Early prediction of pressure injury risk in hospitalized patients using supervised machine learning models based on nursing records

This study aimed to employ supervised models for predicting pressure injuries in hospitalized patients using data collected within the first eight hours after admission, thus providing a tool for early assessment and prevention. The dataset included 446 patients admitted to multiple hospital wards at Félix Bulnes Clinical Service Hospital in Santiago, Chile, between January and December 2022. After preprocessing the data through imputation and feature selection, we evaluated five machine learning models, Decision Tree, Logistic Regression, Random Forest, Extreme Gradient Boosting, and Support Vector Machines, using cross-validation. Their performance was assessed using accuracy, precision, recall, and AUC. The incidence of pressure injuries was 18.8%, with 9.86% occurring in the adult medical-surgical unit. Key risk factors identified included size, weight, total risk score, hospital ward, dependency risk, use of anti-decubitus mattresses, physical restraints, incontinence, and pre-hospital pressure injuries (p-value < 0.01). The Random Forest model showed the best performance, achieving an AUC of 82.4%, an accuracy of 82.5%, a specificity of 86.9%, and an adjusted precision of 93.3%. These results indicate that predictive models based on early nursing records can support clinical decision-making and enable timely prevention of pressure injuries in hospitalized patients.