household Travel survey (312)

Mean distance travelled by car occupants per distance category by age and sex

Lognormal South East Queensland household Travel survey (312)

Total distance travelled by mode N/A South East Queensland household Travel survey (312)

PM2.5 concentration N/A Queensland Goverment (326)

Source apportionment PM2.5 N/A Friend et al. (327)

Road trauma Gamma Crash data: Queensland Department of Transport and Main Roads (328). Assumed Standard Deviation of 20% of the mean.

a Uncertainty distributions around input parameters are presented in the supplementary material. b Breast cancer, colon cancer, tracheal, bronchus and lung cancer, type 2 diabetes, chronic obstructive pulmonary

disease, ischemic heart disease and ischemic stroke.

135

c Note that a modified version of the log of the relative risk function was used as per to avoid a skewed lognormal

distribution (329). d Global Burden of Disease (GBD). ABS: Australian Bureau of Statistics; IHME: Institute of Health Metrics; AIHW: Australian Institute of Health and Wellbeing; WHO: World Health Organization; NNPA: National Nutrition and Physical Activity survey; TMR: Department

of Transport and Main Roads

Physical activity

The ‘relative risk shift’ method for the calculation of population impact fractions (PIF)

(101) was used to estimate incidence of physical inactivity-related diseases due to

changes in PA in the travel targets scenario. The calculation of the PIFs requires data

for prevalence of PA and relative risks (RRs) for physical inactivity-related diseases.

We derived age- and sex-specific baseline PA prevalence estimates. From national

survey data collected for PA surveillance (57), we computed mean minutes spent in the

last week walking for transport, walking for recreation, doing moderate PA (excluding

walking) and doing vigorous PA. We also created a PA MET-minutes/week score by

multiplying the minutes spent in each of these PA types by an assigned metabolic

equivalent value (MET) from a PA compendium (183) and summing them. Following

Danaei et al. (47), these scores were used to categorise participants into highly active

(≥1,600 MET-minutes)/wk. and ≥1h/wk. of vigorous PA), recommended level active (600

to <1,600 MET-minutes/wk. and either ≥1 h of vigorous PA/wk. or ≥2.5 h of moderate

PA/wk.), insufficiently active <600 MET-minutes/wk. or <2.5 h/wk. of moderate PA) and

inactive (0 MET-minutes/wk. of moderate or vigorous PA).

The additional minutes walked, cycled and in public transport were summed to create

an expected additional minutes in active transport per week. To estimate the increase in

minutes walked, the expected increase in the number of walking trips per week was

multiplied by the expected reduction in km driven by car (Table 8-2) and divided by

walking speed (Table S5 supplementary material). The additional minutes cycled per

week was estimated in the same way. The additional minutes per week spent in public

transport walking, defined as walking to/from public transport destinations, was

estimated as the product of the expected additional public transport trips in the travel

targets scenario and the mean time spent walking per public transport trip at baseline

(Table 8-2). Last, the expected additional minutes in active transport were multiplied by

marginal MET-rates to derive the scenario mean energy expenditure by age and sex.

136

RRs were estimated for a four-tier dose response relationship (inactive, low active,

moderately active and highly active) between PA and health outcomes, as done

previously by Danaei et al. (2009). For modelling the relationship, categorical RRs were

converted into continuous functions. Assuming non-linear associations between PA and

health outcomes, we fitted log-linear functions with a power transformation of mean

energy expenditure per PA category (inactive, low active, moderately active and highly

active) serving as the independent variable and RRs reported in the source data (47)

serving as the dependent variable (0.5 for type 2 diabetes, ischemic heart diseases and

breast cancer and 0.25 for ischemic stroke and colon cancer). Next, we used the

estimated parameters (intercept and slope) to estimate the RRs for the baseline and

travel targets scenario per PA category. Those in the highly active group had a RR of

1.00 in the source study, implying no additional benefit from extra PA (47). Because

type 2 diabetes is a risk factor for cardiovascular disease, estimated RRs incorporated

the increased risk of ischemic heart disease and ischemic stroke among those with type

2 diabetes. To avoid double counting we reduced the PIFs for PA with ischemic heart

disease (14%) and ischemic stroke (8%) (184).

PM2.5

Health effects from PM2.5 were estimated at two levels: (1) health effects for the

population from a decrease in exposure attributable to a reduction in private car

kilometres travelled, and (2) health effects at the individual level from exposure during

active travel and accounting for increased inhalation. We used information on the

exposure-response relationship for PM2.5 and health, as well as differential exposure to

PM2.5 between the baseline and travel targets scenarios. We used RRs from a World

Health Organization meta-analysis study on the long-term health effects of exposure to

PM2.5 on cardiovascular and respiratory disease (91). We applied the cardiovascular RR

to ischemic heart disease and ischemic stroke. We also incorporated lung cancer using

the RR from a meta-analysis by Hamra et al. (90). We calculated RRs compared to

baseline exposure (305) to modify incidence rates of PM2.5 related diseases in the

MSLT.

Our calculations required data on background PM2.5 concentrations for the baseline and

travel targets scenarios. We calculated average background levels of PM2.5 for the

Brisbane area from hourly measurements collected between 2006 to 2014 at six

137

regulatory monitoring sites located across Brisbane, which all used standard reference

methods to measure PM2.5 mass concentrations (i.e. tapered element oscillating

microbalances) (326). We calculated the arithmetic mean from all available data

excluding sites with less than 75% of the measurements for a given year. This was to

avoid seasonal bias in the estimates. We used source apportionment data specific to

Brisbane to estimate the proportion of PM2.5 emissions attributable to motor vehicles

(327). Source apportionment data were collected from two sites in Brisbane, one urban

and one suburban, with considerable variation in the proportion of PM2.5 attributable to

motor vehicles (7% and 30%). We took the average and compared our results to other

measurements at the national level, with our average of 18% comparing well with

national estimates of 17% (see Table 9 in reference 330). A range of motor vehicles

contribute to traffic emissions, including passenger vehicles, but source apportionment

data for Brisbane were only available for motor vehicles as a whole, rather than

passenger vs. heavy vehicles. We used data for Queensland to allocate the proportion

of motor vehicle emissions to passenger cars and buses (28% and 10%) (331). We

conducted a sensitivity analysis assuming that passenger vehicles emit 65% of the

motor vehicles related PM2.5 (332). We also tested the sensitivity of our results of using

the two extreme source apportionment values.

Motor vehicle-related PM2.5 for the travel targets scenario was reduced in equal

proportion to the reduction in car passenger cars kilometres travelled per year. The

decrease in car passenger cars kilometres travelled was estimated using data from the

household travel survey (312) as the total distance replaced by active modes and public

transport (Table 8-2). Distance travelled by buses was increased proportionally to the

increase in bus trips in the travel targets scenario.

To estimate the population-level effect we calculated the difference in exposure as the

difference between PM2.5 concentrations for the baseline and travel targets scenarios.

Changes in individual exposure to PM2.5 were estimated by considering mode-specific

concentrations and respiratory ventilation rates. Mode-specific exposure compared to

background exposure and ventilation rates were those used in a recent study by Tainio

et al. (305) (cycle=2 and walk=1.1). Cycling can also take place in designated paths for

pedestrian and bicyclist traffic. In Brisbane it is permitted to cycle on sidewalks (333).

We tested the sensitivity of our results to cycling having the same mode specific

exposure to PM2.5 as walking. Weekly inhaled PM2.5 dose was estimated by multiplying

138

the time during a week spent sleeping, other activities and in a passenger car (baseline)

or walking/cycling/public transport for the travel targets scenario (305). It was assumed

that while not in a passenger car or walking/cycling/public transport, people are exposed

to the background levels of PM2.5.

Road trauma

We used crash data collected by the police and available in the Road Crash database,

maintained by Department of Transport and Main Roads to assess road trauma (328).

We summarised the number of fatalities and road injuries (includes hospitalisation and

medically treated causalities) by victim and striking mode. Our figures are for 2009 to

match travel data. We estimated baseline rate (𝑅0) of fatalities and injuries per

kilometres travelled by victim and striking modes based on methods developed in past

research (195) (Equation S6 supplementary material). We estimated the number of

fatalities and injuries under the travel targets scenario based on the baseline rate and

kilometres travelled per mode involved in a crash (Equation S7 supplementary material).

To reflect the declining risk of injuries with increasing traffic volume (commonly referred

as “safety-in-numbers effect”) (84), we took the square root of kilometres travelled by

victims and striking modes (334). We tested the sensitivity of our results to the

assumption of a linear association between road causalities and traffic volume.

We incorporated the health impact of road fatalities in the MSLT by multiplying baseline

mode specific mortality rates (pedestrian, bicyclist, passenger car occupant and

motorcyclists) by a factor reflecting the change in fatalities in the travel targets scenario

(road fatalities by mode travel targets scenario/Road fatalities by mode baseline). The

same approach was used for injuries; however, the impact was evaluated on mode-

specific years lived with disability.

The supplementary material (Section 2.3) provides further information on the

calculations and details on the data sources used for calculating the health effects of

exposure to PA, PM2.5 and road trauma. Table S14 in the supplementary material

presents a summary of sensitivity scenarios.

139

Results

To achieve the travel targets, trips by car occupants need to decrease for all distance

categories except for trips >17 km by the same percentage points as increases in active

transport modes (Table 8-4).

Table 8-4 Percentage of trips made by distance travelled and transport mode, for baseline and travel target scenarios

Mode

<2km 2-5km 6-16km 17km+

Baseline Target Baseline Target Baseline Target Baseline Target

Car occupant 65% 40% 90% 73% 87% 69% 84% 84%

Walking 34% 59% 4% 4% 0% 0% 0% 0%

Bicycle 1% 1% 1% 18% 1% 1% 0% 0%

Public Transport 0% 0% 5% 5% 12% 30% 16% 16%

We assumed that achieving the travel targets would require an increase in average

weekly walking and cycling across all age and sex categories. Table 8-5 depicts the

change in weekly trips from private car travel to active transport by age and sex. Table

8-6 presents the weekly increase in minutes walked and cycled by age and sex. These

estimates account for baseline travel patterns by age and sex (Table 8-2).

Table 8-5 Mean trips per week (weekdays only) for baseline and travel targets scenarios, by age and sex

Car occupant Walking Bicycle Public Transport Sum

Age (years) and sex Baseline Target Baseline Target Baseline Target Baseline Target

17-49, male 10.34 7.13 1.12 1.91 0.31 1.58 1.40 2.55 13.17

17-49, female 14.01 11.19 1.64 2.88 0.09 0.48 1.44 2.62 17.18

50-74, male 12.57 10.31 1.23 2.19 0.16 0.93 0.63 1.16 14.59

50-74, female 11.89 10.05 1.33 2.32 0.02 0.12 0.91 1.67 14.15

75 plus, male 9.05 8.04 0.71 1.27 - - 0.45 0.90 10.21

75 plus, female 5.83 4.71 0.64 1.13 - - 0.76 1.39 7.23

140

Table 8-6 Additional mean minutes per week of transport physical activity undertaken in the travel targets scenario compared to the baseline scenario (statu-quo), by age and sex

Additional mean minutes per week of physical activity

Age (years) and

sex

Walk for

transport

Bicycle for

transport Public Transport (+ walk)a

17-49, male 13 16 16

17-49, female 21 4 16

50-74, male 15 10 7

50-74, female 16 2 10

75+, male 8 0 5

75+, female 8 0 8 a Estimated as the additional public transport trips in the travel targets scenario multiplied by the mean minutes walking in a public transport trip estimated from the household travel survey.

Road fatalities and injuries per 100 million kilometres travelled decreased for all victim

modes affected by the travel targets scenarios, except for car occupants which remains

nearly the same as at baseline (Table 8-7). Assuming a linear association for kilometres

travelled and road trauma (that is, removing the ‘safety in numbers’ effect) translated

into a less significant reduction of rates (sensitivity scenario).

Table 8-7 Road trauma rates per 100 million kilometres travelled by transport mode

Baseline Base casea Sensitivity scenariob

Fatalities Injuries Fatalities Injuries Fatalities Injuries

Pedestrian 1.43 125.59 1.06 90.67 1.44 119.95

Cyclist 2.24 165.04 1.46 106.21 2.18 156.48

Car occupant 0.09 23.67 0.10 24.37 0.09 23.12

Motorcyclist 3.91 153.38 3.84 149.23 3.77 145.28

a Non-linear association kilometres travelled and road trauma. b. Linear association kilometres travelled and road

trauma

Background PM2.5 decreases marginally in the travel targets scenario, with variations

depending on the attribution of PM2.5 to motor vehicles and passenger cars (Table 8-8).

141

Table 8-8 PM2.5 values baseline and sensitivity scenarios

Baseline Base casea

Sensitivity scenarios

Low level

apportionment MVb High level

apportionment MVc

Passenger cars 65% MV

emissionsd

PM2.5 (µm/m3) 6.964 6.957 6.962 6.940 6.920 Change emissions from passenger cars (%) -0.41% -0.15% -0.66% -0.94% Change emissions from buses (%) 0.31% 0.12% 0.31% 0.31% Total effect (%) -0.10% -0.04% -0.35% -0.63% a 17% of PM2.5 attributable to motor vehicles (MV). Of MV emission, 28% corresponds to passenger cars and 10% to buses. b 7% of PM2.5 attributable to MV. Of MV emission, 28% corresponds to passenger cars and 10% to buses. c 30% of PM2.5 attributable to MV. Of MV emission, 28% corresponds to passenger cars and 10% to buses. d 17% of PM2.5 attributable to MV. Of MV emission, 65% corresponds to passenger cars and 10% to buses.

Health and healthcare cost outcomes

Over the life course of the Brisbane adult population in 2013 (860,000 persons), slightly

over 32,000 HALYs and 28,000 life years could be gained if the proposed travel targets

were achieved by 2026 (Table 8-9). We estimated that significant savings ($AU312

million) in healthcare costs could be accrued in the travel targets scenario. However, an

increase in the number of life years lived would translate into additional healthcare costs

of $AU129 million. Most of the health gains would result from improvements in

population levels of PA, with exposure to PM2.5 and road trauma having small negative

impacts (Figure 8-2). The number of prevalent cases decreased for all modelled

diseases except for respiratory diseases (Figure 8-3 and Table 8-10). We estimated a

reduction in mortality from ischemic heart disease, colon cancer, breast cancer and type

2 diabetes. The uncertainty interval for the reduction for ischemic stroke mortality

includes 0. This can be explained by the weak association between PA and ischemic

stroke and the effect of additional cases from added life years and exposure to PM2.5.

Our results suggest that road trauma may lead to a 30% (95% UI 29% to 32%) increase

in mortality (593 deaths) and a 6.6% (95% UI 6.2% to 7.0%) increase in the number of

years lived with disability (3,339) over the life course of the Brisbane adult population.

The results are most sensitive to the discount factor applied to health outcomes and the

dose-response relationship between road trauma and kilometres travelled by mode. A

higher discount rate for healthcare costs and health outcomes implies a lower present

142

value. Assuming a linear association for road trauma and kilometres travelled translates

into greater negative effects compared to the base case. Supplementary material

section 4 has more information from sensitivity analysis results, and from intermediate

outcomes.

Table 8-9 Healthcare costs and health outcomes for base case by sex over the life course of the Brisbane adult population (95% uncertainty interval)

Health-adjusted life years (thousand)

Life years (thousand)

Healthcare costs total (millions)a

All other healthcare costs

in added LYs total (millions)

Total 32.6 (19.6 to 46.8)

28.1 (13.1 to 44.0)

-$312 (-$463 to -$173)

$129 ($49 to $213)

Females 17.6 (9.2 to 26.3)

16.2 (6.2 to 26.8)

-$139 (-$221 to -$63)

$80 ($22 to $141)

Males 15.0 (9.8 to 20.9)

11.9 (6.8 to 17.8)

-$173 (-$246 to -$107)

$49 ($26 to $76)

a Negative values are savings

Figure 8-2 HALYs by risk factor over the life course of the Brisbane adult population (95% uncertainty interval)

143

Figure 8-3 Change % on disease prevalence and mortality over the life course of the Brisbane adult population (error bars indicate the 95% uncertainty interval)

Table 8-10 Change in prevalent cases and mortality over the life course of the Brisbane adult population (95% uncertainty interval)

Disease Prevalent cases Mortality

Ischemic heart disease -44,902 (-61,765 to -28,463) -1,416 (-2,275 to -624)

Ischemic stroke -14,343 (-30,420 to 182) -1,504 (-4,558 to 1,342)

Colon cancer -19,630 (-26,409 to -13,091) -406 (-552 to -265)

Breast cancer (women) -13,184 (-18,815 to -7,763) -158 (-228 to -091)

Type 2 diabetes -90,440 (-130,002 to -51,905)

-325 (-474 to -169)


7,831 (3,881 to 12,026) 130 (049 to 217)

Tracheal, bronchus and lung cancer

356 (192 to 531) 81 (42 to 122)

Discussion

Significant health gains could be made if government targets for reductions in private

car travel and increases in active transport were achieved in Brisbane, Australia. In this

study we estimated the effect of achieving active transport targets by 2026 (5% cycling,

15% walking and 14% public transport). Achieving the travel targets implies a significant

increase in active transport, which translates in substantial improvements in population

144

PA levels, and gains in terms of health and costs. Health benefits from increases in PA

are significantly higher than the potential negative effects of increases in air pollution

and road trauma exposure. Results from a sensitivity analysis indicated that the greatest

negative variation from the base case is for the discount rate applied to health and costs

outcomes and assuming a linear association (nonlinear for base case) between road

trauma and kilometres travelled by modes involved in a crash.

These results are consistent with those from the limited number of previous studies,

which were conducted in the United States, Spain, England and Australia (195, 334-

336). Similar to our findings, the findings from those studies indicate that the greatest

contributions to health gains from replacing passenger cars kilometres travelled to active

transport were due to improvements in PA. However, it is difficult to make a direct

comparison among studies because of differences in the scenarios analysed, methods

applied and the high context specificity of studies. For example, Maizlish and colleagues

(334) modelled the potential health impact of replacing short car trips with walking and

cycling and introducing driving low-emission cars in the San Francisco Bay Area in

California. The study findings indicate that over 5,000 disability-adjusted life years per

million people annually could be averted by replacing motorised travel with active

modes and the introduction of low–emission driving. To our knowledge, there is only

one previous Australian study that assessed health outcomes of physical activity, air

pollution and road trauma of a shift towards sustainable alternative transport (335).

Projecting travel patterns in 2030 for the City of Adelaide, the authors found that

replacing 40% of motor vehicle kilometres travelled with cycling (10%) and public

transport (30%) would translate into over 7,500 averted DALYs annually for a projected

population of 1.4 million people (1,509 per million people). We calculated annual values

and we found gains of 1,700 HALYs per million people. While in Xia et al’s study 10% of

the car trips were replaced by cycling, in our study, this was only 3%. In addition, as we

explain below, in the Adelaide study a comparative risk assessment (CRA) approach

was used, which tends to overestimate the change in the burden of disease.

Strengths and limitations

To our knowledge, this is the first study to quantify the potential health and healthcare

cost outcomes of a shift towards active transport using the well-established method of

the proportional multi-state life table Markov model (MSLT). The MSLT allows for the

145

long-term estimation of health and economic outcomes (337). By using this approach,

we were able to incorporate a gradual shift from private car trips to active modes. Also,

the interaction between multiple diseases is better accounted for, with proportions of the

population being able to be in more than one disease state (337).This avoids over-

estimating outcomes as a result of summing health outcomes attributable to each

disease individually. Another source of overestimation in past studies arises from the

use of a comparative risk assessment (CRA) approach based on Global Burden of

Disease (GBD) estimates (19). GBD studies estimate their DALYs as the sum of years

of life lost (YLL) and years lost due to disability (YLD) (193). The mortality rates used in

GBD studies to estimate YLL are for a hypothetical population that has the lowest

observed mortality at every age (193), whereas we used the mortality rates for the

population in question for our life years lived component. Furthermore, in GBD studies,

YLLs are not adjusted for disability; hence, their use in estimating intervention effects

results in over-estimation, which our life table approach avoids. Another way of seeing

this is that estimated changes in morbidity using CRA methods do not allow for how

implicit increases in life expectancy impact on morbidity (301). While the changes in

deaths and prevalence using the MSLT are in some ways more accurate than those

from a CRA approach it should be noted that that the average age of death and incident

disease will change and thus the disease burden will be on average be shifted later in

life. Thus changes in HALYs offer a more appropriate measure of gain than changes in

life course prevalence. Past studies using the CRA approach also found that health

gains were about twice as large when predicting the impact of PA on all-cause mortality

compared to on disease specific mortality (195).

Limitations should be highlighted. Our estimates of shifts from private cars to alternative

travel modes in the intervention scenario are compared to a scenario in which travel

patterns by age and sex remain constant over time, and we assume that each group will

become more active in proportion to its current activity level. More sophisticated

methods based on propensity analysis accounting for trip distance and hilliness as

previously done for cycling in England could further refine our active travel estimations

(338). At baseline, the most cycling trips were made by young males (61% of trips), and

hence, this group benefits most from achieving the travel targets. This assumption can

be challenged as where cycling is more common it is also more equitable (339), and

providing safe and direct cycling routes (340) may encourage relatively more women

146

and older people to commute by bicycle. There is also a risk of bias in our baseline

travel estimates as the data collection periods excluded summer months, and people

may be less likely to use active commuting due to the heat. In addition, we did not take

into account that the hot weather characteristic of Brisbane summer months may keep

people from using active transport. While we used the best available road trauma data

for Brisbane, our use of an overall road trauma risk is a limitation that should be

highlighted. Past studies indicate differential road trauma risk by age and sex (341) and

also by road type (334). Because we only used data from Brisbane the number of

events in some categories was small and may be a result of chance rather than

accurately reflecting risk. Also under-reporting of injuries is not uncommon in police data

(342, 343). Another limitation arises from using self-reported data for physical activity

prevalence and travel patterns. Our application of the MSLT assumes that a proportion

of the population that is sufficiently active (≥1,600 MET-minutes)/wk. and ≥1h/wk. of

vigorous PA) receives no benefit from additional PA, which may lead to underestimation

of health impacts. Another source of underestimation arises from an incomplete

inclusion of diseases. There is growing evidence suggesting a causal association

between PA, lung cancer, endometrial cancer and dementia (54, 196). A limitation of

this and similar studies (175) results from not knowing the exact shape of the

association PA-health. Recent studies assessing the association of PA with health and

using continuous exposure indicate greatest benefits at low levels of PA (48, 344). The

level of ambient air pollution in Brisbane is one of the lowest observed in the world

(345), and relative risks for PM2.5 are based on data from more polluted locations,

resulting in uncertainty at the lowest part of the exposure-response relationship (88). In

addition, we did not include improvements in light vehicle fuel efficiency (346). Past

research assessing increases in cycling showed that improvement in the light vehicle

fleet alone would improve population health (238), hence our results for air pollution are

likely to be an overestimation. We incorporated the effect of trends in incidence and

case fatality in our model and assumed that all other model parameters would remain

constant. Variations in these parameters may influence results either upward or

downward. Migration and natural population growth are not considered, though the life

table approach means that the effect of population aging is included. The strong

population growth that is expected for Brisbane would act to increase the health gains

forecast in this paper.

147

Conclusion

Shifting towards active transport is a long-term process that needs investment and

continuity in governments’ strategic planning. In Queensland an updated regional plan

was released early in 2017 however, no specific transport targets were set (347). The

current research indicates that continuing working towards achieving the proposed

mode shares of 5% cycling, 15% walking and 14% public transport would deliver

considerable health and economic gains. From a societal perspective, all would benefit

from improved quality of life and savings in healthcare costs.

Even though our results are highly context specific, they support the international

evidence of the health benefits of investing in active transport.

148

Chapter 9 General discussion

9.1 Introduction

In Australia, nearly 30% of the burden from chronic disease is attributable to five major

modifiable risk factors: tobacco use, high body mass index, alcohol use, physical

inactivity, and high blood pressure (58, 348). Reducing the prevalence of preventable

risk factors would lessen the health and economic burden of years lived with disability

and premature mortality.

The BE can contribute to a healthier population by supporting active lives. The literature

review in Chapter 2 showed that using ecological approaches that target the BE are

feasible ways to facilitating active living. For example, people living in places with well-

connected streets, local destinations, high population density, and a high density of

parks are more likely to do PA compared with those living in areas lacking these

attributes. In addition, evaluations of BE interventions show that health gains from

improvements in PA outweigh the potential negative impacts from greater exposure to

road trauma and air pollution (37). In sum, there is convincing evidence that the BE

facilitates active lifestyles, and thus contributes to a healthier population. However, in

practice, health is at best indirectly considered in decisions about urban infrastructure

developments and in transport evaluations it is mostly considered only in relation to road

trauma. Consequently, investment decisions about BE initiatives are based on

incomplete appraisals of the potential effect of the sector on public health and the

economy.

In Australia, the BE has experienced a number of changes over time, mostly to

accommodate a growing urban population. In the second half of the 20th century,

Australian cities experienced a great expansion to the suburban fringes, encouraged by

affordable large blocks of land and cheap private motorised transport (349). The

resulting urban sprawl presents environmental, social, transport and economic

challenges (349, 350). The negative consequences of sprawling have been recognised

by Australian Governments. For example, programs from the Commonwealth

Government promote inner city densification (e.g. Green Street and Building Better

149

Cities) (Hall (2010) cited in 349). However, density should be supported by urban design

that contribute to the wellbeing of populations (163). Greater density has now been

observed in Australian cities (349). Still, Australian cities face great challenges to house

their growing population and support healthy BEs for all. The population of major cities

(Sydney, Melbourne, Brisbane, Perth and Adelaide) is estimated to grow from 16 million

in 2017 to 27 million in 2053 (351). A growing urban population represents important

infrastructure and housing challenges, but also opportunities to shape cities where

people can be active and healthy.

The overall aim of this thesis was to contribute to the evidence of potential health and

economic benefits of a more active adult population achieved by BE interventions and

policies in the Australian context. A set of five research questions were addressed to

accomplish this aim, each fully developed in a previous chapter (Chapter 4 to 8).

The remainder of this chapter is structured as follows: summary of findings (Section

9.2); implications (Section 9.3); and directions for future research (Section 9.4).

9.2 Summary of findings

What are the attributes of the built environment in Australia that most benefit

physical activity?

A systematic review of Australian contemporary studies addressing the association of

BE attributes with PA outcomes was carried out to address this research question.

Judging by studies that used objective measures of the BE, walkable neighbourhoods

with a wide range of local destinations to go to, as well as a diverse use of land, facilitate

PA among their adult residents. The summarised evidence supports the ability of BE

attributes to encourage PA for transport purposes and, to a lesser extent, total PA

(transport and recreation). Likewise with the evidence from multi-country reviews (see

Section 2.4 literature review), the current Australian evidence is less supportive of the

association between the BE and recreational PA. For self-reported measures of the BE,

convincing evidence of an association with PA for availability and proximity to

destinations and measures of design (e.g. street connectivity, aesthetics and green

spaces) was found. Most of the evidence for self-reported measures of the BE are for

PA for recreational purposes.

150

The majority of studies in Chapter 4 are cross-sectional in design, and as highlighted in

Section 2.4 (literature review), quasi-experiments and natural experiments that control

for confounders are recommended to establish causality. However, evidence from

cross-sectional studies should not be dismissed. Over the years, studies with cross-

sectional designs have been showing consistent findings about what matters for

neighbourhood-based PA.

What are the physical activity-related health externalities and healthcare costs


To address this research question, findings from selected studies from Chapter 4 were

expressed as average minutes of PA per week across the neighbourhood population.

Then, the ACE Prevention approach was used to translate changes in PA into HALYs

and healthcare costs. Pooling results—as per the broader categories presented in

Chapter 4 (density, diversity of land use, availability of destinations, distance to transit,

design and neighbourhood walkability)—was not possible due to the great variability in

statistical methods and definitions of exposure and outcome variables.

Of the 28 modelled scenarios of changes in exposure to BE attributes, 20 indicated

potential health benefits represented by HALYs gained per 100,000 people exposed per

year. Savings in healthcare costs per year for PA-related diseases ranging from

A$2,800 to A$99,600 (2010) per 100,000 adults exposed were estimated. All other

healthcare costs in added life years were approximately 50% higher than the savings

obtained by having to treat fewer cases of PA-related disease in earlier years. The

greatest estimated gains are from improvements in availability of transport and

recreational destinations. Estimates were the most sensitive to the assumption made for

the dose-response association of PA with positive health outcomes. Assuming a non-

linear dose-response as opposed to a linear assumption (main scenario) translated into

a four-fold increase in health benefits. Using estimates from the literature did not allow

for a quantification of the effects of increasing density in combination with other features

of the BE. However, grater health gains would be expected from a larger number of

people exposed to BEs that support PA.

What economic evaluation methods have been used to model future health

outcomes from interventions in active transport?

151

A systematic review of economic evaluations that included PA outcomes was conducted

to address this research question. Most of the included studies (32/36) used cost-benefit

analyses, and a wide range of benefit–cost ratios were reported (-31:91 to 59:1). This

range includes only two interventions for which the cost exceeded the benefits.

However, it is important to highlight that most of the ratios were reported in the grey

literature and refer to hypothetical interventions. In addition, a limitation of the reporting

approach in Table 6-2 is that the lowest ratio per study was reported, instead of the

base case per assessed intervention. Given the wide diversity of interventions evaluated

and methods applied for the quantification of outcomes, results were not pooled. There

is an important question in terms of generalisability of results to other settings.

Interventions within the transport sector are highly context specific. For example,

demographic characteristics (e.g. age structure, income level, education) and levels of

PA and obesity are likely to modify the effect of an intervention on health outcomes. For

PA, research indicated greater benefits for those who do little PA, hence, the impact of

an intervention would differ greatly depending on population levels of PA (48). For

instance, older people tend to be less active than younger ones, therefore, for an equal

increase in PA, the former will benefit more. Similarly with overweight people, evidence

shows that exercise leads to weight loss and is independently related to cardiovascular

disease (4). Hence heavy people would benefit more from a change in PA due to the

positive impacts on health of weight loss and exercise compared to those with a healthy

weight. In addition, for Australia, disadvantaged population groups, characterised by low

income and education, report lower PA levels compared to more advantaged groups.

Hence, promoting PA among the least advantaged groups compared to the most

advantage ones would accrue greater health gains overall for equal intervention

effectiveness (352). In addition, past research showed that health impacts differ across

cities depending on their level of motorisation and air pollution levels (19).

Results were reported following the Consolidated Health Economics Evaluation

Reporting Standards (CHEERS) (222). In general, better reporting was observed in the

peer-reviewed literature compared with the grey literature; this was expected since the

latter is not subject to a formal peer-reviewing process. A significant proportion of the

studies produced in the grey literature originated to inform policymakers on the multiple

benefits of investing in active-transport initiatives. Henceforth, transparency in methods,

152

data sources, as well as complete reporting should be aimed for to ensure that high-

standard evidence is informing decisions within the transport sector.

Can the health impact of changes in physical activity be incorporated more

robustly in cost-benefit analysis of built-environment initiatives in Australia?

This research question builds up on the findings presented in Chapters 4 and 5 and

describes a method that could serve to incorporate PA-related health benefits in cost-

benefit analysis of BE interventions. As an example, monetised values per adult living in

a neighbourhood where the BE changes were estimated for five of the six Ds by Ewing

and Cervero (134) (density, diversity of land use, availability of destinations, distance to

transit, design) plus aggregated neighbourhood measures. Compared to estimates

available from the literature expressed as monetised values per kilometre walked or

cycled, the proposed method here is more flexible and based on a well-established

methodology. The estimated values per Ds and measures of neighbourhood walkability

can be applied by urban planners to who are interested on wider inclusion of outcomes

in the evaluation of projects.

What are the potential health and economic impacts of Brisbane meeting its

targets for active travel?

The potential health and healthcare cost outcomes of Brisbane achieving government

active-travel targets by 2026 (walking 15%, cycling 5% and using public transport 14%)

were quantified using the ACE Prevention approach. An updated version that

incorporated exposure to road trauma and air pollution to exposure to PA was used. In

Brisbane, the greatest majority of overall trips are made in private cars, and the share of

active travel is very small (public transport 8%, bicycle 1%, and walking 9%). Hence,

achieving the travel targets implies significant increases in walking, cycling and public

transport trips (65%, 390%, and 73%). These increases were estimated to accrue

33,000 HALYs and net savings in healthcare costs of A$192 million (2013) over the life-

course of the adult Brisbane population. Results were the most sensitive the discounting

health and healthcare costs.

9.3 Implications

What can be concluded?

153

In Australia, walkable neighbourhoods with a range of local destinations support walking

for transport and walking for all purposes (transport and recreation). The use of

heterogeneous statistical approaches and measures of exposure and outcome in

studies assessing BE-PA associations means that results cannot be pooled. Moving

towards standardised statistical methods and from categorical to continuous exposure

measures would enable pooling results and ensure that information on the exposure

variables is not missed (166). To draw conclusion on strength of effects, pooled results

are preferred to individual study effects—given the greater statistical power resulting

from a larger sample (167). In addition, studies should provide detailed quantitative

information on the changes in the exposure variable. For example, categorisation of the

exposure variable into quantiles or translation into z-scores without giving information

about quantile values or variable distribution (mean and standard deviation) is of little

use. Researchers in the field and policymakers using the evidence would greatly benefit

from clarity as to what exactly needs to change in the exposure variables, rather than

only knowing that more or less of something is better for PA.

Findings from Chapter 5 and 7 suggest that the BE can greatly contribute to population

health by facilitating PA. However, the evidence provided in this research relies on

individual studies, since pooling results was not feasible for the reasons given above.

The estimates for population health and the economic benefit of BEs that facilitate active

living are a starting point for the systematic inclusion of PA health in the appraisal of BE

interventions. Greater refinements of effectiveness measures could be gained from

collaborative work with planners to learn about meaningful alternatives for exposures to

the BE. In addition, researchers should aim for greater precision in measures of

effectiveness by working collaboratively with other researchers investigating the BE-PA

association. Preferably, measures-of-effect would indicate at least the average

proportion of people who would take up PA due to changes in the BE as well as a

measure of the changes. Such complete information was only provided in the study by

McCormack et al (2012) (140). Assumptions were required for the rest of modelled

scenarios in Chapter 5 and 7.

In the transport sector, per-kilometre economic values for walking and cycling are

preferred when including PA-related health benefits in cost-benefit analysis. Estimates

from this thesis are in the lower range of values when compared with values from the

154

literature. Since the BE also incorporates the land use sector, this thesis provides a

method and a range of results for interventions in the realm of urban planning. The

advantage of the approach and values presented in Chapter 7 is the possibility of

directly linking features of the BE (e.g. density, land use mix, destination) to economic

outcomes. This removes the need to estimate the number of kilometres walked or

cycled that result from such interventions, which would be required if only estimates of

the economic value per kilometre of active travel is available. In addition, the range of

estimated values could be used as ballpark estimate of the potential of alternative urban

form design for health and the economy. Further work that uses a consistent set of

measurement and analysis methods on a range of data representative of the Australian

context could deliver more robust values that could function as standard, agreed-upon

values for use in CBAs.

Lastly, findings from this thesis support the case for investment in infrastructure and soft

measures (e.g. work based active travel to work programs) to achieve government

targets aiming at a high share of active transport. The health gain from increased levels

of PA is considerably greater than the potential negative health outcomes from greater

exposure to air pollution and road trauma. Estimates for the effect on PA and air

pollution are more reliable than those for road trauma, since a simple calculation

approach was applied without accounting for age and sex differentials or road types.

What are the main strengths, limitations and validity of results?

Strengths

The research presented in this thesis systematically and quantitatively address the

potential of the BE for health in Australia. By applying methods to quantify health

outcomes from changes in health risk factors, it was possible to provide an overview of

the likely health outcomes of neighbourhood BEs that facilitate PA and policies that

promote active travel. The ACE Prevention approach, which consists of a combination

of the proportional MSLT with the PIF, served to quantify health outcomes and

healthcare costs. Compared with other modelling approaches, ACE Prevention presents

numerous advantages. First, real-life populations were modelled in Chapter 5, 7 and 8,

based on the Australian population and mortality data by single year and sex, and

epidemiological data from burden-of-disease studies. Other approaches, such as the

155

CRA methods based on the Global Burden of Disease framework and the Health

Economic Assessment Tool (HEAT) are not based on real-life populations (103, 353).

Another advantage of ACE Prevention is that it accounts for the time lag between

changes in exposure to a risk factor and health outcomes based on transition

probabilities, rather than assumptions as in other models (354). In addition, with ACE

Prevention the time lag between an intervention and uptake of PA is explicitly modelled.

This allowed for modelling the gradual achievement of the proposed travel targets in

Chapter 8, which is a realistic assumption, as large numbers of people are unlikely to

shift modes from one day to another. The CRA approach based on the Global Burden of

Disease framework is likely to overestimate outcomes, mainly due to the approach for

calculation of the years of life (YLL) component of DALYs based on the highest

attainable life expectancy. In addition, estimates using the CRA and HEAT approach are

static, while ACE-prevention is dynamic. A dynamic approach allows for the inclusion of

inputs trends. In this thesis, future trends were included for disease incidence and case

fatality; however, population growth trends and risk factor trends were not accounted for.

Limitations

Important limitations related to healthcare cost estimates, burden-of-disease data, and

dose-response association of physical inactivity and health outcomes should be

highlighted and addressed in future studies.

Healthcare costs

Considerable caution should be taken in the interpretation of healthcare cost estimates

in this thesis. The macro approach applied here gets the overall costing of diseases

correct; however, it differs greatly according to the estimates of incident and prevalent

cases used to estimate per case costs from total costs of a disease (see section 3.2.2).

Different estimates of prevalent and incident cases used to estimate healthcare costs

per case resulted in important discrepancies. Findings from Chapter 6 indicate greater

healthcare costs in added life years than savings in healthcare costs from physical

inactivity diseases, and findings from Chapter 8 indicated the opposite. In addition, this

approach results in some inconsistencies, such as the lack of a clear trend in costs

(Table 3-2).

156

Better estimates should be aimed for in the future. A starting point would be to have up-

to-date disease cost by age and sex. However, further refinements should be aimed for.

For example, a study demonstrates that healthcare costs of diseases are considerably

higher for decedents compared with survivors (355). These findings indicate that costs

in the last phase of life are higher; hence, an improvement would be to link healthcare

costs to mortality, in addition to prevalence or incidence. Also, micro simulation

approaches linking costs for each disease stage would provide the optimal level of

refinement. However, data requirements would be significant. In addition, methods for

their estimation of all other healthcare costs are improving significantly which would

greatly contribute to the refinement of estimates in future similar studies to the ones in

this thesis. For example, van Baal et al (113) developed a methodology that takes into

account age, sex and time to death, by diseases. This is a significant improvement

compared to the macro estimates in this thesis based on overall all other healthcare

costs by age and sex.

Burden of disease data

New data releases during the development of this thesis from the GBD group and

Australian Burden of Disease (BoD) studies mean that results from prediction studies in

the thesis are not based on the latest available epidemiological data. Results in Chapter

5 and 7 are based on GBD 2010 epidemiological estimates for Australia (356). Results

for Chapter 8 are based on GBD estimates for Australia for 2013 (188) and AIHW data

for 2012 and 2013 (320). The supplementary material for Chapter 8 provides a detailed

explanation of the choice of data sources. At the time, the chosen sources were the best

available. Not having used the latest epidemiological estimates for Australia may be

seen as a limitation. However, updating all models with the latest available data was

outside the scope for timely completion of this thesis.

Dose response physical activity and health outcomes

Another important limitation is in regards to the dose-response associations for PA with

health outcomes. In the original ACE-prevention PA study (94) and in the baseline

scenario for the study in Chapter 5, it was assumed that PA is linearly associated with

positive health outcomes. This is a rather conservative assumption, and there is now

convincing evidence indicating that PA accrues the greatest health gains for the least

157

active (48, 344). Hence, in Chapter 8, results are based on a non-linear dose-response.

Greater certainty of the dose-response for PA and health would improve the predictive

validity of this and similar studies. In a recent study by the GBD group in charge of

modelling the health burden attributable to PA, continuous dose-responses were

presented in a graphical format (48). Future similar studies could be improved by

collaborating with groups (e.g. GBD PA group) to assist with the definition of continuous

dose-response functions for PA with health outcomes.

From evidence to policy

There is now a large body of evidence that supports that the BE can contribute to

population wellbeing. However, health considerations are not included in the planning

and transport sector decision-making process on routine basis (357, 358). Translating

the evidence in this research into practice is a complex process. Public health concerns

are traditionally managed by health departments and organisations. Yet, population

health is determined by multiple sectors (e.g. transport, planning, and education). Past

research suggested that public health researchers need to improve their understanding

of policymaking to be able to generate health-enhancing change in non-health sectors

(359). Also, experience from the United Kingdom suggests that there is a lack of

connection between guidance for health consideration in transport and planning and the

existing policy or legislation (360). However, guidance alone may not be sufficient and

health considerations should be legislated in the planning and transport sectors. This is

not easily achieved, since there are multiple legislations that transport and urban

planners need to follow (360).

In Australia, there is a growing interest in designing healthy and sustainable BE. This is

mostly observed at the local government level, with numerous examples of councils’

initiatives that support healthy behaviours, including PA (361, 362). However, the

legislation structure in the planning and transport sectors in Australia may be hindering

the opportunities for local government to bring about positive change. Australian States

and Territories legislate the objectives of land use planning policies. The

operationalisation is left with local governments (363, 364). Preventive health is not a

legislated objective for the urban planning sectors across all states (358). Therefore,

local governments may adjust their instruments to support healthy BE, subject to these

being aligned with State and Territories’ overarching legislation (363, 365). In the

158

transport sector, the States and local governments share responsibility for roads and

public transport (366-368). Such legislative structures demonstrate the power of State

and Territory BE policy to positively or negatively impact of population wellbeing (364).

The example of Western Australia’s Liveable Neighbourhood guidelines for infill and

greenfield developments shows how State planning policy can support preventive health

(29). An evaluation of the Liveable Neighbourhood guidelines demonstrated that for

every 10% increment in the implementation of the policy, people’s likelihood of walking

for transport increase by 10% (369).

Recent policy-focused research of public policy activity in land use planning for

Australia’s most populous State (New South Wales) shed some light on the complexities

and potential avenues for influencing public policy (357, 358). This research showed

that the lobbying of advocacy groups was key for gaining ministerial support for health

consideration in the planning sector. The researchers highlighted that there is need of

investment in long term advocacy efforts across the health and planning departments as

well as the private planning sector to be ready to submit submissions for Government

Planning Bills (364). In addition, the researchers see as key the active contribution of

health departments in framing population health problems (i.e. physical inactivity,

obesity) as urban planning issues (364, 370).

In Australia, advocacy groups such as the National Heart Foundation and the Obesity

Policy Coalition are active advocates for BEs that support healthy habits (371, 372). For

example, the Obesity Policy Coalition made multiple submissions in Victoria related to

the urban planning environment (http://www.opc.org.au/action-areas/planning-

environments.aspx). The National Heart Foundation funds a comprehensive resource to

guide BE practitioners on the best designs for health and wellbeing

(https://www.heartfoundation.org.au/active-living/healthy-active-communities). In

addition, the National Heart Foundation is an active advocate for healthy planning and

transport policy. These organisations based their submission largely on evidence from

the peer reviewed literature. State health departments have also demonstrated an

active role in influencing planning for a broader inclusion of health outcomes in the

appraisal of BE projects. For example, the Department of Health of New South Wales

strategic plan includes actions targeting the planning and transport sectors to create BE

that are supportive of PA (119).

http://www.opc.org.au/action-areas/planning-environments.aspx

http://www.opc.org.au/action-areas/planning-environments.aspx

159

Australian and international evidence supports that the BE can act as an enabler or

inhibitor of PA (Section 2.4 and Chapter 4). However, Australian context specific

evidence of the potential of the BE for population health and the economy is lacking. A

component of this thesis aims to provide evidence of a method and values for the

inclusion of PA-related health in economic evaluations of the BE. While a direct

inclusion of the values generated in this research may not happen, such information will

strengthen the case for the consideration of health implications when considering

changes in the BE. In addition, evidence of the likely health outcomes of increased

active transport may aid advocates and health departments in their BE related

submissions. Information in this thesis, and the tools it describes, may be useful for

advocacy organisations and State and Territories health departments to lobby for

planning legislation that is conducive to healthy lifestyles.

Validity

The focus of this thesis is on quantitative health predictions of the likely health outcomes

of BE initiatives (Chapter 5, 7 and 8). The validity of results of this thesis can be judged

based on plausibility, formal validity, and predictive validity (373).

Plausibility

Plausibility refers to whether the framework underlying the health prediction models are

understandable, applicable, and plausible (373). The underlying framework for this

thesis is depicted in Figure 2-2 (literature review). In summary, BE initiatives (change in

BE attributes and policies) are associated with exposure to risk factors (PA, air pollution,

and road trauma), which in turn are related to health outcomes. Therefore, in this thesis,

first, a causal association needs to be established for the BE with PA, air pollution, and

road trauma. And second, for PA, air pollution, and road trauma with health outcomes.

In Chapter 5, evidence from the literature for the association BE-PA was used, and the

underlying assumption is that changes in PA follow changes in the BE. In Chapter 8, a

hypothetical scenario of achieving travel targets was explored. In terms of the

relationship of risk factors with health outcomes, the greatest confidence can be placed

on the health benefits of PA (Section 2.2 literature review). PM2.5 was used as a proxy

for transport air pollution (Chapter 8), and there is compelling supporting evidence

indicating that high exposure to PM2.5 is detrimental for health. However, there are still

160

important uncertainties given that the greatest majority of studies assessing the

relationship of PM2.5 with health were based on settings with mean exposure levels that

are higher than those observed in Brisbane. Lastly, the association of active travel with

incidence of road trauma is the least certain (Chapter 8). Based on past research, the

risk of road trauma per active traveller was estimated to decrease when shifting from car

trips to walking, cycling, and using public transport (‘safety in numbers’). Yet, the exact

dose-response of active travel and road trauma is unknown. Another component of

plausibility is the degree of certainty of the results. Extensive uncertainty analyses were

conducted when distribution data for the input variables were available or distribution

assumptions were plausible (e.g. road trauma in Chapter 8). In this thesis, only diseases

for which there is well established evidence of an association with PA and PM2.5 were

included (colon cancer, breast cancer, ischemic heart disease, ischemic stroke, type 2

diabetes, lung cancer and chronic obstructive pulmonary disease). Other research

included an effect of PA on depression (37); however, the direction of the association is

still uncertain (see literature review Section 2.2). Plausibility is also concerned with the

model population. Prediction models in this thesis are for the general adult population of

Australians. No differentiation by subgroup was made (e.g. by socio-economic status),

ignoring health and social equity issues in the analysis. Another component of

plausibility is comparability with other research. Comparisons with other studies were

made in the discussion sections of Chapter 5, 7 and 8. A summary is provided here.

First, the ACE Prevention approach results in more modest outcomes compared with

the widely used methods based on CRA. Second, results in Chapter 8 are in agreement

with past literature indicating greater health benefits from improvements in PA

compared with potential health harms from exposure to air pollution and road trauma.

Lastly, assumptions of the dose-response for included risk factors are common in the

field, with important differences depending on the approach.

Formal validity

Formal validity refers to the methods applied to make the health predictions, which have

to be technically correct (373). The core model used to make health predictions is based

on a macro simulation approach developed for the ACE Prevention project (94). The

ACE Prevention approach is based on the well-established method of the proportional

MSLT Markov model (96) combined with the PIF. Multiple researchers have checked

161

the model, and a large number of studies were produced and published in peer-

reviewed journals using the ACE Prevention approach over the years. In addition, the

models in this body of work were checked by thesis advisors and study collaborators.

Another component of validity is the estimation of magnitude of causal relations. For the

study in Chapter 5, effect sizes for the association BE-PA are from the peer-reviewed

literature. Most of the studies indicated the effect in terms of the likelihood of doing PA

for different BE exposures. Assumptions were made to translate reported effects in the

literature in minutes of PA per week. Relative risks from meta-analysis were applied for

the effect of exposure to PA and PM2.5 on disease incidence. For PM2.5, relative risks

are for mortality; however, in Chapter 8 it was assumed that these apply to incidence.

For road trauma (Chapter 8), rates were estimated based on past literature methods

including the ‘safety in numbers’ effect. Formal validity also includes estimating the

degree of certainty. Outcomes in Chapter 5, 7 and 8 include uncertainty intervals from

Monte Carlo simulations, and distributions for input variables are described in each

chapter.

Predictive validity

Predictive validity refers to whether the findings are aligned with latter developments

(373). Establishing predictive validity based on later developments was not possible for

the prediction studies in this thesis. Generally, establishing predictive validity requires

outcome evaluations, which are seldom feasible in this field. An alternative is to apply

prediction models to historical data; however, the interventions modelled in this thesis

have not been tested elsewhere.

9.4 Directions for future research

Develop standardised methods for the quantification of effects of changes in the BE on PA

This thesis provides projections for the likely health and economic outcomes for a

large set of changes in BE attributes. Similar future work would greatly benefit

from standardised measures of the BE and PA to enable pooling of effect sizes.

Using statistical methods to show the pooled effect of multiple measurements

from diverse studies (systematic review and meta-analysis) would strengthen the

evidence-of-effect used in studies projecting the likely health outcomes of

162

alternative BEs. In addition, moving from categorical measures of exposure to

continuous ones would avoid the loss of exposure information which occurs in

the categorisation process. There is also substantial scope for improvements in

measurements of PA used in studies investigating the association BE-PA.

Moving from self-report measures of PA to objective measures would improve

the accuracy of estimated associations. Furthermore, to date, most of the

evidence is from observational studies estimating PA outcomes for groups of

people exposed to different BE features (e.g. high walkability vs low walkability).

Intervention studies assessing before and after levels of PA with an indication of

who is changing PA as a result of the intervention would provide greater

precision as to the population groups that would benefit the most from BE

interventions. An example of this is a natural experiment of transport

infrastructure in the United Kingdom which found that the least active people at

baseline increased walking, and those who were already active walked more as

a results of the intervention (302). Also, to date, it is unknown whether BE

initiatives that promote PA are increasing total PA or simply replacing PA

previously done for other purposes. For instance, if new transport infrastructure is

in place, people may ride their bicycle to work instead of going for a bike ride in

their leisure time; hence, overall there would be no PA-related health benefit from

the intervention. For health projections also targeting the transport sector,

demand models used in the sector could be applied. For example, the

Department of Transport in the State of Queensland used models to project the

cycling demand of new infrastructure; however, these are not publicity available.

Refine methods for assessing the effect of transport policies on physical activity

In Chapter 8, the potential health benefits of proposed travel targets were

estimated. Likewise, in similar past literature, assumptions were made to quantify

the increase in PA when there was a shift from motorised travel to active travel.

