ML predicts extubation success with pooled AUC 0.90, but clinical readiness is lacking

A meta-analysis of 14 studies (n=34,322) shows ML models predict extubation success with a pooled AUC of 0.90 (95% CI:0.82-0.95). Only 2 studies reported external validation, and high heterogeneity limits routine clinical use.

• Systematic review and meta-analysis of 26 studies; 47 ML models from 14 studies (n=34,322 critically ill adults) were meta-analyzed for predicting planned extubation outcome.
• Pooled AUC was 0.88 for classical ML and 0.85 for deep learning models; the best-performing model per study reached a pooled AUC of 0.90 (95% CI: 0.82-0.95).
• Limitation: Only 2 studies provided externally validated AUCs with confidence intervals; all other pooled estimates derive from at-risk-of-bias internal validation, and between-study heterogeneity was high.

RadPigeon summaries are original and for information only. They are not clinical advice.