Deployment of a Predictive Model for Identifying High-Risk Patients for Hospital-Acquired Infections

Hospital-acquired infections (HAIs) are a significant risk to the sustainability of the healthcare system and patient safety because they impact 5-15 percent of hospitalised patients across the globe and elevate the mortality, morbidity, and healthcare expenses. This research will be focused on developing and deploying a machine learning predictive model capable of assisting in the identification of patients with the highest risks of developing HAIs within the hospital and take proactive interventions. Our approach used a detailed examination of electronic health records of $\mathbf{1 5, 0 0 0}$ patients in various hospital units and performed the ensemble learning algorithm of Random Forest, Gradient Boosting, and XGBoost to compute demographic, clinical, laboratory, and procedural variables. Feature selection was performed using recursive feature elimination and importance scoring to identify the most predictive risk factors. The deployed model achieved an accuracy of 87.3%, sensitivity of 84.6%, specificity of 89.1%, and an area under the ROC curve (AUC) of 0.912, significantly outperforming traditional risk assessment methods. Key predictive factors identified include length of stay, invasive device usage, antibiotic exposure history, and specific laboratory markers. As clinical implementation of this predictive model shows, the potential of decreasing the incidence of HAI by early detection and further prevention approaches is enormous, and patient outcomes and the decrease in the healthcare cost in favor of evidence-based infection control measures will ultimately improve patient outcomes and decrease healthcare expenditures, which is also supported by evidencebased infection control measures.

Hali tarjima qilinmagan