Clinicians Generative AI Usage Guidance

Last edited on January 18, 2024

Introduction

These guidelines are suggested for the responsible and ethical use of Artificial Intelligence (AI) technologies within our EVMS Medical Group. AI and machine learning technologies can enhance patient care, improve operational efficiency, and support clinical decision-making. However, their use must align with ethical principles, legal requirements, and best practices to ensure patient safety and data security. 

Guidelines

1. Ethical considerations:
   1. All AI applications in healthcare must prioritize patient welfare, safety, and privacy.
   2.  Decisions made by AI systems should be transparent and interpretable to clinicians.
   3. Patient data used for training AI models must be de-identified and comply with relevant privacy regulations, such as HIPAA (Health Insurance Portability and Accountability Act).
2. Data governance:
   1. Proper data governance practices must be followed for collecting, storing, sharing and managing healthcare data used in AI applications.
   2. Data quality, integrity, and security must be maintained throughout the AI development lifecycle.
3. Accountability and oversight:
   1. A designated Medical Group AI oversight committee or responsible individual(s) should oversee the development, implementation, performance and monitoring of AI systems.
   2. Regular audits and assessments of AI applications should be conducted to ensure compliance with policies and regulatory requirements.
4. Clinical validation:
   1. AI algorithms intended for clinical decision support must undergo rigorous testing, validation, and peer review before deployment.
   2. The performance of AI systems should be continuously monitored and validated against established benchmarks and to ensure at a minimum data quality and stability.
5. Informed consent:
   1. Patients must be informed about the use of AI technologies in their care, and informed consent should be obtained when AI significantly influences clinical decisions.

Operational guidelines

1. AI development and deployment:
   1. Identify a need for AI application, considering potential clinical benefits and feasibility.
   2. Establish a multidisciplinary team, including clinicians, data scientists, and IT experts.
   3. Select appropriate data sources and ensure data compliance with privacy regulations.
   4. Develop and train AI models, ensuring proper validation and testing.
   5. Deploy AI solutions in a controlled and monitored environment.
   6. Continuously assess and refine AI algorithms based on real-world performance and feedback.
2. Data management:
   1. Implement data governance policies to ensure data quality and security.
   2. Establish mechanisms for data access control and encryption.
   3. Regularly review data sources to maintain accuracy and relevance.
3. Transparency and explainability:
   1. Ensure that AI decision-making processes are transparent and documented.
   2. Provide clinicians with understandable explanations of AI-driven recommendations.
4. Patient privacy and consent:
   1. Implement robust patient data protection measures, adhering to HIPAA and other relevant regulations.
   2. Educate patients about the use of AI in their healthcare and obtain informed consent when necessary.
5. Training and education:
   1. Provide training and education to staff involved in AI development and utilization.
   2. Keep clinicians and staff informed about the benefits and limitations of AI technologies.
6. Monitoring and evaluation:
   1. Regularly monitor the performance and impact of AI applications on patient care and operations.
   2. Conduct periodic audits to ensure compliance with policies and regulations.

AI usage guidance

Specific examples of how AI can be used in different clinical settings (e.g., diagnosis, treatment, decision-making).

• Example 1: Using AI to analyze medical images and help diagnose diseases.
• Example 2: Using AI to personalize treatment plans for patients.
• Example 3: Using AI to monitor patients' health and predict potential complications.

Potential legal and ethical implications of using AI in clinical practice

• Legal implications:
  • Liability: Who is responsible for decisions made by AI-powered systems, especially if they lead to adverse outcomes?
  • Data privacy and security: How can patient data be protected from unauthorized access and misuse?
  • Algorithmic bias: How can bias in AI algorithms be identified and mitigated to ensure fair and equitable care for all patients?
  • Certification and regulation: What regulations and standards should be in place to ensure the safety and effectiveness of AI-powered medical devices and applications?
• Ethical implications:
  • Autonomy and informed consent: How can patients be informed about the use of AI in their care and retain control over their health decisions?
  • Transparency and explainability: Can patients and healthcare professionals understand how AI algorithms arrive at their decisions and have access to relevant information?
  • Human oversight and accountability: How can human judgment and oversight be integrated with AI to ensure responsible and ethical decision-making?
  • Distributional justice: How can access to AI-powered healthcare be equitable and avoid exacerbating existing disparities in healthcare access and outcomes?

Additional resources

• 10 Promising AI Applications in Health Care
• The potential for artificial intelligence in healthcare
• JAMA AI resources for clinicians