Paper Title
Forcasting Hospital Length of Stay Using Medical Systems with Machine-Learning Models

Abstract
This paper evaluates the contribution of medical systems to efficient decision-making by physicians, investigating whether these systems predict well length of stay (LOS) and can lead to more efficient medical care in hospitals. Log-files were retrieved from an integrative medical system that serves a health maintenance organization (HMO). It was found that using an integrative medical system in the hospitals predict well the longer LOS. The longer LOS was used as dichotomous variable (positive for above 5 days and negative for less than 5 days).The reported results were received using three methods for classification: Random forest Gini, Random forest Information Gain and Neural Network. The accuracy levels were between 84%-86%. The PRAUC (Area Under the Precision-Recall Curve (PRAUC)) levels were between 63%-71%. The results show the potential of using real medical data to forecast hospital longer LOS using integrative medical systems with machine-learning models. Keywords - Medical Informatics, Health Care IT, Electronic Medical Record, Integrative Medical Systems.