Applying a BP Neural Network Model to Predict the Length of Hospital Stay

摘要

Length of hospital stay (LOS) is closely related to the control of medical costs and the management of hospital resources. In this study, we implemented a data mining approach based on Back-Propagation (BP) neural net-works to construct a LOS prediction model that can help doctors and nurses individualize patient treatment. We analyzed medical data from 921 patients whowere diagnosed as cholecystitis and treated in a Chinese hospital between 2003and 2007. Our prediction model achieved approximately 80% accuracy, and revealed 5 LOS predictors: days before operation, wound grade, operation approach, charge type and number of admissions. The model can be easily used toprovide suggestions for doctors and nurses determining patient LOS.