Pinnington.ppt - Actuaries

3/19/2009

1

Understanding the Interactions Between Causes of Death – Mortality yProjections by Cause Research GroupAdrian Pinington, TSAP ConsultingScott Reid, Aon Benfield

23rd March 2009

Outline

IntroductionScenarios – flat and steep improvement trendsHeat ChartsHeat ChartsScenario funnel against Lee-Carter PercentilesCaveats and ConclusionsWhere Next

Mortality Projection by Cause Research GroupMembers

Harry Palmes Garth LaneAdrian Pinington Niel Daniels

Scott ReidRichard Morris Neil RobjohnsInput from others:

Dr R Croxson

Dr R Wyse

Dr A Luczak

Dr J Wilden

3/19/2009

2

Mortality Projections “By Cause” Research GroupTerms of Reference – Our Aims

To examine underlying trends in the factors influencing UK population mortality rates, and from these, to assess:

Historical overview of mortality by cause in the UK Review of literature on the key drivers of recent changes in mortality by cause in the UK, including research done by other professionsInvestigate and review data sources availableInvestigate and model past and future mortality improvements by

Introduction

cause. Investigate and model past and future mortality improvements by cause to sub-populations – smoker status, socio-economic categories and immigration patterns (with associated genetic and socio-cultural mortality differencesInvestigate and model the potential for mortality “shocks”, either short-term or via the advent of new diseases, by way of example, reduction in the incidence of heart attack due to the introduction of a smoking ban or obesity.

Formed in December 2007

Overview I

All Causes

Introduction

Overview II

All Causes

Cardiovascular

Cancer

Introduction

Other Causes

Respiratory

3/19/2009

3

Overview III

All Causes

Cardiovascular

Cancer

Other Causes

Forms of Chronic Ischaemic Heart Disease

Acute Myocardial Infarction

trachea, bronchus and lung

Other Cardiovascular

Introduction

Heart Failure

Aneurysm

Breast

Oesophagus

Uterus

Respiratory

Chronic obstructive pulmonary Disease

Pneumonia

prostate

Other Cancer

Other Respiratory Condition

Colon

Ovary

rectal

Stomach

Bladder

Pancreas

All causes, Male, Age 75-79Introduction

Age 75-79 - All Causes

800 00

1,000.00

1,200.00

1,400.00

1,600.00

Projection Observed

70-74 80-84

0.00

200.00

400.00

600.00

800.00

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055

Source: ONS data using EASui Software (TSAP)


Age 75-79 All Cause

800.00

1,000.00

1,200.00

1,400.00

1,600.00

Other AllMentalNervousStrokeRespiratoryConditionCancer

-

200.00

400.00

600.00

800.00

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055

CardiovascularObservedProjection


3/19/2009

4


Age 75-79 All Cause

50%

60%

70%

80%

90%

100%

0%

10%

20%

30%

40%

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055Cardiovascular Cancer RespiratoryConditionStroke MentalNervous Other All


Outline


Scenarios – flat and steep

Explain terminology and techniqueComparison International mortality by causeIllustration approach for a few specific causesIllustration approach for a few specific causes

3/19/2009

5

Approach

By eye fit using the historical data as a guideFlat mortality improvement trendSteep mortality improvement trend


Carried out at group cause and sub-cause level for chosen conditions

Further medical input for scenarios still requiredVariation in Mortality rate by cause internationally

Variation in Mortality rate by cause internationally –Mortality Rates for Cardiovascular condition across 24 EU counties


France is 65% of UK

Bulgaria is 305% of UK

Source: WHO

Cardiovascular – Medical History

Myocardial Infarction has been the key driver of mortality improvement over the last 30 yearsThese improvements may well continue for some time to come – scenarios attempt to capture this uncertainty


come – scenarios attempt to capture this uncertaintyBut, Cardiovascular is a much smaller cause of death now, andIt will stop (reincarnations aside)

3/19/2009

6

Cardiovascular – Medical History

Medical input (Dr Croxson)Pollypill is not on market but clinical trials show this drug may reduce illnesses by more than 80%Expected to impact Heart Attack and “Forms of Ischaemic


Expected to impact Heart Attack and Forms of Ischaemic Heart Disease” equally

Smoking ban in UK is also expected to have a beneficial effect

Cardiovascular – Flat ScenarioScenarios – flat and steep

Age 75-79 Group Cause - Cardiovascular

400

500

600

700 Projection Observed

70-74 80-84

0

100

200

300

400

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055


Cardiovascular – Steep ScenarioScenarios – flat and steep


400

500

600

700Projection Observed

70-74 80-84

0

100

200

300

400

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055


3/19/2009

7

Cardiovascular – Funnel of DoubtScenarios – flat and steep


400

500

600

700Steep

Observed

Flat

0

100

200

300

400

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055


Cardiovascular – All Ages (Flat)Scenarios – flat and steep

Cardiovascular Group Cause - All SubCause (M)

800

1,000

1,200

85plus

Observed

80-84

Observed

75-79

Observed

70-74

Observed

65-69

0

200

400

600

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055

Observed

60-64

Observed

55-59

Observed

50-54

Observed

45-49

Observed


Cardiovascular – All Ages (Steep)Scenarios – flat and steep

Cardiovascular Group Cause - All SubCause (M)

800

1,000

1,200

85plus

Observed

80-84

Observed

75-79

Observed

70-74

Observed

65-69

0

200

400

600

1965 1975 1985 1995 2005 2015 2025 2035 2045 2055

65 69

Observed

60-64

Observed

55-59

Observed

50-54

Observed

45-49

Observed


3/19/2009

8

Outline


Heat Charts

ApproachHeat Charts specific Group causes:

Cardiovascular Disease Breast and Lung CancerCardiovascular Disease, Breast and Lung CancerAggregate Flat Scenarios-

With Cardiovascular Disease, Breast and Lung cancer switched to steep

Observations

Heat Chart

Covers range of ages pictoriallyProjections against ActualSmoothing has been applied to Actual “Gaussian

Heat Charts

smoothing”Projected 20 years into the futureInvestigate the Heat Charts for group cause and sub-causeInvestigate the impact of scenarios on the aggregate level

3/19/2009

9

Heat Chart – Cardiovascular (m) - FlatHeat Charts

Diagonal to follow birth cohort(1930)

Calendar year across horizontal axis

Age on 1 5% b d

Colour scale

Age on vertical axis

Hot diagonal –exceptional

improvements

1.5% boundarycooler colours <

warmer colours >

Heat chart based on smoothed actuals (Gauss) – disregard

edges due to distortions

Source: XL Re Heat chart tool, ONS data using EASui software (TSAP)

Heat Chart – Cardiovascular (m) - Steep

Heat Charts


Heat Chart – Cardiovascular(m)Heat Charts


3/19/2009

10

Aggregate Flat - MalesHeat Charts


Aggregate Flat with Cardiovascular Steep (m)

Heat Charts


Heat Chart – Breast Cancer (f) - FlatHeat Charts


3/19/2009

11

Heat Chart – Breast Cancer (f) - SteepHeat Charts


Heat Chart – Breast Cancer (f)Heat Charts


Aggregate Flat (f)Heat Charts


3/19/2009

12

Aggregate Flat with Breast Cancer Steep (f)

Heat Charts


Heat Chart – Lung Cancer (m) - FlatHeat Charts


Heat Chart – Lung Cancer (m) - SteepHeat Charts


3/19/2009

13

Heat Chart – Lung Cancer (m)Heat Charts


Aggregate Flat - MalesHeat Charts


Aggregate Flat with Lung Cancer steep (m)

Heat Charts


3/19/2009

14

Heat Charts - Observations

CardiovascularImprovements have some way to goBut diluted at aggregate level as becoming smaller cause and trend flattening

Scenario funnel against Lee-Carter Percentiles

trend flattening Breast Cancer

Screening has created improvements in mortality trendFuture improvements more at younger ages (50-60) rather than the older ages (60+)At Aggregate level impact small

Heat Charts - Observations

Lung CancerImpact of steep much more significant at younger ages.But steep scenario impacts all ages


Females benefit slightly more compared to males –female smoking epidemic started later compared to males and therefore has further to go in steep scenarioAt aggregate level small impact

Outline


3/19/2009

15

Background: Lee-Carter

Lee-Carter stochastic model Using ReMetrica software (Aon Benfield) for stochastic outputBasic model Calibrated using HMD from 1979 to 2006


Calibrated using HMD from 1979 to 2006Used 20 intervals (for each year), 10000 trialsProjected in single ages but grouped post simulation

No jumps to allow for extreme mortality eventsi.e. Pandemics

UK – caveat – cohort effect is not captured in this model

Graphic of scenario funnel against Lee-Carter Percentiles, all cause mortality, age 75-79 males


0.08

0.1

0.12

0

0.02

0.04

0.06

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

Source: ONS data, EASui software (TSAP) and ReMetrica software (Aon Benfield)



0.08

0.1

0.1297.5 to 99.5

90 to 97.5

80 to 90

70 to 80

60 to 70

50 to 60

0

0.02

0.04

0.06

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

40 to 50

30 to 40

20 to 30

10 to 20

2.5 to 10

0.5 to 2.5

Median


3/19/2009

16



0.08

0.1

0.12

0.08

0.1

0.1297.5 to 99.5

90 to 97.5

80 to 90

70 to 80

60 to 70

50 to 60

40 to 50

0

0.02

0.04

0.06

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

0

0.02

0.04

0.06

40 to 50

30 to 40

20 to 30

10 to 20

2.5 to 10

0.5 to 2.5

Median

Group_Steep

Group_Flat




0.08

0.1

0.12

0.08

0.1

0.12 97.5 to 99.5

90 to 97.5

80 to 90

70 to 80

60 to 70

50 to 60

40 to 50

0

0.02

0.04

0.06

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

0

0.02

0.04

0.06

30 to 40

20 to 30

10 to 20

2.5 to 10

0.5 to 2.5

Median

Subgroup_Steep

Group_Steep

Subgroup_Flat

Group_Flat




0.08

0.1

0.12

0.08

0.1

0.12 97.5 to 99.5

90 to 97.5

80 to 90

70 to 80

60 to 70

50 to 60

40 to 50

30 to 40

20 to 30

0

0.02

0.04

0.06

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

0

0.02

0.04

0.06

20 to 30

10 to 20

2.5 to 10

0.5 to 2.5

Median

Medium Cohort witha 1.5% underpinQiS4 Scenario

Subgroup_Steep

Group_Steep

Subgroup_Flat

Group_Flat

1 year shock or permanent ??


3/19/2009

17



0.015

0.02

0.025

0.015

0.02

0.025 97.5 to 99.5

90 to 97.5

80 to 90

70 to 80

60 to 70

50 to 60

40 to 50

30 to 40

20 to 30

0

0.005

0.01

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

2026

0

0.005

0.01

20 to 30

10 to 20

2.5 to 10

0.5 to 2.5

Median

Medium Cohort witha 1.5% underpinQiS4 Scenario

Subgroup_Steep

Group_Steep

Subgroup_Flat

Group_Flat


Observations

Age group 75-79Output is consistent with funnelsCohort effect is strong for this age group and the Lee-Carter variant allowing for cohort effect will be steeper


variant allowing for cohort effect will be steeperAge group 60-64

Output is steeper compared to scenario funnelsLee-Carter will replicate historical data into future

Lee-Carter fans rapidly in first few years and then widens very gradually thereafter

Observations

QiS 4 is extremely conservative and significantly outside our funnel scenariosOur funnel scenarios are extreme as we are assuming

ll diti ith fl t t


all conditions are either flat or steep“The longevity shock to be applied is a (permanent) 25% decrease in mortality rates for each age” (CEIOPS (2008))

Does permanent make sense in context of Solvency II ?Is it a 1 year shock or permanent ?

3/19/2009

18

Outline


Caveats

Based on past extrapolation of dependent ratesCorrelations poorly understoodCause of death codification subject to changes

Conclusions

Cause of death codification subject to changes in practice and errorBase projection has no deliberate allowance for new medical advances accelerating pace of improvement....

Conclusions

Scenarios by cause give great insightReasonable feel for the extreme boundariesCan allow for medical input

Conclusion

pAllows lucid Communication of variability

Allows validation against stochastic outputGives a reasonable shape for future trends compared to short, medium and long with an under pin.

3/19/2009

19

Outline


Where Next

Consolidation of research workPeer review within Actuarial ProfessionPublication of findingsPublication of findingsContinuing research and model refinementDecision maker education

Documents

Pinnington.ppt - Actuaries