Background Medical safety huddles are short, structured meetings for physicians to proactively discuss and respond to profession-specific patient safety concerns, with the goal of decreasing future adverse events. Prior observational studies found associations with improved patient safety outcomes, but no randomised controlled studies have been conducted.

Objective The primary objective was to determine the impact of medical safety huddles on adverse events. Secondary objectives included the fidelity of huddle implementation and the impact on patient safety culture among physicians.

Design Stepped-wedge cluster randomised trial with four sequences, and each hospital site was a cluster.

Setting Inpatient oncology, surgery and rehabilitation programmes in four academic hospitals.

Participants Physicians in participating programmes.

Intervention Medical safety huddles were adapted for local context and implemented sequentially based on a computer-generated random sequence every 2 months after a 4-month control period. All sites remained in the intervention phase for at least 9 months.

Main outcome and measures The primary outcome was the rate of adverse events, as determined through blinded chart audits of 912 randomly selected patients. The fidelity of implementation was assessed through the huddle attendance rate, number of safety issues raised in the huddles and number of actions taken in response. Patient safety culture was assessed using the Agency for Healthcare Research and Quality Hospital Survey on Patient Safety.

Results The adjusted rate of adverse events (per 1000 patient days) in the postintervention phase was 12% lower compared with preintervention (RR: 0.88; 95% CI: 0.80 to 0.98; p=0.016). The odds of having adverse events posthuddle implementation were 17% lower in the postintervention period compared with preintervention (OR intervention vs control: 0.83; 95% CI: 0.80 to 0.87; p<0.001). The mean huddle attendance rate at each site ranged from 30% to 85%, and the mean number of issues raised per huddle and the mean number of actions taken per huddle ranged from 1.6 to 3.1. The mean (SD) overall patient safety rating increased from 2.3 (0.53) to 2.8 (0.88), p=0.010. The mean per cent (SD) positive score for the composite measures of ‘Organisational learning’ increased significantly from 35% (26%) to 54% (23%), p=0.00, ‘Response to error’ 37% (24%) to 52% (22%), p=0.025 and ‘Communication about error’ 36% (28%) to 64% (42%), p=0.016 after implementation.

Conclusions and relevance Medical safety huddles decreased adverse events and may improve patient safety culture through engaging physicians.