Key result
Machine learning models significantly enhanced predictive discrimination for in-hospital mortality following cardiac arrest compared to the APACHE III score (AUROC 0.87 vs 0.80, p<0.001).
Why the study?
Resuscitated cardiac arrest has high mortality, but existing illness severity scores have limited ability to estimate risk of adverse outcomes.
Do machine learning models improve the prediction of in-hospital mortality in patients resuscitated from cardiac arrest compared to existing illness severity scores?
Observational (n=39,566)
Yes
Do machine learning models improve the prediction of in-hospital mortality in patients resuscitated from cardiac arrest compared to existing illness severity scores?
Effect estimate: AUROC 0.87 (95% CI 0.86-0.88)
Absolute Event Rate: 0.87% vs 0.8%
p-value: p=<0.001
No takes yet. Share an insight, caveat, or question.
ML models may improve post-arrest mortality prediction; hypothesis-generating and requires prospective validation before clinical adoption.
Nanayakkara et al. (2018) conducted an observational in Cardiac arrest admitted to intensive care unit (n=39,566). Machine learning models (ensemble and gradient boosting machine) vs. APACHE III and ANZROD scores was evaluated on Discrimination for in-hospital mortality (AUROC) (AUROC 0.87, 95% CI 0.86-0.88, p=<0.001). Machine learning models significantly enhanced predictive discrimination for in-hospital mortality following cardiac arrest compared to the APACHE III score (AUROC 0.87 vs 0.80, p<0.001).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: