The Ethics of Leveraging Routinely Collected Patient Data for AI Development: Mixed Methods Study

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Bioethics

Background:

• Electronic health records (EHR) offer vast potential for medical research and AI development.
• However, EHR data quality and ethical considerations for AI training remain significant challenges.
• The intersection of AI and EHR data ethics is an underexplored research area.

Purpose of the Study:

• To examine ethical challenges at the nexus of EHR data and AI development.
• To propose practical recommendations using the Dutch LEAPfROG project as a case study.
• To ensure ethical and effective use of EHR data for AI innovation.

Main Methods:

• Mixed-methods design combining scoping literature review and systematic search.
• Two stakeholder workshops with patients, clinicians, ethicists, and AI developers.
• Guidance ethics approach applied to iterative project development phases.

Main Results:

• Identified four key themes: data privacy/consent, public trust/regulation, fair representation/generalizability, and responsible AI integration.
• Highlighted risks of re-identification, fragmented governance, health inequities, and overreliance on AI.
• Emphasized decontextualization risks and the need for clear data reuse purposes.

Conclusions:

• Responsible AI development necessitates understanding EHR data's production, interpretation, and governance within clinical contexts.
• Technical solutions and top-down regulation are insufficient; stakeholder-led, context-sensitive approaches are vital.
• Ethical and effective AI innovation relies on transparency, fairness, and clinical relevance, prioritizing patient and clinician perspectives.