Digital Approaches for Understanding Behaviors Prior to Suicide

30% of suicides occur in public places, including railways, cliffs, bridges and shopping centres. These suicides can adversely affect bystanders, and increased media attention may lead to further suicides in these locations. Closed circuit television (CCTV) cameras have been installed in some locations frequently used for suicide, increasing the likelihood of third-party intervention. Some degree of automatic monitoring is currently possible, for example initiating a response when someone has climbed a safety fence – however this is often tragically too late. There is an urgent need to identify the earlier behaviours that precede a suicide attempt to allow sufficient time for intervention.

 Behaviours which precede a suicide attempt can include swapping between platforms within a railway station, pacing back-and-forth, remaining stationary in one location for extended periods of time, or leaving personal possessions or clothing behind on the ground. This project will use data from different sources to further our understanding of these behaviours, and incorporate this knowledge into an automated system to detect these behaviours.

This project examines CCTV data from public locations frequently used for suicide. A unique partnership has been established, which is representative of two-thirds of public hotspot locations in Australia and which will provide insights into similar locations internationally. CCTV data will be manually reviewed to determine what behaviours can be observed prior to a suicide attempt, and whether behaviours from one location are generalizable to other settings. A previously developed proof-of-concept automated detection system will also be retrained to identify visual features associated with these behaviours, and evaluated to determine the accuracy of the algorithm. These insights will inform a knowledge translation and future co-design strategy to develop appropriate responses to help save lives.