Two recent events in this winter’s panoply of D.C. health policy conferences stand out, largely because they invite us to think about a host of problems that beset our health care system in a new way; but also because they raise two nettlesome issues. One event was convened by Health Affairs to highlight a core set of Web-Exclusive papers on the ways electronic health records can promote a “Rapid-Learning Health System.” [Editor’s Note: Today, April 3, Health Affairs published a new Web Exclusive paper on the wide gap between the vision and reality for e-prescribing.]
The other event, a roundtable on evidence-based medicine, was convened by the Institute of Medicine to focus attention on a “learning health care system” in which the process of medical discovery follows naturally and systematically as an outgrowth of care delivery. Each dealt with a problem that lies at the core of some of the most difficult and vexing problems in health care policy and practice today. Given rapid innovation in medical technology, waste and uneven quality of care, patient safety problems, a chronic access crisis, and cost increases as far as the eye can see, how can we expedite knowledge generation and systematically translate what we know into what we do? How do we create evidence-based, high-value, efficient delivery systems?
Lessons for health reform. The answer suggested by these thought-provoking events is that national health care reform may take some lessons from the playbooks of the Veterans Health Administration and Kaiser Permanente, but reform of a fragmented system will require something akin to a Kuhnian paradigm shift in the way we think about knowledge creation, dissemination, and adoption. As Walter Stewart and his Geisenger colleagues write, we must find new ways to “bridge the inferential gap . . . between the paucity of what is proved to be effective for selected groups of patients versus the infinitely complex clinical decisions required for individual patients.” To this I would add a key message derived from the IOM event: We need to learn how to deal with the fact that the “evidence pyramid” — which assigns the randomized controlled trial most-favored scientific status and which many thought-leaders in the scientific community consider the golden rule for producing gold-standard evidence–sacrifices external for internal validity and thereby shrinks the range of real-world problems to which science provides sorely needed insight and guidance.
In a forward-looking lead essay, Lynn Etheredge argues that reorganization of knowledge production to accelerate the translation of knowledge into practice critically depends on large public and private initiatives that use information technology (IT). “With large, computer-searchable EHR databases and new research software, studies that would now take years will be doable, at low expense, in a matter of weeks, days, or hours,” he predicts. Other authors in the Health Affairs series describe examples of such initiatives; consider the possibility and potential benefits of undertaking rapid-learning initiatives in Medicare; elaborate means of linking electronic health records (EHRs) to large-scale simulation models (David Eddy’s Archimedes model); and enumerate some of the barriers, concerns, and challenges to advancing a rapid-learning health system.
Managing expectations for health IT. I applaud the visionary, innovative work that is necessary to reform a system that is, if not broken, in serious need of overhaul. I also think that those who take a less-than-sanguine view about the prospects of creating a rapid-learning health system deserve serious attention. The nettlesome problem the Health Affairs event raised for me is that to a significant degree, the case for creating a transformational rapid-learning system was made so firmly on the back of the pressing need for increasing the adoption of health IT generally and expanding the use of EHRs specifically. If we are to prevent health IT and electronic health information exchange from going the way of the community health information networks (CHINs) of a decade ago, we must reasonably manage expectations for what health IT can deliver, at what cost, with the collaboration of which stakeholders, and in what time frame. Thus, if we portray EHRs to policy- and decisionmakers as the answer to a host of problems that beset our health care system, build one expectation upon another (quality improvement, cost savings, error reduction, etc.), do we not risk bringing the whole house of cards down by creating high hopes and building unrealistically high expectations?
I suspect that nationwide adoption of IT in the health care sector will require, first and foremost, a strong business case. However, even if a value proposition can be demonstrated for core stakeholders, federal action remains necessary to pave the way for broad adoption, through creation of standards around privacy, interoperability, and nomenclature. In short, the rapid-learning health system envisioned by Etheredge and others is exactly what we need. And who can argue that it is not time to modernize the health care system through application of IT? I just don’t think that it is a good strategy to sell them together. I would hate to see the failure of one to cause the demise of the other.
Expanding the tent. My concern with the approach to developing a learning health care system that unfolded at the recent IOM event is of an entirely different nature. Participants, without a doubt, were experienced and highly respected members of the health care research and policy communities. The problem, I fear, is that these communities, although their membership is large and diverse, are too limited by the natural constraint that afflicts most of us — disciplinary nearsightedness — to take the bold steps needed to create a rapid-learning health care system. Granted, much credit is due participants who challenged the preeminence of the evidence pyramid and engaged in serious conversation about how to broaden evidentiary standards in support of clinical decision making. However, when discussion turned to the need to incorporate data collected through means other than the randomized control trial, concerns were raised about the lack of scientific standards. To me, at least, this concern expressed either poor information about or the undervaluation of the scientific rigor incorporated into the best empirical work produced by members of the economic, social, and behavioral science disciplines.
If a rapid-learning health system needs to expand its cache of tools to produce knowledge on which practice reasonably can be based, where better to look than to those disciplines where scientific studies that randomize the most important variables are either impractical, unethical, or both? As the IOM moves forward with the important task of focusing attention on a learning health care system, I think it would be beneficial to expand the tent; to take the best from the “lesser” sciences, which, nevertheless, may prove quite useful to the critical goals of rapid knowledge generation, dissemination, and adoption.