Predictors and variability of antibiotic prescribing amongst family physicians

BACKGROUND: Rising rates of antimicrobial resistance are driven by overuse of antibiotics. Characterizing physician antibiotic prescribing variability can inform interventions to optimize antibiotic use. OBJECTIVES: To describe predictors of overall antibiotic prescribing as well as the inter-physician variability in antibiotic prescribing amongst family physicians. METHODS: We conducted a 5 year cohort study of antibiotic prescribing rates by family physicians in Ontario, Canada using a repository of electronic medical records. Using multilevel logistic regression models fitted with random intercepts for physicians, we evaluated the association of patient-, physician- and clinic-level characteristics with antibiotic prescribing rates. RESULTS: We included 3956921 physician-patient encounters, 322129 unique patients and 313 physicians from 41 family medicine clinics. Physicians prescribed a median of 54 (interdecile range 28-95) antibiotics per 1000 encounters. Female children aged 3-5 years were most likely to receive antibiotics compared with men ≥65 years (OR 4.01, 95% CI 3.89-4.13). The only significant physician-level predictor was median daily patient visits of ≥20 compared with <10 (OR 1.28, 95% CI 1.06-1.55). The median ORs without and with patient characteristics were 1.68 and 1.69, respectively. Thus, the odds of receiving an antibiotic varied by 1.7-fold for the same patient simply by virtue of encountering two different physicians. CONCLUSIONS: We observed substantial inter-physician variability in antibiotic prescribing that could not be explained by sociodemographic and clinical patient differences, suggesting that risk adjustment of antibiotic prescribing practices may not be required for audit and feedback of family physicians working in similar practice settings.