Abstract

OBJECTIVE— The purpose of this article was to evaluate the impact of self-reported patient factors on quality assessment of Veterans Health Administration medical centers in achieving glycemic control.

RESEARCH DESIGN AND METHODS— We linked survey data and administrative records for veterans who self-reported diabetes on a 1999 national weighted survey. Linear regression models were used to adjust A1C levels in fiscal year 2000 for socioeconomic status (education level, employment, and concerns of having enough food), social support (marital status and living alone), health behaviors (smoking, alcohol use, and exercise level), physical and mental health status, BMI, and diabetes duration. Medical centers were ranked by deciles, with and without adjustment for patient characteristics, on proportions of patients achieving A1C <7 or <8%.

RESULTS— There was substantial medical center level variation in patient characteristics of the 56,740 individuals from 105 centers, e.g., grade school education (mean 15.3% [range 2.3–32.7%]), being retired (38.3% [19.9−59.7%]) or married (65.2% [43.7–77.8%]), food insufficiency (13.9% [7.2–24.6%]), and no reported exercise (43.2% [31.1–53.6%]). The final model had an R² of 7.8%. The Spearman rank coefficient comparing the thresholds adjusted only for age and sex to the full model was 0.71 for <7% and 0.64 for <8% (P < 0.0001). After risk adjustment, 4 of the 11 best-performing centers changed at least two deciles for the <7% threshold, and 2 of 11 changed two deciles for the <8% threshold.

CONCLUSIONS— Adjustment for patient self-reported socioeconomic status and health impacts medical center rankings for glycemic control, suggesting the need for risk adjustment to assure valid inferences about quality.