Estimating physician effects on glycemic control in the treatment of diabetes: methods, effects sizes, and implications for treatment policy

Abstract

Objective: Researchers have only just begun to investigate physician-related effects on medical outcomes. Such research is necessary for developing empirically informed practice guidelines and policy. The primary goal of this study was to investigate whether glucose management in type 2 diabetes varies by randomly assigned physicians over the course of a year in treatment. A second goal of the study was to investigate whether physician-related effects vary across differential patient characteristics. A tertiary goal was to investigate potential patient-level effects on glucose management.

Research design and methods: Hierarchical linear models were used to investigate A1C among 1,381 patients, nested within 42 randomly assigned primary care physicians at a Veterans Affairs medical center in the southeastern U.S. The primary outcome measure was change in A1C over the course of 1 year in treatment. On average, each study physician had 33 patients with diabetes.

Results: Overall, physician-related factors were associated with statistically significant but modest variability in A1C change (2%), whereas patient-level factors accounted for the majority of variation in A1C change (98%). Physician effects varied by patient characteristics, mattering more for black patients, patients aged 65 years, and patients whose glucose management improved over the treatment year.

Conclusions: The results of this study indicate that differential physician effects have minimal impact on glycemic control. Results suggest that it is logical to support policies encouraging the development of patient-level behavioral interventions because that is the level that accounts for the majority of variance in glycemic control.