Background A growing number of studies have sought to detect clusters of all suicides, but few have sought to identify clusters of method-specific suicides.

Methods Data on railway suicides occurring in Victoria, Australia, between 2001 and 2012 were obtained from the National Coronial Information System. We used the Poisson discrete scan statistic to identify railway suicides that occurred close together in space and/or time. We then used a case–control design to compare clustered railway suicides with non-clustered railway suicides on a range of individual and neighbourhood factors.

Results We detected four spatial clusters that accounted for 35% of all railway suicides. Railway suicides by individuals who were hospitalised for mental illness had nearly double the odds of being in a cluster compared with those individuals who had never been hospitalised (OR 1.80, 95% CI 1.02 to 3.18). Higher frequency train services were associated with increased odds of being in a cluster (OR 1.11, 95% CI 1.03 to 1.19). No other predictors were associated with being in a cluster.

Conclusions Railway suicides that occur in clusters warrant particular attention because of the ripple effect they can have for communities and the risk that they may lead to copycat acts. Railway suicide prevention strategies should consider the fact that these suicides can occur in clusters, particularly among individuals who had previous hospitalisations for mental illness or live in areas with high-frequency train services.