Predicting forensic admission among the mentally ill in a multinational setting: A Bayesian modelling approach