On Oct. 25, I wrote something about the NIH Health Care Systems Collaboratory’s work to identify best practices for using EHR phenotypes in research applications. That blog post was based on a conversation with Rachel Richesson, Ph.D., M.P.H., an associate professor of the Duke University School of Nursing, whose research involves work on standard EHR queries for finding patients with a given condition (e.g., diabetes or chronic pain) from EHRs.
Just after that was published, I attended a roundtable discussion presentation co-hosted by two health policy groups, NEHI and the California Healthcare Institute, focused on developments in big data and patient-centered outcome research. One of the speakers was Richard Platt, M.D., M.S., professor and chair of the Harvard Medical School Department of Population Medicine at the Harvard Pilgrim Health Care Institute in Boston. He leads the FDA's Mini-Sentinel program, which performs surveillance of the safety of marketed medical products using a distributed data network that includes electronic health data from over 125 million people.
Dr. Platt, who is a co-leader in the NIH Health Care Systems Collaboratory, had some very interesting things to say about the potential of big data in the evolving learning health system.
It’s becoming clear that done right, big data can support meaningful observational analyses about patient safety, Platt said. “Done right, that will be true of comparative effectiveness research, but it will be harder,” he added. There has been significant progress, he noted, with both the Agency for Healthcare Research and Quality (AHRQ) and the Patient-Centered Outcomes Research Institute (PCORI) putting money into it. “The future will be terrific,” he said. Data that is now routinely collected in the care process will be used to address a wide array of questions. “We are on a journey. Every five years we are in better shape than we were.” But researchers still have some problems to overcome involving completeness and quality of data for the purposes they want to use it for. “Data collection may be perfect for clinical care but problematic for research purposes,” he said. Clinicians may differ in how they gather and record conditions. “We have to be very careful in understanding where we can make solid inferences on data we have. What we have found is that it is essential to have the people who created the data at the table when you are trying to use it in order to know whether it represents what we expect.”
Dr. Platt also suggested that it might be unreasonable to ask that all data collected in care processes be made available for secondary uses. The healthcare systems have to weigh what’s it worth to get it cleaned up for research purposes. And do we want to ask this of every provider or just some?