pitfalls and realities of working with big data

1

Plenary II

Pitfalls and Realities of Working with Big Data

Karim Keshavjee MD, MBA, CCFP, CPHIMS

NAPCRG PBRN Jun 30, 2014Bethesda, Maryland

2

Disclosure

I am a provider of commercial services that may be alluded to in this CME activity

I do not intend to discuss an unapproved or investigative use of a commercial product or device in my presentation

I will not be discussing any use of products used on patients

3

German Proverb

“You have to take life as it happens, but you should try to make it happen the way you

want to take it”

Outline

The Problem Experience with Solutions/Lessons

Learned Stakeholder Engagement Solution Design Brief Data Collection Architecture Key Barriers and resolution Advantages Conclusion

5

The Problem

>Physician use of EMR|EHR in North America has increased substantially in the last 4 years

Current EMRs are not able to Capture standardized data across

EMRs Transmit data to a central

repository Present guideline

recommendations at point of care

1National Physician Survey 2013

6

Big Data Assumptions

Requires structured data

Ability to conduct many small experiments rapidly (Amazon phenomenon)

Greatest benefit lies in speeding up feedback cycle between research findings and bedside application

7

Methods

Review of previous projects, experiences, lessons learned

Stakeholder Analysis Identified 8 distinct groups

Stakeholder interviews N = 90, 8-12 per stakeholder group

Iterative process of asking about problems and designing solutions

8

N=90

9

Experience to Date

• Data collection projects are costly

• EMR vendors not able to focus on data projects

• Not scalable to multiple diseases

• Difficult updating to new evidence

10

Highly SuccessfulHigh Blood Pressure

Management Initiative

Now used in 40 Clinics across Ontario

11

RA Form

Used by 60

RheumatologistsIn Ontario

12

Design Challenge

Design clinical forms

Usability tested (once)(with researchers, policy makers, patients and

providers)

Evidence-based

incorporated into EMRs|EHRs instantly or almost instantly?

Independent of the vendor

13

A Browser Window in Every EMR|EHR

• Vendors include browser window in EMR

• Provider selects template the usual way

• EMR gets the form from a website

• Form data provided to EMR using XML

Structured Data Collection Architecture

14

Standard Data

Forms Authoring System

EMR 1 EMR 2 EMR 3

Forms Server

Research Database

Ability to Intervene at Point of Care

15

Standard

Data

Forms Authoring System

EMR 1 EMR 2 EMR 3

Forms and Knowledge Server

Research Database

Clinical Knowledge Server

16

Issues Identified and Solved

Privacy –Privacy Architecture

Governance –Governance infrastructure

Balancing needs of various stakeholders Usability for clinicians vs. structured/coded for

research

17

Advantages of Solution

Faster updates to forms and evidence

Faster time to data collection and research

Less onus on vendor

A/B testing of forms

Ability to deliver decision support into form

18

Longer Term Goals

Make research easier, faster and cheaper

BUILD IN

REB approvals and Privacy Patient involvement Usability testing A/B testing and forms feedback eHealth and mHealth Testing

19

Conclusion – The Zen of Design Big data pitfalls can only be solved by new

designs, not by accepting the limitations of current EHRs

New designs need to balance the needs of multiple stakeholders to be successful

New designs need easy data capture at the point of care, guideline recommendations in real-time, analyze provider and patients behaviors quickly, reject hypotheses daily

20

Acknowledgements