Artificial intelligence bias in the prediction and detection of cardiovascular disease

Area of Science:

• Artificial Intelligence in Medicine
• Cardiovascular Disease Epidemiology
• Health Equity Research

Background:

• Artificial intelligence (AI) algorithms show promise for identifying individuals at risk of cardiovascular disease (CVD), enabling early interventions.
• However, AI bias can emerge during algorithm development, validation, and evaluation.
• Biased algorithms may perform inadequately in historically marginalized populations, exacerbating existing healthcare disparities based on age, sex, race, ethnicity, and socioeconomic status.

Purpose of the Study:

• To discuss the origins and impacts of AI bias in the context of CVD prediction and detection.
• To introduce an AI health equity framework.
• To review strategies for mitigating AI bias throughout the AI lifecycle.

Main Methods:

• This perspective synthesizes current knowledge on AI bias in healthcare.
• It reviews literature on the sources and consequences of bias in CVD prediction algorithms.
• It proposes a framework for promoting health equity in AI development and implementation.

Main Results:

• AI bias can stem from data, algorithm design, and evaluation processes.
• Consequences include poorer performance in underrepresented groups, leading to amplified healthcare inequities.
• Existing bias mitigation strategies can be applied across the AI lifecycle.

Conclusions:

• Addressing AI bias is crucial for ensuring equitable cardiovascular disease detection and prevention.
• Implementing an AI health equity framework can guide the development of fair and effective algorithms.
• Proactive bias mitigation strategies are essential for realizing the full potential of AI in improving cardiovascular health outcomes for all populations.