Heart failure & cardiomyopathies
Race, Sex, and Age Disparities in the Performance of ECG Deep Learning Models Predicting Heart Failure
Authors' conclusion
The biases found in this study warrant caution against perpetuating disparities through the development of machine learning tools for the prognosis and management of heart failure.
Study context
- Population
- Patient encounters with 12-lead ECGs for standard clinical indications at Stanford Hospital, 2008–2018
- Participants
- 326,518 patient encounters
- Exposure
- ECG deep learning model performance across race, ethnicity, age and sex
- Outcome
- Model performance (AUC) for predicting incident HF, stratified by race, ethnicity, age and sex
- Follow-up
- Within 5 years
Methodological detailssingle-centre
- Single-centre
Main finding
As reported in the abstract
AUC of primary model across protected attributes
- Performance of the primary model declined with age
- No significant differences between racial groups overall
- Significantly worse performance in Black patients aged 0 to 40 years vs all other racial groups in this age group, most pronounced among young Black women
Observational data: associations, not effects.
Further findings
Incident HF within 5 years
- Cases of incident HF
- 59 817
Bias mitigation strategies
- Disparities did not improve with integrating race, ethnicity, sex and age into model architecture, separate models per racial group, or equal racial representation
- Probability thresholds individualized for race, age and sex offered substantial improvements in F1 scores
Original publication
Race, Sex, and Age Disparities in the Performance of ECG Deep Learning Models Predicting Heart Failure. Circ Heart Fail. Published 21 December 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →