Predicting Forensic Admission among the Mentally Ill: A Bayesian Approach

Our objective was to explore protective and risk factors for a later forensic admission among the mentally ill in a multinational setting. Between 1998-2000, 308 forensic and general psychiatry patients were recruited for a case-control study in four countries. A greedy Bayesian algorithm was utilized to search for generalized factors from datasets and create a merger model (naive Bayesian fusion) using a new type of triangulation. The evidence were assessed with various measures and the data included independent training and test sets. The most influential risk factors included violent crime prior to the crime that lead to the index hospitalization (PO 11.8, 95% CrI 5.9 to 30.3), conviction of the biological father (9.4, 4.7 to 20.0), and no use of psychotropic medications before the age of 18 (7.7, 4.3 to 16.5). The merger model indicated high discriminative power, robustness and accuracy.