Data mining of hospital suicidal and self-harm presentation records using a tailored evolutionary algorithm

The assessment of outcomes for the Gold Coast Mental Health and Specialist Services Suicide Prevention Strategy implementation required data on suicidal and self-harm presentations to be captured from the Emergency Department Information System (EDIS) database. Suicidal and self-harm presentations are not uniformly coded in the EDIS and require human assessment to differentiate these presentations from other cases (e.g., accidental injuries). A novel evolutionary algorithm was used to learn weighting variables from a psychiatrist-rated training dataset in order to generate an appropriate cut-off score for identifying suicidal and self-harm presentations from EDIS. The resulting Searching EDIS for Records of Suicidal Presentations (SERoSP) program was then run on a psychiatrist-rated validation dataset using the weights generated by the algorithm. SERoSP is optimised to be able to detect suicidal and self harm
presentations with a high degree of accuracy (a sensitivity of 0.95 and a
specificity of 0.92). The SERoSP program is a reliable and cost-effective tool for the identification of suicidal and self-harm presentations from EDIS data, and is currently being successfully used in the suicide prevention strategy evaluation.