Spatiotemporal and behavioural correlates of cycling near misses: evidence from helmet-mounted video data

This study examines cycling near misses in London using naturalistic data collection methods. Sixty commute cyclists were equipped with helmet-mounted 360-degree cameras and GPS devices to record their journeys over two weeks. Participants verbally reported near misses as they occurred, with 94 events recorded across 317 h of footage. Close passes were the most common near miss type (25 events), followed by near left/right hooks (22 events) and pulling out incidents (17 events). Most near misses occurred during peak hours and on roads without cycling infrastructure. Poisson regression modelling revealed that higher cycling speeds, time spent on shared lanes, and not having previous solo crashes were associated with fewer near misses, while time spent on 30mph roads and morning peak cycling increased near miss rates. The findings suggest that cycling infrastructure plays a protective role and highlight the need for targeted interventions to address close passes and junction conflicts.