Context: Physician profiling is widely used by many health care systems, but little is known about the reliability of commonly used profiling systems.
Objectives: To determine the reliability of a set of physician performance measures for diabetes care, one of the most common conditions in medical practice, and to examine whether physicians could substantially improve their profiles by preferential patient selection.
Design and setting: Cohort study performed from 1990 to 1993 at 3 geographically and organizationally diverse sites, including a large staff-model health maintenance organization, an urban university teaching clinic, and a group of private-practice physicians in an urban area.
Participants: A total of 3642 patients with type 2 diabetes cared for by 232 different physicians.
Main outcome measures: Physician profiles for their patients' hospitalization and clinic visit rates, total laboratory resource utilization rate and level of glycemic control by average hemoglobin A1c level with and without detailed case-mix adjustment.
Results: For profiles based on hospitalization rates, visit rates, laboratory utilization rates, and glycemic control, 4% or less of the overall variance was attributable to differences in physician practice and the reliability of the median physician's case-mix-adjusted profile was never better than 0.40. At this low level of physician effect, a physician would need to have more than 100 patients with diabetes in a panel for profiles to have a reliability of 0.80 or better (while more than 90% of all primary care physicians at the health maintenance organization had fewer than 60 patients with diabetes). For profiles of glycemic control, high outlier physicians could dramatically improve their physician profile simply by pruning from their panel the 1 to 3 patients with the highest hemoglobin A1c levels during the prior year. This advantage from gaming could not be prevented by even detailed case-mix adjustment.
Conclusions: Physician "report cards" for diabetes, one of the highest-prevalence conditions in medical practice, were unable to detect reliably true practice differences within the 3 sites studied. Use of individual physician profiles may foster an environment in which physicians can most easily avoid being penalized by avoiding or deselecting patients with high prior cost, poor adherence, or response to treatments.