A Machine Learning Approach to Predict Health Care–Acquired Urinary Tract Infections From Electronic Nursing Documentation

Reducing health care-acquired urinary tract infections is a common goal among health care providers and organizations. This study took an initial step toward this goal by exploring context-based variables contributing to HAUTI. We included a comprehensive list of nursing assessments and applied multiple machine learning methods to process the datasets and manage missing data. Among the models tested, eXtreme Gradient Boosting (XGBoost) emerged as the most effective in predicting HAUTI, identifying associations between improved skin integrity, mobility, and neurological status monitoring, which may be linked to lower HAUTI rates. However, our results should be carefully interpreted, given the study's significant missing data. The findings of this study reinforce the necessity of high-quality data to support the interpretation of machine learning (ML) models in clinical settings.