ECMO PAL: using deep neural networks for survival prediction in venoarterial extracorporeal membrane oxygenation
Authors' conclusion
ECMO PAL demonstrated high generalisability across ECMO regions and outperformed existing, widely used scores.
Study context
- Population
- Patients on venoarterial extracorporeal membrane oxygenation (VA-ECMO) in the international Extracorporeal Life Support Organisation (ELSO) registry; training 2017–2020, external validation in adult patients from 2021external validation N = 5015
- Participants
- 18,167 patients
- Comparison
- ECMO PAL vs SAVE, Modified SAVE and ECMO ACCEPTS scores for predicting in-hospital mortality
- Outcome
- Prediction of in-hospital mortality
Methodological detailsretrospective cohort, deep neural network, fivefold cross-validation, external validation, international registry
- Retrospective cohort
- Deep neural network
- Fivefold cross-validation
- External validation
- International registry
Performance
As reported in the abstract
In-hospital mortality prediction accuracy, external validation (2021 cohort)
| Test or model | Value | Accuracy |
|---|---|---|
| ECMO ACCEPTS | 54.7% | 54.7% |
| SAVE | 61.1% | 61.1% |
| Modified SAVE | 62% | 62% |
Value
ECMO ACCEPTS54.7%
SAVE61.1%
Modified SAVE62%
Accuracy
ECMO ACCEPTS54.7%
SAVE61.1%
Modified SAVE62%
- Validation accuracy (ECMO PAL), 99% CI 71.1-74.3%; training accuracy 75.5%
- 72.7%
Original publication
ECMO PAL: using deep neural networks for survival prediction in venoarterial extracorporeal membrane oxygenation. Intensive Care Med. Published 7 August 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →