Clinical Prediction Models for Prognostication After Out-of-Hospital Cardiac Arrest: A Systematic Review and Meta-Analysis

OBJECTIVES: Summarize the prognostic performance of existing clinical prediction models (CPMs) for neuroprognostication after out-of-hospital cardiac arrest (OHCA). DATA SOURCES: We searched Medline and Embase databases from inception to June 1, 2025. STUDY SELECTION: We selected English-language studies that included adults with OHCA and evaluated a CPM for the prediction of poor functional outcome. We excluded derivation cohorts for prognostic scores and excluded models without at least two external validation cohorts. DATA EXTRACTION: Two authors performed citation screening and data extraction. Where possible, we pooled the sensitivity and specificity of poor functional outcome, and the area under the receiver operating characteristic curve (AUROC) values for each CPM. We assessed risk of bias using the Prediction model study Risk of Bias Assessment Tool, and rated the certainty of evidence using the Grading of Recommendations, Assessment, Development, and Evaluation. DATA SYNTHESIS: We included 39 observational cohorts (95,037 patients) evaluating 11 different CPMs, with the two most common scores being the OHCA and Cardiac Arrest Hospital Prognosis (CAHP) scores. An OHCA score greater than or equal to 17 had a pooled sensitivity of 81.8% (95% CI, 65.8-91.4%) and specificity of 74.2% (95% CI, 58.8-85.3%), while a score of greater than or equal to 32 had a pooled sensitivity of 64.9% (95% CI, 44.0-81.3%) and specificity of 89.5% (95% CI, 75.9-95.8%) for poor functional outcome (low certainty). A CAHP score greater than or equal to 150 had a pooled sensitivity of 81.3% (95% CI, 77.7-84.4%) and specificity of 77.0% (95% CI, 70.6-82.4%) for poor functional outcome (moderate certainty). Pooled AUROCs across the 11 CPMs varied from 0.75 to 0.88, with substantial heterogeneity. CONCLUSIONS: CPMs for neuroprognostication after OHCA demonstrate only moderate accuracy, with substantial heterogeneity across validation cohorts. These limitations restrict their clinical utility, particularly for irreversible decisions such as withdrawal of life-sustaining therapy.