Generative AI equity in cardiovascular care needs lifecycle governance

Generative AI is being positioned as a tool to improve cardiovascular care through clinical decision support, automated interpretation of ECGs and echocardiograms, risk prediction, patient education, telemedicine, remote monitoring, and trial recruitment. Cardiovascular disease accounts for over 36% mortality worldwide and 50% mortality in the United States, with higher burdens among Non-Hispanic Black, Indigenous, rural, and lower-income populations.

Equity risks can emerge across the AI lifecycle. Representation bias can leave minority groups and women underrepresented in training data, while measurement and labeling bias can arise from flawed EHR data, device inaccuracies, or poor proxies such as health care costs. Deployment bias can occur when systems validated in academic centers are used in rural or resource-limited settings. A study of 56 automated arrhythmia detection algorithms found a 39% accuracy difference between Black and Asian subjects despite similar aggregate metrics.

An equity-focused framework emphasizes diverse and contextualized data, transparent documentation, accountability, community engagement, digital access, and continuous recalibration. Governance should include pre-deployment validation, real-time monitoring, subgroup performance reporting, vendor disclosure of limitations, and post-deployment surveillance. Effective use of generative AI in cardiovascular medicine depends not only on technical safeguards but also on infrastructure investment, trust-building, and attention to the structural conditions that shape health outcomes.