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 if 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 if 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 (HLM) were used to investigate Hemoglobin A1c (HbA1c) among 1,381 patients, nested within 42 randomly assigned primary-care physicians at a VA medical center in the Southeastern United States. The primary outcome measure was change in HbA1c over the course of a 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 HbA1c change (2%), while patient-level factors accounted for the majority of variation in HbA1c change (98%). Physician effects varied by patient characteristics, mattering more for Black patients, patients over 65, and patients whose glucose management improved over the treatment year.

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