A statistical method for evaluating suicide clusters and implementing cluster surveillance.

The absence of any standard definition of suicide cluster events hinders understanding of the prevalence of the problem, hinders the development of appropriate public health responses to observed clusters, and ultimately hinders investigation of the mechanisms underlying contagious communication of suicidal behavior. The authors introduce a Poisson mixture model for assessing potential clusters of adolescent suicide, apply that model to the monthly incidence rates of adolescent suicide for one populous US county over the last 11 years, and generate 99% tolerance limits with 95% confidence for the number of suicides which may occur by chance within specific intervals of time in that county. The suicide incidence data showed a remarkable fit to a single Poisson distribution, suggesting it is not unreasonable to consider the cases as randomly-distributed and independent events. The authors conclude that there is no evidence that adolescent suicides occurred in clusters in the place and in the time frame under study, and recommend the Poisson mixture model for ascertaining clusters as well as implementing cluster surveillance.