Prediction of Healthcare Associated Infections in an Intensive Care Unit Using Machine Learning and Big Data Tools

Abstract

Healthcare associated infections (HAIS) can be acquired by patients during their stay in a hospital. HAIS are very endemic, causing a huge burden for the patients and for the health care system. We propose a machine learning approach to predict HAIS in an intensive care unit (ICU), combining heterogeneous data from longitudinal electronic health records and from microbiology laboratory. A NoSQL database, mongoDB, was developed to consider a big data environment. Results show that the fusion of these heterogeneous data sources provides 82% accuracy when a random forest algorithm was considered. In this study, the age, the length of stay, the bed where the patient stayed, and the admission month, are the most relevant risk factors to predict HAIS in the ICU.