Article Title

Evaluating Unwarranted Variation in Treatment Patterns Using Unblinded Data

Publication Date



program evaluation, quality of care, evaluation research, health care organizations, communication of research findings, quality improvement


Background: Unwarranted practice variation is an issue for most health care systems and is sometimes caused by external factors. We examined the peer group Variation Reduction (VR), a program in which clinicians are shown data about their treatment patterns alongside other clinicians and are usually surprised by the variation. The information drives a desire to understand the variation in how they care for their patients and can lead to a change in clinician behavior. We sought to evaluate VR projects aimed at reducing brand name prescribing and increasing appropriate documentation of end-of-life (EOL) wishes in a large integrated health care system.

Methods: In this physician-controlled process, providers reviewed and discussed data in a safe environment ruled by principles that confirm confidentiality, safety and group control. Verbal consent of all group members is required for each VR step, including topic selection and discussion of relevant data. For analyzing projects, we chose a pre-post parallel design. Providers having at least 12 months of data prior to the VR program and 3 months of data after the VR program were included in the analysis. The control group was comprised of clinicians not involved in the VR program who had enough data for the selected projects. Orders in the 3 months after the VR program were compared to the 12 months before the VR program for the intervention and control physicians. Repeated measures within clinicians were modeled using generalized estimating equations. Standard errors were estimated using nonparametric bootstrap with 1,000 iterations, and 95% confidence intervals (CI) are reported.

Results: Postintervention, accounting for the 12-month pre-period, physicians in the intervention group prescribed generics for corticosteroids and nasal steroids significantly more than physicians in the control group, and EOL wishes were documented more frequently in the intervention group: 9.4% (CI: 5.4–14.3) corticosteroids; 3.2% (CI: 1.8–4.6) nasal steroids; 0.5% (CI: 0.2–0.8) documenting EOL wishes.

Conclusion: For all three projects presented, the VR intervention was associated with a reduction in variation in the practices. These results indicate that clinician behavior can be changed with a peer-group process, without the need for additional incentives.




June 28th, 2017


August 10th, 2017