Desarrollo de un algoritmo para determinar el riesgo de muerte en pacientes dentro de una Unidad de Cuidado Intensivo utilizando Regresión Múltiple no Lineal

Abstract

Systems for scoring the severity of illness had been used for decades on the Intensive Care Unit (ICU) of Health Care Institutions as indicators of the patient’s health and risk of death. Even after being accepted worldwide, it has been proved that these systems have limitations and do not provide the most accurate results, it is because of this that scientists and engineers have tried different techniques in order to improve these systems. This article presents the results of the development of an algorithm that uses non-linear multiple regression to establish the death risk for patients in the ICU, and the comparison of those results with the ones given by the SAPS I traditional scoring system. Parting from a database of physiological variables measurements for 4000 patients, an extended processing of this database is made together with data analysis, to finally apply the nonlinear multiple regression techniques: Regression trees, Logistic Regression, Support Vector Machine and Artificial Neural Networks. The best results were obtained with the Support Vector Machine technique, having a better performance in comparison with the SAPS I score