Learning Healthcare System

Professor Charles Friedman Interview


Dr Tom Foley, Dr Fergus Fairmichael



Professor Charles Friedman is Chair of the Department of Learning Health Sciences at the University of Michigan Medical School. He is the former Deputy National Coordinator and Chief Scientific Officer, Office of the National Coordinator for Health IT in the U.S. Department of Health and Human Services.

Professor Friedman first became interested in the concept of Learning Health Systems in 2009, whilst working at the Office of the National Coordinator for Health IT, where he became aware of the work of the Institute of Medicine into the field. 

Interview Synopsis

US Government policies such as the HITECH act and Meaningful Use have been a necessary foundation but not a complete solution to developing the learning health system.

There are three elements required to achieve the learning health system:
Element 1: Adoption
Element 2: Interoperability
Element 3: Socio-technical factors, over and above elements 1 and 2, that are required to achieve an LHS

In addition to his university role Prof Friedman is involved in the Learning Health Community, a grassroots movement that grew out of a national summit funded by the Joseph H. Kantor foundation in 2012.  This summit brought together stakeholders to agree upon what a LHS is and to create a consensus about what the core values of an LHS are.  This succeeded in generating a draft set of core values that were subsequently refined until a consensus was reached (http://www.learninghealth.org/about-the-community/).  This has since been endorsed in a formal way by around 70 entities.  The Learning Health community is politically neutral and not dominated by any particular stakeholder. It has 2 groups that provide its focus:
1. standards and structures
2. policy and governance

Professor Friedman has continued to explore the science of the Learning Health System at The University of Michigan. He has recently championed an effort to identify the challenges to developing an LHS.  The National Science Foundation helped organise a workshop to identify scientific problems to developing a high functioning LHS.  This identified 106 questions to be addressed in the development of learning health systems. This work highlights that current understanding and infrastructure are inadequate for an effective system and provides direction for further work.

The learning cycle

Funding for the technical elements of the learning health system is necessary but not sufficient.  Big Data research has been relatively well funded but there is a need to recognise that the Learning Health System is more than just “Big Data” and that this only represents perhaps only a third of what is needed.  The other two thirds include completing the cycle, learning lessons, implementing change and scaling the system.  The Learning Health System can be viewed as a fractal, an idea that could be implemented in the same way at any level of scale – global/national/state/institutional. Clinical research is increasingly becoming an international collaboration.

Figure 1.The learning cycle, as described in “Toward Complete & Sustainable Learning Systems” by Professor Charles Friedman, available at http://medicine.umich.edu/sites/default/files/2014_12_08-Friedman-IOM%20LHS.pdf (accessed 24/02/2015)

This cycle is made up of three components
•Afferent (Blue) side
     ◦Gathering and analysing data (big data)
•Efferent (Red) side
     ◦Feeding back into the system what has been learned and implementing change

“This cycle is the double helix of the Learning Health System”

A significant challenge to the learning health system is one of scaling this cycle. The red/efferent side becomes an enormous interdisciplinary problem incorporating behavioural psychology, communication science, implementation science, behavioural economics, policy science and organisational theory.

If we continue to rely on journal articles as the means of dissemination, there will continue to be a 10, 15, 17 year latency before learning makes its way into practice.  We need to find a way to interpret findings and codify knowledge.  Automated systems could codify knowledge with machine executable guidelines and representations of information.  These are potential methods that could be used to translate knowledge back into the efferent aspect of the cycle.  However there needs to be research into behaviour change to create efficient platforms to complete this learning cycle.

Currently, no part of the cycle is working as well as it could.

The Future

Ultimately, a system whereby treatment decisions are based upon data from “patients like me”, where clinicians could search a database to identify similar patients, their treatments and outcomes, to inform choice for the patient would be the key use case to empower the consumer.  However this will pose the “fuzzy proposition” of how similar patients within this ‘dynamic cohort’ would need to be to draw valid conclusions. No other patient is truly like me.

Fundamentally difficult to predict the progress that will be made, but in the next 5 years we may see:
• This will be the institutional stage.
• The maturation of inter-organisational networks that are now forming (e.g. PCORI CDRNs, CancerLinQ, Cancer Research Network)
These will address problems requiring multiple institutions e.g. public health and rare diseases
• Intraorganisational networks or learning organisations will evolve
     ◦ Large healthcare delivery systems will become learning entities
• Developments of the platform components for the LHS
     ◦ Standards
     ◦ Technical components
     ◦ Implementations
     ◦ Best practice
     ◦ Knowledge base of how to learn
• Privacy concerns will continue to feature

The next 5-10 years are much less clear:
• This will likely be the consumer stage, where the consumer will emerge as the most important stakeholder and the concept will creep into consumer awareness. “The consumer will become the glue that will hold the system together”
• Interoperability will be tackled, this is an Office of the National Coordinator 10 year goal but this on its own will not be sufficient to deliver a learning health system
• LHS will scale, becoming true networks of networks.  Unclear if this will be a top down or bottom up process
• Governance will emerge
• Privacy concerns will continue.  Best practice may emerge for managing issues of privacy and it will become accepted that this is the best method to improve
• We may have moved to a safety culture where we see errors as a learning opportunity

Workforce Implications

In an era of ubiquitous information, medical (and other clinical) education will need to instil three core competencies
1. Able to know when you are right or wrong
2. Able to ask a good question
3. Able to deal with fuzzy information

The knowledge cloud will not provide clear answers. Doctors will become organisers and question askers. They will provide the scaffold for patients and carers.  IT systems are unlikely to replace this role in the foreseeable future. 

Key Points

• The learning health system approach is a shift in our thinking 
• There is huge enthusiasm for the field of learning health systems
• Adoption and interoperability of Electronic Health Records are foundational elements to the learning health system but it requires a third element beyond these
• If each patient or doctor only uses their data for immediate needs then we are missing an opportunity
• We need just the right amount of standardization in systems, too much may stifle innovation
• We are currently in a position to create the first generation of “clunky” learning health systems but this won’t get us to the finish line of improving care and Public Health
• The efferent feedback of knowledge into change in practice is a key element in the learning cycle
• There are concerns regarding the validity of conclusions drawn from observational data, however this may be the best available data to inform choice




Recent Comments
No Comments found...

Leave a Comment

Please login to give your comment