Modelling the length of hospital stay in medicine and surgical departments

Abstract

Healthcare Associated Infections are among the world's leading public health problems and the most serious complications for hospitalized patients that can impact length of stay (LOS). In this work, medical record data of 24365 patients admitted to general surgery and clinical medicine wards were used collectively with the aim of creating models capable of predicting overall LOS, measured in days, considering clinical information. Multiple linear regression analysis was performed with IBM SPSS, the coefficient of determination (R2) was equal to 0,288. A regression analysis with ML algorithms was performed with the Knime Analysis Platform. The R2 were quite low for both multiple linear regression and ML regression analyses. The use of these techniques showed that there is a relationship between clinical variables and overall LOS. The results constitute a valid support tool for decision makers to provide the turnover index for the benefit of health policy in the management of departments.