Integration of an electronic hand hygiene auditing system with electronic health records using machine learning to predict hospital-acquired infection in a health care setting
André Luís Franco Cotia,Anderson Paulo Scorsato,Elivane da Silva Victor,Marcelo Prado,Guilherme Gagliardi,José Edgar Vieira de Barros,José R Generoso Jr,Fernando Gatti de Menezes,Mariana Kim Hsieh,Gabriel O V Lopes,Michael B Edmond,Eli N Perencevich,Michihiko Goto,Sérgio B Wey,Alexandre R Marra
DOI: https://doi.org/10.1016/j.ajic.2024.09.012
2024-09-21
Abstract:Background: Hospital-acquired infections (HAIs) increase morbidity, mortality, and health care costs. Effective hand hygiene (HH) is crucial for prevention, but achieving high compliance remains challenge. This study explores using machine learning to integrate an electronic HH auditing system with electronic health records to predict HAIs. Methods: A retrospective cohort study was conducted at a Brazilian hospital during 2017-2020. HH compliance was recorded electronically, and patient data were collected from electronic health records. The primary outcomes were HAIs per CDC/National Healthcare Safety Network surveillance definitions. Machine learning algorithms, balanced with Random Over Sampling Examples (ROSE), were utilized for predictive modeling, including generalized linear models (GLM); generalized additive models for location, scale, and shape (GAMLSS); random forest; support vector machine; and extreme gradient boosting (XGboost). Results: 125 of 6,253 patients (2%) developed HAIs and 920,489 HH opportunities (49.3% compliance) were analyzed. A direct correlation between HH compliance and HAIs was observed. The GLM algorithm with ROSE demonstrated superior performance, with 84.2% sensitivity, 82.9% specificity, and a 93% AUC. Conclusions: Integrating electronic HH auditing systems with electronic health records and using machine learning models can enhance infection control surveillance and predict patient outcomes. Further research is needed to validate these findings and integrate them into clinical practice.