Variations in the use of cardiac procedures in the Veterans Health Administration.

ABSTRACT

BACKGROUND: Considerable variability exists in the use of cardiac procedures for patients with heart disease. One cause for this variability is the availability of local facilities to perform these procedures. This study was initiated to identify health system features that are related to rates of catheterization, percutaneous coronary angioplasty, and coronary artery bypass graft surgery in the Veterans Affairs health care system in which structured referral systems are intended to compensate for variation in local resource availability. METHODS: Medical records of 30,901 patients admitted to a Veterans Affairs medical center with coronary artery disease were analyzed. Odds ratios (OR) and 95% confidence intervals (CI) for undergoing each procedure, based on clinical variables (age, sex, race, coronary artery disease type, and a computed comorbidity score), and local Veterans Affairs facility features (geographic region, primary service area size, and hospital complexity) were estimated by logistic regression. RESULTS: Regression models demonstrated significant associations between the odds of undergoing each procedure and medical center geographic and complexity features, after adjustment for clinical variables. Associations included the presence of a cardiac catheterization laboratory for undergoing catheterization (OR 1.86, CI 1.76 to 1.95) and the presence of a cardiac surgical program for angioplasty (OR 1.46, CI 1.36 to 1.57) and bypass grafting (OR 1.43, CI 1.34 to 1.53). Including health system variables in addition to clinical variables in the regression models improved the discriminating ability of the models by 44.2% to 51.4%. CONCLUSIONS: Geographic location and the complexity of the local Veterans Affairs hospital are important determinants of the use of cardiac procedures in the Veterans Affairs health care system, even though referral networks are intended to correct for local differences in hospital complexity.