AI models that analyse electrocardiograms (ECGs) could help identify people at high risk of heart failure early, and potentially reduce racial disparities in heart failure outcomes. But whether these models perform equally well for different groups of patients has been little studied. This study used ECGs from 326,518 patient encounters at Stanford Hospital to evaluate a model trained to predict heart failure within five years, comparing its performance by race, ethnicity, age and sex.
Overall, there were no significant differences in performance between racial groups, and performance declined with age. However, among patients aged 0 to 40, the model performed significantly worse for Black patients than for any other racial group, with the largest differences among young Black women.
The researchers tested several ways of reducing these disparities: including race, ethnicity, sex and age in the model, training separate models for each racial group, and training it on data with equal racial representation. None of these improved the disparities. Adjusting the model’s decision thresholds individually by race, age and sex did lead to substantial improvements. The authors warn that machine learning tools for heart failure risk perpetuating existing disparities, and suggest these individualised thresholds as one way to reduce them.
Learn more about this study here: https://doi.org/10.1161/CIRCHEARTFAILURE.123.010879
Reference:
Kaur, D., Hughes, J.W., Rogers, A.J., Kang, G., Narayan, S.M., Ashley, E.A., & Perez, M.V. (2024). Race, Sex, and Age Disparities in the Performance of ECG Deep Learning Models Predicting Heart Failure. Circulation: Heart Failure, 17(1), e010879.