|| Checking for direct PDF access through Ovid
We sought to assess which interventions are most effective at improving the prescribing of recommended antibiotics for acute outpatient infections.We undertook a systematic review with quantitative analysis of the Cochrane Registry Effective Practice and Organization of Care (EPOC) database, supplemented by MEDLINE and hand-searches. Inclusion criteria included clinical trials with contemporaneous or strict historical controls that reported data on antibiotic selection in acute outpatient infections. The effect size of studies with different intervention types were compared using nonparametric statistics. To maximize comparability between studies, quantitative analysis was restricted to studies that reported absolute changes in the amount of or percent compliance with recommended antibiotic prescribing.Twenty-six studies reporting 33 trials met inclusion criteria. Most interventions used clinician education alone or in combination with audit and feedback. Among the 22 comparisons amenable to quantitative analysis, recommended antibiotic prescribing improved by a median of 10.6% (interquartile range [IQR] 3.4–18.2%). Trials evaluating clinician education alone reported larger effects than interventions combining clinician education with audit and feedback (median effect size 13.9% [IQR 8.6–21.6%] vs. 3.4% [IQR 1.8–9.7%], P = 0.03). This result was confounded by trial sample size, as trials having a smaller number of participating clinicians reported larger effects and were more likely to use clinician education alone. Active forms of education, sustained interventions, and other features traditionally associated with successful quality improvement interventions were not associated with effect size and showed no evidence of confounding the association between clinician education-only strategies and outcome.Multidimensional interventions using audit and feedback to improve antibiotic selection were less effective than interventions using clinician education alone. Although confounding may partially account for this finding, our results suggest that enhancing the intensity of a focused intervention may be preferable to a less intense, multidimensional approach.