“Learning” Health Care Organizations

“Learning” Health Care Organizations 150 150 Kelly Kelleher, MD, MPH

In the push toward innovative health care, “learning” systems may offer solutions for evaluating new technologies without the challenges of randomized trials.

“The only constant is change.” – Heraclitus

The pace of innovation in health care is breathtaking and unlikely to slow in our lifetimes. The former secretary of the U.S. Department of Health and Human Services, Kathleen Sebelius, even challenged providers to speed up their acceptance of new technologies. Although this sounds constructive, it creates a problem for the evaluation of new interventions and technologies because the speed at which these products and services are being developed precludes adequate assessment.

The traditional gold standard for evaluation of new technologies is the randomized controlled trial. Unfortunately, randomized trials require relatively short time periods for effects, relatively large effect sizes, large amounts of time for recruitment and sufficient money for follow-up. These challenges are even more substantial in pediatrics, where developmental changes in children require larger sample sizes and longer follow-up periods.

In response, pediatric experts across the country have called for hospitals and provider groups to become “learning health care organizations,” whereby we study all the patients and data routinely collected by our labs and clinics to learn what works best for whom and under what circumstances. Many new statistical and epidemiological methods have been developed for simulating or expanding upon the results of randomized trials, including for pediatrics. These methods include meta-analyses (the synthesis of multiple studies), simulated randomized trials of observational data and cross-design synthesis (the combining of RCTs and observational databases). These methods use large-scale observational data to evaluate benefits and risks of interventions.

In pediatrics, most database studies are conducted with national surveys by the federal government or large insurance claims databases. One largely untapped source of data is the burgeoning electronic health record repositories.

PEDSnet: A National Learning Health Care System

Nationwide Children’s Hospital is part of a cross-country network of eight pediatric hospitals known as PEDSnet that will, in part, employ electronic health record data to answer questions efficiently that no single center easily could. Funded by the Patient-Centered Outcomes Research Institute (PCORI) and led by the Battelle Center for Mathematical Medicine in The Research Institute here at Nationwide Children’s, PEDSnet will create an infrastructure for simulated trials, rapid prospective enrollment trials and other tools for improving interventions. Our hospital and five other founding member institutions already demonstrated the potential of such collaborative learning in a pilot study of body mass index among child patients of the six hospitals. The network generated precise growth curves and noted specific opportunities to improve care.

If the initial PEDSnet funding develops a secure and functioning electronic exchange among the participating pediatric hospitals over the next year, PCORI, National Institutes of Health and the Agency for Health Care Research and Quality are poised to release contracts and grant announcements for projects employing the networks to improve care. PEDSnet is the latest sign of the increased demand for multi-site collaborative research from federal and national funders. Given the large and diverse patient population at Nationwide Children’s, our investigators will be well-positioned for these efforts, but such studies require special training for our faculty and staff. A functional multi-site information exchange for observational learning and large practical trials across several large hospitals is one important step on the way to leadership as a learning health care system.

New drugs, gene therapies, software tools, medical devices and other treatments will radically alter the practice of pediatrics over the next decade. Most of these tools will not be evaluated using traditional randomized trials funded by the NIH because of patient and community demand, lack of resources and the ethics of trials. Instead, innovative collaboratives like PEDSnet will require new training techniques, improved sharing of data, new models for incentivizing collaboration in academia and new ways of thinking about who our patients and partners are. And our hospital, for one, is committed to being part of these solutions.



  1. Gardner W, Kelleher K. A learning health care system for pediatricsJAMA Pediatrics. 2014 Apr;168(4):303-4.
  2. Forrest CB, Margolis PA, Baily LC, Marsolo K, Del Beccaro MA, Finkelstein JA, Milov DE, Vieland VJ, Wolf BA, Yu FB, Kahn MG. PEDSnet: a national pediatric learning health systemJournal of the American Medical Informatics Association.2014 Jul;21(4):602-6.
  3. Bailey LC, Milov DE, Kelleher K, Kahn MG, Del Beccaro M, Yu F, Richards T, Forrest CB. Multi-institutional sharing of electronic health record data to assess childhood obesityPLoS One. 2013 Jun 18;8(6):e66192.

About the author

Dr. Kelleher is a pediatrician whose research interests focus on accessibility, effectiveness and quality of health care services for children and their families, especially those affected by mental disorders, substance abuse or violence. He has a longstanding interest in formal outcomes research for mental health and substance abuse services. Dr. Kelleher is director of the Center for Innovation in Pediatric Practice and vice president of Health Services Research at The Research Institute at Nationwide Children’s Hospital. Dr. Kelleher is also professor in the Department of Pediatrics of The Ohio State University College of Medicine.