Cardiac arrest & resuscitation
Large-scale automated phenotyping of cardiac arrest and withdrawal of life-sustaining therapy using electronic health record data
Authors' conclusion
These models establish a reproducible framework for automated cohort identification. Nearly half of comatose post-arrest patients died following WLST, with 42 % of these deaths occurring within 72 h, highlighting the impact of early prognostication decisions.
Study context
- Population
- Patients with cardiac arrest and coma, structured and unstructured EHR data from two major hospitals
- Comparison
- Cardiac arrest vs withdrawal of life-sustaining therapy (WLST)
- Outcome
- Model performance (AUROC, AUPRC, other precision metrics)
Methodological detailsmachine learning, two hospitals, holdout internal testing, cross-hospital evaluation
- Machine learning
- Two hospitals
- Holdout internal testing
- Cross-hospital evaluation
Performance
As reported in the abstract
Holdout (internal) testing
| Test or model | Area under the receiver operating characteristic (AUROC) | AUPRC |
|---|---|---|
| Cardiac arrest | 0.984 | 0.968 |
| WLST | 0.992 | 0.991 |
Area under the receiver operating characteristic (AUROC)
Cardiac arrest0.984
WLST0.992
AUPRC
Cardiac arrest0.968
WLST0.991
Further findings
- Population-level error rates — Cardiac arrest
- <0.5 %
Cross-hospital evaluation
- Strong performance for the cardiac arrest phenotype; variable generalizability for WLST, sensitivity depending on training site
Population-level error rates
- WLST estimates varied by hospital
Deaths after WLST among comatose post-arrest patients
- Of these deaths, occurring within 72 h
- 42 %
- Nearly half died following WLST
Original publication
Large-scale automated phenotyping of cardiac arrest and withdrawal of life-sustaining therapy using electronic health record data. Resuscitation. Published 8 December 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →