50
1 Clinical research based on EHR systems Why is it so hard and what can be done about it ? Gunnar O Klein professor in Health Informatics at NSEP Norwegian Centre for EHR Research Plenary presentation at HelseIT in Trondheim 2012-09-20

Clinical research based on EHR systems - Kith.no · PDF file1 Clinical research based on EHR systems – Why is it so hard and what can be done about it ? Gunnar O Klein professor

Embed Size (px)

Citation preview

1

Clinical research based

on EHR systems –

Why is it so hard and what can be done about it ?

Gunnar O Klein professor in Health Informatics

at NSEP – Norwegian Centre for EHR Research

Plenary presentation at HelseIT

in Trondheim 2012-09-20

2

We had a workshop yesterday

• Together with some very interesting invited experts

we got an update on some recent projects that in

various ways provide insights into the future

possibilities for research using clinical data in EHR-

systems (Electronic Health Record) – or EPJ in Norwegian

• In this presentation I will attempt to give some highlights from

these presentations with the kind permission of the authors

3

The panel • Gerard Freriks, Netherlands, former GP and medical

scientist, past convenor of the CEN working group

that developed the EHR standard. Now working for

the EN13606 Association

• Arnulf Langhammer, Associate Professor, NTNU,

The Nord-Trøndelag health study (HUNT)

• Rong Chen MD, PhD, Sweden, Chief Medical

Informatics Officer, Cambio HealthCare Systems &

Karolinska Institutet, Stockholm

• Damon Berry, PhD, Dublin Institute of Technology,

Ireland

4

Who is Gunnar Klein

• Professor of Health informatics at NTNU Jan 2012

• Have worked with ICT for health since 1975 in

different roles, often from Karolinska Institutet

• Chairman of European standardization of Health

Informatics in Europe 1997-2006 (CEN/TC 251)

• Leader and participant of a number of European R&D

projects, particularly in Information Security and for

communication of EHRs with semantic interoperabilty

• Physician, mainly in Primary care but 2009 at the

Karolinska University hospital

• Also a background as a Cancer researcher and in

Biotech industry in the 1980ies

5

Why should we attempt to

use data from clinical records?

• There is so much we do not know in medicine – and about health systems effectiveness and efficiency

• A lot has been found in the past using records, even

paper records – but very inefficiently

• With electronic records it should be much easier –

piece of cake

Or …

Helseinformatikk - Introduksjon

6

Is the EHR data only

garbage?

7

If we put garbage in a vault

8

Datatilsynet

Protected as gold

9

Do we expect to get a treasure?

10

Is the ocean empty?

Studies have shown that in routine use a lot of things never become documented

11

Is the ocean empty?

Or is it a gold mine?

12

How can we turn EHRs

into gold mines ?

13

There is so much we do not know

• Evaluations of health outcomes related to various

interventions, including medication – On real life patient groups in large scale, at all locations

– With multiple diseases and treatments

– In all age groups

• Comparing biomedical laboratory data, genotypic and

phenotypic with outcomes and treatments - IRL

• Generate and test new hypotheses for basic

biomedical functions – compared with genetics –

Functional genomics

• Results for management of quality and planning of

health services. Eg. Do we follow guidelines?

14

The requirements for EHR information and

some of the problems

in routine record information for research

Arnulf Langhammer

2012 09 19 AL EHR

15 15 AL 05

General practitioner

Høvdinggården Legekontor, Steinkjer

HUNT Research Centre, Levanger

Project leader of the Lung and Osteoporosis Study

Head of HUNT Databank

16

Oslo

Trondheim

The Nord-Trøndelag Health Study

HUNT

County of Nord-Trøndelag 24 Municipalities

Inhabitants: N=130,000

Age 20-100 yrs: n = 94,000

Age 13- 19 yrs: n = 10,000

17

EHR sources for HUNT

• Hospitals – Levanger and Namsos

– St Olavs Hospital

• General practices – All use electronic patient records

– Linked to Helsenett

– Most communication with hospitals electronically

– Electronic prescription handling

18

Data from hospital records

Challenges were discovered during the

HUNT studies over a long period of time

– Change in ICD-codes

• ICD 9 replaced by ICD 10

– Validity of ICD codes

• Diagnostic uncertainty – code + ? (e.g. fracture maybe)

• Precision – Different according to level of speciality

– Change of diagnostic criteria :

• Myocardial infarction

• COPD

19

The alternatives: Registries • Special health registries on a national or local level

that has collected certain data for certain purposes.

The general registry of all causes of deaths and the

cancer registries are such examples but also the

more recent quality registries in relation to certain

diseases or procedures. – Has generated a lot of useful information despite very limited in

information content

– Cumbersome to get data, often increased work for health

professionals and double registrations also in EHRs.

– A limited and predetermined set of questions that may be asked

even if a lot remains to be explored

• One question of today – How can we improve

collection of data from EHRs to these registries?

20

The alternatives: Questionaires

• Questionaires to the persons included. This has often

been performed in conjunction with the collection of

the biological sample but may be repeated over the

years. More and more examples from various

countries are using web based surveys for easy data

collection. The method has several weaknesses in

addition to the ethical consequences related to

disturbing repeatedly possibly healthy persons with

intimate questions on their health. The answers are

subjective and may often lack the accuracy of a

professional assessment that may be needed to

achieve the desired results.

21

The alternatives: Examniations

• Special clinical and laboratory examinations of the

study group for the sole purpose of obtaining

research data.

• This is the typical means of conducting clinical trials

e.g. for the approval of new medicines – Very time consuming and expensive

– Interfering with the daily lives of the study population

• Will be necessary for a long time – But how do we

find the interesting patients if they have a particular

health problem ( excl. a general population study)

22

Obstacles to EHR

based research Scattered EHRs

The records over time of one

individual may be scattered in

several institutions:

- geographic location

- specialty

- legal entity c.f. the division

between primary care and

specialist health care, in Norway

23

Obstacles to EHR

based research Various formats and terminologies

The data of the EHRs exists in various

formats with regard to information

structure and terminology used.

- partly follows various EHR products

- Whereas the exchange of some

limited data in the form of electronic

messages has some good results,

essentially no attention has been

given to the task of long term

harmonization of EHR structure of

terminology in order to create a

better infrastructure for clinical

research

24

Obstacles to EHR

based research

Lack of structure

Often there is very little structure in

the EHR systems of today.

Typewriters.

Many health care organisations and

thus systems have focused on the

perceived easiness for the physicians

to record data, with the use of free text

dictation as the solution, more and

more often combined with automatic

speech recognition software.

25

Obstacles to EHR

based research

Privacy concerns

Concerns about protecting the

confidentiality of sensitive

personal information must also

be addressed. Ethical approval

and patient consent is

necessary. New systems may

facilitate the latter using

electronic means and the net.

26

Obstacles are challenges

«Obstacles are those frightful things you see when you

take your eyes off the goal» (Henry Ford)

Sarah Louise Rung

27

Gerard Freriks showed us impressive figures on

the business case for the pharmaceutical industry

When conducting clinical trials using EHR data

there are potential savings for one big company alone

2.000.000.000 EUR/year

28

Reduce time needed for:

• Study Design

• Site selection

• Site initiation

Reduce time needed for:

•Patient recruitment

•Study execution

Less attrition

Less Site closure

Less effort by investigator

Reduce time needed for:

•Post processing

Better data quality

Less data curation

29

Pilot experiences were quite promising

30

Overview of the EHR4CR

project Electronic Health Record systems for Clinical Research

Selected presentation slides kindly provided by Mats

Sundgren (AstraZeneca, coordinator) and prof

Georges De Moor, univ Gent.

Gunnar O Klein NTNU/NSEP (member of the advisory board)

31 31

Project Objectives

• To promote the wide scale data re-use of EHRs to

accelerate regulated clinical trials, across Europe

• EHR4CR will produce:

– A requirements specification

• for EHR systems to support clinical research

• for integrating information across hospitals and countries

– The EHR4CR Technical Platform (tools and services)

– Pilots for validating the solutions

– The EHR4CR Business Model, for sustainability

RDLT meeting July 2012

32 32

Project Facts

• The IMI EHR4CR project runs over 4 years (2011-2014) with a budget of +16 million €

– 10 Pharmaceutical Companies (members of EFPIA)

– 22 Public Partners (Academia, Hospitals and SMEs)

– 5 Subcontractors

• The EHRCR project is to date- one of the largest public-private partnerships aiming at providing adaptable, reusable and scalable solutions (tools and services) for reusing data from Electronic Health Record systems for Clinical Research.

• Electronic Health Record (EHR) data offer large opportunities for the advancement of medical research, the improvement of healthcare, and the enhancement of patient safety.

33

Protocol Feasibility Pilot

• Pilot ready October-November 2012 with 11 Hospitals

RDLT meeting July 2012

34

Vision

35

Rong Chen, MD, Ph.D. chief medical informatics officer at Cambio Healthcare Systems and affiliated with Karolinska Institutet, Stockholm, Sweden

EHR Data Reuse through

openEHR Archetypes

36

Quality Registers Background

• About 80+ quality registers (QR) in Sweden

– National or regional ones

– Usually single condition based

• Common challenges/issues with QR data report

– (Aggregated) data sets do not exist in EHRs

– Unsynchronized data structures among QRs

– Mismatched terminology bindings

– Some QR are guideline based, some not

– Multiple integrations, multiple data entries

– Clinical decision support from QRs (?!)

37

IFK2 – Pilot with the Swedish Heart Failure register

38

IFK2 Results - Archetypes • Total 21 archetypes

• 7 international archetypes – openEHR-EHR-OBSERVATION.blood_pressure.v2

– openEHR-EHR-OBSERVATION.body_weight.v2

– openEHR-EHR-OBSERVATION.ecg_12_lead_standard_recording.v1

– openEHR-EHR-OBSERVATION.heart_rate.v2

– openEHR-EHR-OBSERVATION.height.v2

– openEHR-EHR-OBSERVATION.lab_test.v1

– openEHR-EHR-OBSERVATION.waist_hip.v2

• Expected generally reusable – openEHR-EHR-OBSERVATION.eq_5d.v2

– openEHR-EHR-OBSERVATION.heart_failure_stage.v2

• Some expected to be reusable in QR reports – openEHR-EHR-EVALUATION.review_of_conditions.v1

– openEHR-EHR-EVALUATION.review_of_procedures.v1

39

A L Rector PD Johnson S Tu C Wroe and J Rogers (2001) Interface of inference models with concept and

medical record models. in S Quaglini, P Barahona and S Andreassen (eds) Proc Artificial Intelligence in

Medicine Europe (AIME-2001 ) Springer:314-323

openEHR Archetype

SNOMED CT ???

Clinical Decision Support

40

Rong Chen showed a world premiere of the

new Guide Definition Language (GDL)

• A sub-language of dADL, driven by an object

model

• The object model consists of

– Header: Id, concept, language, description, translation

– Archetype binding

– Guide definition, pre-condition and list of rules

– Each rule has when and then expressions

– Term definition for language-dependent labels

Extensive reuse of existing openEHR specifications Aiming to release through openEHR as open Source

41

Clinical Decision Support Workbench

(GDL implementation)

• A tool to import, export and author clinical rules

• A rule engine to execute the rules

• Linked to COSMIC (EHR) Intelligence for verification, simulation and compliance checking

• An extension of Cambio COSMIC (EHR)

2. Model new or find

existing clinical rules

using evidence based

guidelines

3. Analyze EHR data in

CDS workbench

4. Confirm the clinical

gaps and find areas for

improvements

5. Deploy Runtime

CDSS inside COSMIC

(EHR)

1. Identify or monitor

the clinical problems

42

Case Study: Antithrombotic Management in Atrial

Fibrillation

• 20% of strokes caused by atrial fibrillation

• Evidence-based European guideline on management of

atrial fibrillation, European Heart Journal (2010) 31, 2369–2429

doi:10.1093/eurheartj/ehq278

43

Compliance Checking

44

Compliance Checking Results

45

Archetype Research in Ireland (with a focus on records to support

biomedical research)

Damon Berry Dublin Institute of Technology

46

Example 1: Archetype-based

shared assessment tool (Hussey 2010)

• Using archetype tools and services in the development

of a shared assessment tool between – Community care nurses

– Public health nurse

– Community intervention team

– Respite care

– Primary care

– Acute care

47

Example 2: Archetypes for CF

review records (Corrigan 2009)

• Cystic Fibrosis (CF) has high incidence in Ireland

• An assessment of how archetypes could be applied for representation of CF record for multi-disciplinary teams

• Starting point, CF Registry of Ireland

• Develop archetypes, through to user interface to experience development process.

• Feed back archetypes to openEHR org.

48

Example 3: Archetypes for

wound care (Gallagher – 2012)

• MSc (HI) student who is an experienced tissue viability

nurse.

• Recognised wound care documentation issues in Irish

health system

• Studied doc. practices “on the ground”

• Researched best practice re documentation

• Incorporated ideas based on this study into draft archetype

and submitted to CKM.

49

Conclusions

• Yes – We can turn EHR data into a goldmine for

Clinical Research

• To fully exploit the possibilities for secondary use of

data for research and quality management we need

structured data – Using standardised structures EN ISO 13606/openEHR with

archetypes modelled by the clinical professionals and defined

terminologies (for international use SNOMED CT is preferable)

– This also gives new possibilities for decision support

– Very encouraging support from DIPS the major Norwegian EHR

supplier to hospitals

• It is possible to start building infrastructures for

clinical research using archetype methodology and

conversions of legacy data

50

Strukturert EPJ

Gunnar O Klein professor i helseinformatikk

Presentation for Helse Midt-Norge, IKT- strategigruppa

13 september, 2012

The road to better health goes through research and

structured EHR systems based on standards

A bridge to the future

It starts now!