AI Outshines Traditional Scores in Predicting Patient Deterioration
JAMA Network Open
|
Contributed by: Drex DeFord
Summary
A study in JAMA Network Open evaluated the effectiveness of early warning scores (EWS) in detecting clinical deterioration in hospitalized patients, analyzing data from 362,926 encounters across seven hospitals. The research compared three AI-based scores—eCARTv5, Rothman Index, and Epic Deterioration Index—with three traditional scores—Modified Early Warning Score, National Early Warning Score, and NEWS2—focusing on their ability to predict transfers to ICU or death within 24 hours. Results indicated that eCARTv5 had the highest predictive accuracy, with an AUROC of 0.895, while the EDI score underperformed significantly with an AUROC of 0.808. The findings underscore notable differences in the predictive power of these scores, particularly in identifying at-risk patients.
