How can health information governance be advanced in the context of Artificial Intelligence (AI) in healthcare?

Advancing Health Information Governance in AI-Driven Healthcare

Building robust governance architectures that create trust in AI/ML technologies is essential for advancing health information governance in healthcare, focusing on transparency, accountability, fairness, and patient safety to improve clinical outcomes and reduce health disparities. 1

Core Governance Framework Components

Trust-Building Architectures

• AI governance in healthcare requires comprehensive architectures that protect individual rights while promoting public benefit 1
• Key goals include empowering patients (especially from underrepresented groups), ensuring affordable digital health, protecting digital rights, and regulating the digital-health ecosystem 1
• Governance models must adapt to different societal contexts while accounting for implications on individual health and well-being 1

Data Quality and Representation

• Data used for AI development must be adequate, representative, well-characterized, and reusable 1
• Standardized methodologies for data quality improvement should be implemented, including plan-do-study-act (PDSA) or define-measure-analyze-improve-control (DMAIC) cycles 2
• Key data quality dimensions include accuracy, consistency, security, timeliness, completeness, reliability, accessibility, objectivity, relevancy, and understandability 2

Transparency and Documentation

• Formal assessments of bias and societal impact should be conducted and reported, including equality impact assessments, algorithmic impact assessments, and medical algorithmic audits 1
• Algorithm "auditing" processes should recognize groups or individuals for which decisions may not be reliable, reducing implications of bias 1
• Transparent documentation of datasets is critical for mitigating algorithmic bias and promoting health equity 1

Implementation Strategies

Risk Mitigation and Monitoring

• Data users must identify uncertainties or variable performance in groups and clearly state clinical implications as risks 1
• Strategies to monitor, manage, and reduce risks should be documented as part of AI implementation 1
• Post-market surveillance and clinical follow-up are essential, especially when risk of harm differs between groups 1

System Maintenance and Security

• Decision support systems need regular updates to mitigate effects of changing data quality, population characteristics, and clinical practices 1
• Cybersecurity measures must be implemented, including firewalls, secure transmission modes, and encryption to protect electronic protected health information 2
• A coordinated national approach to data protection is more effective than relying solely on health systems and vendors 2

Multidisciplinary Governance

• Governance should involve clinical, technical, and administrative stakeholders 2
• Establish data sharing review committees with appropriate stakeholder representation 2
• Clearly define roles for data stewards, managers, and users within organizations 2

Addressing Bias and Promoting Inclusivity

Inclusive Development Approaches

• Open-source software improves transparency and participation in AI technology design 1
• Citizen science involves non-professional scientists in research, broadening perspectives 1
• Increase diversity of data by promoting involvement of people familiar with potential bias, context, and regulations throughout algorithm development 1

Bias Mitigation Strategies

• When necessary, implement debiasing techniques to decrease variation in performance across subgroups 1
• Reevaluate race correction practices that may exacerbate inequities in disease outcomes and treatments 1
• Make the purpose of data sharing activities transparent to all stakeholders, including patients 2

Regulatory and Legal Considerations

Liability and Oversight

• AI governance architectures should engage all stakeholders (developers, clinicians, researchers) to continuously evaluate safety and effectiveness 1
• Companies should file applications with regulatory bodies like FDA to allow marketing of algorithms 1
• Post-market safety monitoring similar to phase IV drug trials should be implemented 1

Regulatory Frameworks

• Proactive regulatory approaches are needed to mitigate AI harms before they happen 1
• The lack of transparency in how AI mechanisms formulate clinical recommendations creates challenges in establishing standards of care 1
• International collaboration on AI governance is essential to ensure coherent solutions and allow countries to benefit from each other's work 3

Common Pitfalls and Challenges

• Data Quality Issues: Unstructured medical data lacking uniform standardization directly affects AI algorithm quality 4
• Algorithmic Bias: Can affect clinical predictions and exacerbate health disparities if not properly addressed 4
• Opacity: Affects patients' and doctors' trust in medical AI 4
• Security Vulnerabilities: Can pose significant risks and harm to patients 4
• Responsibility Attribution: Unclear attribution when accidents occur with medical AI 4

By implementing these governance strategies, healthcare organizations can advance health information governance in the context of AI, ensuring that these technologies improve clinical outcomes while maintaining patient safety, privacy, and equity.