Informatics for the Nationwide Learning Health

SystemCharles P. FriedmanProfessor of Information and Public

HealthUniversity of Michigan

University of UtahJanuary 27, 2014

PreambleI believe that people are drawn to ideas that reflect imagination and vision.Up until recently, the field of health IT has lacked such an idea, one that would unite all the stakeholders and attract the brightest minds from a range of disciplines.By the end of this hour, I hope to convince you that the Learning Health System is that key idea.

2

Today’s MenuPrimi• The plight of nation’s health system

(briefly)• The vision of a nationwide Learning Health

System (LHS)• Widespread calls for the LHS and early

progress toward a “first system”

Secondi• Features of a high-functioning and

sustainable system• The perspective needed to achieve it 3

The Plight in a Figure

4Expenditure Per Capita

The Plight in Words

• Spending 18% of GDP on health, which is unsustainable– 25% of which is “wasted”

• 45th in infant mortality. Japan and Sweden have a rate 40% of ours.

• New estimate: 200,000 to 400,000 die each year in hospitals due to medical errors

• Among five highly developed nations, the U.S. is last or next-to-last on five indicators of a “high functioning” health system 5

6Slide courtesy of Kenneth Mandl

And We’re Virtually Blind: We Can’t Monitor the System in Real Time

We Need a Hefty Rudder, Now…

7

Today’s MenuPrimi• The plight of nation’s health system

(briefly)• The vision of a nationwide Learning Health

System (LHS)• Widespread calls for the LHS and early

progress toward a “first system”

Secondi• Features of a high-functioning and

sustainable system• The perspective needed to achieve it 8

9

9

The Rudder: A National-Scale Learning Health System (LHS)

Pharma

Beacon Community

IntegratedDelivery System

Patient-centered Groups

Health Information Organization

Health CenterNetwork

FederalAgencies

State Public Health

GovernancePatient EngagementTrustAnalysisDissemination

The LHS Enables “Virtuous Cycles” of Study, Learning and

Improvement

10

AssembleChange

Interpret

AnalyzeFeedback

Example: Reducing Falls in Nursing Homes

11

Assemble Data:How do we preventfalls?What is the fall rate?

Change Current Practice:In whole or part…

Interpret findings:Are the results credible?What advice should be given?

Analyze DataWhat practicesassociate withlower fallrates?

Feedback:Based on your currentpractice, you might want to consider…

Decision to study falls

Assemble Data:How do we preventfalls?What is the fall rate?

LHS: A Platform that Supports Multiple “Virtuous Cycles”

12

A High Functioning LHS Will Make These (and Other) Things Possible“17 years to 17 months, or maybe 17 weeks or even 17 hours…“

• Over time, the best “personalized” dosage of a new drug is learned based on patient experience. The current optimal dosage algorithm is automatically implemented nationwide in EHR systems.

• During an epidemic, new cases are reported directly from EHRs and the spread of the disease is predicted. Clinicians are alerted as the epidemic approaches their practice areas.

• A patient facing a difficult medical decision describes her case. She discovers the experiences of other patients like her. 13

14

The Rudder’s Raw Material: Our Health System Goes

Digital

Source: National Center for Health Statistics, Centers for Disease Control, NAMC (National Ambulatory Medical Care) Survey

The LHS: One Infrastructure that Supports

• Research– Clinical– Comparative effectiveness– Translational

• Public Health– Surveillance– Situational Awareness

• Quality Improvement– Health process and outcomes research– Best practice dissemination

• Consumer Engagement– Knowledge-driven decision making

15

The LHS as Currently Envisioned

• A federation– Not a centralized database

• Grounded in public trust and patient engagement

• Participatory governance• An “Ultra Large Scale” System

– “Just enough” standardization– Supports innovation around standards

1616

The LHS as a Fractal

17

• At every level of scale, it looks pretty much the same

• Local, regional, national, global• A system of like sub-systems

17

Today’s MenuPrimi• The plight of nation’s health system

(briefly)• The vision of a nationwide Learning Health

System (LHS)• Widespread calls for the LHS and early

progress toward a “first system”

Secondi• Features of a high-functioning and

sustainable system• The perspective needed to achieve it 18

Institute of Medicine ReportsDigital Infrastructure for the Learning Health System: The Foundation for Continuous Improvement in Health and Health Care

19

Best Care at Lower Cost: The Path to Continuously Learning Health Care in America

• Perspective: Jan 3, 2013• “Code Red and Blue — Safely Limiting Health

Care’s GDP Footprint”Arnold Milstein, M.D., M.P.H.

…U.S. health care needs to adopt new work methods, outlined in the Institute of Medicine’s vision for a learning health system… Such methods would enable clinicians and health care managers to more rapidly improve value by continuously examining current clinical workflows, management tools from other service industries, burgeoning databases, and advances in applied sciences (especially health psychology and information, communication, and materials technologies). They could then use the insights gained to design and test innovations for better fulfilling patients’ health goals with less spending and rapidly scaling successful innovations. 20

The 2011 Federal Health IT Strategic Plan

Better Technology

Better Information

Transform Health Care

Goal V: Achieve Rapid Learning and Technological Advancement

Goal IV: Empower Individuals with Health IT to Improve their Health and the Health Care System

Goal III: Inspire Confidence and Trust in Health IT

Goal II: Improve Care, Improve Population Health, and Reduce Health Care Costs through the Use of Health IT

Goal I: Achieve Adoption and Information Exchange through Meaningful Use of Health IT

21

Progress: Learning “Islands”

• Organizations that have become Learning Health Systems at their level of scale.

