Antibiotic Review Kit for Hospitals (ARK-Hospital): a stepped wedge cluster randomised controlled trial

Background: Strategies to reduce antibiotic overuse in hospitals depend on clinicians reviewing antibiotics which have been started empirically. There is a lack of evidence on how to do this effectively. We evaluated a multifaceted behaviour change intervention (ARK) aimed at reducing antibiotic consumption in hospitals by increasing prescriber decisions to stop antibiotics at clinical review. Methods: We performed a stepped-wedge, hospital-level, cluster -randomised controlled trial using computer-generated sequence randomisation of 39 acute hospitals to 7 calendar-time blocks (12/February/2018–01/July/2019). Co-primary outcomes were monthly antibiotic defined-dailydoses (DDD) per acute/medical admission (organisation-level, superiority) and all-cause 30-day mortality (patient-level, non-inferiority, margin 5%). Clusters were eligible if they admitted nonelective medical patients, could identify an intervention “champion” and provide pre-intervention data from February/2016. Sites were followed up for a minimum of 14 months. Intervention effects were assessed using interrupted time series analyses in each cluster. Overall effects were derived through random-effects meta-analysis, using meta-regression to assess heterogeneity in effects across prespecified factors. Trial registration was ISRCTN12674243. Findings: Adjusted estimates showed a year-on-year reduction in antibiotic consumption (-4.8%, 95%CI: -9.1%,-0.2%, p=0.042) following the ARK intervention. Among 7,160,421 acute/medicaladmissions, we observed a -2.7% (95%CI: -5.7%,+0.3%, p=0.079) immediate and +3.0% (95%CI: - 0.1%,+6.2%, p=0.060) sustained change in adjusted 30-day mortality. This mortality trend was not related to the magnitude of antibiotic reduction achieved (Spearman’s ?=0.011, p=0.949). Whilst 90- day mortality odds appeared to increase over time (+3.9%, 95%CI:+0.5%,+7.4%, p=0.023), this was not observed among admissions before COVID-19 onset (+3.2%, 95%CI:-1.5%,+8.2%, p=0.182). Length of hospital stay was unaffected. Interpretation: The weak, inconsistent effects of the intervention on mortality are likely to be explained by the COVID-19 pandemic onset during the post-implementation phase. We conclude that the ARK-intervention resulted in sustained, safe reductions in hospital antibiotic use.