Improving quality improvement using achievable benchmarks for physician feedback - A randomized controlled trial

Context Performance feedback and benchmarking, common tools for health care improvement, are rarely studied in randomized trials. Achievable Benchmark s of Care (ABCs) are standards of excellence attained by top performers in a peer group and are easily and reproducibly calculated from existing perfo rmance data. Objective To evaluate the effectiveness of using achievable benchmarks to e nhance typical physician performance feedback and improve care. Design Group-randomized controlled trial conducted in December 1996, with f ollow-up through 1998. Setting and Participants Seventy community physicians and 2978 fee-for-serv ice Medicare patients with diabetes mellitus who were part of the Ambulator y Care Quality Improvement Project in Alabama. Intervention Physicians were randomly assigned to receive a multimodal impr ovement intervention, including chart review and physician-specific feedbac k (comparison group; n=35) or an identical intervention plus achievable ben chmark feedback (experimental group; n=35). Main Outcome Measure Preintervention (1994-1995) to postintervention (1997- 1998) changes in the proportion of patients receiving influenza vaccination ; foot examination; and each of 3 blood tests measuring glucose control, ch olesterol level, and triglyceride level, compared between the 2 groups. Results The proportion of patients who received influenza vaccine improved from 40% to 58% in the experimental group (P<.001) vs from 40% to 46% in th e comparison group (P=.02). Odds ratios (ORs) for patients of achievable be nchmark physicians vs comparison physicians who received appropriate care a fter the intervention, adjusted for preintervention care and nesting of pat ients within physicians, were 1.57 (95% confidence interval [CI], 1.26-1.96 ) for influenza vaccination, 1.33 (95% CI, 1.05-1.69) for foot examination, and 1.33 (95% CI, 1.04-1.69) for long-term glucose control measurement. Fo r serum cholesterol and triglycerides, the achievable benchmark effect was statistically significant only after additional adjustment for physician ch aracteristics (OR, 1.40 [95% CI, 1.08-1.82] and OR, 1.40 [95% CI, 1.09-1.79 ], respectively). Conclusion Use of achievable benchmarks significantly enhances the effectiv eness of physician performance feedback in the setting of a multimodal qual ity improvement intervention.