Arrhythmias & electrophysiology
Prediction of early death after atrial fibrillation diagnosis using a machine learning approach: A French nationwide cohort study
Authors' conclusion
Machine learning algorithms predict early death after atrial fibrillation diagnosis and may help clinicians to better risk stratify atrial fibrillation patients at high risk of mortality.
Study context
- Population
- Newly diagnosed AF patients seen in French hospitals, 2011–2019
- Participants
- 2,435,541 participants
- Exposure
- Deep neural network vs clinical risk scores
- Outcome
- Death within the first year following AF diagnosis; discrimination by C index (validation set)
- Follow-up
- Within the first year following AF diagnosis
Methodological detailsnationwide, machine learning prediction model development and validation, 70% training / 30% validation split
- Nationwide
- Machine learning prediction model development and validation
- 70% training / 30% validation split
Main finding
As reported in the abstract
Prediction of death within the first year (C index, validation set)
| Group | C index |
|---|---|
| Deep neural network | 0.785 (95% CI, 0.781-0.789) |
| CHA2DS2-VASc | 0.597 |
| HAS-BLED | 0.562 |
| Charlson Comorbidity Index | 0.643 |
| Hospital Frailty Risk Score | 0.626 |
C index
Deep neural network0.785 (95% CI, 0.781-0.789)
CHA2DS2-VASc0.597
HAS-BLED0.562
Charlson Comorbidity Index0.643
Hospital Frailty Risk Score0.626
- Deep neural network vs CHA2DS2-VASc: P < 0.0001 — model superior
- Deep neural network vs HAS-BLED: P < 0.0001 — model superior
- Deep neural network vs Charlson Comorbidity Index: P < 0.0001 — model superior
- Deep neural network vs Hospital Frailty Risk Score: P < 0.0001 — model superior
Observational data: associations, not effects.
Further findings
Deaths within the first year after AF diagnosis (whole cohort)
- Patients who died
- 342,005 (14.4%)
- Time to death, days
- 83 (SD 98); median 37 [10-129]
Original publication
Prediction of early death after atrial fibrillation diagnosis using a machine learning approach: A French nationwide cohort study. Am Heart J. Published 16 August 2023. doi:10.
Summarised from the published abstract; every number is checked against it