Dandelion Releases Report on AI Bias and Health Equity

Driving Health Equity through Inclusive Clinical AI Validation

While many AI models perform equitably, significant exceptions highlight how algorithmic bias—especially against older adults and rural populations—can deepen health disparities if left unaddressed. This report insights highlight the critical need for rigorous AI validation to ensure that clinical AI algorithms perform equitably across diverse patient populations, particularly those historically underserved in healthcare.

As AI increasingly has the potential to guide clinical decision-making and transform care, concerns over algorithmic bias remain a barrier to adoption. Dandelion Health has developed a validation framework that evaluates AI models against social determinants of health (SDoH) and demographic measures. The report shares key findings from conducting these validations for electrocardiogram (ECG)-based AI models, and provides actionable insights to ensure AI advances equitable health outcomes rather than reinforcing disparities.

Among the report’s key findings:

• AI often underperforms for older adults. Many models struggle to accurately assess patients over 65 due to gaps in training data, leading to misdiagnoses and delayed care.

• Rural populations face unique AI challenges. Algorithms trained on urban data often fail to generalize to rural healthcare settings, limiting their effectiveness for nearly 20% of Americans.

• Bias linked to socioeconomic factors is less common but carries high stakes. When errors disproportionately affect lower-income or socially vulnerable patients, they risk deepening existing healthcare inequities.

To read the full report, download it today.