Classification Models for Early Prediction of Surgical Site Infections

Abstract

Nosocomial infections and antimicrobial resistance are problems of enormous magnitude that impact the morbidity and mortality of hospitalized patients as well as the cost of healthcare. Implementation of control policies and programs assumes an important role in reducing the negative impacts on patient welfare and hospital management. In this context, this study aims to analyze how the integration of data from different sources and analysis of risk factors and history pre- and post-surgery can help in the early detection of surgical site infection. For this project, data was collected from Hospital Senhora de Oliveira in Guimarães (HSOG), which concerns surgeries performed, admissions, emergency room visits, antibiotic prescriptions, and recorded infections. The results showed that it is possible to use predictive models for better infection management (models with SE above 88% and PC and AUC above 90%).