Acute cardiac care & resuscitation
A pilot study to predict cardiac arrest in the pediatric intensive care unit
Authors' conclusion
We have created high-performing models that identify signatures of in-hospital cardiac arrest (IHCA) that may not be evident to clinicians.
Study context
- Population
- Children in the pediatric ICU at the Johns Hopkins Hospital15 with cardiac arrest
- Participants
- 1,145 children
- Comparison
- Six machine learning models for prediction of cardiac arrest
- Outcome
- Prediction of cardiac arrest up to three hours in advance
- Follow-up
- Up to three hours in advance
Methodological detailspilot study, single-centre
- Pilot study
- Single-centre
Performance
As reported in the abstract
Cardiac arrest prediction (independent test set)
| Test or model | auROC | auPRC | Sensitivity | Specificity |
|---|---|---|---|---|
| XGBoost | 0.971 | 0.797 | 99.5% | 69.6% |
XGBoost
auROC0.971
auPRC0.797
Sensitivity99.5%
Specificity69.6%
Original publication
A pilot study to predict cardiac arrest in the pediatric intensive care unit. Resuscitation. Published 16 February 2023. doi:10.
Summarised from the published abstract; every number is checked against it