Greater refinements should be aimed for in future research by using methods

that enable investigating who is likely to replace car travel with active travel—

based on sociodemographic, trip distance, and topographic information. Such

methods already exist, for example, in England, a propensity-to-cycle tool was

developed which accounts for trip distance and hilliness (338). A similar

163

approach could be applied in Brisbane with the available transport information.

This could add an additional layer to the work in this thesis by investigating

socioeconomic differences. Refinement in projections for travel target policies

would serve to better inform policymakers as to where infrastructure and

community programs (e.g. awareness of active transport) need to take place to

support active travel. A fairly simple approach was taken in this research, which

may have resulted in inaccurate estimation of health outcomes from road trauma.

First, age and sex differences in the risk of road trauma should be accounted for.

Second, past research indicated that there is variability in the incidence of road

accidents depending on the road type. However, data for Brisbane is very

sparse; hence, stratification would result in very small or no observations for

many of the road trauma rates for striking and victim modes. Pooling road trauma

data from multiple Australian cities with similar characteristics would allow for

stratifications to better reflect road trauma patterns. Further refinements should

also be aimed for to account for trends in improvements in motor vehicles

technical requirements which will impact on emissions and safety (346, 374).

Develop quality guidelines for health projections of built environment interventions

The field evaluating health and economic outcomes of BE interventions is

relatively new. Hence, to date, there are no specific recommendations to guide

researchers about the minimum requirements for conducting and reporting

evaluation of BE interventions. In this research, the CHEERS checklist was used

to evaluate the quality of economic evaluations and a checklist from the literature

was adapted to evaluate the quality of studies reporting the association BE-PA.

In the past, and used in this thesis, a validation checklist was recommended to

assess the quality of predictions in health impact assessments (373). Greater

transparency could be gained from a developing a checklist specific to

projections of BE interventions, considering distinctiveness of the field, such as

the type of evidence of effectiveness from observational studies.

Improve dose-response for the association of PA and health outcomes

Research investigating the protective effect of PA is extensive. However, there

are still considerable uncertainties as to the dose-response of PA with health

164

outcomes. Most of the studies suggest non-linear associations, with greater

gains for lower levels of PA and diminishing returns at higher levels. Still, the

exact shape of the dose-response associations is uncertain. Also, from the

literature reviewed in this thesis, it may not just be that all PA is equally beneficial

for health, but the intensity may play an important role. A recent meta-analysis of

the protective effect of total and leisure time PA for people with type 2 diabetes

shows that greater health gains are accrued from increasing leisure-time PA

compared with total PA (344). In addition, gains from vigorous PA were found to

be higher than from moderate and total PA. In this meta-analysis, non-linear

functions were fitted to the data using cubic splines. In the future, similar

approaches should be followed for other diseases related to physical inactivity.

Greater information regarding intensity and domain would help to improve the

predictive validity of models like the ones in this thesis.

Incorporate health and social inequities

Health and economic predictions in this thesis are for the average population,

without accounting for important health and social inequities. Future similar

research would greatly contribute to the literature by estimating the concentration

of health gains/losses by socioeconomic group. For example, a recent study of

sugar-sweetened beverage tax found that most of the health gains would accrue

to the most disadvantage according to an index of economic and social

disadvantage (375).

9.5 Conclusions

This thesis systematically analysed the potential of the BE for health, primarily by its

impact on population levels of PA. The developments in this thesis contributed to the

CRE in Healthy, Liveable Communities and also served to inform the Health

Department of New South Wales as to the likely cost and benefits of BE interventions.

The studies included in Chapter 4 to 8 were presented at local, national, and

international conferences and workshops and have been well-received by the research

community. In addition, the study conducted for the Health Department of New South

Wales (adapted in Chapter 4, 5 and 7) caught significant attention among researchers,

165

practitioners and advocates. Dr Lennert Veerman was invited to present this work at

national and international conferences and workshops.

Furthermore, the candidate’s participation in the CRE set the basis for further

developments with collaborators in Australia and England. In Australia, the research

candidate is working with researchers at the Healthy Liveable Cities Group, Centre for

Urban Research at the Royal Melbourne Institute of Technology University in a project

that aims to estimate health and economic outcomes of realistic land-use developments.

Stakeholders from the planning sector are involved in contributing to the definition of

realistic scenarios and costs of alternatives. This is an important step towards producing

relevant scientific evidence that allows planners to consider health in the design of land

developments. In England, the aim is to estimate health outcomes for a transport

initiative based on a quasi-experiment of a real intervention.

166

Chapter 10 References

1. World Health Organization. Prevalence of insufficient physical activity [Internet].

2015 [cited 2014 6 June]. Available from:

http://www.who.int/gho/ncd/risk_factors/physical_activity/en/.

2. Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, et al. A

comparative risk assessment of burden of disease and injury attributable to 67 risk

factors and risk factor clusters in 21 regions, 1990-2010: a systematic analysis for the

Global Burden of Disease Study 2010. The Lancet. 2012;380(9859):2224-60.

3. Woodcock J, Franco OH, Orsini N, Roberts I. Non-vigorous physical activity and

all-cause mortality: systematic review and meta-analysis of cohort studies. Int J

Epidemiol. 2011;40(1):121-38.

4. Lee I, Shiroma E, Lobelo F, Puska P, Blair S, Katzmarzyk P. Effect of physical

inactivity on major non-communicable diseases worldwide: an analysis of burden of

disease and life expectancy. Lancet. 2012;380(9838):219 - 29.

5. Physical Activity Guidelines Advisory Committee. Physical Activity Guidelines

Advisory Committee Report, 2008. Washington, DC: US Department of Health and

Human Services, 2008.

6. Jakicic JM, Davis KK. Obesity and physical activity. Psychiatr Clin North Am.

2011;34(4):829-40.

7. Chang JT, Morton SC, Rubenstein LZ, Mojica WA, Maglione M, Suttorp MJ, et

al. Interventions for the prevention of falls in older adults: systematic review and meta-

analysis of randomised clinical trials. BMJ 2004;328(7441):680-.

8. Giles-Corti B, Sallis JF, Sugiyama T, Frank LD, Lowe M, Owen N. Translating

active living research into policy and practice: One important pathway to chronic disease

prevention. J Public Health Pol. 2015;36(2):231-43.

9. Sallis J, Owen N, Fisher E. Ecological models for health behaviour. In: Glanz K,

Rimer BK, Viswanath K, editors. Health behavior and health education: theory,

research, and practice. San Francisco: Jossey-Bass; 2008. p. 465-86.

167

10. World Health Organization. Interventions on diet and physical activity: what

works: summary report [Internet]. 2009 [cited 2015 15 October]. Available from:

http://www.who.int/dietphysicalactivity/summary-report-09.pdf.

11. McCormack G, Shiell A. In search of causality: a systematic review of the

relationship between the built environment and physical activity among adults. Int J

Behav Nutr Phys Act. 2011;8(1):125-.

12. Sallis JF, Cerin E, Conway TL, Adams MA, Frank LD, Pratt M, et al. Physical

activity in relation to urban environments in 14 cities worldwide: a cross-sectional study.

The Lancet. 2016;387(10034):2207-17.

13. Van Cauwenberg J, De Bourdeaudhuij I, De Meester F, Van Dyck D, Salmon J,

Clarys P, et al. Relationship between the physical environment and physical activity in

older adults: a systematic review. Health Place. 2011;17(2):458-69.

14. Van Holle V, Deforche B, Van Cauwenberg J, Goubert L, Maes L, Van de

Weghe N, et al. Relationship between the physical environment and different domains

of physical activity in European adults: a systematic review. BMC Public Health.

2012;12(1):807-.

15. Sallis JF, Bull F, Burdett R, Frank LD, Griffiths P, Giles-Corti B, et al. Use of

science to guide city planning policy and practice: how to achieve healthy and

sustainable future cities. The Lancet. 2016;388(10062):2936–47.

16. Giles-Corti B, Vernez-Moudon A, Reis R, Turrell G, Dannenberg AL, Badland H,

et al. City planning and population health: a global challenge. The Lancet.

2016;388(10062):10-6.

17. Leyden KM. Social capital and the built environment: the importance of walkable

neighborhoods. Am J Public Health. 2003;93(9):1546-51.

18. Giles-Corti B, Foster S, Shilton T, Falconer R. The co-benefits for health of

investing in active transportation. NSW Public Health Bull. 2010;21:122 - 7.

19. Stevenson M, Thompson J, de Sá TH, Ewing R, Mohan D, McClure R, et al.

Land use, transport, and population health: estimating the health benefits of compact

cities. The Lancet. 2016;388(10062):2925-35.

168

20. World Health Organization. Ottawa charter for health promotion. Health

Promotion. 1986;1(4):i-v.

21. WHO Regional Office for Europe. Healthy Cities [Internet]. N/D [cited 2016 13

October]. Available from: http://www.euro.who.int/en/health-topics/environment-and-

health/urban-health/activities/healthy-cities.

22. United Nations. Sustainable Development Goals [Internet]. N/D [cited 2016 13

October]. Available from: http://www.un.org/sustainabledevelopment/cities/.

23. United Nations. HABITAT III [Internet]. 2016 [cited 2016 16 December]. Available

from: https://habitat3.org/about.

24. Australian Local Government Association, Bus Industry Confederation, Cycling

Promotion Fund, National Heart Foundation of Australia, International Association of

Public Transport. An Australian vision for active transport 2010 [cited 2014 15

November]. Available from:

http://www.heartfoundation.org.au/SiteCollectionDocuments/Active-Vision-for-Active-

Transport-Report.pdf.

25. Connecting SEQ 2031 – An Integrated Regional Transport Plan for South East

Queensland [Internet]. 2011. Available from:

http://www.ppt.asn.au/pubdocs/connecting_seq2031+(1).pdf.

26. Queensland Cycle Strategy 2011-2021 [Internet]. 2015 [cited 2016 2 March].

Available from: http://www.tmr.qld.gov.au/Travel-and-transport/Cycling/Strategy.aspx.

27. Victoria State Goverment. Active Transport Victoria [Internet]. 2016 [cited 2017

19 January]. Available from: http://economicdevelopment.vic.gov.au/transport/cycling-

and-walking/active-transport-victoria.

28. New South Wales Goverment. Walking and Cycling [Internet]. Roads and

Maritime; 2016 [cited 2017 19 January]. Available from:

http://www.rms.nsw.gov.au/business-industry/partners-suppliers/lgr/active-

transport/index.html.

29. Planning Western Australia. Liveable neighbourhoods Western Australia. 2015

[cited 2015 27 February ]. Available from: http://www.planning.wa.gov.au/650.asp.

169

30. Mulley C, Tyson R, McCue P, Rissel C, Munro C. Valuing active travel: Including

the health benefits of sustainable transport in transportation appraisal frameworks. Res

Transp Bus Manag. 2013;7:27-34.

31. University of Melbourne. NHMRC Centre of Research Excellence in Healthy,

Liveable Communities [Internet]. N/D [cited 2017 31 January]. Available from:

http://mccaughey.unimelb.edu.au/programs/cre.

32. Zapata-Diomedi B, Brown V, Veerman J. An evidence review and modelling

exercise: The effects of urban form on health: costs and benefits. The Australian

Prevention Partnership Centre 2015.

33. Jackson RJ, Dannenberg AL, Frumkin H. Health and the built environment: 10

years after. Am J Public Health. 2013;103(9):1542-4.

34. Preventive Health Taskforce. Australia: The healthiest country by 2020 [Internet].

Australian Goverment; 2009 [cited 2014 10 November]. Available from:

http://www.health.gov.au/internet/preventativehealth/publishing.nsf/Content/nphs-

overview.

35. Giles-Corti B, Timperio A, Bull F, Pikora T. Understanding physical activity

environmental correlates: Increased specificity for ecological models. Exerc Sport Sci

Rev. 2005;33(4):175-81.

36. Sallis J, Cervero R, Ascher W, Henderson K, Kraft M, Kerr J. An ecological

approach to creating active living communities. Annu Rev Public Health. 2006;27:297 -

322.

37. Mueller N, Rojas-Rueda D, Cole-Hunter T, de Nazelle A, Dons E, Gerike R, et al.

Health impact assessment of active transportation: a systematic review. Prev Med.

2015;76:103-14.

38. Mansfield TJ, Rodriguez DA, Huegy J, MacDonald Gibson J. The Effects of

Urban Form on Ambient Air Pollution and Public Health Risk: A Case Study in Raleigh,

North Carolina. Risk Analysis. 2015;35(5):901-18.

39. United States Department of Health and Human Services. Physical activity and

health: a report of the Surgeon General. President's Council on Physical F, Sports.

Atlanta, Georgia: U.S. Department of Health and Human Services, CDC, National

Center for Chronic Disease Prevention and Health Promotion; 1996.

170

40. U.S. Department of Health and Human Services. 2008 Physical Activity

Guidelines for Americans-Be active, healthy and happy! Washington, D.C. USA: 2008.

41. Australian Goverment-Department of Health. Australia's Physical Activity and

Sedentary Behaviour Guidelines 2014 [cited 2016 3 Mar]. Available from:

http://www.health.gov.au/internet/main/publishing.nsf/Content/health-pubhlth-strateg-

phys-act-guidelines#apaadult.

42. World Health Organization. Global recommendations on Physical Activity for

Health [Internet]. 2010 [cited 2015 3 February]. Available from:

http://whqlibdoc.who.int/publications/2010/9789241599979_eng.pdf?ua=1.

43. Australian Bureau of Statistics. 4363.0.55.001 - Australian Health Survey: Users'

Guide, 2011-13 [Internet]. 2013 [cited 2015 10 November]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/Lookup/59FF6FE5D62BABE1CA257BAC002

25A07?opendocument.

44. Kelly P, Kahlmeier S, Götschi T, Orsini N, Richards J, Roberts N, et al.

Systematic review and meta-analysis of reduction in all-cause mortality from walking

and cycling and shape of dose response relationship. Int J Behav Nutr Phys Act.

2014;11(1):132.

45. Samitz G, Egger M, Zwahlen M. Domains of physical activity and all-cause

mortality: systematic review and dose-response meta-analysis of cohort studies. Int J

Epidemiol. 2011;40(5):1382-400.

46. Bull FC, Armstrong TP, Dixon T, Ham S, Neiman A, Pratt M. Physical inactivity.

Comparative quantification of health risks: Global and regional burden of disease due to

selected major risk factors Volume 1. 2004:729 - 881.

47. Danaei G, Ding EL, Mozaffarian D, Taylor B, Rehm J, Murray CJL, et al. The

preventable causes of death in the United States: comparative risk assessment of

dietary, lifestyle, and metabolic risk factors. PLoS Med. 2009;6(4):e1000058.

48. Kyu HH, Bachman VF, Alexander LT, Mumford JE, Afshin A, Estep K, et al.

Physical activity and risk of breast cancer, colon cancer, diabetes, ischemic heart

disease, and ischemic stroke events: systematic review and dose-response meta-

analysis for the Global Burden of Disease Study 2013. BMJ. 2016;354.

171

49. Goodwin RD. Association between physical activity and mental disorders among

adults in the United States. Prev Med. 2003;36(6):698-703.

50. Harvey SB, Hotopf M, Overland S, Mykletun A. Physical activity and common

mental disorders. Br J Psychiatry. 2010;197(5):357-64.

51. Ströhle A. Physical activity, exercise, depression and anxiety disorders. J Neural

Transm (Vienna). 2009;116(6):777-84.

52. Steinmo S, Hagger-Johnson G, Shahab L. Bidirectional association between

mental health and physical activity in older adults: Whitehall II prospective cohort study.

Prev Med 2014;66:74-9.

53. Azevedo Da Silva M, Singh-Manoux A, Brunner E, Kaffashian S, Shipley M,

Kivimäki M, et al. Bidirectional association between physical activity and symptoms of

anxiety and depression: the Whitehall II study. Eur J Epidemiol. 2012;27(7):537-46.

54. Blondell SJ, Hammersley-Mather R, Veerman JL. Does physical activity prevent

cognitive decline and dementia?: A systematic review and meta-analysis of longitudinal

studies. BMC Public Health. 2014;14(1):510.

55. Center for Disease Control and Prevention. State Indicator Report on Physical

Activity 2014 [Internet]. Atlanta, GA: U.S: Department of Health and Human Services;

2014 [cited 2015 1 September]. Available from:

http://www.cdc.gov/physicalactivity/downloads/pa_state_indicator_report_2014.pdf.

56. Statistics Canada. Directly measured physical activity of Canadian adults 2007 to

2011, [Internet]. Ottawa, Ontario, Canada 2013 [cited 2015 28 October]. Available from:

http://www.statcan.gc.ca/pub/82-625-x/2013001/article/11807-eng.htm.

57. Australian Health Survey: Physical Activity, 2011-12 [Internet]. 2015 [cited 2015

22 September]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/Lookup/D4495467B7F7EB01CA257BAC001

5F593?opendocument.

58. Australia's health 2016—in brief [Internet]. AIHW. 2016 [cited 2017 1 February].

Available from: http://www.aihw.gov.au/publication-detail/?id=60129555545.

59. Australian Institute of Health and Welfare. Australia's health 2014. Canberra:

AIHW, 2014.

172

60. Australian Institute of Health and Welfare. Chronic disease and participation in

work. Canberra: AIHW, 2009.

61. Australia's commuting distance: cities and regions [Internet]. Bureau of

Infrastructure, Transport and Regional Economics, 2015 [cited 201616 October].

Available from: https://bitre.gov.au/publications/2015/files/is_073.pdf.

62. Participation in Sport and Physical Recreation, Australia, 2009-10 [Internet].

Australian Bureau of Statistics. 2010 [cited 2017 29 March]. Available from:

http://www.abs.gov.au/AUSSTATS/[email protected]/Lookup/4177.0Main+Features12009-

10?OpenDocument.

63. Doorley R, Pakrashi V, Ghosh B. Quantifying the health impacts of active travel:

assessment of methodologies. Transport Rev. 2015;35(5):559-82.

64. Ding D, Gebel K. Built environment, physical activity, and obesity: What have we

learned from reviewing the literature? Health Place. 2012;18(1):100-5.

65. Frank LD, Schmid TL, Sallis JF, Chapman J, Saelens BE. Linking objectively

measured physical activity with objectively measured urban form: findings from

SMARTRAQ. Am J Prev Med. 2005;28.

66. Craig P, Cooper C, Gunnell D, Haw S, Lawson K, Macintyre S, et al. Using

natural experiments to evaluate population health interventions: new Medical Research

Council guidance. J Epidemiol Commun H. 2012;66(12):1182-6.

67. Giles-Corti B, Bull F, Knuiman M, McCormack GR, Van Niel K, Timperio A, et al.

The influence of urban design on neighbourhood walking following residential relocation:

longitudinal results from the RESIDE study. Soc Sci Med. 2013;77:20-30.

68. Panter J, Heinen E, Mackett R, Ogilvie D. Impact of New Transport Infrastructure

on Walking, Cycling, and Physical Activity. Am J Prev Med. 2016;50(2):e45-e53.

69. Welle B, Liu Q, Li W, Adrialoza-Steil A, King R, Sarmiento C, et al. Cities safer by

design Washington, DC: World Resources Institute, 2015.

70. Lalive R, Luechinger S, Schmutzler A. Does supporting passenger railways

reduce road traffic externalities? University of Zurich, 2013.

71. Chen P, Zhou J. Effects of the built environment on automobile-involved

pedestrian crash frequency and risk. J Transp Health. 2016;3(4):448-56.

173

72. Frank L, Giles-Corti B, Ewing R. The influence of the built environment on

transport and health. J Transp Health. 2016;3(4):423-5.

73. Reynolds CC, Harris MA, Teschke K, Cripton PA, Winters M. The impact of

transportation infrastructure on bicycling injuries and crashes: a review of the literature.

Environ Health. 2009;8(1):47.

74. Dumbaugh E, Rae R. Safe Urban Form: Revisiting the Relationship Between

Community Design and Traffic Safety. J Am Plann Assoc. 2009;75(3):309-29.

75. Pucher J, Dijkstra L. Promoting safe walking and cycling to improve public health:

lessons from The Netherlands and Germany. Am J Public Health. 2003;93(9):1509-16.

76. Blaizot S, Papon F, Haddak MM, Amoros E. Injury incidence rates of cyclists

compared to pedestrians, car occupants and powered two-wheeler riders, using a

medical registry and mobility data, Rhone County, France. Accid Anal Prev. 2013;58:35-

45.

77. McAndrews C. Traffic risks by travel mode in the metropolitan regions of

Stockholm and San Francisco: a comparison of safety indicators. Inj Prev.

2011;17(3):204-7.

78. OECD/International Transport Forum. Cycling, Health and Safety [Internet].

OECD Publishing/ITF; 2013. Available from: 10.1787/9789282105955-en.

79. Pucher J, Buehler R. Making cycling irresistible: lessons from the Netherlands,

Denmark and Germany. Transp Rev. 2008;28.

80. Jacobsen P. Safety in numbers: more walkers and bicyclists, safer walking and

bicycling. Inj Prev. 2003;9(3):205 - 9.

81. Tin ST, Woodward A, Thornley S, Ameratunga S. Regional variations in pedal

cyclist injuries in New Zealand: safety in numbers or risk in scarcity? Aust N Z J Public

Health. 2011;35(4):357-63.

82. Robinson DL. Safety in numbers in Australia: more walkers and bicyclists, safer

walking and bicycling. Health Promot J Austr. 2005;16(1):47-51.

83. Bhatia R, Wier M. "Safety in Numbers" re-examined: can we make valid or

practical inferences from available evidence? Accid Anal Prev. 2011;43(1):235-40.

174

84. Elvik R, Bjørnskau T. Safety-in-numbers: A systematic review and meta-analysis

of evidence. Safety Science. 2017;92:274-82.

85. Valavanidis A, Fiotakis K, Vlachogianni T. Airborne particulate matter and human

health: toxicological assessment and importance of size and composition of particles for

oxidative damage and carcinogenic mechanisms. J Environ Sci Health C.

2008;26(4):339-62.

86. World Health Organization. Ambient (outdoor) air quality and health [Internet].

WHO; 2016 [cited 2017 27 January]. Available from:

http://www.who.int/mediacentre/factsheets/fs313/en/.

87. De Nazelle A, Nieuwenhuijsen MJ, Antó JM, Brauer M, Briggs D, Braun-

Fahrlander C, et al. Improving health through policies that promote active travel: a

review of evidence to support integrated health impact assessment. Environ Int.

2011;37(4):766-77.

88. Burnett RT, Pope CA, Ezzati M, Olives C, Lim SS, Mehta S, et al. An integrated

risk function for estimating the Global Burden of Disease attributable to ambient fine

particulate matter exposure. Environ Health Perspect. 2014;122(4):397-403.

89. Hoek G, Krishnan RM, Beelen R, Peters A, Ostro B, Brunekreef B, et al. Long-

term air pollution exposure and cardio- respiratory mortality: a review. Environ Health

2013;12(1):43-.

90. Hamra GB, Guha N, Cohen A, Laden F, Raaschou-Nielsen O, Samet J, et al.

Outdoor particulate matter exposure and lung cancer: a systematic review and meta-

analysis. Environ Health Perspect. 2014;122:906-11.

91. World Health Organization. WHO Expert Meeting: Methods and tools for

assessing the health risks of air pollution at local, national and international level.

Copenhagen: WHO Regional Office for Europe, 2014.

92. Gan WQ, Koehoorn M, Davies HW, Demers PA, Tamburic L, Brauer M. Long-

term exposure to traffic-related air pollution and the risk of coronary heart disease

hospitalization and mortality. Environ Health Perspect. 2011;119(4):501-7.

93. Tsai D-H, Wang J-L, Chuang K-J, Chan C-C. Traffic-related air pollution and

cardiovascular mortality in central Taiwan. Sci Total Environ. 2010;408(8):1818-23.

175

94. Cobiac LJ, Vos T, Barendregt JJ. Cost-effectiveness of interventions to promote

physical activity: a modelling study. Plos Med. 2009;6(7):e1000110-e.

95. Vos T, Carter R, Barendregt J, Mihalopoulos C, Veerman J, Magnus A, et al.

Assessing Cost-Effectiveness in Prevention (ACE-Prevention): Final Report 2010 [cited

2014 1 February]. Available from: https://public-health.uq.edu.au/files/571/ACE-

Prevention_final_report.pdf.

96. Barendregt JJ, Van Oortmarssen GJ, Van Hout BA, Van Den Bosch JM. Coping

with multiple morbidity in a life table. Math Popul Stud. 1998;7(1):29-49.

97. Drummond MF, Sculpher MJ, Torrance GW, O'Brien BJ, Stoddart GL. Methods

for the economic evaluation of health care programmes. 3rd ed. United States: Oxford

University Press; 2005.

98. Cobiac LJ. Cost-effectiveness of interventions to prevent lifestyle-related disease

and injury in Australia [PhD thesis]. St Lucia, Qld: University of Queensland; 2010.

99. Briggs AH, Claxton K, Sculpher MJ. Decision modelling for health economic

evaluation. Oxford: Oxford University Press; 2006.

100. Barendregt JJ, Oortmarssen vGJ, Murray CJ, Vos T. A generic model for the

assessment of disease epidemiology: the computational basis of DisMod II. Popul

Health Metr. 2003;1(1):4-.

101. Barendregt JJ, Veerman JL. Categorical versus continuous risk factors and the

calculation of potential impact fractions. J Epidemiol Community Health.

2010;64(3):209-12.

102. Barendregt JJ. EpiGear International 2012 [cited 2015 1 Mar]. Available from:

http://www.epigear.com/index_files/prevent.html.

