Suicide clusters in Australian youth: A comparison of cluster detection methods

Aims: Suicide clusters are significantly more common in young people. Yet, there is currently no gold-standard method for detecting suicide clusters and there is some evidence that the different methods for detecting clusters give inconsistent results. Our aim was to conduct a comparative analysis of suicide clusters in young people using 1) The scan statistic; 2) A systematic search of coronial inquests into suicide clusters and; 3) Descriptive network analysis. We sought to identify similarities and differences in cluster detection methods and to quantify rates of exposure to suicide among cluster members. Methods: Suicide data were obtained from the National Coronial Information System from 2006-2015 for Australians aged 10-24 years. We included N=3027 suicides from seven Australian state and territories. Suicide clusters were determined using: 1) Poisson discrete scan statistics; 2) A systematic search of coronial inquests; and 3) Descriptive network analysis involving psychosocial links between three or more cluster members. We analysed the prevalence of suicide clusters, the geospatial overlap between clusters, the proportion of overlap among cluster members and quantified rates of exposure to suicide for each cluster method. We examined the narrative text of police and coronial reports for evidence exposure to suicide and psychosocial links between cluster members. Results: Eight suicide clusters (69 suicides) were identified using the scan statistic; seven (40 suicides) from coronial inquests into suicide clusters; and 11 (37 suicides) using descriptive network analysis. Of the eight clusters detected using the scan statistic, two suicide clusters were identified using descriptive network analysis and one was identified in coronial inquest reports. Of the seven coronial inquests into suicide clusters, four suicide clusters were detected using descriptive network analysis and one was detected using the scan statistic. Geospatial congruence among overlapping clusters ranged from 25 to 100%. Overall, 9.2% (12 suicides) of individuals were identified using more than one cluster method. Prior exposure to suicide was 10.1% (N=7) for suicide clusters identified using the scan statistic; 32.5% (N=13) for clusters identified using coronial inquests reports; and 56.8% (N=21) for clusters identified using descriptive network analysis. Conclusion: Different methods for determining suicide clusters identified different suicide clusters and cluster members. The use of multiple cluster detection methods has the potential to increase cluster response activities and suicide prevention interventions in communities that would not otherwise be detected by a single cluster method.