Using Machine Learning to Predict and Reduce Central Line Associated Bloodstream Infections in Hospitals

What this Trial Is About

Central Line-Associated Bloodstream Infections CLABSIs are a serious problem in U.S. hospitals, leading to higher death rates, longer hospital stays, and increased costs. This research evaluates whether a machine learning ML model that predicts possible CLABSI risk can help Infection Preventionists IPs reduce infection rates compared to usual practices. The study is a prospective, multi-center, cluster-randomized trial conducted in 20 hospitals with the highest CLABSI rates. In this trial, hospitals are split into early and late groups. Early hospitals get access to the ML model via a daily dashboard that flags high-risk patients. Infection Preventionists use this information to provide targeted education and recommend best practices for central line care, including line removal when appropriate. The model was built using extensive patient data and predicts infection risk with high accuracy. The control hospitals continue with routine clinical practice without access to the model during the study period. Participants are hospitalized adults with central lines, and the model scores patients daily if their line has been in place for over 48 hours. Researchers monitor CLABSI rates per 1,000 central line-days, central line removals within 48 hours of alerts, positive blood culture rates, and IP intervention frequencies. Safety outcomes such as pneumothorax and hemorrhage are also tracked. The study duration is about five months with interim and final analyses planned to assess the impact of providing the ML model to Infection Preventionists.

Machine Learning Prediction of Possible Central Line Associated Blood Stream Infections and Rate of Reduction