103. Ezzati M, World Health O. Comparative quantification of health risks: global and

regional burden of disease attributable to selected major risk factors. Geneva: World

Health Organization; 2004.

104. Global Burden of Disease Study 2015 (GBD 2015) Results [Internet]. 2016.

Available from: http://ghdx.healthdata.org/gbd-results-tool.

105. Vos T, Carter R, Barendregt JJ, C. M, Veerman J, Magnus A, et al. Assessing

Cost-Effectiveness in the Prevention of Non-Communicable Disease (ACE Prevention)

176

Project 2005–09-Economic Evaluation Protocol Brisbane, Australia2007 [cited 2015 13

January]. Available from: http://public-health.uq.edu.au/filething/get/1833/ACE-

P_Econ_Protocol_no_append.pdf.

106. Australian Institute of Health and Welfare. Disease costs and impact study data.

Australian Institute of Health and Welfare, 2001.

107. Begg S, Vos T, Barker B, Stevenson C, L S, Lopez A. The burden of disease

and injury in Australia 2003. Canberra: AIHW: 2007 Cat. no. PHE 82 Contract No.: 4

March.

108. Götschi T, Hadden Loh T. Advancing project-scale health impact modeling for

active transportation: A user survey and health impact calculation of 14 US trails. J

Transp Health. Forthcoming 2017. 2017.

109. Van Baal PHM, Feenstra TL, Polder JJ, Hoogenveen RT, Brouwer WBF.

Economic evaluation and the postponement of health care costs. Health Econ.

2011;20(4):432-45.

110. Gros B, Soto Álvarez J, Ángel Casado M. Incorporation of future costs in health

economic analysis publications: current situation and recommendations for the future.

Expert Rev Pharmacoecon Outcomes Res. 2015;15(3):465-9.

111. Barendregt JJ, Bonneux L, van der Maas PJ. The Health Care Costs of

Smoking. N Engl J Med. 1997;337(15):1052-7.

112. Van Baal PHM, Feenstra TL, Hoogenveen RT, Ardine De Wit G, Brouwer WBF.

Unrelated medical care in life years gained and the cost utility of primary prevention: in

search of a ‘perfect’ cost–utility ratio. Health Econ. 2007;16(4):421-33.

113. Van Baal PHM, Wong A, Slobbe LCJ, Polder JJ, Brouwer WBF, de Wit GA.

Standardizing the Inclusion of Indirect Medical Costs in Economic Evaluations.

Pharmacoeconomics. 2011;29(3):175-87.

114. Rappange DR, van Baal PH, van Exel NJ, Feenstra TL, Rutten FF, Brouwer WB.

Unrelated medical costs in life-years gained: should they be included in economic

evaluations of healthcare interventions? Pharmacoeconomics. 2008;26(10):815-30.

177

115. Morton A, Adler AI, Bell D, Briggs A, Brouwer W, Claxton K, et al. Unrelated

Future Costs and Unrelated Future Benefits: Reflections on NICE Guide to the Methods

of Technology Appraisal. Health Econ. 2016;25(8):933-8.

116. Nyman JA. Should the consumption of survivors be included as a cost in cost-

utility analysis? Health Econ. 2004;13(5):417-27.

117. International Society for Pharmacoeconomics and Outcomes Research.

Pharmacoeconomic Guidelines around the world 2017 [cited 2017 15 August]. Available

from: https://www.ispor.org/PEguidelines/index.asp.

118. Torrance GW. Framing and designing the cost-effectiveness analysis. Cost-

effectiveness in health and medicine. 1996:54-81.

119. Centre for Population Health. NSW Healthy Eating and Active Living Strategy.

NSW Ministry of Health: 2013.

120. Van Wee B, Börjesson M. How to make CBA more suitable for evaluating cycling

policies. Transp Pol. 2015;44:117-24.

121. Transport for New South Wales. Principles and guidelines for economic appraisal

of transport investment and initiatives Sydney: Transport for NSW; 2013 [cited 2015 3

March]. Available from:

http://www.transport.nsw.gov.au/sites/default/files/b2b/publications/tfnsw-principles-and-

guidelines-for-economic-appraisal-of-transport-initiatives.pdf.

122. Dallat MA, Soerjomataram I, Hunter RF, Tully MA, Cairns KJ, Kee F. Urban

greenways have the potential to increase physical activity levels cost-effectively. Eur J

Public Health. 2014;24(2):190-5.

123. Mansfield TJ, Gibson JM. Health Impacts of Increased Physical Activity from

Changes in Transportation Infrastructure: Quantitative Estimates for Three

Communities. BioMed Res. Int. 2015;2015:812325.

124. Lhachimi SK, Nusselder WJ, Smit HA, van Baal P, Baili P, Bennett K, et al.

DYNAMO-HIA–A Dynamic Modeling Tool for Generic Health Impact Assessments.

PLoS One. 2012;7(5):e33317.

178

125. Davison KK, Lawson CT. Do attributes in the physical environment influence

children's physical activity? A review of the literature. Int. J. Behav. Nutr. Phys. Act..

2006;3:19-.

126. Grasser G, Van Dyck D, Titze S, Stronegger W. Objectively measured walkability

and active transport and weight-related outcomes in adults: a systematic review. Int J

Public Health. 2013;58(4):615-25.

127. McCormack GR, Giles-Corti B, Lange A, Smith T, Martin K, Pikora T. An update

of recent evidence of the relationship between objective and self-report measures of the

physical environment and physical activity behaviours. J Sci Med Sport. 2004;7(1,

Supplement 1):81-92.

128. Fraser SDS, Lock K. Cycling for transport and public health: a systematic review

of the effect of the environment on cycling. Eur J Pub Health. 2011;21(6):738-43.

129. Ferdinand AO, Sen B, Rahurkar S, Engler S, Menachemi N. The relationship

between built environments and physical activity: a systematic review. Am J Public

Health. 2012;102(10):e7-e13.

130. Renalds A, Smith TH, Hale PJ. A systematic review of built environment and

health. Fam Community Health. 2010;33(1):68-78.

131. Saelens BE, Handy SL. Built Environment Correlates of Walking: A Review.

Med. Sci. Sports Exerc.. 2008;40(7 Suppl):S550-S66.

132. Dobbins M, Tirilis D. Built Environment: A Synthesis of Review of Evidence.

2011.

133. Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for

systematic reviews and meta-analyses: the PRISMA statement. Ann. Intern. Med..

2009;151(4):264-9.

134. Ewing R, Cervero R. Travel and the built environment: a meta-analysis. J Am

Plann Assoc. 2010;76(3):265-94.

135. Giles-Corti B. Personal communication 2015.

136. Ding D, Sallis JF, Kerr J, Lee S, Rosenberg DE. Neighborhood Environment and

Physical Activity Among Youth: A Review. Am. J. Prev. Med.. 2011;41(4):442-55.

179

137. Owen N, Humpel N, Leslie E, Bauman A, Sallis JF. Understanding

environmental influences on walking: Review and research agenda. Am. J. Prev. Med..

2004;27(1):67-76.

138. Saelens B, Sallis J, Frank L. Environmental correlates of walking and cycling:

Findings from the transportation, urban design, and planning literatures.

Ann Behav Med. 2003;25(2):80-91.

139. Wilson LAM, Giles-Corti B, Burton NW, Giskes K, Haynes M, Turrell G. The

association between objectively measured neighborhood features and walking in

middle-aged adults. Am J Health Promot. 2011;25(4):e12-e21.

140. McCormack GR, Shiell A, Giles-Corti B, Begg S, Veerman JL, Geelhoed E, et al.

The association between sidewalk length and walking for different purposes in

established neighborhoods. Int J Behav Nutr Phys Act. 2012;9:92.

141. Christian H, Bull F, Middleton NJ, Knuiman MW, Divitini ML, Hooper P, et al.

How important is the land use mix measure in understanding walking behaviour?

Results from the RESIDE study. Int. J. Behav. Nutr. Phys. Act.. 2011;8:55-.

142. Foster S, Knuiman M, Villanueva K, Wood L, Christian H, Giles-Corti B. Does

walkable neighbourhood design influence the association between objective crime and

walking? Int. J. Behav. Nutr. Phys. Act.. 2014;11(1):100-.

143. Astell-Burt T, Feng X, Kolt GS. Identification of the impact of crime on physical

activity depends upon neighbourhood scale: Multilevel evidence from 203,883

Australians. Health Place. 2015;31:120-3.

144. Astell-Burt T, Feng X, Kolt GS. Green space is associated with walking and

moderate-to-vigorous physical activity (MVPA) in middle-to-older-aged adults: findings

from 203 883 Australians in the 45 and Up Study. Br J Sports Med. 2014;48(5):404-6.

145. Gebel K, Ding D, Foster C, Bauman AE, Sallis JF. Improving Current Practice in

Reviews of the Built Environment and Physical Activity. Sports Medicine.

2015;45(3):297-302.

146. Sugiyama T, Giles-Corti B, Summers J, du Toit L, Leslie E, Owen N. Initiating

and maintaining recreational walking: a longitudinal study on the influence of

neighborhood green space. Prev. Med.. 2013;57(3):178-82.

180

147. Shimura H, Sugiyama T, Winkler E, Owen N. High neighborhood walkability

mitigates declines in middle-to-older aged adults' walking for transport. J Phys Act

Health. 2012;9(7):1004-8.

148. Knuiman MW, Christian HE, Divitini ML, Foster SA, Bull FC, Badland HM, et al. A

longitudinal analysis of the influence of the neighborhood built environment on walking

for transportation: the RESIDE study. Am J Epidemiol. 2014;180(5):453-61.

149. Duncan MJ, Winkler E, Sugiyama T, Cerin E, duToit L, Leslie E, et al.

Relationships of land use mix with walking for transport: do land uses and geographical

scale matter? J Urban Health. 2010;87(5):782-95.

150. Koohsari MJ, Sugiyama T, Lamb KE, Villanueva K, Owen N. Street connectivity

and walking for transport: role of neighborhood destinations. Prev. Med.. 2014;66:118-

22.

151. Owen N, De Bourdeaudhuij I, Sugiyama T, Leslie E, Cerin E, Van Van Dyck D,

et al. Bicycle use for transport in an Australian and a Belgian city: associations with built-

environment attributes. J Urban Health. 2010;87(2):189-98.

152. Sugiyama T, Leslie E, Giles-Corti B, Owen N. Physical activity for recreation or

exercise on neighbourhood streets: associations with perceived environmental

attributes. Health Place. 2009;15(4):1058-63.

153. Titze S, Giles-Corti B, Knuiman MW, Pikora TJ, Timperio A, Bull FC, et al.

Associations Between Intrapersonal and Neighborhood Environmental Characteristics

and Cycling for Transport and Recreation in Adults: Baseline Results From the RESIDE

Study. J Phys Act Health. 2010;7(4):423-31.

154. Learnihan V, Van Niel KP, Giles-Corti B, Knuiman M. Effect of Scale on the Links

between Walking and Urban Design. Geographical Research. 2011;49(2):183-91.

155. Koohsari MJ, Kaczynski AT, Giles-Corti B, Karakiewicz JA. Effects of access to

public open spaces on walking: Is proximity enough? Landsc Urban Plan. 2013;117:92-

9.

156. Koohsari MJ, Karakiewicz JA, Kaczynski AT. Public Open Space and Walking:

The Role of Proximity, Perceptual Qualities of the Surrounding Built Environment, and

Street Configuration. Environ Behav. 2013;45(6):706-36.

181

157. Cleland VJ, Ball K, Crawford D. Is a perceived supportive physical environment

important for self-reported leisure time physical activity among socioeconomically

disadvantaged women with poor psychosocial characteristics? An observational study.

BMC Public Health. 2013;13:280-.

158. Heesch KC, Giles-Corti B, Turrell G. Cycling for transport and recreation:

associations with socio-economic position, environmental perceptions, and

psychological disposition. Prev Med. 2014;63:29-35.

159. McKibbin M, editor The influence of the built environment on mode choice -

Evidence from the journey to work in Sydney. ATRF 2011 - 34th Australasian Transport

Research Forum; 2014.

160. Frank LD, Sallis JF, Saelens BE, Leary L, Cain L, Conway TL, et al. The

development of a walkability index: Application to the neighborhood quality of life study.

Br J Sports Med. 2010;44(13):924-33.

161. Gebel K, Bauman AE, Sugiyama T, Owen N. Mismatch between perceived and

objectively assessed neighborhood walkability attributes: prospective relationships with

walking and weight gain. Health Place. 2011;17(2):519-24.

162. Ma L, Dill J. Associations between the objective and perceived built environment

and bicycling for transportation. J Transp Health. 2015;2(2):248-55.

163. Giles-Corti B, Foster S, Ryan K. Increasing density in Australia: maximising the

health benefits and minimising harm. Melbourne: National Heart Foundation of

Australia, 2012.

164. National Health and Medical Reserach Council of Australia. A guide to

development, implementation and evaluation of clinical practice guidelines. Canberra:

1999.

165. BMJ. Chapter 4 Measurement error and bias: BMJ Publishing Group Ltd; 2016

[cited 2016 9 March]. Available from: http://www.bmj.com/about-bmj/resources-

readers/publications/epidemiology-uninitiated/4-measurement-error-and-bias.

166. Lamb K, White S. Categorisation of built environment characteristics: the trouble

with tertiles. Int J Behav Nutr Phys Act. 2015;12(1):19.

182

167. Fagard RH, Staessen JA, Thijs L. Advantages and disadvantages of the meta-

analysis approach. J Hypertens Suppl. 1996;14(2):S9-12; discussion S3.

168. Australian Bureau of Statistics. National Health Survey: First Results, 2014-15

Canberra. 2015 [cited 2015 9 December]. Available from:

http://www.abs.gov.au/AUSSTATS/[email protected]/Lookup/4364.0.55.001Main+Features320

14-15?OpenDocument.

169. GBD Compare [Internet]. IHME, University of Washington. 2015 [cited 2016 15

July]. Available from: https://vizhub.healthdata.org/gbd-compare/.

170. Pratt M, Norris J, Lobelo F, Roux L, Wang G. The cost of physical inactivity:

moving into the 21st century. Br J Sports Med. 2012.

171. Sallis J, Floyd M, Rodriguez D, Saelens B. The role of built environments in

physical activity, obesity, and CVD. Circulation. 2012;125:729 - 37.

172. Eichinger M, Titze S, Haditsch B, Dorner TE, Stronegger WJ. How Are Physical

Activity Behaviors and Cardiovascular Risk Factors Associated with Characteristics of

the Built and Social Residential Environment? PLoS ONE. 2015;10(6):e0126010.

173. Kramer D, Maas J, Wingen M, Kunst AE. Neighbourhood safety and leisure-time

physical activity among Dutch adults: A multilevel perspective. Int J Behav Nutr Phys

Act.. 2013;10(1):11.

174. Centre for Diet and Activity Research. Integrated Transport and Health Impact

Modelling Tool (ITHIM) Centre for Diet and Activity Research (CEDAR) 2015 [cited 2015

27 January]. Available from: http://www.cedar.iph.cam.ac.uk/research/modelling/ithim/.

175. Woodcock J, Edwards P, Tonne C, Armstrong BG, Ashiru O, Banister D, et al.

Public health benefits of strategies to reduce greenhouse-gas emissions: urban land

transport. The Lancet. 2009;374(9705):1930-43.

176. MacDonald Gibson J, Rodriguez D, Dennerlein T, Mead J, Hasch T, Meacci G,

et al. Predicting urban design effects on physical activity and public health: A case

study. Health Place. 2015;35:79-84.

177. Australia: the healthiest country by 2020-National Preventative Health Strategy

[Internet]. Commonwealth of Australia. 2009 [cited 2016 5 November]. Available from:

183

http://www.preventativehealth.org.au/internet/preventativehealth/publishing.nsf/Content/

CCD7323311E358BECA2575FD000859E1/$File/nphs-roadmap.pdf.

178. Thompson S, Kent J, Lyons C. Building partnerships for healthy environments:

research, leadership and education. Health Promot J Austr. 2014;25(3):202-8.

179. Zapata-Diomedi B, Veerman JL. The association between built environment

features and physical activity in the Australian context: a synthesis of the literature. BMC

Public Health. 2016;16(1):1-10.

180. Grant RL. Converting an odds ratio to a range of plausible relative risks for better

communication of research findings. BMJ. 2014;348.

181. Zhang J, Yu KF. What's the relative risk?: A method of correcting the odds ratio

in cohort studies of common outcomes. JAMA. 1998;280(19):1690-1.

182. Gold MR. Cost-effectiveness in health and medicine. New York: Oxford


183. Ainsworth BE, Haskell WL, Herrmann SD, Meckes N, Bassett DR, Jr., Tudor-

Locke C, et al. 2011 compendium of physical activities: a second update of codes and

MET values. Med Sci Sports Exerc. 2011;43(8):1575-81.

184. GBD 2013 Risk Factors Collaborators. Global, regional, and national

comparative risk assessment of 79 behavioural, environmental and occupational, and

metabolic risks or clusters of risks in 188 countries, 1990-2013: A systematic analysis

for the Global Burden of Disease Study 2013. The Lancet. 2015;386(10010):2287-323.

185. Australian Institute of Health and Welfare. Health expenditure Australia 2012–13

Canberra: AIHW2014. Available from:

http://www.aihw.gov.au/WorkArea/DownloadAsset.aspx?id=60129548869.

186. Australian Bureau of Statistics. 3302.0 - Deaths, Australia-2012 Canberra2013

[cited 2015 5 Mar]. Available from:

http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3302.02012?OpenDocument

.

187. Australian Bureau of Statistics. 3101.0 - Australian Demographic Statistics, Dec

2013 [cited 2015 5 March]. Available from:

184

http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202013?OpenD

ocument.

188. Global Burden of Disease (GBD) [Internet]. 2015 [cited 2016 25 July]. Available

from: http://www.healthdata.org/gbd.

189. Asia Pacific Cohort Studies Collaboration. The Effects of Diabetes on the Risks

of Major Cardiovascular Diseases and Death in the Asia-Pacific Region. Diabetes Care.

2003;26(2):360-6.

190. Sattelmair J, Pertman J, Ding EL, Kohl HW, Haskell W, Lee I-M. Dose Response

Between physical activity and risk of coronary heart disease: a meta-analysis.

Circulation. 2011;124(7):789-95.

191. Tan-Torres Edejer T. Making choices in health: WHO guide to cost-effectiveness

analysis. Geneva: World Health Organization; 2003.

192. Boarnet MG, Greenwald M, McMillan TE. Walking, urban design, and health:

toward a cost-benefit analysis framework. J Plann Educ Res. 2008;27(3):341-58.

193. Murray CJL, Ezzati M, Flaxman AD, Lim S, Lozano R, Michaud C, et al. GBD

2010: design, definitions, and metrics. The Lancet. 2012;380(9859):2063-6.

194. Commonwealth of Australia. Handbook of Cost Benefit Analysis, 2006.

195. Woodcock J, Givoni M, Morgan AS. Health impact modelling of active travel

visions for England and Wales using an Integrated Transport and Health Impact

Modelling Tool (ITHIM). PLoS One. 2013;8(1):e51462.

196. Hamer M, Chida Y. Physical activity and risk of neurodegenerative disease: a

systematic review of prospective evidence. Psychol Med. 2009;39(1):3-11.

197. Cavill N, Kahlmeier S, Rutter H, Racioppi F, Oja P. Economic analyses of

transport infrastructure and policies including health effects related to cycling and

walking: a systematic review. Transp Pol. 2008;15(5):291-304.

198. World Health Organisation. Global Health Estimates: WHO; 2014 [cited 2014 1

July]. Available from: http://www.who.int/healthinfo/global_burden_disease/en/.

199. Spence JC, Lee RE. Toward a comprehensive model of physical activity. J Sport

Exerc Psychol. 2003;4(1):7-24.

185

200. Cordain L, Gotshall R, Eaton SB, Eaton 3rd S. Physical activity, energy

expenditure and fitness: an evolutionary perspective. Int J Sports Med. 1998;19(5):328-

35.

201. Hallal PC, Andersen LB, Bull FC, Guthold R, Haskell W, Ekelund U. Global

physical activity levels: surveillance progress, pitfalls, and prospects. The Lancet.

2012;380(9838):247-57.

202. Egger G, Swinburn B. An ''ecological'' approach to the obesity pandemic. BMJ.

1997;315(7106):477-80.

203. Stokols D. Establishing and maintaining healthy environments: toward a social

ecology of health promotion. Am. Psychol. 1992;47(1):6-22.

204. Sallis JF, Cervero RB, Ascher W, Henderson KA, Kraft MK, Kerr J. An ecological

approach to creating active living communities. Annu Rev Public Health. 2006;27:297-

322.

205. Heath GW, Parra DC, Sarmiento OL, Andersen LB, Owen N, Goenka S, et al.

Evidence-based intervention in physical activity: lessons from around the world. Lancet.

2012;380(9838):272-81.

206. Sahlqvist S, Song Y, Ogilvie D. Is active travel associated with greater physical

activity? The contribution of commuting and non-commuting active travel to total

physical activity in adults. Prev Med. 2012;55(3):206-11.

207. Reynolds R, McKenzie S, Allender S, Brown K, Foulkes C. Systematic review of

incidental physical activity community interventions. Prev Med. 2014;67:46-64.

208. Pratt M, Sarmiento OL, Montes F, Ogilvie D, Marcus BH, Perez LG, et al. The

implications of megatrends in information and communication technology and

transportation for changes in global physical activity. The Lancet. 2012;380(9838):282-

93.

209. Bassett DR, Jr., Pucher J, Buehler R, Thompson DL, Crouter SE. Walking,

cycling, and obesity rates in Europe, North America, and Australia. J Phys Act Health.

2008;5(6):795-814.

210. Sugiyama T, Ding D, Owen N. Commuting by car: weight gain among physically

active adults. Am J Prev Med. 2013;44(2):169-73.

186

211. Ding M, Gebel K, Phongsavan P, Bauman A, Merom D. Driving: a road to

unhealthy lifestyles and poor health outcomes. PLOS One. 2014;9(6).

212. Frank LD, Andresen MA, Schmid TL. Obesity relationships with community

design, physical activity, and time spent in cars. Am J Prev Med. 2004;27(2):87-96.

213. Jacobson SH, King DM, Yuan R. A note on the relationship between obesity and

driving. Trans Policy. 2011;18(5):772-6.

214. Saunders LE, Green JM, Petticrew MP, Steinbach R, Roberts H. What are the

health benefits of active travel? A systematic review of trials and cohort studies. PLoS

ONE. 2013;8(8).

215. Litman T. Evaluating active transport benefits and costs. Canada: Victoria

Transport Policy Institute, 2014.

216. Pate RR, Pratt M, Blair SN, Haskell WL, Macera CA, Bouchard C, et al. Physical

activity and public health: A recommendation from the Centers for Disease Control and

Prevention and the American College of Sports Medicine. J. Am. Med. Assoc..

1995;273(5):402-7.

217. Rutter H. Mortality benefits of cycling in London. London, England: Transport for

London, 2006.

218. Krag T. Cost Benefit Analysis of Cycling. Copenhagen: Nordic Council of

Ministers, 2007.

219. Sælensminde K. Cost–benefit analyses of walking and cycling track networks

taking into account insecurity, health effects and external costs of motorized traffic.

Transp Res Part A Policy Pract. 2004;38(8):593-606.

220. Kahlmeier S, Kelly P, Foster C, Gotschi T, Cavill N, Dinsdale H, et al. Health

economic assessment tools (HEAT) for walking and for cycling. Methodology and user

guide. Economic assessment of transport infrastructure and policies. 2014 update.

Copenhagen, Denmark: WHO Regional office for Europe, 2014.

221. Higgins J, Green S. Cochrane Handbook for Systematic Reviews: The Cochrane

Collaborative; 2011.

222. Husereau D, Drummond M, Petrou S, Carswell C, Moher D, Greenberg D, et al.

Consolidated health economic evaluation reporting standards (CHEERS)—explanation

187

and elaboration: a report of the ISPOR health economic evaluation publication

guidelines good reporting practices task force. Value Health. 2013;16(2):231-50.

223. AECOM. Inner Sydney regional bicycle network - demand assessment and

economic appraisal. Sydney: City of Sydney, 2010.

224. Beale SJ, Bending MW, Trueman P, Naidoo B. Should we invest in

environmental interventions to encourage physical activity in England? An economic

appraisal. Eur J Pub Health. 2012;22(6):869-73.

225. Buis J, Wittink R. The economic significance of cycling. The Hague: VNG, 2000.

226. Co S, Vautin D. Benefit-Cost Analysis of public health outcomes in long range

transportation planning in the San Francisco Bay Area. Transportation Research Board

93rd Annual meeting; Washington D.C.2014.

227. Cope A, Kennedy A, Ledbury M, Cambery R, Cavill N, Parkin J, et al. Cycling

demonstration towns - an economic evaluation. Association for European transport and

contributors, 2010.

228. COWI and the City of Copenhagen. Economic evaluation of cycle projects -

methodology and unit prices. Denmark: City of Copenhagen, n.d.

229. Deenihan G, Caulfield B. Estimating the health economic benefits of cycling. J

Transp Health. 2014;1(2):141-9.

230. Department for Transport. TAG Unit A5.1 Active Mode Appraisal. London:

Department for Transport, 2014.

231. Fishman E, Ker I, Garrard J, Litman T, Rissel C. Cost and Health Benefit of

Active Transport in Queensland. Queensland: Prepared by CATALYST for Health

Promotion Queensland, 2011.

