Heart failure & cardiomyopathies
Machine Learning-Based Prediction of in-Hospital Death for Patients with Takotsubo Syndrome: The InterTAK-ML Model
Authors' conclusion
A ML-based approach for the identification of TTS patients at risk of adverse short-term prognosis is feasible and effective. The InterTAK-ML model showed unprecedented discriminative capability for the prediction of in-hospital death.
Study context
- Population
- Patients with Takotsubo syndrome (TTS), International Takotsubo (InterTAK) Registryexternal cohort 1037
- Participants
- 3,482 patients
- Comparison
- Internal validation vs External validation
- External validationn = 1,037
- Outcome
- Model performance (AUC, sensitivity, specificity) for predicting in-hospital death
- Follow-up
- In-hospital
Methodological detailsregistry-based, ridge logistic regression machine learning (ML) model, random 75%/25% train/internal validation split, external validation
- Registry-based
- Ridge logistic regression ML model
- Random 75%/25% train/internal validation split
- External validation
Performance
As reported in the abstract
In-hospital death prediction (InterTAK-ML model)
| Test or model | AUC | Sensitivity | Specificity |
|---|---|---|---|
| Internal validation | 0.89 (0.85-0.92) | 0.85 (0.78-0.95) | 0.76 (0.74-0.79) |
| External validation | 0.82 (0.73-0.91) | 0.74 (0.61-0.87) | 0.79 (0.77-0.81) |
AUC
Internal validation0.89 (0.85-0.92)
External validation0.82 (0.73-0.91)
Sensitivity
Internal validation0.85 (0.78-0.95)
External validation0.74 (0.61-0.87)
Specificity
Internal validation0.76 (0.74-0.79)
External validation0.79 (0.77-0.81)
Secondary outcomes
Overall in-hospital death (incidence)
- Overall incidence of in-hospital death
- 5.2%
In-hospital death by phenotypic cluster (K-medoids clustering on the 10 most important features)
- Cluster in-hospital death rate
- 28.8%
- Cluster in-hospital death rate
- 15.5%
- Cluster in-hospital death rate
- 5.4%
- Cluster in-hospital death rate
- 0.5%
- Six clusters with different risks of in-hospital death, consistent in the external cohort
- Cluster in-hospital death rate (value printed illegibly in the abstract)
- Abstract prints five rates for six clusters; one rate is printed illegibly
Original publication
Machine Learning-Based Prediction of in-Hospital Death for Patients with Takotsubo Syndrome: The InterTAK-ML Model. Eur J Heart Fail. Published 18 September 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →