Physician and Patient Influences on Provider Performance

Abstract

Efforts to improve the quality of care for patients with cardiovascular disease frequently target the decrease of physician-level performance variability. We assessed how variability in providing β-blockers to ambulatory postmyocardial infarction (MI) patients was influenced by physician and patient level characteristics. β-Blocker prescription and patient characteristics were abstracted from charts of post-MI patients treated by 133 primary care physicians between 2003 and 2007 and linked to physician and practice characteristics. Associations of β-blocker prescription with physician- and patient-level characteristics were examined using mixed-effects models, with physician-level effects as random. Mean physician-specific predicted probabilities and the intraclass correlations, which assessed the proportion of variance explainable at the physician level, were estimated. Of 1901 patients without major contraindication, 69.1% (range across physicians, 20% to 100%) were prescribed β-blockers. Prescription varied with comorbidity from 78.3% in patients with chronic kidney disease to 54.7% for patients with stroke. Although physician characteristics such as older physician age, group practice, and rural location were each positively associated with β-blocker prescription, physician factors accounted for only 5% to 8% of the variance in β-blocker prescription; the preponderance of the variance, 92% to 95%, was at the patient level. The mean physician-specific probability of β-blocker prescription (95% confidence interval) in the fully adjusted model was 63% (61% to 65%). β-Blocker prescription rates were surprisingly low. The contribution of physician factors to overall variability in β-blocker prescription, however, was limited. Increasing evidence-based use of β-blockers may not be accomplished by focusing mostly on differential performance across physicians.