232. Foltynova H, Kohlova MB. Cost-Benefit Analysis Of Cycling Infrastructure: A

Case Study Of PILSEN. Charles University Environment (Prague); 2002.

233. Gotschi T. Costs and Benefits of Bicycling Investments in Portland, Oregon. J

Phys Act Health. 2011;8:S49-S58.

234. Guo JY, Gandavarapu S. An economic evaluation of health-promotive built

environment changes. Prev Med. 2010;50 Suppl 1:S44-S9.

188

235. Li M, Faghri A. Cost-Benefit analysis of added cycling facilities. 93rd Annual

Meeting of the Transportation Research Board; Washington D.C.2014.

236. Lind G, Hyden C, Persson U. Benefits and costs of bicycle infrastructure in

Sweden. Copenhagen: Nordic Council of Ministers, 2005.

237. Meggs J, Schweizer J. Effects of bicycle facility provision on mortality prevention

and GHG reduction: Cost-Benefit Analyses within the BICY Project. n.d.

238. Macmillan A, Connor J, Witten K, Kearns R, Rees D, Woodward A. The societal

costs and benefits of commuter bicycling: simulating the effects of specific policies using

system dynamics modeling. Environ Health Perspect. 2014;122(4):335-44.

239. PricewaterhouseCoopers. Evaluation of the costs and benefits to the community

of financial investment in cycling programs and projects in New South Wales, Final

Report. Sydney: Roads and Traffic Authority of NSW and Department of Environment

and Climate Change, 2009.

240. Rabl A, De Nazelle A. Benefits of shift from car to active transport. Trans Policy.

2012;19(1):121-31.

241. Saari R, Metsaranta H, Tervonen J. Finnish guidelines for the assessment of

walking and cycling projects. Nordic Council of Ministers: 2007.

242. Schweizer J, Rupi F. Performance evaluation of extreme bicycle scenarios. In:

DeSousa JF, DeSousa JP, Costa A, Farias T, Melo S, editors. Transportation: can we

do more with less resources? - 16th Meeting of the Euro Working Group on

Transportation - Porto 2013. Procedia Social and Behavioral Sciences. 1112014. p.

508-17.

243. Sinclair Knight and PriceWaterhouseCoopers (SKM & PWC). Benefits of the

inclusion of active transport in infrastructure projects. Queensland, Australia: Transport

and Main Roads, 2011.

244. Sinnett D, Powell J. Economic evaluation of Living Streets' Fitter for Walking

project. Bristol: University of the West of England, 2012.

245. SQW Consulting. Valuing the benefits of cycling: A report to Cycling England.

UK: SQW, 2007.

246. SQW Consulting. Planning for Cycling: Report to Cycling England. UK: 2008.

189

247. Stokes RJ, MacDonald J, Ridgeway G. Estimating the effects of light rail transit

on health care costs. Health Place. 2008;14(1):45-58.

248. Sustrans Scotland. Sustrans Scotland: walking and cycling outcomes, Sustrans'

Scottish Government Grant 2012-2015: First Interim Report. Edinburgh, Scotland:

Sustrans Scotland, 2013.

249. Transport for Greater Manchester. Local Sustainable Transport Fund - Key

Component Bid, Greater Manchester Commuter Cycle Project. Manchester: Transport

for Greater Manchester, 2011.

250. Transport for London. A Business Case and Evaluation of the Impacts of Cycling

in London-Draft. London, UK: Cycling Centre of Excellence Surface Transport, TfL,

2004.

251. Wang G, Macera CA, Scudder-Soucie B, Schmid T, Pratt M, Buchner D. A cost-

benefit analysis of physical activity using bike/pedestrian trails. Health Promotion

Practice. 2005;6(2):174-9.

252. Wilson AM, Cope A. Value for money of walking and cycling interventions:

making the case for investment in active travel. UK: Sustrans Research and Monitoring

Unit, 2011.

253. Moodie M, Haby M, Galvin L, Swinburn B, Carter R. Cost-effectiveness of active

transport for primary school children - walking school bus program.

Int J Behav Nutr Phys Act. 2009;6.

254. Moodie M, Haby MM, Swinburn B, Carter R. Assessing Cost-Effectiveness in

Obesity: Active Transport Program for Primary School Children--TravelSMART Schools

Curriculum Program. J Phys Act Health. 2011;8(4):503-15.

255. Sinclair Knight and PriceWaterhouseCoopers. Benefits of the inclusion of active

transport in infrastructure projects. Queensland, Australia: Transport and Main Roads,

2011.

256. Warburton DER, Nicol CW, Bredin SSD. Health benefits of physical activity: the

evidence. CMAJ. 2006;174(6):801-9.

257. Genter JA, Donovan S, Petrenas B, Badland H. Valuing the health benefits of

active transport modes. 2008.

190

258. Nordic Council of Ministers. CBA of Cycling. Copenhagen: Nordic Council of

Ministers, 2005.

259. McCabe C, Claxton K, Culyer AJ. The NICE cost-effectiveness threshold: what it

is and what that means. Pharmacoeconomics. 2008;26(9):733-44.

260. Briggs AH, Weinstein MC, Fenwick EA, Karnon J, Sculpher MJ, Paltiel AD.

Model parameter estimation and uncertainty analysis a report of the ISPOR-SMDM

Modeling Good Research Practices Task Force Working Group–6. Med Decis.

2012;32(5):722-32.

261. Giles-Corti B, Donovan RJ. The relative influence of individual, social and

physical environment determinants of physical activity. Soc. Sci. Med..

2002;54(12):1793-812.

262. National Institute for Health and Clinical Excellence. Physical Activity and the

Environment: Guidance on the promotion and creation of physical environments that

support increased levels of physical activity. London: 2008.

263. Ogilvie D, Egan M, Hamilton V, Petticrew M. Promoting walking and cycling as

an alternative to using cars: systematic review. BMJ. 2004;329(7469):763-6B.

264. Sallis JF, editor Changing physical activity environments - how realistic and how

effective? International Conference on Obesity; 2014; Kuala Lumpur.

265. Oja P, Titze S, Bauman A, de Geus B, Krenn P, Reger-Nash B, et al. Health

benefits of cycling: a systematic review. Scand J Med Sci Sports. 2011;21(4):496-509.

266. Dalton AM, Jones AP, Panter JR, Ogilvie D. Neighbourhood, route and

workplace-related environmental characteristics predict adults' mode of travel to work.

PLoS ONE. 2013;8(6):e67575.

267. Andersen LB, Schnohr P, Schroll M, Hein HO. All-cause mortality associated

with physical activity during leisure time, work, sports, and cycling to work. Arch Intern

Med.. 2000;160(11):1621-8.

268. WHO Regional Office for Europe, editor Development of guidance and a

practical tool for economic assessment of health effects from walking. Consensus

workshop; 2010 1-2 July 2010; Oxford, United Kingdom: WHO Regional Office for

Europe.

191

269. Mindell JS, Watkins SJ, Cohen JM, editors. Health On The Move 2; Policies for

health-promoting transport. Stockport: Transport and Health Study Group; 2011.

270. Kahlmeier S, Racioppi F, Cavill N, Rutter H, Oja P. "Health in All Policies" in

Practice: Guidance and Tools to Quantifying the Health Effects of Cycling and Walking.

J Phys Act Health. 2010;7:S120-S5.

271. Krizek KJ, Handy SL, Forsyth A. Explaining changes in walking and bicycling

behavior: challenges for transportation research. Environ Plann B Plann Des.

2009;36(4):725.

272. Galea S, Riddle M, Kaplan GA. Causal thinking and complex system approaches

in epidemiology. Int. J. Epidemiol.. 2010;39(1):97-106.

273. Ogilvie D, Foster CE, Rothnie H, Cavill N, Hamilton V, Fitzsimons CF, et al.

Interventions to promote walking: systematic review. BMJ. 2007;334(7605):1204-7.

274. Yang L, Sahlqvist S, McMinn A, Griffin SJ, Ogilvie D. Interventions to promote

cycling: systematic review. BMJ. 2010;341.

275. Prest AR, Turvey R. Cost-benefit analysis: a survey. Econ. J. 1965;75(300):683-

735.

276. Dobes L, Argyrous G, Leung J. Social cost-benefit analysis in Australia and New

Zealand. The state of current practice and what needs to be done: ANU Press; 2016.

277. Campbell HF, Brown RPC. Benefit-cost analysis: financial and economic

appraisal using spreadsheets / Harry F. Campbell and Richard P.C. Brown. Cambridge:

Cambridge : Cambridge University Press; 2003.

278. Lange O. The foundations of welfare economics. Econometrica.

1942;10(3/4):215-28.

279. Litman T. Evaluating active transport benefits and costs: Victoria Transport Policy

Institute; 2017 [cited 2017 1 February]. Available from: http://www.vtpi.org/nmt-tdm.pdf.

280. Mackie P, Worsley T, Eliasson J. Transport appraisal revisited. Res Transp

Econ. 2014;47:3-18.

281. Australian Transport Assessment and Planning (ATAP). Parameter Values 2016

[cited 2017 16 March]. Available from: https://atap.gov.au/parameter-values/index.aspx.

192

282. PriceWaterhouseCoopers (PWC). Evaluation of the costs and benefits to the

community of financial investment in cycling programs and projects in New South

Wales, Final Report. Sydney: Roads and Traffic Authority of NSW and Department of

Environment and Climate Change, 2009.

283. Zapata-Diomedi B, Mantilla Herrera AM, Veerman JL. The effects of built


outcomes in Australia. Health Place. 2016;42:19-29.

284. Giles-Corti B, Timperio A, Cutt H, Pikora TJ, Bull FC, Knuiman M, et al.

Development of a reliable measure of walking within and outside the local

neighborhood: RESIDE’s Neighborhood Physical Activity Questionnaire. Prev Med.

2006;42.

285. Gold MR, Stevenson D, Fryback DG. HALYs and QALYs and DALYs, oh my:

similarities and differences in summary measures of population health. Annu Rev Public

Health. 2002;23(1):115-34.

286. Vos T, Carter R, Barendregt JJ, C. M, Veerman J, Magnus A, et al. Assessing

Cost-Effectiveness in Prevention (ACE Prevention): Final Report. University of

Queensland, Brisbane and Deakin University, Melbourne: 2010.

287. Australian Institute of Health and Welfare. Health expenditure Australia 2014–15.

Canberra: AIHW, 2016.

288. Ackerman F, Heinzerling L. Pricing the priceless: cost-benefit analysis of

environmental protection. University of Pennsylvania Law Review. 2002;150(5):1553-

84.

289. Johannesson M, Jönsson B. Economic evaluation in health care: Is there a role

for cost-benefit analysis? Health Policy. 1991;17(1):1-23.

290. Ryen L, Svensson M. The Willingness to Pay for a Quality Adjusted Life Year: A

Review of the Empirical Literature. Health Econ.. 2015;24(10):1289-301.

291. Mason H, Baker R, Donaldson D. Willingness to pay for a QALY: past, present

and future. Expert Rev Pharmacoecon Outcomes Res. 2008;8(6):575-82.

292. Mason H, Jones‐Lee M, Donaldson C. Modelling the monetary value of a QALY:

a new approach based on UK data. Health Econ.. 2009;18(8):933-50.

193

293. Access Economics. The health of nations: the value of statistical life. Australian

Goverment: Australian Safety and Compensation Council, 2008.

294. Banzhaf HS. Retrospectives: The Cold-War Origins of the Value of Statistical

Life. J Econ Perspect. 2014;28(4):213-26.

295. Hirth RA, Chernew ME, Miller E, Fendrick AM, Weissert WG. Willingness to pay

for a quality-adjusted life year: in search of a standard. Med Decis Making.

2000;20(3):332-42.

296. Office of Best Practice Regulation (OBPR). Best Practice Regulation Guidance

Note: Value of statistical life: Department of Prime Minister and Cabinet. Australian

Goverment; 2014 [cited 2015 5 May]. Available from:

https://www.dpmc.gov.au/sites/default/files/publications/Value_of_Statistical_Life_guida

nce_note.pdf.

297. Abelson P. Establishing a monetary value for lives saved: issues and

controversies Canberra: Office of Best Practice Regulation, 2008 2008-02.

298. Abelson P. The value of life and health for public policy. Economic Record.

2003;79(Special Issue):S2-S13.

299. Australian Transport Assessment and Planning (ATAP). Active Travel 2016 [cited

2017 5 January]. Available from: http://atap.gov.au/mode-specific-guidance/active-

travel/1-introduction.aspx.

300. PriceWaterhouseCoopers (PWC). Estimating the benefits of walking, a cost

benefit methodology. Australia: PWC, 2010.

301. Mytton O, Tainio M, Ogilvie D, Panter J, Cobiac LJ, Woodcock J. Impact of

increases in physical activity on disease burden, considering the effect of

increasing life expectancy and reduced incidence of disease: a multi-state lifetable

modelling study. Eur J Epidemiol. 2017;32(3):235-50.

302. Panter J, Ogilvie D. Can environmental improvement change the population

distribution of walking? J Epidemiol Community Health Forthcoming 2017. 2017.

303. Global report of urban health: equitable, healthier cities for sustainable

development [Internet]. 2016. Available from:

194

http://www.who.int/kobe_centre/measuring/urban-global-

report/ugr_full_report.pdf?ua=1.

304. Brown V, Zapata-Diomedi B, Moodie M, Veerman JL, Carter R. A systematic

review of economic analyses of active transport interventions that include physical

activity benefits. Transport Policy. 2016;45:190-208.

305. Tainio M, de Nazelle AJ, Götschi T, Kahlmeier S, Rojas-Rueda D,

Nieuwenhuijsen MJ, et al. Can air pollution negate the health benefits of cycling and

walking? Prev Med. 2016;87:233-6.

306. GBD Compare Data Visualization [Internet]. 2016 [cited 19 January 2017].

Available from: http://vizhub.healthdata.org/gbd-compare.

307. Environmental Issues: Waste Management, Transport and Motor Vehicle Usage

[Internet]. 2012 [cited 2016 15 August]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/mf/4602.0.55.002.

308. Brisbane Active Transport Strategy 2012-2026 [Internet]. 2012 [cited 2015 15

February]. Available from:

https://www.brisbane.qld.gov.au/sites/default/files/active_transport_strategy_2012-

2026.pdf.

309. Brisbane suburbs [Internet]. 2015 [cited 2015 25 February]. Available from:

http://www.brisbane.qld.gov.au/about-council/council-information-rates/brisbane-

suburbs.

310. Brisbane Community Profiles [Internet]. 2016 [cited 2016 25 February]. Available

from: http://www.brisbane.qld.gov.au/about-council/governance-strategy/business-

brisbane/business-opportunities/brisbane-community-profiles.

311. Population by Age and Sex, Regions of Australia [Internet]. 2014 [cited 2016 10

April]. Available from:

http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3235.02013?OpenDocument

.

312. 2009 South East Queensland Household Travel Survey [Internet]. Queensland

Goverment. 2009 [cited 2015 15 December]. Available from:

https://data.qld.gov.au/dataset/2009-south-east-queensland-household-travel-survey.

195

313. Department of Transport and Main Roads (TMR). South-East Queensland Travel

Survey 2009-Survey Procedures and Documentation. 2010;V2.2. Unpublished.

314. 2009-12 South East Queensland Household Travel Survey [Internet]. 2016 [cited

2016 15 November]. Available from: https://data.qld.gov.au/en/dataset/2009-12-south-

east-queensland-household-travel-survey/resource/2086c5cd-966e-49e2-ad53-

f2a9d789cbab.

315. 2901.0 - Census Dictionary, 2006 (Reissue) [Internet]. 2006 [cited 2016 22

December]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/0/413876F3BAE9CC70CA25720A000C428B

?opendocument.

316. Walking, Riding and Access to Public Transport-Suporting Active Travel in

Australian Communities [Internet]. 2013. Available from:

https://infrastructure.gov.au/infrastructure/pab/active_transport/files/infra1874_mcu_acti

ve_travel_report_final.pdf.

317. Queensland Household Travel Survey summary reports [Internet]. Queensland

Goverment. 2016 [cited 2016 16 December]. Available from:

http://www.tmr.qld.gov.au/Community-and-environment/Research-and-

education/Queensland-Household-Travel-Survey-summary-reports.aspx.

318. Deaths, Australia, 2015 [Internet]. Australia Bureau of Statistics. 2016 [cited 15

October 2016]. Available from: http://www.abs.gov.au/ausstats/[email protected]/mf/3302.0.

319. Estimated Resident Population By Single Year of Age Australia [Internet]. 2016

[cited 2016 10 October]. Available from:


ocument.

320. Australian Cancer Incidence and Mortality (ACIM) books [Internet]. 2016 [cited

2016 4 August]. Available from: http://www.aihw.gov.au/acim-books.

321. National Health Survey-Expanded Confidentialised Unit Record File 2011-2012

[Internet]. 2014 [cited 2015 22 September]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/Lookup/4324.0.55.001main+features12011-

12.

196

322. How much is spent on COPD? [Internet]. 2016 [cited 2016 12 October]. Available

from: http://www.aihw.gov.au/copd/expenditure/.

323. Australian Institute of Health and Welfare. Health expenditure Australia 2009–

2010. Canberra: 2011 Health and welfare expenditure series no. 46. Cat. no. HWE 55.


Canberra: 2015 Cat. no. HWE 63.

325. Global Burden of Disease Study 2015 (GBD 2015) Results [Internet]. 2016 [cited

2016 19 October]. Available from: http://ghdx.healthdata.org/gbd-results-tool.

326. Air Quality Monitoring [Internet]. Queensland Goverment, . 2015 [cited 2016 10

August]. Available from: https://data.qld.gov.au/dataset?q=air+quality.

327. Friend AJ, Ayoko GA, Stelcer E, Cohen D. Source apportionment of PM at two

receptor sites in Brisbane, Australia. Envir. Chem. 2011;8(6):569-80.

328. Crash data from Queensland roads [Internet]. Queensland Government 2014

[cited 2016 1 September]. Available from: https://data.qld.gov.au/dataset/crash-data-

from-queensland-roads.

329. Barendregt JJ. The effect size in uncertainty analysis. Value Health.

2010;13(4):388-91.

330. Australian Motor Vehicle Emission Inventory for the National Pollutant Inventory

(NPI) [Internet]. 2014. Available from: http://www.npi.gov.au/resource/australian-motor-

vehicle-emission-inventory-national-pollutant-inventory-npi.

331. Air Emissions Inventory-South East Queensland Region [Internet]. Queensland

Government. 2004. Available from:

http://s3.amazonaws.com/zanran_storage/www.epa.qld.gov.au/ContentPages/1552129

3.pdf.

332. Kristensson A, Johansson C, Westerholm R, Swietlicki E, Gidhagen L, Wideqvist

U, et al. Real-world traffic emission factors of gases and particles measured in a road

tunnel in Stockholm, Sweden. Atmos. Environ.. 2004;38(5):657-73.

333. Bicycle road rules and safety [Internet]. 2016 [cited 2016 28 September].

Available from: https://www.qld.gov.au/transport/safety/rules/wheeled-

devices/bicycle/#riding.

197

334. Maizlish N, Woodcock J, Co S, Ostro B, Fanai A, Fairley D. Health Cobenefits

and Transportation-Related Reductions in Greenhouse Gas Emissions in the San

Francisco Bay Area. Am. J. Public Health. 2013;103(4):703-9.

335. Xia T, Nitschke M, Zhang Y, Shah P, Crabb S, Hansen A. Traffic-related air

pollution and health co-benefits of alternative transport in Adelaide, South Australia.

Environ Int. 2015;74:281-90.

336. Rojas-Rueda D, de Nazelle A, Teixidó O, Nieuwenhuijsen MJ. Replacing car

trips by increasing bike and public transport in the greater Barcelona metropolitan area:

A health impact assessment study. Environ Int. 2012;49:100-9.

337. Briggs ADM, Wolstenholme J, Blakely T, Scarborough P. Choosing an

epidemiological model structure for the economic evaluation of non-communicable

disease public health interventions. Popul Health Metr. 2016;14(1):17.

338. Lovelace R, Goodman A, Aldred R, Berkoff N, Abbas A, Woodcock J. The

Propensity to Cycle Tool: An open source online system for sustainable transport

planning. 2016.

339. Aldred R, Woodcock J, Goodman A. Does More Cycling Mean More Diversity in

Cycling? Transport Rev.. 2015;36(1):28-44.

340. Aldred R, Elliott B, Woodcock J, Goodman A. Cycling provision separated from

motor traffic: a systematic review exploring whether stated preferences vary by gender

and age. Transport Rev.. 2016;37(1):29-55.

341. Santamariña-Rubio E, Pérez K, Olabarria M, Novoa AM. Gender differences in

road traffic injury rate using time travelled as a measure of exposure. Accid Anal Prev.

2014;65:1-7.

342. Yannis G. Modeling road accident injury under-reporting in Europe. European

transport research review. 2014;6(4):425-38.

343. Watson A, Watson B, Vallmuur K. Estimating under-reporting of road crash

injuries to police using multiple linked data collections. Accid Anal Prev. 2015;83:18-25.

344. Smith AD, Crippa A, Woodcock J, Brage S. Physical activity and incident type 2

diabetes mellitus: a systematic review and dose–response meta-analysis of prospective

cohort studies. Diabetologia. 2016:1-19.

198

345. World Health Organization. Global Ambient Air Pollution 2016 [cited 2016 18

October]. Available from: http://maps.who.int/airpollution/.

346. Australian Government. Improving the efficiency of new light vehicle. 2016.

347. Shaping SEQ-Draft South East Queensland Regional Plan [Internet].

Queensland Government. 2016. Available from:

http://www.dilgp.qld.gov.au/noindex/shapingseq/draft-south-east-queensland-regional-

plan.pdf.

348. Global Burden of Disease Study 2015 (GBD 2015) Results [Internet]. 2016.

349. Coffee NT, Lange J, Baker E. Visualising 30 years of population density change

in Australia’s major capital cities. Australian Geographer. 2016;47(4):511-25.

350. Trubka R, Newman P, Bilsborough D. The costs of urban sprawl - Infrastructure

and Transportation. 2010.

351. Australian Bureau of Statistics. Capital cities: past, present and future 2014 [cited

2017 28 February]. Available from:

http://www.abs.gov.au/AUSSTATS/[email protected]/featurearticlesbyCatalogue/AC53A071B4

B231A6CA257CAE000ECCE5?OpenDocument#PARALINK3.

352. Turrell G, Haynes M, Wilson L-A, Giles-Corti B. Can the built environment reduce

health inequalities? A study of neighbourhood socioeconomic disadvantage and walking

for transport. Health Place. 2013;19.

353. WHO Regional Office for Europe. Development of the Health Economic

Assessment Tools (HEAT) for walking and cycling. Consensus workshop: background

document. Copenhagen, Denmark: 2013.

354. Kahlmeier S, Kelly P, Foster C, Gotschi T, Cavill N, Dinsdale H, et al. Health

economic assessment tools (HEAT) for walking and for cycling. Methodology and user

guide. Economic assessment of transport infrastructure and policies. 2014 update.

Copenhagen, Denmark.: WHO Regional office for Europe; 2014 [cited 2014 7 July].

Available from:

http://www.euro.who.int/__data/assets/pdf_file/0010/256168/ECONOMIC-

ASSESSMENT-OF-TRANSPORT-INFRASTRUCTURE-AND-POLICIES.pdf.

199

355. Polder JJ, Barendregt JJ, van Oers H. Health care costs in the last year of life—

The Dutch experience. Soc Sci Med. 2006;63(7):1720-31.

356. The Lancet. Global Burden of Disease Study 2010 2012 [cited 2017 4 March].

Available from: http://thelancet.com/gbd/2010.

357. Harris P, Friel S, Wilson A. ‘Including health in systems responsible for urban

planning’: a realist policy analysis research programme. BMJ Open. 2015;5(7).

358. Harris P, Kent J, Sainsbury P. The mysterious disappearance of health from New

South Wales planning laws 2017 [cited 2017 6 March]. Available from:

https://theconversation.com/the-mysterious-disappearance-of-health-from-new-south-

wales-planning-laws-72098.

359. Carey G, Crammond B. Action on the social determinants of health: Views from

inside the policy process. Soc Sci Med. 2015;128:134-41.

360. Allender S, Cavill N, Parker M, Foster C. ‘Tell us something we don't already

know or do!’ — The response of planning and transport professionals to public health

guidance on the built environment and physical activity. J Public Health Policy.

2009;30(1):102-16.

361. National Heart Foundation of Australia. Heart Foundation Local Government

Awards N/D [cited 2017 15 August]. Available from:

https://www.heartfoundation.org.au/programs/heart-foundation-local-government-

awards.

362. Giles-Corti B, Eagleson S, Lowe M. Securing Australia's Future-Sustainable

Urban Mobility. 2014.

363. Mills C. Planning law and public health at an impasses in Australia: the need for

targeted law reforms to improve local food environments to reduce overweight and

obesity. Journal of Law and Medicine. 2014;22(1):179-87.

364. Harris P, Kent J, Sainsbury P, Thow AM. Framing health for land-use planning

legislation: A qualitative descriptive content analysis. Soc. Sci. Med.. 2016;148:42-51.

365. .

200

366. Australian Local Government Association. ROADS AND TRANSPORT

STRATEGY 2006-2016 N/D [cited 2017 15 August]. Available from:

http://alga.asn.au/?ID=240.

367. Australian Transport Council. National Guidelines for Transport System in

Management in Australia-Urban Transport. Canberra: Bureau of Transport and

Regional Economics, 2006.

