Diagnostic or prognostic study
Optimizing patient selection for primary prevention implantable cardioverter-defibrillator implantation: utilizing multimodal machine learning to assess risk of implantable cardioverter-defibrillator non-benefit
- Prognostic model development and external validation study
- multicentre
- machine learning
Authors' conclusion
ML models combining ECG time-series features and clinical variables were able to predict non-arrhythmic mortality within three years after device implantation in a primary prevention population, with robust performance in an independent cohort.
Study context
- Population
- Patients with ischaemic, dilated or non-ischaemic cardiomyopathy and left ventricular ejection fraction (LVEF) ≤ 35% implanted with a primary prevention implantable cardioverter-defibrillator (ICD) 2007–2021, two academic hospitals
- Participants
- 1,010 patients
- Comparison
- Development vs External
- Developmentn = 550
- Externaln = 460
- Outcome
- ICD non-arrhythmic mortality (mortality without prior appropriate ICD therapy), predicted by machine learning (ML) models combining clinical variables and 12-lead electrocardiograms (ECG) time-series features
- Follow-up
- three-year follow-up
Performance
As reported in the abstract
Prediction of non-arrhythmic mortality (XGBoost models) · 3-year follow-up
AUROC, internal validation (Hospital A development cohort)
0.90 (95% CI 0.80-1.00)
AUROC, external cohort (Hospital B)
0.79 (95% CI 0.75-0.84)
Further findings
Mortality · three-year follow-up
- Patients who had died
- 16.0%
- Deaths meeting criteria for non-arrhythmic mortality
- 72.8%
Original publication
Optimizing patient selection for primary prevention implantable cardioverter-defibrillator implantation: utilizing multimodal machine learning to assess risk of implantable cardioverter-defibrillator non-benefit. Europace. Published 15 September 2023. doi:10.1093/europace/euad271
Summarised from the published abstract; every number is checked against it