Predictors of Antibiotic Overuse in Febrile Children

BACKGROUND: Antibiotic overuse contributes to antimicrobial resistance. Mechanisms that drive overuse, including diagnostic uncertainty, are thought to be complex. We aimed to investigate predictive factors for antibiotic overuse in a large pediatric cohort. METHODS: This is a subanalysis from the AutoPilot-Dx study, which enrolled febrile children 90 days to 18 years old presenting to the pediatric emergency department with respiratory tract infection or fever without source, including both hospitalized children and those discharged home. We analyzed patient flow from being treated with antibiotics at recruitment to definitive antibiotic treatment after assessment, depending on physician judgment. Variables significantly associated with antibiotic overuse in univariate analysis were entered into a multivariable logistic regression model to identify independent predictors of antibiotic overuse, defined as receipt of antibiotic treatment following clinical assessment in cases with viral panel-adjudicated diagnosis. Diagnostic uncertainty was assessed by treating physicians. RESULTS: Upon recruitment, 127/1008 (12.6%) were receiving antibiotic treatment. Overall, 186 of 628 cases with a viral panel-adjudicated diagnosis were judged to have antibiotic overuse. Factors with significant association in univariate analysis were diagnostic uncertainty, nonexudative tonsillitis, rales and younger age. Presence of a diagnostic test result upon presentation to the hospital and current antibiotic treatment were found not to be associated with overuse. Among 566 of 1008 cases with complete data on all model covariates, multivariable logistic regression yielded diagnostic uncertainty [odds ratio (OR), 2.09; 95% confidence interval (CI): 1.18-3.93; P = 0.015], nonexudative tonsillitis (OR, 2.78; 95% CI: 1.72-4.59; P < 0.001) and rales (OR, 1.82; 95% CI: 1.15-2.86; P = 0.01) to be positively associated with antibiotic overuse, while age showed a negative association in complete-case analysis (OR, 0.93; 95% CI: 0.85-0.99; P = 0.038), which attenuated and was no longer significant in a multiple imputation sensitivity analysis (OR, 0.96; 95% CI: 0.91-1.01; P = 0.150). CONCLUSIONS: Diagnostic uncertainty contributes to antibiotic overuse, alongside clinical factors. Future studies should investigate the impact of improved diagnostic algorithms on antibiotic prescribing in patients with respiratory tract infections in whom physicians experience diagnostic uncertainty.