368. Library of Congress. National Funding of Road Infrastructure: Australia 2015

[cited 2017 15 August]. Available from: https://www.loc.gov/law/help/infrastructure-

funding/australia.php.

369. Hooper P, Giles-Corti B, Knuiman M. Evaluating the implementation and active

living impacts of a state government planning policy designed to create walkable

neighborhoods in Perth, Western Australia. Am J Health Promot. 2014;28(3 Suppl):S5-

18.

370. Kent J, Harris P, Sainsbury P. A Healthy approach: how to turn what we know

about liveable cities into public policy 2015 [cited 2017 15 Aug]. Available from:

https://theconversation.com/a-healthy-approach-how-to-turn-what-we-know-about-

liveable-cities-into-public-policy-50185.

371. Australian Heart Foundation. Healthy Built Environmnets 2015 [cited 2015 6

January]. Available from: http://www.heartfoundation.org.au/active-living/healthy-built-

environments/Pages/default.aspx.

372. Obesity Policy Coalition. Action Areas N/D [cited 2017 15 August]. Available

from: http://www.opc.org.au/.

373. Veerman JL, Mackenbach JP, Barendregt JJ. Validity of predictions in health

impact assessment. J Epidemiol Community Health. 2007;61(4):362-6.

374. Australian Government. Vehicle Regulation 2015 [cited 2017 16 August].

Available from: https://infrastructure.gov.au/roads/vehicle_regulation/index.aspx.

375. Lal A, Mantilla-Herrera AM, Veerman L, Backholer K, Sacks G, Moodie M, et al.

Modelled health benefits of a sugar-sweetened beverage tax across different

socioeconomic groups in Australia: A cost-effectiveness and equity analysis. PLoS Med.

2017;14(6):e1002326.

201

Chapter 11 Appendix 1: Supplementary material in support of paper

in Chapter 4

Zapata-Diomedi, B & Veerman, JL 2016, 'The association between built environment



doi: 10.1186/s12889-016-3154-2

Appendices are available as online material accessible via this link:



202


in Chapter 5

Zapata-Diomedi, B, Herrera, AMM & Veerman, JL 2016, 'The effects of built

environment attributes on physical activity-related health and healthcare costs outcomes

in Australia', Health & Place, vol. 42, pp. 19-29.

doi: 10.1016/j.healthplace.2016.08.010




203


in Chapter 6

Brown, V, Zapata-Diomedi, B, Moodie, M, Veerman, JL & Carter, R 2016, 'A



doi: http://dx.doi.org/10.1016/j.tranpol.2015.10.003





204


in Chapter 7

205

Introduction

This document aims to support the main manuscript by presenting existing estimates for

monetised physical activity-related health benefits from the Australian literature. In

addition, we briefly discuss alternative population health measures to represent length

and quality of life. Lastly, we present a table that supplement the results in the main

manuscript.

206

Literature review

We searched Australian contemporary literature for monetised values for physical

activity (PA)-related health benefits that could be included in cost-benefit analysis

(CBAs) of built environment (BE) interventions. This search was initially conducted for a

report prepared for the Centre for Population Health of the New South Wales Ministry of

Health) (1).

Existing estimates of PA-related health benefits for the incorporation in CBAs of the BE

are limited to the transport sector. To our knowledge, there are no estimates for the

appraisal of urban planning initiatives more broadly (e.g. mixed land use, aesthetics,

safety, density). There are two approaches for monetising PA-health benefits in

transport appraisal (2). One option is to estimate the monetised value per person who

takes up PA as a result of the intervention. The other is to produce estimates on a per

kilometre basis. The per kilometre approach offers the advantage of its easy inclusion

into the stream of benefits in the economic evaluation of transport infrastructure, as

these are commonly performed on a per kilometre basis (2, 3).

In Table S1 we present a summary of the included studies. In six studies mortality and

morbidity outcome were quantified (2, 4-8). Estimates of the value of PA-related health

benefits per additional kilometre cycled ranged from less than A$0.02 to A$1.46 (2016)

(Figure S1). For walking, values ranged from A$1.79 to A$2.92 (2016) (Figure S2). The

range of values can be explained by differences in benefits included attributable to a

more active population (mortality and morbidity, healthcare costs and productivity gains)

and methods used for the quantification of those benefits. In the study by PWC (9) only

mortality was evaluated, which explains the comparatively low estimate. ATAP (2016)

and SMK & PWC (2011) adapted methods developed by Genter et al (3) for valuing

health benefits of active travel in New Zealand to the Australian context and for the

Australian state of Queensland respectively. Genter et al (3) quantified benefits from

reducing costs to the health system of physical inactivity related diseases as well as

benefits of prolonged life and avoided suffering measured in disability-adjusted life years

(DALYs). The population attributable fraction (PAF) was used to assign the proportion of

healthcare costs and DALYs to physical inactivity. DALYs were monetised using the

value of a statistical life year (VSLY) (10). Annual costs per person (healthcare costs +

monetised DALYs) were translated into per kilometre costs applying the concept of

207

“weighted per kilometre health benefit,” which is based on categorisation of physical

activity into inactive, insufficiently active and sufficiently active (3). This method assumes

that previously inactive or insufficiently active users of the new infrastructure become

physically active (in Australia 30 minutes of moderate PA at least 5 times per week) as a

consequence of the infrastructure (2). Given the robust evidence supporting a non-linear

dose-response relationship for PA and health outcomes, those in the inactive category

receive full benefits while the insufficiently active category receives 85% of the benefits

(3). For the sufficiently active group different approaches have been taken, for instance,

SKM & PWC (6) assumed that sufficiently active people receive no benefits while ATAP

(4) allocated a benefit of 15% as in the original research by Genter et al (3).

Mulley et al. (2), PWC (7, 9) and Transport for NSW (5) presented values based on past

literature. Mulley et al (2) recommended values for walking and cycling from SKM &

PWC (6) (middle value and range). Transport for NSW’s (5) value for walking is an

average of four studies and for cycling the upper value of a range was selected

(Transport for NSW 5 Tables 57 & 58 p 275-76). PWC’s (7) proposed value is based on

their own unpublished work. In the study by AECOM (8) the annual probability of

mortality was estimated using HEAT and monetised by applying the values of value of a

statistical life year (VSLY) recommended by Abelson (11). Morbidity outcomes in

AECOM (8) were derived from the cost of absenteeism due to sick days attributable to

physical inactivity. The AECOM value is in the lower range due to the use of the VSLY

over the VLS. This is a point of differentiation with the HEAT tool which recommends the

use of the VSL, whereas the justification by AECOM is that deaths rates represent the

probability of death in a given year, hence, the VSLY should be applied (8). However,

this is not correct, since one death leads to the loss of multiple life years, not a single life

year, hence, as proposed by the HEAT tool, the VSL should be used. Lastly, PWC (9)

provided two values. The first was based on healthcare costs of heart attack mortality

due to physical inactivity based on Road Traffic Authority estimates for 2003 (reference

not available). The second value recommended for sensitivity analysis was estimated

using HEAT.

There are several challenges with the per kilometre valuation. One is that information on

the baseline level of PA for the population exposed to the new or improved infrastructure

is often unavailable, hence assumptions are required. In the literature cited here, the

most common assumption is that local (SKM & PWC 6) or national (ATAP 4) PA

208

prevalence estimates apply. Further, values per kilometre do not account for the fact

that benefits differ by demographic and health characteristics (12). Another source of

uncertainty stems from the lack of information on the substitution effect of the

intervention (12). For instance, new users of the infrastructure may replace their leisure

time PA for walking and cycling for transport. Moreover, assumptions on the speed of

walking and cycling are required.

Values for walking are approximately twice those for cycling. This can be explained by

the greater efficiency of cycling compared to walking: faster speed (about 4 times) and

greater energy expenditure (about twice) (1/4*2=1/2) (3).

Given the lack of peer-review of most of these studies, the confidence we can have in

the size of the benefits of walking and cycling is limited. However, it is well accepted that

PA is associated with health benefits, and thus represents an economic value.

Table S1 Summary of studies measuring the economic value per kilometre walked/cycled

Source (year)

Physical Activity

Initiativesa Health Values per km (A$ 2016)b

ATAP (4) Walking and cycling for transport

Independent active travel projects

Outcomes: mortality and morbidity

Methods based on Genter et al (2008):

Healthcare costs attributable to physical inactivity.

The annual value of a DALY was calculated by dividing the VSL by the remaining average life expectancy calculated by subtracting the average age (37 years) from the average life expectancy in 2010 (82.1 years) to get 45.1 years.

VSL: Office of Best Practice Regulation (OBPR) (10) of $4.2 million (2014 values).

Walking: $2.92

Cycling: $1.46

Population:

Australians

≥18 years old

Transport for NSW (13)

Walking and cycling for transport

Not specified Outcomes: mortality and morbidity.

Methods: values are from the literature.

Walking: $1.79 ($0.47 to $2.52)

Cycling: $1.26 ($0.08 to $1.30) For walking an average from 4 values is calculated. For cycling a value is chosen and a range provided

SKM & PWC (6)


Active transport projects in combination with larger projects

Outcomes: mortality and morbidity.

Methods based on Genter et al. (2008)c:

Healthcare costs attributable to physical inactivity.

The annual value of a DALY was calculated by dividing the VSL by the remaining average

Walking: $2.03

($1.36 to 2.77)

Cycling: $1.35 ($0.91 to $1.52)

Population:

209

Source (year)

Physical Activity

Initiativesa Health Values per km (A$ 2016)b

life expectancy calculated by subtracting the average age (37.2 years) from the average life expectancy in 2010 (82.6 years) to get 45.4 years.

(DALYs for Queensland including cardiovascular diseases, cancer and type 2 diabetes).

VSL for main analysis: Office of Best Practice Regulation (OBPR) (10) of $4.2 million (2014 values).

Australians

≥14 years old

Linear increase over 5 years up to fully achieving benefits.

Mulley et al (2)


Active travel projects and transport projects with and active travel component

Outcomes: mortality and morbidity.

Methods: values are from the literature (PWC &

SKM 6).

Walking: $2.03

($1.36 to 2.77)

Cycling: $1.35 ($0.91 to $1.52)

PWC (7) Walking for transport and leisure

Not specified Outcomes: mortality and morbidity.

Methods: values are from unpublished workd.

Walking: $2.36

AECOM (2010)

Cycling for transport

Separated cycle ways

Outcomes: mortality and absenteeism.

Methods:

Mortality: reduction of probability of dying in one year calculated using WHO HEAT and monetised with the VSLY.

VSLY: Abelson (11).

Absenteeism: loss of working days attributable to physical inactivity.

Cycling: $0.26

PWC (9) Cycling for transport

Cycling missing links

Outcomes:

Main analysis: healthcare costs of mortality from heart attacks based on Road Traffic Authority own unpublished work (2003) recommended value.

Sensitivity analysis: all-cause mortality using

WHO HEATf.

Cycling: (1)

$0.01 and (2) $0.71

a Projects can be independent or part of an integral project. Integral projects such as a new public transit centre may

include pedestrian facilities, however, these were not explicitly addressed. b Indexed using the Consumer Price Index (CPI) for indirect costs (14) and the Health Price Index (15) when studies

presented full calculation allowing for us to convert each component of the per km value. CPI was applied when only a per km value was presented. c Conservative approach where sufficiently active individuals receive no health benefits. d Source document mentioned that the value is from a report prepared by PwC in 2010 titled “Evaluation of the costs and

benefits of investment in the Naremburn to Harbour Bridge Active Transport Corridor”. It is clarified that Genter et al. (2008) methods were used, hence, outcomes evaluated include mortality and morbidity. f The study does not show calculations and mentions that the WHO HEAT tool uses an indicative value of $0.60. No year

is specified, therefore assume 2008 as with RTA value.

210

Figure S1 Economic value per kilometre cycled (error bars are for the range recommended in source study)

Figure S2 Economic value per kilometre walked (error bars are for the range recommended in source study)

For the inclusion of these values in CBAs of transport interventions, estimates of

transport demand are needed. Australian guidance recommends six methods for active

211

transport: (1) comparison studies, (2) aggregate behaviour studies, (3) sketch planning

methods, (4) discrete choice models, (5) traditional demand models and (6)

Geographical Information Systems (GIS) based models (ATAP 4). The choice of

approach will depend on the complexity of the initiative, data availability, resources

available and required levels of accuracy. In Table S2 we present a summary of all six

methods based on ATAP (4). Methods 1, 3 and 4 are also recommended by the United

Kingdom Department of Transport for the appraisal of active transport interventions (16).

Accurate estimation of demand resulting from an active transport intervention is of

central relevance (12, 16). Demand estimates indicate the initiative’s effectiveness in

terms of changes in PA and therefore health.

Table S2 Methods for the estimation of demand for active transport generated by infrastructure developmenta

Methods Brief description

Comparison studies Based on a comparable intervention in an area with similar land use characteristics. Appropriate for small-scale interventions (e.g. path widening).

Aggregate behaviour studies Regression analysis based on population, land use and other determinants of active travel, including those modified by the intervention.

Sketch planning methods Simple calculations based on existing knowledge for mode share, trips’ length and other aspects of active travel. Estimates are based on data that may not be transferable to the area of interest. Appropriate for middle size interventions (e.g. off road share path).

Discrete choice models Prediction model for individual’s trip decisions based on several variables (e.g. sex, availability of facility, time differential between modes). Stated preferences survey data can be used to calibrate the models. An example is the demand forecast component of the study by Mulley et al (2).

Traditional demand models Four-step demand models used in the transport sector to predict travel patterns based on land use, transport network and travel behaviour characteristics. The steps are: (1) trip generation, (2) trips distribution, (3) mode choice, and (4) trip assignment (ATAP 17). The evaluation of active transport initiatives using this method requires zone systems and networks considerably smaller to those used in major transport projects for motorised transport.

Geographical Information Systems (GIS) based models

GIS methods used to forecast active travel demand.

a Source: Australian Transport Assessment and Planning (ATAP) (4), Australian Transport Assessment

and Planning (ATAP) (17)

HALYs, QALYs and DALYs

The measure for health outcomes used in this study, is the ‘health-adjusted life year’

(HALY). As ‘summary measure of population health’ it measures both quantity and

quality of life, where one HALY represent the equivalent of one year in full health (which

could be two years with a quality of life of 0.5, for example). Specific types of HALY are

212

the quality-adjusted life year (QALY) and the disability-adjusted life year (DALY). The

QALY derives from economics and was first used in the 1960s as a measure of health

gain (18). The DALY was developed for use in burden of disease studies as a measure

of health loss due to disease (18). Our calculated HALYs are neither QALYs not DALYs,

but something in between. They are similar to QALYs in that they represent health

gains. However, the main difference is in the calculation of the health-related quality of

life component. QALYs use measures of utility weights that traditionally represent

individual experiences of health, whereas our estimated HALYs use disability weights

linked to specific diseases, which were developed for the Global Burden of Disease

study (18). As discussed in past research (19, 20) the main advantage of using disability

weights over utility weights is that disability weights refer to specific diseases rather than

health states. Past research using similar methods to the ones applied for this study

used the term DALYs (21). Yet, we opted to use the more general terms HALYs given

that the use of the DALYs terminology may lead to think that our calculations are similar

to those in burden of diseases studies (22, 23). In our study, our model does not

explicitly separate years of life lost (YLL) and years lived with disability (YLD)

components, but instead calculates the total number of life years lived, adjusted for the

average health-related quality of life in those years (by age and sex). In burden of

disease studies, DALYs are defined as the sum Years of Life Lost (YLL) and Years

Lived with Disability (YLD). The main difference is related to the way the YLL

component is calculated in burden of disease studies relative to our use of HALYs. First,

the mortality rates we used are those expected for the population in question, not those

for a hypothetical population that has the lowest observed mortality at every age, as is

the case with the YLL component of DALYs. Second, in contrast to our HALYs, YLLs

used in the burden of disease studies are not adjusted for quality of life, and implicitly

assume all added life years are in full health. This leads to an over-estimation in health

gain when examining interventions that postpone death.

Complete results

In Table S3 below we present results that complement those presented in the main

manuscript in Figure 1.

213

Table S3 Complete outcomes mean (95% Uncertainty Intervals)

Scenarios/Outcomes

Minutes PA HALYs Healthcare costs savings

(A$ 2016)

All other healthcare costs in added Life

Years (A$ 2016)

Monetised HALYs (A$ 2016)

Total Economic Value (A$ 2016)

Density. + 1 SD (Christian et al. 2011)

0.79 (-1.72 to 3.32)

0.000054 (-0.000028 to 0.000176)

-$0.31 (-$1.14 to $0.16)

$0.47 (-$0.2 to $1.56)

$5.86 (-$2.87 to $19.14)

$5.7 (-$2.76 to $18.68)

Density. + 1 SD (Knuiman et al. 2014)

-0.19 (-0.88 to 0.54)

0.000002 (-0.000015 to 0.000039)

-$0.01 (-$0.23 to $0.08)

$0.02 (-$0.11 to $0.34)

$0.2 (-$1.52 to $4.23)

$0.19 (-$1.5 to $4.1)

Density. From 650 sqm2 to 9205

sqm2 average size of residential zone land within 1 km radius of residence. (Wilson et al. 2011)

1.75 (0.26 to 3.3)

0.000098 (0.000022 to 0.000177)

-$0.56 (-$1.23 to -$0.11)

$0.84 ($0.18 to $1.72)

$10.56 ($2.36 to $19.26)

$10.27 ($2.32 to $18.75)

Land use mix. + 1 SD (Christian et al. 2011)

0.85 (0.19 to 1.5)

0.000055 (0.000017 to 0.000093)

-$0.31 (-$0.67 to -$0.08)

$0.48 ($0.14 to $0.9)

$5.92 ($1.8 to $10.2)

$5.76 ($1.75 to $9.95)

Land use mix. + 1 SD (Knuiman et al. 2014)

1.2 (0.61 to 1.79)

0.000072 (0.00004 to 0.000109)

-$0.42 (-$0.78 to -$0.16)

$0.63 ($0.3 to $1.1)

$7.85 ($4.32 to $12.01)

$7.63 ($4.18 to $11.71)

Connectivity. + 1 SD (Christian et al. 2011)

0.85 (0.18 to 1.52)

0.000054 (0.000016 to 0.000093)

-$0.31 (-$0.65 to -$0.07)

$0.47 ($0.13 to $0.89)

$5.89 ($1.72 to $10.19)

$5.72 ($1.67 to $9.96)

Connectivity. Increase in 10

intersections (3-way or more) (Koohsari et al. 2014)

2.7 (2.58 to 2.83)

0.000141 (0.000107 to 0.000176)

-$0.81 (-$1.37 to -$0.37)

$1.22 ($0.73 to $1.83)

$15.29 ($11.33 to $19.47)

$14.88 ($10.91 to $19.01)

Connectivity. + 1 SD (Knuiman et al.

2014) 0.51

(0.04 to 0.99) 0.000036

(0.000004 to 0.000065) -$0.2

(-$0.45 to -$0.02) $0.31

($0.03 to $0.64) $3.87

($0.45 to $7.22) $3.76

($0.44 to $7.02)

Connectivity. From 4 to 51 four-way

intersections (Wilson et al 2011) 2.09

(0.63 to 3.73) 0.000113

(0.000041 to 0.000195) -$0.65

(-$1.35 to -$0.18) $0.98

($0.33 to $1.89) $12.28

($4.37 to $21.45) $11.95

($4.25 to $20.87)

Sidewalk. 10 km increase in sidewalk. (McComack et al. 2012)

4.22 (2.8 to 7.18)

0.000203 (0.00013 to 0.000335)

-$1.16 (-$2.25 to -$0.49)

$1.74 ($0.91 to $3.07)

$21.95 ($13.76 to $35.67)

$21.37 ($13.35 to $35.19)

Bikeways. From 0 km. to 7km. (mean highest quintile) of off road bikeways. (Wilson et al 2011)

1.65 (0.37 to 3.05)

0.000093 (0.000029 to 0.000167)

-$0.53 (-$1.18 to -$0.13)

$0.8 ($0.22 to $1.61)

$10.07 ($3.15 to $18.16)

$9.8 ($3.04 to $17.78)

Street lights. From 315 to 783 street lights. (Wilson et al 2011)

1.24 (-0.04 to 2.67)

0.000073 (-0.000001 to 0.000145)

-$0.42 (-$0.96 to $0.)

$0.63 ($0. to $1.37)

$7.91 (-$0.06 to $15.8)

$7.69 (-$0.05 to $15.43)

Destinations. + 1 transport destination (Giles-Corti et al. 2013)

5.8

0.000266 (0.000202 to 0.00033)

-$1.51 (-$2.54 to -$0.71)

$2.26 ($1.37 to $3.38)

$28.67 ($21.41 to $36.15)

$27.91 ($20.53 to $35.3)

214

Scenarios/Outcomes


(A$ 2016)


Years (A$ 2016)



Destinations. + 1 recreational destination (Giles-Corti et al. 2013) 17.6

0.00067

(0.000513 to 0.000828) -$3.79

(-$6.3 to -$1.8) $5.62

($3.43 to $8.33) $71.97

($54.06 to $90.2) $70.14

($52.02 to $88.14)

Destinations. From retail zone land within >1km to >0.2km. (Wilson et al. 2011)

2.16 (0.65 to 3.75)

0.000117 (0.000042 to 0.000196)

-$0.66 (-$1.35 to -$0.19)

$1 ($0.33 to $1.84)

$12.61 ($4.73 to $21.3)

$12.27 ($4.6 to $20.74)

Destinations. From park zone land within >1km to >0.2km. (Wilson et al. 2011)

0.4 (-0.92 to 1.85)

0.000032 (-0.000015 to 0.000107)

-$0.18 (-$0.68 to $0.08)

$0.28 (-$0.11 to $1.02)

$3.43 (-$1.54 to $11.82)

$3.34 (-$1.49 to $11.41)

Destinations. From =<3 to 4-7 (Knuiman et al. 2014) 0.29

(-0.9 to 1.56) 0.000026

(-0.000015 to 0.000094) -$0.15

(-$0.58 to $0.08) $0.23

(-$0.11 to $0.85) $2.81

(-$1.51 to $10.31) $2.73

(-$1.49 to $10.)

Destinations. From =<3 to 8-15 (Knuiman et al. 2014) 1.4

(-0.41 to 3.35) 0.00008

(-0.000007 to 0.000175) -$0.46

(-$1.19 to $0.04) $0.69

(-$0.05 to $1.65) $8.67

(-$0.71 to $18.97) $8.43

(-$0.7 to $18.41)

Bus stops. From 0-14 to 15-19 (Knuiman et al. 2014) 3.03

(1.66 to 4.46) 0.000155

(0.000084 to 0.000232) -$0.88

(-$1.65 to -$0.35) $1.33

($0.64 to $2.29) $16.72

($9.1 to $25.4) $16.27

($8.84 to $24.8)

Bus stops. From 0-14 to =>30

(Knuiman et al. 2014) 3.75

(1.83 to 5.6) 0.000184

(0.000101 to 0.000276) -$1.05

(-$1.97 to -$0.39) $1.58

($0.74 to $2.71) $19.91

($10.74 to $30.) $19.39

($10.44 to $29.28)

Train station. Railway station present within 1.6km compared to no rail way station. (Knuiman et al. 2014)

2.39 (-0.21 to 5.1)

0.000125 (-0.000003 to 0.000245)

-$0.71 (-$1.69 to $0.02)

$1.07 (-$0.02 to $2.34)

$13.47 (-$0.36 to $26.64)

$13.11 (-$0.35 to $25.92)

Transit stop. From one within >1km to one within >0.2km. (Wilson et al. 2011)

1.62 (-0.27 to 3.64)

0.000091 (-0.000004 to 0.00019)

-$0.52 (-$1.28 to $0.02)

$0.78 (-$0.03 to $1.73)

$9.82 (-$0.45 to $20.77)

$9.56 (-$0.44 to $20.25)

Walkability index. + 1SD (Christian at al. 2011)

1.28 (0.29 to 2.29)

0.000076 (0.000023 to 0.00013)

-$0.43 (-$0.9 to -$0.1)

$0.66 ($0.18 to $1.25)

$8.22 ($2.54 to $14.26)

$8 ($2.47 to $13.92)

Walkability index neighbourhood scale. High walkable compared to low. (Learnihan et al. 2011)

3.02 (1.24 to 4.93)

0.000154 (0.000069 to 0.000248)

-$0.88 (-$1.73 to -$0.28)

$1.32 ($0.52 to $2.38)

$16.63 ($7.33 to $27.04)

$16.18 ($7.13 to $26.39)

215

Scenarios/Outcomes


(A$ 2016)


Years (A$ 2016)



Walkability index CCD scale. High walkable compared to low. (Learnihan et al. 2011)

4.54 (2.8 to 6.28)

0.000217 (0.000134 to 0.000308)

-$1.23 (-$2.19 to -$0.52)

$1.85 ($0.98 to $3.01)

$23.37 ($14.46 to $33.41)

$22.75 ($13.95 to $32.65)

Walkability index 15 minutes area scale. High walkable compared to low. (Learnihan et al. 2011)

6.46 (4.59 to 8.18)

0.00029 (0.000197 to 0.000392)

-$1.65 (-$2.85 to -$0.74)

$2.47 ($1.41 to $3.88)

$31.27 ($21.04 to $42.37)

$30.46 ($20.41 to $41.32)

Walkability index. + 1SD (McCormack et al. 2012) 2.94

(1.62 to 8.38) 0.000148

(0.00008 to 0.000379) -$0.84

(-$2.24 to -$0.3) $1.26

($0.57 to $3.24) $15.94

($8.55 to $41.11) $15.52

($8.28 to $39.7)

Walkability index. High walkable compared to low (Owen et al. 2010)

0.87 (0.25 to 1.59)

0.000081 (0.000031 to 0.000144)

-$0.47 (-$0.95 to -$0.13)

$0.7 ($0.23 to $1.33)

$8.81 ($3.28 to $15.61)

$8.57 ($3.11 to $15.23)

a Walking refers to any walking (transport and recreation).

