Forty-three studies were included (n greater than 320,177). These included 22 RCTs (n greater than 47,873), three quasi-RCTs (n=4,076) and 19 controlled before-and-after studies (n greater than 268,228). One study was listed both as an RCT and as a controlled before-and-after study. These studies provided 55 separate comparisons.
The authors stated that the overall study quality was fair; most studies failed to meet most or all of the internal validity and generalisability criteria. Most of the included studies were clustered studies in which the unit of allocation differed from the unit of analysis. Only nine studies reported the number of clusters and only one study reported the intracluster correlation.
The median reduction in the proportion of patients receiving antibiotics was 9.7 per cent (interquartile range 6.6 per cent to 13.7 per cent; 30 comparisons from 20 studies). No single quality improvement strategy appeared to be more effective than the others (p=0.85 for comparison across all strategies). Active clinician education strategies were associated with a non statistically significant decrease in antibiotic prescribing compared to passive education strategies (12.9 per cent versus 7.0 per cent, p=0.096).
Non quantitative analysis (18 comparisons from 16 studies): three of four studies of active clinician education plus mass media campaign and two large studies of active clinician education plus patient education were associated with significant reductions in antibiotic prescribing. Three trials evaluating reminders reported no reductions in prescribing.
Delayed prescriptions (seven studies): The median rate of antibiotic use was lower in intervention compared to control groups (37.5 per cent, interquartile range: 27.3 per cent to 39.7 per cent versus 75.0 per cent, interquartile range 62.3 per cent to 87.0 per cent, six studies, p not reported).