Evaluating AI Risks in Healthcare: Strategies for Boards to Mitigate Bias and Legal Issues

The use of AI in healthcare is no longer just for testing but is quickly becoming a key part of diagnosing diseases, managing patients, and running daily tasks. This fast change means that boards need good ways to manage AI at the highest level.

Holly J. Gregory, a partner at Sidley Austin LLP, says boards must look at how AI affects company plans and risk management. AI relies a lot on good data. There is always a chance of bias or mistakes if the data or algorithms are wrong, which can cause unexpected problems. She also says boards should keep up with how AI is used in the company and any outside AI products they use. Knowing this helps boards prepare for risks in operations, money, reputation, and law.

At a recent Yale meeting, 42% of CEOs worried about the bad effects AI could have on society. These worries show that leaders, including those in healthcare, have to manage AI risks to stop serious issues.

Common Sources of Bias in Healthcare AI

Bias in AI systems is a big problem. Healthcare AI tools help with things like diagnosing diseases, planning treatment, and deciding which patients need care first. If AI is biased, it can give wrong or unfair results. This hurts patients and can make health differences worse.

Research shows three main kinds of bias in healthcare AI and machine learning:

• Data Bias: This happens when the data used to train AI is missing important groups or is not complete. For example, if data does not include enough people of different ages, races, or genders, the AI might be unfair to those groups. Also, data can become outdated if medical practices or diseases change over time.
• Development Bias: Bias can come up when making the AI if the wrong features are chosen or data is handled poorly. Sometimes, the AI is accidentally made to prefer certain results.
• Interaction Bias: Differences in how doctors or hospitals use AI tools can change how well the AI works. This means AI may not work the same everywhere.

Careful checks are needed at every step—when developing and using AI. This helps make sure AI tools are accurate, fair, and clear. It also builds trust among doctors and patients who use these tools.

Legal and Regulatory Challenges in AI Healthcare Tools

Healthcare groups in the U.S. have to follow many rules to use AI responsibly and fairly. Boards need to know about these rules to help with following laws and managing risks.

The National Artificial Intelligence Initiative Act of 2020 defines AI as a machine system that can make predictions or decisions based on human goals. This law supports federal AI research and policies but also shows the risks if AI is not properly controlled.

Regulators are focusing on bias, transparency, data privacy, and responsibility:

• The Equal Employment Opportunity Commission (EEOC) started a program in 2021 to make sure AI used in hiring follows civil rights laws. This matters for healthcare employers using AI for hiring or staff reviews.
• New York City’s Local Law 144 (2021) requires employers to do yearly bias checks on AI hiring tools. This law is one of the first in the U.S. and might influence other states and healthcare groups.
• The National Institute of Standards and Technology (NIST) created the Artificial Intelligence Risk Management Framework (AI RMF 1.0) in 2023. It gives advice on how to manage AI risks by being clear, measuring outcomes, and being responsible.
• The European Union’s AI Act (June 2023) groups AI by risk level and sets rules for high-risk AI, like medical devices. Healthcare groups working in the EU need to follow these rules.
• In the UK, AI rules focus on safety, fairness, and responsibility. These ideas are becoming popular worldwide.

Boards must make sure there is clear reporting about AI use and get regular updates on risks, benefits, and law compliance. If rules are not followed, it could lead to lawsuits, fines, and harm to the healthcare provider’s reputation.

Ethical Considerations in Healthcare AI

Besides following laws, boards should keep strong ethical standards when using AI to help all patients fairly and avoid harm.

Writers like Matthew G. Hanna, Brian Jackson, and Joshua Pantanowitz say AI bias and ethical problems often come from limits in data or how AI is built. To use AI ethically, ongoing checks are needed to find bias and keep up with medical changes.

Main ethical concerns are:

• Fairness: AI must treat all groups equally and not cause harm to anyone.
• Transparency: Doctors need to understand how AI makes decisions to know if it is right for their patients.
• Accountability: Organizations must take responsibility for AI results and protect patient rights.
• Ongoing Evaluation: AI needs regular updates to fix outdated information and keep up with changes in healthcare.

Following these ethical ideas helps keep patient trust and good care.

AI Risk Management in Healthcare Workflows and Automation

Boards and managers should also think about how AI can help run healthcare tasks through automation, while watching out for risks.

AI tools that answer phone calls and do scheduling show how AI helps healthcare offices. These tools can ease staff workload and improve patient communication.

But automated tools also bring risks to watch for:

• Data Privacy: Patient information must be safely handled during automated calls to follow laws like HIPAA.
• Bias and Accessibility: Automated systems must correctly understand different voices and languages so nobody is left out.
• System Reliability: AI must be accurate and quickly pass tough cases to human staff.
• Cybersecurity: Automated phone systems need strong protection against cyber attacks.

Boards should guide policies to keep AI tools legal, ethical, and working well. Regular checks and reviews help keep these tools fair and effective.

The Role of Boards in AI Oversight and Compliance

Boards play a key part in managing AI in healthcare. This includes:

• Strategic Oversight: Knowing how AI fits with company goals and medical care.
• Risk Assessment: Finding out about risks to operations, money, reputation, and law from AI use.
• Compliance Assurance: Making sure policies follow rules on bias audits and data privacy.
• Transparency: Asking for clear explanations on how AI makes decisions to support doctors.
• Monitoring and Reporting: Setting up regular reports about AI performance, errors, and changes from teams or vendors.

Board members’ involvement helps stop unexpected problems, supports new ideas, and keeps patients safe. Gregory also says boards should think about environmental impacts, because AI training uses a lot of computer power that can affect sustainability.

Preparing Healthcare Organizations for Responsible AI Use

Healthcare leaders in the U.S. should take these steps to get ready for AI governance:

• Educate Board Members: Train them to understand AI’s strengths, limits, and rules in healthcare.
• Develop Bias Mitigation Policies: Work with AI builders and clinical teams to check data and training for fairness.
• Implement AI Risk Frameworks: Use tools like the NIST AI RMF for ongoing risk and compliance management.
• Establish Ethical Guidelines: Set up groups to watch ethics in AI use and check patient effects regularly.
• Engage Legal Counsel: Keep up with changing laws and get advice on needed changes to stay legal.
• Integrate Workflow Automation Carefully: Test new AI tools for patient satisfaction and workflow changes, and fix bias or errors as needed.

By involving boards and working with many experts, healthcare groups can use AI well while cutting down risks.

Key Takeaways

Artificial Intelligence brings both chances and challenges for healthcare in the United States. Board members must use frameworks like NIST’s and stay aware of new rules to manage bias, ethics, and legal duties well. Good leadership can help healthcare providers use AI to improve patient care and operations without causing unfairness or losing patient trust.