216

References

1. Zapata-Diomedi B, Brown V, Veerman J. An evidence review and modelling

exercise: The effects of urban form on health: costs and benefits. 2015.

2. Mulley C, Tyson R, McCue P, et al. Valuing active travel: Including the health

benefits of sustainable transport in transportation appraisal frameworks. Res Transp

Bus Manag. 2013;7:27-34.

3. Genter JA, Donovan S, Petrenas B, et al. Valuing the health benefits of active

transport modes. 2008.

4. Australian Transport Assessment and Planning (ATAP). Active Travel 2016 [cited

2017 5 Jan]. Available from: http://atap.gov.au/mode-specific-guidance/active-travel/1-

introduction.aspx.

5. Transport for New South Wales. Principles and guidelines for economic appraisal

of transport investment and initiatives Sydney: Transport for NSW; 2013 [cited 2015 3

March ]. Available from:

http://www.transport.nsw.gov.au/sites/default/files/b2b/publications/tfnsw-principles-and-

guidelines-for-economic-appraisal-of-transport-initiatives.pdf.

6. Sinclair Knight and PriceWaterhouseCoopers (SKM & PWC). Benefits of the

inclusion of active transport in infrastructure projects. Queensland, Australia: Transport

and Main Roads, 2011.

7. PriceWaterhouseCoopers (PWC). Estimating the benefits of walking, a cost

benefit methodology. Australia: PWC, 2010.

8. AECOM. Inner Sydney regional bicycle network - demand assessment and

economic appraisal. Sydney: City of Sydney, 2010.

9. PriceWaterhouseCoopers (PWC). Evaluation of the costs and benefits to the

community of financial investment in cycling programs and projects in New South

Wales, Final Report. Sydney: Roads and Traffic Authority of NSW and Department of

Environment and Climate Change, 2009.

10. Office of Best Practice Regulation (OBPR). Best Practice Regulation Guidance

Note: Value of statistical life: Department of Prime Minister and Cabinet. Australian

Goverment; 2014 [5 May 2015]. Available from:

217

https://www.dpmc.gov.au/sites/default/files/publications/Value_of_Statistical_Life_guida

nce_note.pdf.

11. Abelson P. Establishing a monetary value for lives saved: issues and

controversies Canberra: Office of Best Practice Regulation, 2008 2008-02.

12. van Wee B, Börjesson M. How to make CBA more suitable for evaluating cycling

policies. Transp Pol. 2015;44:117-24.

13. Transport for NSW. Principles and Guidelines for Economic Appraisal of

Transport Investment and Initiatives. Sydney, NSW: Transport for NSW,, 2013.

14. Australian Bureau of Statistics. 6401.0 - Consumer Price Index, Australia, Sep

2016. In: Australian Bureau of Statistics, editor. Consumer Price Index. Canberra,

Australia2016.


Canberra: AIHW, 2016.

16. Department for Transport. Transport analysis guidance: WebTAG. London:

Department for Transport, 2014.

17. Australian Transport Assessment and Planning (ATAP). Model design 2016

[cited 2017 10 Jan]. Available from: http://atap.gov.au/tools-techniques/travel-demand-

modelling/3-model.aspx.

18. Gold MR, Stevenson D, Fryback DG. HALYs and QALYs and DALYs, oh my:

similarities and differences in summary measures of population health. Annual review of

public health. 2002;23(1):115-34.

19. Roux L, Pratt M, Tengs TO. Cost effectiveness of community-based physical

activity interventions. Am J Prev Med. 2008;35.



21. Vos T, Carter R, Barendregt JJ, et al. Assessing Cost-Effectiveness in the

Prevention of Non-Communicable Disease (ACE–Prevention) Project 2005–09-

Economic Evaluation Protocol Brisbane, Australia2007 [cited 2015 13 Jan ]. Available

from: http://public-health.uq.edu.au/filething/get/1833/ACE-

P_Econ_Protocol_no_append.pdf.

218

22. Murray CJL. Understanding DALYs. Journal of health economics.

1997;16(6):703-30.

23. Murray CJL, Ezzati M, Flaxman AD, et al. GBD 2010: design, definitions, and

metrics. The Lancet. 2012;380(9859):2063-6

219


in Chapter 8

220

1 Introduction

In this document we expand on the methods description and results in the main

manuscript. Section two explains the methods and input parameters for the multi-cohort

proportional multi-state life table Markov Model (MSLT). In section three we present

results for the sensitivity analyses. Section four expands on results not presented in the

main manuscript. Lastly, we present the results of validity tests performed.

221

2 Proportional multi-state multi-cohort life table Markov model

A schematic description of the proportional multi-state life table is presented below

(Figure S1) only for the travel targets scenario population in the model (as it is derived

from the baseline population).

Figure S1 Schematic description of a proportional MSLT

This figure describes the interaction between life-table parameters and disease parameters. All the parameters are age specific denoted with x, i is incidence, p is prevalence and m is mortality, w is disability adjustment, q is probability of dying, l is number of survivors, L is life years, Lw is disability adjusted life years and HALE is health adjusted life expectancy, ‘-‘denotes parameter related to diseases or causes that specifically excludes modelled diseases or injuries and ‘+‘ relates to all modelled diseases in the model. A change in the determinant of health (physical activity and PM2.5) translates into changes in incidence (ix), which changes disease specific prevalence (px) and mortality (mx). For presentation purposes we only depict two diseases processes, however, in this study we modelled 7 diseases (ischemic heart disease, ischemic stroke, breast cancer, colon cancer, type 2 diabetes, COPD and tracheal, bronchus and lung cancer). Road fatalities (mi) impact directly on mortality and road injuries (ri) impact on years lived with disability, which are captured in Lw in the schematic description.

2.1 Changes in disease frequency

The seven included diseases are modelled applying a set of differential equations to

describe the transition between four states (healthy, diseased, dead from the disease

and dead from all other causes) (1) (Figure S2). Transition probabilities among the four

states are based on rates of incidence, remission, case fatality and background

mortality. A change in exposure to the risk factors of interest (physical activity and PM2.5)

m x - m x + q x l x L x e x w x - w x + Lw x HALE x

i x p

x m x

w

Disease process 1 Disease process 2

i x p

x m x

w

Risk factor

m x - m x + q x l x L x e x w x - w x + Lw x HALE x

i x p

x m x

w

Life table

Disease process 1 Disease process 2

i x p

x m x

w

mi

Risk factor

ri

Trach

eal,

bronc

hus

and

lung

cance

r

cance

rs

222

modifies incidence via the potential impact fraction (PIF) (Equation S1) or relative risk

(RR) (Equation S4) (Figure S2). To simplify the process, remission is set to zero.

Figure S2 Conceptual disease model

The model was used for each of the physical activity and PM2.5 related diseases. The disease conceptual model, applied to each disease separately, has four health states (healthy, diseased, dead from the disease and dead from other causes) and transition hazards between health states (1). PIF calculations for PA and RR for PM2.5 related diseases are explained in section 2.3. The diseases conceptual model is that of a multi-state life table model (box in figure).

2.2 Data sources

Data from the Global Burden of Disease 2013 (GBD 2013) (2), Australian Institute of

Health and Welfare (AIHW) (3) and Australian Bureau of Statistics (ABS) (4, 5) were

used to populate the MSLT model. We used DisMod II to enforce internal consistency in

the epidemiological estimates, and to derive parameters not provided in the data

sources (6). For the MSLT model we needed estimates for incidence and case fatality

rates for diseases causally associated with exposure to low levels of physical activity

(breast cancer, colon cancer, type 2 diabetes, ischemic heart disease and ischemic

stroke) and fine particles with a diameter of 2.5 μm or less (PM2.5) (chronic obstructive

pulmonary disease, tracheal, bronchus and lung cancer, ischemic heart diseases and

ischemic stroke). The conceptual model of DisMod II is that of a multi-state life table

(Figure S2): “Healthy people, defined as people unaffected by the disease being

modelled, are subject to an incidence hazard, and may become diseased. When

223

diseased they are subject to a hazard of dying from the disease, the case fatality, and to

a hazard of recovery from the disease, called remission. Both healthy and diseased

people are subject to the same mortality hazard from all other causes” (1, p2 ). DisMod

II generates age and sex specific and internally consistent estimates for the disease

modelled based on a set of differential equations. Population and mortality data were

obtained from the GBD 2013 and the Australian Bureau of Statistics (see Table S1). At

least three of the following input parameters are required for DisMod II to estimate

required input values: incidence, prevalence, remission, case fatality, duration, mortality

and relative risks on all-cause mortality (for the modelled disease). Cancers were

modelled with incidence and mortality from the AIHW. Remission was set to zero. For

the rest of the diseases (ischemic heart disease, ischemic stroke, type 2 diabetes and

COPD) we used prevalence and mortality from GBD 2013 and assumed a remission

hazard of zero (same as in GBD 2013). Prevalence and disease specific mortality

figures were obtained from the GBD 2013. Disease specific mortality from the GBD

2013 study are publicly available from the IHME webpage (7), whereas prevalence data

was requested from IHME. The AIHW provides data (numbers and rates) for incidence

and mortality for cancers for the year 2012 (3). In Table S1 we provide all data sources

used for DisMod II as well as expand on the procedures used to generate the data. In

the following section we provide more detail on the methods used for cancers.

224

Table S1 DisMod II input data and procedures for modelled diseases

Disease Collection input parameters

Dataset input parameters

ICD-10 codes Proceduresa

Breast Cancer (females)

Population: ABS

(5) Mortality rates:

ABS (4)

Incidence and Mortality: (AIHW 3)

AIHW: C50

Remission set to exact zero. Manual adjustment to incidence in older ages. Moving average smoothing to incidence. Higher weight to incidence and mortality.

Colon cancer

Population: ABS


ABS (4)


AIHW: C18

Cubic spline interpolation for incidence and mortality (inputs) for females and cubic spline interpolation for incidence for males. Higher weight to incidence and mortality.

Tracheal, bronchus and lung cancer cancers

Population: ABS


ABS (4)


AIHW: C33-C34

Fitted a sigmoid curve to incidence and interpolated mortality using cubic spline. Higher weight to mortality.


Population and mortality: GBD

2013** (IHME 2)

Prevalence and disease specific mortality: GBD

2013 (IHME 2)

GBD 2013: J40-J44.9,

J47-J47.9

Moving averages smoothing to prevalence and mortality (input). Fitted a sigmoid curve to incidence (output). Higher weight to prevalence and mortality.

Type 2 Diabetes


2013 (IHME 2)


2013 (IHME 2)

GBD 2013: E10-E10.11,

E10.3-E11.1, E11.3-E12.1, E12.3-E13.11, E13.3-E14.1, E14.3-E14.9, P70.0-P70.2, R73-R73.9, Z13.1, Z83.3

Manually adjusted prevalence for older ages (80+) as there was an inconsistent drop. Fitted a sigmoid curve to prevalence data after manual adjustment for females. For males manual adjustment of prevalence at older ages and smoothing of curve with moving averages



2013 (IHME 2)


2013 (IHME 2)

GBD 2013: I20, I21, I22,

I23, I24, I25

Cubic spline interpolation to prevalence and mortality. Fitted a sigmoid curve to case fatality (output).

Ischemic stroke


2013 (IHME 2)


2013 (IHME 2)

GBD 2013: G45, G46.8,

I63-I63.9, I64.0, I66.9, I67.2, I67.3, I67.5, I67.6, I69.3-I69.398

Cubic spline fitting to prevalence and mortality. Higher weight to prevalence and mortality.

a These procedures are based on options given by DisMod II.

2.2.1 Breast cancer, colon cancer and tracheal, bronchus and lung cancer

Input parameters from the AIHW for incidence and mortality were used. This decision

was made because the data for prevalence from the GBD 2013 study accounted for

prevalent cases for up to 10 years after becoming an incident case plus cases with long

life sequelae (supplementary material for reference 8).The GBD definition of prevalence

is different from that in our model, where we assume no remission, and ‘prevalence’

therefore includes all cases that ever experienced an incidence event in prior years. To

avoid underestimating we used Australian data from AIHW for incidence and mortality.

In the following graphs we present outputs for the modelled cancers, comparing

prevalence and incidence obtained with AIHW (incidence and mortality) and GBD 2013

(prevalence and mortality). These graphs show that prevalence generated only with

225

GBD data is considerable lower than that based on AIHW data. Under estimation of

prevalence forces DisMod II to generate lower estimates for incidence (for most ages).

Furthermore, estimates generated with GBD 2013 data result in unrealistic drops of

incidence in older ages followed by steep increases.

Figure S3 Breast cancer incidence and prevalence for females

Prevalence on the left vertical axis, incidence rates on the right.

226

Figure S4 Colon cancer prevalence and incidence females and males

Prevalence on the left vertical axis, incidence rates on the right. GBD 2013 data is for

Colon and rectum cancer.

Figure S5 Tracheal, bronchus and lung cancer

Prevalence on the left vertical axis, incidence rates on the right.

227

2.3 Disability weights

We derived disability weights (DW) from disease specific years lived with disability

(YLD) and disease specific prevalence by age group (5 years) and sex. Data for YLDs

were obtained from the online tool GBD Compare (7) and prevalence was requested

from the IHME. Our calculations are based on the GBD methods for estimating YLDs as

the sum of sequelae prevalence multiplied by sequelae disability weights (9). For this

study we had data at the cause level (e.g. ischemic heart disease) instead of squelae

level (e.g. myocardial infarction, angina and heart failure). An age and sex specific-

correction was introduced to couteract the effects of accumulating comorbid illnesses in

the older age groups.

DW adjusted for total YLDs = (YLDd/Pd)/(1-YLDt)

Where YLDd is the YLD mean number per age and sex for a given disease, Pd is the

prevalence (as reported in GBD 2013) for a given disease by age and sex and YLDt is

total YLD rate per age and sex. For the modelled cancers we further adjust the

calculated DWs by the ratio of our estimated prevalence and GBD reported prevalence.

This adjustment is due to the potential over estimation of disease specific YLDs in our

model given our higher estimate for prevalence. Hence in the base situation, our model

reproduces the YLD burden estimated by the GBD study.

2.4 Included risk factors

2.4.1 Physical activity

We estimated the prevalence of physical activity for the Brisbane adult population (aged

>20 years) from a representative sample of Australian adults. The data were collected

by the Australian Bureau of Statistics and published as the Basic-CURF National

Nutrition and Physical Activity Survey 2011-2012 (NNPAS) (10). A sample of 9,435

adults was available to derive physical activity prevalence estimates.

Survey respondents were asked about the time they spent doing physical activity in the

last week, including walking for transport, walking for recreation, moderate physical

activity (excludes walking) and vigorous physical activity. A walking session was only to

be recorded if the duration of the session was at least 10 minutes.

228

We computed Metabolic Equivalent of Tasks minutes per week (MET-mins/wk.) as

minutes per week spent in each of the above mentioned activities multiplied by MET

values used in the NNPAS (10) (walking min x 3.5, moderate min x 5 and vigorous min

x 7.5 (11)). A MET represents the ratio of the work metabolic rate to the resting

metabolic rate (11). For each observation in the dataset we derived a measure of total

MET-mins/wk. and grouped observations as per the categories by Danei et al. (12):

highly active (≥1,600 MET-minutes)/wk. and 1h/wk of vigorous PA), recommended level

active (600≤MET-minutes/wk. ≤1,600 and 1 h of vigorous PA/wk. or 2.5 h of moderate

PA/wk.), insufficiently active (0<MET-minutes/wk. ≤600 or <2.5 h/wk of moderate PA)

and inactive (no moderate or vigorous PA). Population weighted prevalence estimates

by age and sex are presented in Figures S6.

Figure S6 Prevalence physical activity

Mean energy expenditure in MET-minutes per week was also derived from the NNPAS

and used to fit relative risk (RR) functions for the included physical activity-related

diseases (Table S2). Relative risks used in this study for the association physical

activity-health outcomes are presented in Table S3. As discussed in the main

manuscript we used the “Relative Risk Shift” methods to modify incidence of physical

activity diseases (Equation S1).

229

Equation S1 Relative risk shift

The ‘relative risk shift’ method for the calculation of the PIF (13) was used to estimate new levels of incidence due to changes in physical activity, where 𝑝𝑖 is physical activity prevalence at level i (4 levels in

this research), 𝑅𝑅𝑖 is the relative risk of physical activity for each of the diseases associated with PA level i

and 𝑅𝑅𝑖′ is the relative risk of physical activity for each disease after the intervention.

Table S2 Mean (SE) MET-minutes/wk. per physical activity category Physical Activity categorya Mean (SE)

Inactive 0 (0) Insufficiently active 281.7 (4.3) Recommended level active 1024.7 (7.2) Highly active 3100.4 (57)

a Highly active (≥1,600 MET-minutes)/wk. and ≥1h/wk of vigorous PA), recommended level active (600 to <1,600 MET-minutes/wk. and either ≥1 h of vigorous PA/wk. or ≥2.5 h of moderate PA/wk.), insufficiently active <600 MET-minutes/wk. or <2.5 h/wk of moderate PA) and inactive (0 MET-minutes/wk of moderate or vigorous PA).

Table S3 Relative risks (SE log (RR)) of disease at different levels of physical activity Outcome (mortality)

Inactive Insufficiently Active

Recommended Level Active

Highly Active

Ischaemic heart diseasea

15-69 1.97 (0.12) 1.66 (0.19) 1.15 (0.05) 1

70-79 1.73 (0.12) 1.51 (0.21) 1.15 (0.07) 1

80+ 1.50 (0.14) 1.38 (0.24) 1.15 (0.10) 1

Ischaemic strokea

15-69 1.72 (0.23) 1.23 (0.56) 1.12 (0.30) 1

70-79 1.55 (0.24) 1.21 (0.62) 1.12 (0.37) 1

80+ 1.39 (0.28) 1.18 (0.83) 1.12 (0.59) 1

Type 2 diabetes

15-69 1.76 (0.10) 1.50 (0.26) 1.21 (0.12) 1

70-79 1.60 (0.11) 1.43 (0.30) 1.21 (0.16) 1

80+ 1.45 (0.14) 1.34 (0.39) 1.21 (0.25) 1

Breast cancer 15-44 1.56 (0.09) 1.41 (0.26) 1.25 (0.17) 1

45-69 1.67 (0.08) 1.41 (0.26) 1.25 (0.17) 1

70-79 1.56 (0.08) 1.36 (0.33) 1.25 (0.23) 1

80+ 1.45 (0.11) 1.32 (0.45) 1.25 (0.36) 1

Colon Cancer 15-69 1.80 (0.11) 1.27 (0.20) 1.07 (0.06) 1

70-79 1.59 (0.11) 1.21 (0.21) 1.07 (0.08) 1

80+ 1.39 (0.11) 1.16 (0.26) 1.07 (0.12) 1

a Relative risks of ischaemic heart disease and ischaemic stroke due to diabetes are 2.19 (1.81-2.66) and 2.64 (1.78-3.92) respectively (14). N.B. Values shown are the mean and SE

𝑃𝐼𝐹 =∑ 𝑝𝑖𝑅𝑅𝑖

𝑛𝑖=1 − ∑ 𝑝𝑖𝑅𝑅𝑖

′𝑛𝑖=1

∑ 𝑝𝑖𝑅𝑅𝑖𝑛𝑖=1

230

In past research (15), a linear assumption was made for the association of physical

activity with health outcomes. However, we assumed a curvilinear association with the

greatest gains at low levels of physical activity. In this section we demonstrate the more

realistic fit of a log-linear curve to the association of physical activity with health

outcomes. Figure S7 shows the estimated relative risks with the log-linear and linear

functions for the baseline and travel targets scenario as well as the prevalence of

physical activity per category. This example is for males in the age group 60-69 for

ischemic heart disease. We added the relative risks as reported in the literature for

reference and also estimates assuming a linear association. As can be observed,

estimated RRs with the log–linear function of a power transformation of MET-minutes

per week corresponds better to the RRs reported in the literature. The RRs for the

intervention show that for the linear assumption, risk decreases equally in all categories

with increasing PA (by approx. 2%), except for the highly active group which receives no

benefits (Table S4). In contrast, using the log-linear dose response function results in

greater reduction of risk at lower levels of PA (approx. 13% for inactive group, 4% for

insufficiently active and 2% for recommended level active).

Table S4 Relative risks comparison example

60-69 males Inactive Insufficiently Active

Recommended Level Active

Highly Active

RR IHD log-linear baseline 1.95 1.57 1.29 1.00 RR IHD log-linear intervention 1.70 1.51 1.26 1.00

Change in risk -13% -4% -2% 0%

RR IHD linear baseline 1.95 1.87 1.64 1.00 RR IHD linear intervention 1.92 1.83 1.61 1.00

Change in risk -2% -2% -2% 0%

Observed (Danaei et al. 2009) 1.97 1.66 1.15 1

231

Figure S7 Relative Risks fitting function example

232

2.4.1.1 Travel targets incremental physical activity

Additional walking and cycling minutes per week (Table 8-6 in the main manuscript)

were estimated by age and sex using data from the South East Queensland Travel

survey (16), as follows:

∆ 𝒘𝒂𝒍𝒌𝒊𝒏𝒈 𝒎𝒊𝒏𝒔. =∆ 𝒘𝒂𝒍𝒌𝒊𝒏𝒈 𝒕𝒓𝒊𝒑𝒔 𝒑𝒆𝒓 𝒅𝒂𝒚 ∗ 𝑴𝒆𝒂𝒏 𝒅𝒊𝒔𝒕𝒂𝒏𝒄𝒆 𝒕𝒓𝒊𝒑 ∗ 𝟓

𝑺𝒑𝒆𝒆𝒅 𝒑𝒆𝒓 𝒉𝒐𝒖𝒓 ∗ 𝟔𝟎

Equation S2 Additional minutes walking in travel targets scenario

The same equation applies for cycling. Speed per hour is the corresponding values for the MET values for walking and cycling from the physical activity compendium (11) (Table S5).

Table S5 Active transport speed

Baseline Sensitivity

km/h lowerc km/h upper

Walka 4.48 5.12

Bicycleb 16 19.04 a Walking for pleasure in the PA compendium (3.5 MET-minutes) b Bicycling, to/from work, self-selected pace; bicycling (6.8 MET-minutes) c Values in source are reported per miles

For additional walking of public transport trips (Table 6 in the main manuscript), the

following formula was applied, by age and sex:

∆𝒘𝒂𝒍𝒌𝒊𝒏𝒈 𝒎𝒊𝒏𝒔. 𝒐𝒇 𝑷𝑻 = ∆ 𝑷𝑻 𝒕𝒓𝒊𝒑𝒔 𝒑𝒆𝒓 𝒅𝒂𝒚 ∗ 𝑴𝒆𝒂𝒏 𝒘𝒂𝒍𝒌𝒊𝒏𝒈 𝒊𝒏 𝑷𝑻 𝒕𝒓𝒊𝒑𝒔 ∗ 𝟓

Equation S3 Additional minutes walking for public transport trips in travel targets scenario

2.4.2 PM2.5

With the help of the statistical software Stata (StataCorp, 2013.Stata Statistical

Software: Release 13. College Station, TX: StataCorp LP) we derived mean

background PM2.5 measured in each of the monitoring sites and estimated the arithmetic

mean of all sites with more than 75% of observation in a year (Table S6) (17). Our

overall estimated mean of 6.96 (SE 0.02) µg/m3 was decreased in the travel targets

scenario proportionally to the change in the contribution to ambient PM2.5 from

passenger cars and buses (Table 8-8 main manuscript).

233

Table S6 Level of data completeness (%) by monitoring site and estimated annual mean for background PM2.5 (µg/m3)

Year Rocklea South Brisbane Woolloongabba Wynnum North Lutwyche

% Mean % Mean % Mean % Mean % Mean

2006 90 6.23 N/A N/A N/A N/A N/A N/A N/A N/A 2007 93 6.77 N/A N/A N/A N/A N/A N/A N/A N/A 2008 85 5.61 N/A N/A 54 9.27 39 4.93 N/A N/A 2009 90 10.72 75 10.79 95 8.61 94 5.66 N/A N/A 2010 97 8.25 84 6.84 95 8.30 89 4.30 N/A N/A 2011 3 6.66 99 7.04 95 8.68 N/A N/A 67 10.42 2012 56 6.93 95 6.93 89 7.78 N/A N/A 46 5.51 2013 86 6.57 94 7.84 93 8.03 91 4.76 N/A N/A

N/A=not available

As described in the main manuscript, we incorporated the health effects of exposure to

PM2.5 via two mechanisms: the societal and individual effects. For both effects we used

the relative risks presented in Table S7.

Table S7 Relative Risks for exposure to PM2.5

Outcome (Mortality) RR Exposure change

Cardiovascular (WHO 18) a 1.10 (1.05, 1.15) 10 μg/m3

Respiratory (WHO 18)b 1.10 (0.98, 1.24) 10 μg/m3

Lung cancer (19) 1.09 (1.04, 1.14) 10 μg/m3 a Used for IHD and ischemic stroke. b Used for chronic obstructive pulmonary disease

The following formula was used to calculate the relative risks for the travel targets

scenario:

RRtravel targets = exp (ln(𝑅𝑅2.5) ∗ 𝐶ℎ𝑎𝑛𝑔𝑒 𝑖𝑛 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒

10)

Equation S4 RR travel targets

Change in exposure for the societal effect was estimated as the difference in the mean PM2.5

between baseline and travel targets scenario (Table 8-8 main manuscript).

