Using Nudges to Implement Comparative Effectiveness

Historically, many interventions have been studied to improve the quality, safety, and effectiveness of medical care, particularly through the new focus on comparative effectiveness research. Unfortunately, sustained provider and system uptake of these interventions has been severely lacking, to the serious detriment of patient health. The most commonly tried tools to increase uptake, including pay-for-performance, have substantially fallen short of expectations. Moreover, often these interventions are created in highly artificial settings, and we have not come up with ways to implement them in the long-term. The challenge, therefore, is to create sustainable change that impacts care in meaningful ways.

In contrast, behavioral economics represents a powerful tool by which to influence provider and systems behavior in a large-scale, meaningful, and sustainable way. Briefly, behavioral economics recognizes that individuals often are not fully "rational" in the purely economic sense, but are subject to the influence of various social, environmental and cognitive factors in their decision making. And, one can take advantage of these findings to "nudge" individuals, in our case physicians, towards more optimal choices. While the application of behavioral economics has been incredibly successful in altering behavior outside the health sphere, surprisingly little attention has been given to health.

We have chosen to focus on physician behavior in prescribing HMG-CoA reductase inhibitors (statins) to patients with elevated cardiac risk and elevated low density lipoprotein cholesterol (LDL-C) as recommended by cholesterol management guidelines. In a cluster randomized trial at several private, community-based, multispecialty practices, we propose to compare usual care to a system of automated, default, opt-out clinical decision support that prescribes statins as appropriate.

We propose to use a cluster randomized trial design in several multispecialty private practices to examine the comparative effectiveness of an EHR-based lipid management protocol based on ATP III guidelines vs. usual care. Cluster randomization of participating physicians is useful when blinding is impossible and "contamination" might be a problem, i.e. more aggressive management among a physician's non-intervention patients as a result of experience with intervention patients. Of an estimated 150 primary care physicians at the recruited private practices, we expect at least 100 to consent to participate. Physicians will be clustered for randomization based on the number of patients in their panel that meet ATP III guidelines for statin. Physicians in each cluster will then be individually randomized to the intervention or control arm.

Physicians randomized to usual care will not get the intervention or decision support. Physicians randomized to the automated clinical decision support "nudge" will see the new "optout" prescribing procedure as part of their EHR interface. This will include initially prescribing the guideline-based medication, simvastatin 20mg. Nearly six months after this visit, physicians will receive a reminder via EHR to schedule a follow-up fasting lipid profile as recommended by ATP III guidelines.