Predicting High Urinary Tract Infection Rates in Skilled Nursing Facilities: A Machine Learning Approach.

Area of Science:

• Healthcare-associated infections
• Infectious disease epidemiology
• Machine learning in healthcare

Background:

• Urinary tract infections (UTIs) are prevalent healthcare-associated infections in skilled nursing facilities (SNFs).
• High UTI rates are linked to increased healthcare costs, prolonged stays, and mortality.
• Predictive models are needed to identify SNFs at high risk for UTIs.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting high UTI rates in SNFs.
• To identify key SNF characteristics associated with increased UTI risk.
• To inform targeted infection prevention strategies.

Main Methods:

• Analysis of 94,877 SNF-year observations (2019-2024).
• Utilized Random Forest, XGBoost, and LightGBM machine learning models.
• Employed Shapley Additive exPlanations for model interpretation and feature importance.

Main Results:

• Machine learning models significantly outperformed logistic regression in predicting SNF UTIs.
• The Random Forest model identified rural SNFs and number of staffed beds as primary predictors.
• Average length of stay and geographic location were also significant predictors of high UTI rates.

Conclusions:

• Facility-level characteristics can be effectively used with machine learning to predict UTI risk in SNFs.
• Identifying high-risk SNFs allows for proactive infection prevention measures.
• These findings support data-driven approaches to enhance patient safety in post-acute care settings.