A similar formula as with the societal effect was used for the calculation of the relative

risk, however, the calculaltion of the change in exposure differs. Following procedures

developed in past studies (18, 20, 21), we calculated the change in exposure as follows:

∆ 𝐢𝐧𝐝𝐢𝐯𝐢𝐝𝐮𝐚𝐥 𝐞𝐱𝐩𝐨𝐬𝐮𝐫𝐞 =((𝐓𝐨𝐭𝐚𝐥 𝐝𝐨𝐬𝐞 𝐭𝐫𝐚𝐯𝐞𝐥 𝐭𝐚𝐫𝐠𝐞𝐭𝐬/𝐓𝐨𝐭𝐚𝐥 𝐝𝐨𝐬𝐞 𝐛𝐚𝐬𝐞𝐥𝐢𝐧𝐞) − 𝟏)

∗ 𝒃𝒂𝒄𝒌𝒈𝒓𝒐𝒖𝒏𝒅 𝒆𝒙𝒑𝒐𝒔𝒖𝒓𝒆

Equation S5 Change in individual exposure to PM2.5

In the MSLT model we multiplied estimated relative risks for the travel targets scenario

by incidence of IHD, ischemic stroke, chronic obstructive pulmonary disease and

tracheal, bronchus and lung cancers.

234

Please note that in the original document by the WHO the term used is “equivalent

change”. Here we just use ‘change’ as we found this terminology confusing.

To estimate the total dose per week for the baseline (status quo) and travel targets

scenario we needed activity-specific information for: hours per week spent, ventilation

rates, concentration of PM2.5 and total inhaled dose of PM2.5 (Table S8 example for

males aged 17 to 49). We assumed that people sleep eight hours per day and spend

the rest of the day in activities (other activities) that have a ventilation rate equivalent to

resting and are exposed to average background PM2.5. The only difference between

scenarios is that car occupant time at baseline is replaced by active transport for the

travel targets scenario (hrs/wk. in activity in Table S8). Using ventilation rates per minute

and concentration of PM2.5 for each of the activites previously applied in a similar study

(21) we calculated the total inhaled dose per week for each activity. Sleep and resting

time have the equivalent to the background PM2.5 concentration, whereas for the rest of

the activities have higher concentration levels.

Table S8 Parameters and calculations for the change in exposure at the individual level for the travel targets scenario

Scenario Activity Hrs/wk. in

activity

Ventilation rates

(m3/hr)

Concentration PM2.5 (µg/m3)a

Inhaled dose PM2.5

(µg/m3)c

Total dose per week

Baseline Sleep 56 0.27 6.96b 105.20 Car

Occupant 0.75 0.61 8.35 3.84 Other

activities 111.25 0.61 6.96 472.14 581.18

Travel targets Sleep 56 0.27 6.96 105.20 Other

activities 111.25 0.61 6.96 472.14 Walk 0.22 1.37 7.65 2.36 Cycle 0.27 2.55 13.92 9.47 PT

(walking) 0.26 1.37 7.65 2.75 591.92 a Car occupant (1.2), walking (1.1) and bicycling (2) (21). b Note that this includes the reduction of background PM2.5 from achieving the travel targets. c Inhaled dose PM2.5=Hrs per week in activity*ventilation rates*Concentration

2.4.3 Road trauma

A simple distance based model was used to model the effect of road trauma of

achieving the travel targets scenario. Firstly, baseline injuries (hospitalisations and

medically treated) and fatalities were summarised by combinations of victim and striking

mode, using data from the Queensland Road Crash Database maintained by Transport

235

and Main Roads (22) (Table S9 and S10). The database provides information on the

unit counts for modes involved in a crash. However, there is no indication of striking and

victim modes involved. We therefore assumed than in a two modes crash, the largest

was the striking vehicle. When more than two modes were involved in a crash, the

largest was assumed the striking vehicle and the rest victims. Road crashes only

involving one mode are presented as “Single mode”. A fatality is a person who dies

within 30 days as a result of injuries sustained in a road traffic crash. A hospitalised

causality is a person who is transported to hospital as a result of a road traffic crash who

does not die from injuries sustained in the crash within 30 days of the crash (23). A

medically treated casualty is a person requiring medical treatment (but not

hospitalisation) as a result of a road traffic crash (23).

Table S9 Baseline road fatalities

Striking Single vehicle

Total

Victim Pedestrian Cyclist Motorcycle Car Bus Truck Other Victim mode

Pedestrian 0 0 0 2 1 0 0 0 3

Cyclist 0 0 0 1 0 0 1 1 3

Motorcycle 0 0 0 5 0 2 0 4 11

Car 0 0 0 0 0 3 0 8 11

Bus 0 0 0 0 0 0 0 0

Truck 0 0 0 0 0 1 0 1 2

Other 0 0 0 0 0 0 0 0 0

Total 0 0 0 8 1 6 1 14 30

Table S10 Baseline road injuries

Striking Single vehicle

Total

Victim Pedestrian Cyclist Motorcycle Car Bus Truck Other Victim mode

Pedestrian 0 2 17 216 16 12 0 0 263

Cyclist 0 2 2 186 4 9 1 17 221

Motorcycle 0 0 3 303 7 13 2 103 431

Car 0 0 0 2105 67 273 28 412 2885

Bus 0 0 0 0 4 0 0 19 23

Truck 0 0 0 0 0 4 3 23 30

Other 0 0 0 0 0 0 0 4 4

Total 0 4 22 2810 98 311 34 578 3857

We used an application of the Integrated Transport and Health Impact model (ITHIM) for

the calculation of baseline injury and fatality rates per pairwise combination of victim and

236

striking mode (24, 25). Calculating baseline rates per pairwise combination requires

data on the number of injuries/fatalities, person-kilometres travelled for the victim mode,

and vehicle-kilometres travelled for the striking mode per year (Equation S6). Person-

kilometres travelled were derived from the South East Queensland Household Travel

Survey (TMR 16) for pedestrians, bicyclists and cars (Table S11). For car occupants,

motorcycle/moped, bus, trucks and other modes, vehicle- kilometres travelled were

those reported in the Survey of Motor Vehicle Use (26). For pedestrians and bicyclist

person- kilometres travelled are the same as vehicle kilometres travelled. For

motorcycle/moped, vehicle-kilometres travelled were the same as person-kilometres

travelled. Number of injuries and fatalities for the travel targets scenario were estimated

using equation S7.

𝑅0 =𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑗𝑢𝑟𝑖𝑒𝑠𝑉𝑖𝑐𝑡𝑖𝑚

(𝑃𝐾𝑀𝑉𝑖𝑐𝑡𝑖𝑚 ∗ 𝑉𝐾𝑇𝑆𝑡𝑟𝑖𝑘𝑖𝑛𝑔 𝑣𝑒ℎ𝑖𝑐𝑙𝑒)0.5

Equation S6 Rate of injuries per pairwise combination of victim and striking modes

𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑗𝑢𝑟𝑖𝑒𝑠𝑉𝑖𝑐𝑡𝑖𝑚0= 𝑅0 ∗ (𝑃𝐾𝑀𝑉𝑖𝑐𝑡𝑖𝑚 ∗ 𝑉𝐾𝑇𝑆𝑡𝑟𝑖𝑘𝑖𝑛𝑔 𝑣𝑒ℎ𝑖𝑐𝑙𝑒)0.5

Equation S7 Number of injuries per pairwise combination of victim and striking modes

Table S11 Person and vehicle kilometres travelled per capita for baseline and travel targets scenario

Baseline Travel targets scenario

Mode Person Vehicles Person Vehicle

Pedestrian 108 108 197 197

Cyclist 69 69 157 157

Car occupant 10,050 8,200 9,555 7,795.57

Motorcycle 5,100 5,100 5,100 5,100

Bus 24,000 24,000 24,000 28,038

Truck 28,800 28,800 28,800 28,800

Othersa 36,400 36,400 36,400 36,400

a Light commercial vehicles

2.5 Healthcare costs and costs in added life years

In tables S12 and S13 we present healthcare costs and cost in added life years used in

the MSLT.

237

Table S12 Disease cost per case

Age and sex Ischemic

heart diseasea

Strokea Type 2

diabetesa Breast

Cancerb Colon

cancerb

Lung, tracheal

and bronchus cancerb

COPDa

Road traffic injuriesc

Male 24-34 $65,414 34-44 $22,776 44-55/>55 $9,007 $10,056 $613 - $20,354 $19,962 $2,064 $14,574 55–64 $8,309 $13,459 $974 - $23,030 $19,682 $2,064 $13,769 65–74 $7,243 $18,270 $1,175 - $23,348 $20,221 $2,064 $15,184 75+ $8,384 $24,047 $1,596 - $25,213 $19,210 $2,064 $37,576

Female 24-34 $47,890 34-44 $18,150 44-55/>55 $6,021 $6,848 $537 $16,365 $21,121 $23,489 $2,064 $13,236 55–64 $7,209 $9,225 $938 $15,064 $20,252 $21,124 $2,064 $12,219 65–74 $8,041 $11,664 $1,118 $16,223 $22,130 $22,725 $2,064 $16,337 75+ $8,052 $33,981 $1,596 $17,336 $22,578 $21,222 $2,064 $30,913

a Cost per prevalent case of disease. b Cost per incident case of disease. c Road traffic injuries costs are per prevalent year lived with disability. N.B. Costs are in Australian dollars, from the Disease Costs and Impact Study 2001 prepared by the Australian Institute of Health and Welfare and indexed to the year 2013 using the health price index (27) for all diseases except COPD. COPD costs are estimates from the AIHW for 2008-2009 without disaggregation by age and sex (28). Incidence, prevalence and YLDs are for 2000 and 2010 (COPD) from the GBD 2015 study (29).

Table S13 Costs for all other diseases in added life years

Age and sex

Cost of all other diseases

Male 15–24 $1,792 25–34 $1,750 35–44 $1,925 45–54 $2,474 55–64 $3,867 65–74 $6,702 75-84 $10,758 85+ $17,537

Female 15–24 $2,429 25–34 $2,984 35–44 $2,698 45–54 $3,186 55–64 $4,365 65–74 $6,826 75-84 $10,974 85+ $20,563

N.B. Costs are in Australian dollars, from the Disease Costs and Impact Study 2001 prepared by the Australian Institute of Health and Welfare and indexed to the year 2013 using the health price index (27). Includes overall healthcare costs per person minus costs of diseases and injuries included in the model (Table S12).

238

3 Sensitivity analysis

Table S14 is a summary of sensitivity analyses described in the main manuscript.

Table S14 Univariate sensitivity analysis parameters Parameter Base case Sensitivity

Baseline and travel targets scenario

Discount rate health outcomes and healthcare costs

0% health and 3% healthcare costs per annum/3% health and 3% healthcare costs per annum

3% health and 5% healthcare costs per annum (30, 31)

Travel targets scenario

Time taken per km walked and cycleda

Walking 4.48 km/h, cycling 16 km/h Walking 5.12km/h, cycling 19.04km/h (upper limits from PA compendium (11)

PM2.5 source apportionment

18% 7% and 30% (32)

Passenger vehicles and buses contribution to traffic related PM2.5

28% and 10% 65% passenger vehicles

Bicycling exposure to PM2.5

2 relative to ambient PM2.5 1.1 relative to ambient PM2.5 (assumption, same as walking)

Road trauma-km travelled

Non-linear Linear

a Distances per km walked and cycled are based on the PA compendium’s lower value for the corresponding MET-rates. MET are the energy expenditure of an activity compared to resting.

4 Additional results to the main manuscript

In Table S15 we present results for the sensitivity scenarios discussed in the main

manuscript.

Table S15 Healthcare costs and health outcomes for sensitivity scenarios (95% uncertainty interval)

Health adjusted

life years (thousand)

Life years (thousand)

Healthcare costs total (millions)a

Other healthcare costs in added Lys

total (millions)

1. Walking speed 5.12km/h and cycling 19.4km/ha 30.3

(17.7 to 44.2) 26.3

(11.6 to 41.8) -$294

(-$442 to -$159) $122

($43 to $204)

2. PM2.5 source apportionment high

32.8 (19.8 to 47.0)

28.3 (13.3 to 44.3)

-$313 (-$464 to -$174)

$130 ($49 to $214)

3. PM2.5 source apportionment low

32.5 (19.5 to 46.7)

28.1 (13.1 to 43.9)

-$312 (-$463 to -$173)

$129 ($48 to $213)

4. Passenger vehicles PM2.5

attribution highc 33.0

(20.1 to 47.1) 28.5

(13.5 to 44.5) -$315

(-$465 to -$176) $131

($50 to $215)

5. Cycling exposure to PM2.5 similar to walking

32.7 (19.8 to 46.9)

28.3 (13.3 to 44.2)

-$313 (-$463 to -$174)

$129 ($49 to $213)

6. Linear association road trauma

21.6 (8.6 to 35.8)

19.0 (4.1 to 34.9)

-$264 (-$415 to -$125)

$093 ($12 to $177)

7. Discount health 3% 10.0 (6.0 to 14.3)

7.8 (3.6 to 12.3)

-$312 (-$463 to -$173)

$43 ($14 to $74)

8. Discount healthcare costs 5% 32.6 (19.6 to 46.8)

28.1 (13.1 to 44.0)

-$175 (-$257 to -$099)

$61 ($21 to $103)

9. Discount health 3% and healthcare costs 3%

10.0 (6.0 to 14.3)

7.8 (3.6 to 12.3)

-$312 (-$463 to -$173)

$43 ($14 to $74)

a Negative values are savings. b Base case: Walking speed of 4.48 km/h and cycling speed of 16 km/h. c Base case 28% and sensitivity scenario 65%.

239

Intermediate outputs used to modify incidence of physical activity and PM2.5 related

diseases are presented in tables S16 and S17.

Table S16 Population impact fraction for PA-related diseases for travel targets scenario compared to baseline by age and sex used to modify incidence rates in MSLTa

Age and sex IHD Ischemic

stroke Type 2

diabetes Colon cancer

Breast cancer

20-25, male 0.04 0.05 0.04 0.07

25-29, male 0.05 0.06 0.05 0.08

30-34, male 0.05 0.07 0.05 0.08

35-39, male 0.06 0.07 0.05 0.09

40-44, male 0.06 0.07 0.05 0.09

45-49, male 0.06 0.08 0.06 0.10

50-54, male 0.05 0.07 0.04 0.08

55-59, male 0.05 0.07 0.04 0.09

60-64, male 0.04 0.06 0.04 0.08

65-69, male 0.05 0.07 0.04 0.09

70-74, male 0.04 0.07 0.04 0.08

75-79, male 0.02 0.05 0.02 0.06

80-84, male 0.02 0.04 0.02 0.05

85-89, male 0.02 0.05 0.02 0.05

90-94, male 0.02 0.05 0.02 0.05

95-100, male 0.02 0.05 0.02 0.05

20-25, female 0.04 0.04 0.03 0.05 0.03

25-29, female 0.04 0.05 0.04 0.06 0.03

30-34, female 0.04 0.05 0.04 0.07 0.03

35-39, female 0.04 0.06 0.04 0.07 0.03

40-44, female 0.05 0.07 0.05 0.08 0.03

45-49, female 0.05 0.07 0.05 0.09 0.04

50-54, female 0.04 0.06 0.03 0.07 0.03

55-59, female 0.03 0.05 0.03 0.07 0.03

60-64, female 0.04 0.06 0.03 0.08 0.03

65-69, female 0.03 0.05 0.03 0.07 0.02

70-74, female 0.03 0.06 0.03 0.07 0.03

75-79, female 0.03 0.06 0.03 0.08 0.03

80-84, female 0.02 0.05 0.02 0.06 0.02

85-89, female 0.03 0.06 0.03 0.07 0.03

90-94, female 0.03 0.06 0.03 0.07 0.03

95-100, female 0.03 0.06 0.03 0.07 0.03 a Only for baseline scenario. Incidence was modified by 1-PIF.

240

Table S17 Relative risks for PM2.5-related diseases for travel targets scenario compared to baseline by age and sex used to modify incidence rates in MSLT

Age and sex COPD soc.

effecta

TBL cancersb

soc. effect.

IHD soc.

effect

Ischemic stroke soc.

effect

COPD ind.

Effectc

TBL cancers

ind. effect.

IHD ind.

effect

Ischemic stroke

ind. effect

20-25, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

25-29, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

30-34, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

35-39, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

40-44, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

45-49, male 0.9999 0.9999 0.9999 0.9999 1.0012 1.0011 1.0012 1.0012

50-54, male 0.9999 0.9999 0.9999 0.9999 1.0008 1.0007 1.0008 1.0008

55-59, male 0.9999 0.9999 0.9999 0.9999 1.0008 1.0007 1.0008 1.0008

60-64, male 0.9999 0.9999 0.9999 0.9999 1.0008 1.0007 1.0008 1.0008

65-69, male 0.9999 0.9999 0.9999 0.9999 1.0008 1.0007 1.0008 1.0008

70-74, male 0.9999 0.9999 0.9999 0.9999 1.0008 1.0007 1.0008 1.0008

75-79, male 0.9999 0.9999 0.9999 0.9999 1.0001 1.0001 1.0001 1.0001

80-84, male 0.9999 0.9999 0.9999 0.9999 1.0001 1.0001 1.0001 1.0001

85-89, male 0.9999 0.9999 0.9999 0.9999 1.0001 1.0001 1.0001 1.0001

90-94, male 0.9999 0.9999 0.9999 0.9999 1.0001 1.0001 1.0001 1.0001

95-100, male 0.9999 0.9999 0.9999 0.9999 1.0001 1.0001 1.0001 1.0001

20-25, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

25-29, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

30-34, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

35-39, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

40-44, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

45-49, female 0.9999 0.9999 0.9999 0.9999 1.0007 1.0006 1.0007 1.0007

50-54, female 0.9999 0.9999 0.9999 0.9999 1.0003 1.0003 1.0003 1.0003

55-59, female 0.9999 0.9999 0.9999 0.9999 1.0003 1.0003 1.0003 1.0003

60-64, female 0.9999 0.9999 0.9999 0.9999 1.0003 1.0003 1.0003 1.0003

65-69, female 0.9999 0.9999 0.9999 0.9999 1.0003 1.0003 1.0003 1.0003

70-74, female 0.9999 0.9999 0.9999 0.9999 1.0003 1.0003 1.0003 1.0003

75-79, female 0.9999 0.9999 0.9999 0.9999 1.0002 1.0003 1.0002 1.0003

80-84, female 0.9999 0.9999 0.9999 0.9999 1.0002 1.0003 1.0002 1.0003

85-89, female 0.9999 0.9999 0.9999 0.9999 1.0002 1.0003 1.0002 1.0003

90-94, female 0.9999 0.9999 0.9999 0.9999 1.0002 1.0003 1.0002 1.0003

95-100, female 0.9999 0.9999 0.9999 0.9999 1.0002 1.0003 1.0002 1.0003 a Societal effect. b Tracheal, bronchus and lung cancer cancers TBL. c Individual effect.

241

5 Validation tests

We conducted a number of tests to check for any inconsistencies in the MSLT (Table

S15). Excluding all risk factors results in zero changes. Applying an extreme discount

rate of 20% to health and healthcare costs significantly reduces the outcomes.

Table S18Model testsa HALYs LYs Savings in healthcare

costs Healthcare costs of

added LYs

Complete model 33,455 28,723 -$ 319,094,510 $ 131,530,927

Exclude all risk factors - - - -

Discount healthcare costs at 20%

36,725 32,447 -$ 11,387,186 $ 1,898,921

Discount health at 20% 166 105 -$ 319,094,510 $ 1,237,623

a Results from deterministic analysis.

242

References

1. Barendregt JJ, Oortmarssen vGJ, Murray CJ, et al. A generic model for the

assessment of disease epidemiology: the computational basis of DisMod II. Popul

Health Metr. 2003;1(1):4-.

2. Global Burden of Disease (GBD) [Internet]. 2015 [cited 25 July 2016]. Available

from: http://www.healthdata.org/gbd.

3. Australian Cancer Incidence and Mortality (ACIM) books [Internet]. 2016 [cited 4

August 2016]. Available from: http://www.aihw.gov.au/acim-books.

4. Deaths, Australia, 2015 [Internet]. Australia Bureau of Statistics. 2016 [cited 15

October 2016]. Available from: http://www.abs.gov.au/ausstats/[email protected]/mf/3302.0.

5. Estimated Resident Population By Single Year of Age Australia [Internet]. 2016

[cited 10 October 2016]. Available from:


ocument.

6. Barendregt JJ. EpiGear International 2012 [cited 2015 1 Mar]. Available from:

http://www.epigear.com/index_files/prevent.html.

7. GBD Compare [Internet]. IHME, University of Washington. 2015 [cited 15 July

2016]. Available from: https://vizhub.healthdata.org/gbd-compare/.

8. GBD 2013 Risk Factors Collaborators. Global, regional, and national

comparative risk assessment of 79 behavioural, environmental and occupational, and

metabolic risks or clusters of risks in 188 countries, 1990-2013: A systematic analysis

for the Global Burden of Disease Study 2013. The Lancet. 2015;386(10010):2287-323.

9. Global Burden of Diseases Study 2013 Collaborators. Global, regional, and

national incidence, prevalence, and years lived with disability for 301 acute and chronic

diseases and injuries in 188 countries, 1990–2013: a systematic analysis for the Global

Burden of Disease Study 2013. The Lancet. 2015;386(9995):743-800.

10. Australian Health Survey: Physical Activity, 2011-12 [Internet]. 2015 [cited 22

September 2015]. Available from:

http://www.abs.gov.au/ausstats/[email protected]/Lookup/D4495467B7F7EB01CA257BAC001

5F593?opendocument.

11. Ainsworth BE, Haskell WL, Herrmann SD, et al. 2011 compendium of physical

activities: a second update of codes and MET values. Med Sci Sports Exerc.

2011;43(8):1575-81.

243

12. Danaei G, Ding EL, Mozaffarian D, et al. The preventable causes of death in the

United States: comparative risk assessment of dietary, lifestyle, and metabolic risk

factors. PLoS Med. 2009;6(4):e1000058.

13. Barendregt JJ, Veerman JL. Categorical versus continuous risk factors and the

calculation of potential impact fractions. J Epidemiol Community Health.

2010;64(3):209-12.

14. Asia Pacific Cohort Studies Collaboration. The Effects of Diabetes on the Risks

of Major Cardiovascular Diseases and Death in the Asia-Pacific Region. Diabetes Care.

2003;26(2):360-6.



16. 2009 South East Queensland Household Travel Survey [Internet]. Queensland

Goverment, . 2009 [cited 15 December 2015]. Available from:

https://data.qld.gov.au/dataset/2009-south-east-queensland-household-travel-survey.

17. Air Quality Monitoring [Internet]. Queensland Goverment, . 2015 [cited 10 August

2016]. Available from: https://data.qld.gov.au/dataset?q=air+quality.

18. World Health Organization. WHO Expert Meeting: Methods and tools for

assessing the health risks of air pollution at local, national and international level.

Copenhagen: WHO Regional Office for Europe, 2014.

19. Hamra GB, Guha N, Cohen A, et al. Outdoor particulate matter exposure and

lung cancer: a systematic review and meta-analysis. Environ Health Perspect.

2014;122:906-11.

20. Rojas-Rueda D, de Nazelle A, Teixidó O, et al. Replacing car trips by increasing

bike and public transport in the greater Barcelona metropolitan area: A health impact

assessment study. Environment International. 2012;49:100-9.

21. Tainio M, de Nazelle AJ, Götschi T, et al. Can air pollution negate the health

benefits of cycling and walking? Preventive Medicine. 2016;87:233-6.

22. Crash data from Queensland roads [Internet]. Queensland Goverment 2014

[cited 1 September 2016]. Available from: https://data.qld.gov.au/dataset/crash-data-

from-queensland-roads.

23. Department of Transport and Main Roads (TMR). Fatal Road Traffic Crashes in

Queensland 2011. 2012.

244

24. Woodcock J, Givoni M, Morgan AS. Health impact modelling of active travel

visions for England and Wales using an Integrated Transport and Health Impact

Modelling Tool (ITHIM). PLoS One. 2013;8(1):e51462.

25. Maizlish N, Woodcock J, Co S, et al. Health Cobenefits and Transportation-

Related Reductions in Greenhouse Gas Emissions in the San Francisco Bay Area.

American Journal of Public Health. 2013;103(4):703-9.

26. Australian Bureau of Statistics. Survey of Motor Vehicle Use, Australia, 12

months ended 31 October 2010. Canberra2011.

27. Australian Institute of Health and Welfare. Disease costs and impact study data.

Australian Institute of Health and Welfare, 2001.

28. How much is spent on COPD? [Internet]. 2016 [cited 12 October 2016]. Available

from: http://www.aihw.gov.au/copd/expenditure/.

29. Global Burden of Disease Study 2015 (GBD 2015) Results [Internet]. 2016 [cited

19 October 2016]. Available from: http://ghdx.healthdata.org/gbd-results-tool.

30. Gold MR. Cost-effectiveness in health and medicine. New York: Oxford


31. Tan-Torres Edejer T. Making choices in health: WHO guide to cost-effectiveness

analysis. Geneva: World Health Organization; 2003.

32. Friend AJ, Ayoko GA, Stelcer E, et al. Source apportionment of PM at two

receptor sites in Brisbane, Australia. Environmental Chemistry. 2011;8(6):569-80.