Using a Bayesian belief network model to categorize length of stay for radical prostatectomy patients

A clinical pathway implements best medical practices and represents sequencing and timing of interventions by clinicians for a particular clinical presentation. We used a Bayesian belief network (BBN) to model a clinical pathway for radical prostatectomy and to categorize patient's length of stay (LOS) as being met or delayed given the patient's outcomes and activities. A BBN model constructed from historical data collected as part of a retrospective chart study represents probabilistic dependencies between specific events from the pathway and identifies events directly affecting LOS. Preliminary evaluation of a BBN model on an independent test sample of patients' data shows that model reliably categorizes LOS for the second and third day after the surgery (with overall accuracy of 82% and 84%, respectively).