• But don’t routinely connect with other islands.

22

Progress: Inter-organizational Data Federations and Networks

2323

Progress: Grant Programs

• NIH “Big Data to Knowledge”• PCORI Research Networks• NSF Smart and Connected

Health

24

Progress: Grassroots Movement• National “Summit” convened in May 2012 to

envision LHS as set of shared beliefs• A Dumbarton Oaks conference for the LHS• ~ 70 organizations represented at the National

Press Club• Resulted in 10 consensus Core Values• 56 organizations have formally endorsed

• Giving rise to a Learning Health Community

25

26

Approaching a “First” National System

1. Words: calls and reports

2. “Learning Islands”

3. Data federations and networks

4. Grant programs

5. A grassroots coalition of the willing

But this “first” system will not turn the Titanic. We need to set our sights on a higher-functioning system.

27

Today’s MenuPrimi• The plight of nation’s health system

(briefly)• The vision of a nationwide Learning Health

System (LHS)• Widespread calls for the LHS and early

progress toward a “first system”

Secondi• Features of a high-functioning and

sustainable system• The perspective needed to achieve it 28

LHS Research Challenges Workshop: April 2013

• A national workshop to explore the research challenges inherent in achieving a high functioning LHS

• Computer science to epidemiology to economics• 45 invited participants plus Federal liaisons• Report (“Toward a Science of Learning Systems”)

at healthinformatics.umich.edu/lhs/nsfworkshop

29

A High-Functioning LHS

Would meet four system-level requirements:1. Trusted and Valued by All Stakeholders2. Economically Sustainable and Self-

Governing3. Stable, Rapidly-Functioning, Certifiable,

Adaptable, and Self-Improving4. Capable of Engendering a Virtuous

Cycle of Health Improvement

30

Workshop Findings at Two Levels

1. What we were asked to do: the research questions that must be addressed to meet LHS system level requirements

– 106 questions organized into four categories and 19 sub-categories

2. Something transcendent: A vision of a science of learning systems necessary to address these questions (and achieve the LHS) 31

Sample Questions (2/106)• An LHS that is Trusted and Valued by All All

Stakeholders– What properties of heterogeneous, routinely

collected clinical data, including methods in place to manage and curate them, will engender confidence and trust in the knowledge generated by the LHS?

• An LHS that is Economically Sustainable and Self-Governing– What ingredients essential to standing up and

sustaining the LHS have no private rationale for funding or are unlikely to be funded privately, and therefore should be considered for catalysis stemming from public funding? 32

Sample Questions (2/106)• An LHS that is Stable, Rapidly-Functioning,

Certifiable, Adaptable, and Self-Improving– How do we make data sufficiently self-

describing so that, for example, the system might be able to identify, without human intervention, data that are most relevant to addressing a question?

• An LHS that is Capable of Engendering a Virtuous Cycle of Health Improvement– How do we build an LHS that is smart enough

to explain its own inferences: how it learns, what it learns, and what it has already learned?

33

Today’s MenuPrimi• The plight of nation’s health system

(briefly)• The vision of a nationwide Learning Health

System (LHS)• Widespread calls for the LHS and early

progress toward a “first system”

Secondi• Features of a high-functioning and

sustainable system• The perspective needed to achieve it 34

Workshop Findings at Two Levels

1. What we were asked to do: the research questions that must be addressed to meet LHS system level requirements

– 106 questions organized into four categories and 19 sub-categories

2. Something transcendent: A vision of a science of learning systems necessary to address these questions (and achieve the LHS). 35

The LHS Can’t be Framed Purely as a Technical Problem

36

=

(Advice for Utah and the nation.)

History Lesson: the Panama Canal

37

They tried to build the Panama Canal the same way, and failed. They didn’t have the right perspective on the problem.

The French built the Suez Canal as a ditch in the desert.

The Socio-Technical Approach that Built the Panama Canal

38

The Transcendent Workshop Product

• The right perspective for the LHS challenge• The LHS makes the irrefutable case for a science of

cyber-social ecosystems that will be a science of learning systems closely aligned with informatics

• Fundamental principles to enable the effective design, analysis, construction, governance, operation and evolution of ultra-large systems such as the LHS

39

Key Precepts of Cyber-Social Ecosystems

• Complex learning processes are effected by networks of people, institutions, and digital computing machines.

• These networks self-organize and evolve.• The system components are loosely coupled and

generally operate in their own self-interest• The system as a whole, not just the digital

infrastructure, but also networks of people and institutions, will have to be understood as parts of an information processing ecosystem.

40

This Framing is Mission Critical• The cyber-social ecosystem perspective must

drive the achievement of the LHS– Otherwise, we will fail like the French attempt to

build the Panama Canal• Key implications for Utah and the nation:

– Take an evolutionary approach (less is more)– Design for adaptation– Embrace the fractal– Attend to all parts of the virtuous cycle– Attend to all domains (care, empowerment, public

health, research)– Engage all stakeholders

41

In Sum

• Our health system is headed directly for the iceberg

• A high-functioning LHS, a single multi-purpose infrastructure, can turn the “Titanic”

• A science of cyber-social ecosystems, a science of learning systems, can guide achievement of the high functioning LHS 42

And…

As the University of Utah becomes a learning entity, it should:• Adopt the cyber-social ecosystem perspective

to achieve a high-functioning system• Engage with the national movement (and not

be an island)

43

Thanks & Write to Me cpfried@umich.edu