UCSF Informatics Day 2014 - Mark Pletcher, "Making EHR Data Useful for the Learning Healthcare System"
1. Making EHR Data Useful:
The Learning Healthcare System
UCSF Informatics Day, June 10, 2014
Mark J. Pletcher, MD MPH
Dept of Epidemiology and Biostatistics
2. EHR Data Learn
• The Learning Healthcare System
– Concept, IOM vision, PCORI vision
– Barriers
• Getting started at UCSF
• Randomized quality improvement
– Rationale, example
– How to make this easy at UCSF
3. Learning Healthcare System
• Institute of Medicine concept
– A system of healthcare delivery that “drives the
process of discovery as a natural outgrowth of
patient care”
– Electronic Health Records are a major facilitator
2007: The Learning Healthcare System: Workshop Summary
2012: Best Care at Lower Cost: The Path to Continuously Learning Health
Care in America
4. Learning Healthcare System
• Major barriers
– Mixing research with clinical care
• Can slow down clinical care
• Ethical considerations
– Clinical care is not delivered at random
• We treat some and don’t treat others for good reasons
(confounding by indication)
• We don’t measure things systematically (selection bias and
measurement bias)
– EHR systems
• Not designed for research data collection
• Different across institutions
5. Learning Healthcare System
• PCORI’s answer: PCORnet
– National network geared for comparative
effectiveness research
– Clinical Data Research Networks
• Gather common data elements from EHRs
– Patient-Powered Research Networks
• Gather engaged patients excited to do research
– Support very large and low-cost research studies
– Phase I: Just started!
6. Getting started at UCSF
• Why we need to get started here:
– PCORnet needs help!
– Pilot data for PCORnet studies
– For many studies, 1 center is enough
– Local issues may trump
– We can get “deep” into our EHR data
– Improving care HERE
7. Randomized quality improvement
• Quality improvement
– Doing things we THINK will improve care
– Evidence-based…but not entirely
– The law of unintended consequences
– Show that it works disseminate more broadly
• Randomized controlled trials better evidence
• Randomized quality improvement (RQI) trials
– Disseminate, but cautiously (to a randomly selected subset)
– Make sure it’s working
8. Randomized quality improvement
• Example: The Statin Guidelines Project
• New Statin Guidelines are:
– VERY evidence-based
– VERY controversial
9. Randomized quality improvement
• Randomized dissemination of statin guidelines
– Find people who “should” be on a statin
– Randomly assign them to
• Usual care
• Encouragement to take a statin
– Compare outcomes
• Statin use
• Statin side effects and quality of life
• CHD outcomes (pilot for PCORnet)
10. Randomized quality improvement
• Let’s make this easy at UCSF
– Build infrastructure
• Data systems, regulatory processes, expertise
– Identify EHR-measurable outcomes
• Engage patients, providers, administrators
• Which are important, which can be improved
• Start measuring them
– Find providers with good ideas for how to improve
• Invite, support, partner
– Randomly disseminate and evaluate with RQI Trial
11. Summary
• Let’s turn UCSF into a Learning Healthcare
System
– Do quality improvement, but “cautiously”
– Invest in infrastructure so we can study as we go
– Start doing it!
13. Barriers to getting started at UCSF
• Why we “can’t” just start doing this at UCSF…
– Wait for PCORnet to figure this out
– Small sample size
– UCSF is not a “closed” system
– Infrastructure investment
– Ethical issues
14. Ethical issues
• Randomized dissemination
– Do we need informed consent?
– Is it EVER ethical to randomize without consent?
– Is it more ethical to do “cluster-randomization”?
– Different approaches to informed consent
• Informational, opt out, after-the-fact, etc
Faden 2013: An ethics framework for a learning health care system
Pletcher 2014: Informed consent in randomized quality improvement trials: A
critical barrier for learning health systems
15. The Health eHeart Study
• Online platform for engaging patients
– Cardiovascular cohort study
– Leverage emerging technology
– Platform to support trials and other studies
• Health eHeart participants are consented to
participate in research
– Answer surveys
– Special data collection
– Special interventions (sensors, social networks, etc)