Predicting Patient’s Length of Stay By Mining Hospital Data

Abstract

In the last few years, prediction of patient’s length of stay (LOS) in a hospital is one of the major research areas in healthcare and data mining. Here, the primary objective is to develop a predictive model that can help to reduce the unnecessary hospitalization and readmissions based on the historical data. Such predictive model will also help to reduce the resources spent on unnecessary hospitalization and readmissions. In this thesis, we have developed a predictive model that is based on the ensemble or blending of different individual predictors. Here, performance of individual predictors and their ensemble was evaluated on Health Heritage Prize (HHP) dataset that contains 2.6 million historical claims of 218, 415 patients. It was observed that an ensemble is significantly more effective than any individual models used in this study.