OBJECTIVES: To determine association of recommended pharmacotherapy in CHF with geographic variation in clinical outcomes. METHODS: Veterans with at least one outpatient visit with CHF diagnosis from FY2000-2001 were included in a cohort. Patients with non-utilization of VA for two years were excluded. Twenty-two Veteran Integrated Service Network (VISNs) of the VA were geographic units of analyses. Outcomes evaluated were: death, all-cause hospitalization and heart failure hospitalization. Recommended pharmacotherapy was measured by use of ACE inhibitors, ARBs and Beta-blockers. Follow up period was one year post index date of CHF diagnosis. We accounted for patient factors (gender, race, age, co-morbidities, severity of illness, co-medications, adherence) and process of care measures (prescription of recommended pharmacotherapy). Multivariate logistic regression models were used to calculate observed-to-expected (O/E) ratios across VISNs. Separate models were constructed accounting for patient factors, process of care measures and structure (VISNs). We ranked the outliers and interpreted, in context of, the quality of pharmaceutical care provided.RESULTS: We analyzed 125,239 CHF cases with enough power to detect differences across VISNs. About 60% patients were on ACE inhibitors, 8% on ARBs and 49% on Beta Blockers. Factor of variance in prescription rates ranged from 1.11 to 2.79. Strong correlation was observed between prescription rates and outcome rates (-0.48 to -0.54). c-indexes for all the models were above 0.7. Change in the outlier status, from the model containing patient factors versus the model with patient factors and process of care factors, indicated change in the quality of pharmaceutical care provided by the VISNs. CONCLUSIONS: In 1998, VA implemented Quality Enhancement Research Initiative (QUERI) to systemize quality improvement. During FY2000-02, significant variations were observed in prescription and outcome rates, despite efforts to synchronize quality of care. Differential prescription rates and differential adherence rates led to the variations in the outcomes.