External validation and comparative performance of ACLS, SARICA, NULL-PLEASE, BDR scores for in-hospital mortality after out-of-hospital cardiac arrest: a multicenter retrospective cohort study.

BACKGROUND

Prognostic scores are used to estimate outcomes after out-of-hospital cardiac arrest (OHCA), but their comparative performance in external cohorts remains unclear. We externally validated the ACLS, SARICA, NULL-PLEASE, and BDR scores for predicting in-hospital mortality after return of spontaneous circulation.

METHODS

We conducted a retrospective multicenter cohort study using the OHCA Research Database. Adult patients with emergency medical services-attended OHCA who achieved return of spontaneous circulation were included. Patients without data required for score calculation, including brain computed tomography, blood gas analysis, or lactate measurement, were excluded. Discrimination, calibration, and diagnostic performance were assessed using area under the receiver operating characteristic curve, decile-based calibration plots, and standard diagnostic indices.

RESULTS

Among 1036 included patients, 718 (69.3%) died during hospitalization. The BDR score showed the highest discrimination for in-hospital mortality, with an area under the receiver operating characteristic curve of 0.775 (95% CI, 0.745-0.805), followed by NULL-PLEASE (0.756; 95% CI, 0.723-0.788), modified SARICA (0.746; 95% CI, 0.714-0.779), and modified ACLS (0.662; 95% CI, 0.627-0.697). BDR, NULL-PLEASE, and modified SARICA all significantly outperformed modified ACLS, whereas differences among these three scores were not statistically significant. At Youden-index thresholds, BDR demonstrated the highest sensitivity and negative predictive value, but specificity was only moderate across all models.

CONCLUSIONS

BDR, NULL-PLEASE, and modified SARICA demonstrated fair discrimination for in-hospital mortality after OHCA, whereas modified ACLS showed lower discrimination. Because clinically practical thresholds had limited specificity, these scores should be interpreted cautiously and used only as adjuncts to serial multimodal assessment.