Artificial Intelligence: Issues for Medical Scheme Policy and Governance

Let’s Get Practical – and Local

Issues raised by AI, particularly its newest form, generative AI, are global, but context is key. Medical schemes will need to consider implications for strategy, operations and governance across a range of AI-related issues. We highlight four issues for understanding and action.

Data Governance and Compliance

Data privacy and security  

AI system development and operation requires access to data; medical schemes and their administrators are holders of the biggest repositories of health data in the country. Ensuring that this data, much of it personal and sensitive, is handled securely and in compliance with existing privacy laws, such as POPIA, is essential. Robust data governance frameworks will be needed to ensure compliance with legal standards, regular audits, and employee training on data handling and privacy.

Regulatory compliance                                                                                          

The international regulatory environment for AI in healthcare and its clinical use is evolving rapidly. The US Food and Drug Administration has approved hundreds of AI-based applications, but SA’s health technology licensing environment is immature, and clinical implementation outside the US is so far quite limited. In that country, health insurers will establish compliance teams dedicated to monitoring and implementing AI-related regulatory changes. In SA, this should be on the agenda, and schemes should engage with industry bodies (e.g., CMS, BHF, PHISC) and regulatory agencies to stay informed and participate in shaping policy and regulations.

Image generated for this article using Midjourney as a generative AI program.

Ethical and Responsible AI Use

Ethical use

AI applications raise ethical concerns, such as biases in decision-making, which could lead to unfair treatment of certain groups. The best way to avoid such bias is local development on representative local data sets. Schemes should develop and enforce ethical guidelines for AI use, focusing on fairness, transparency and accountability and ensuring decision-making that uses AI algorithms is explainable.

Managing AI risks     

AI technologies come with their own set of risks, including technical failures, the aforementioned algorithmic biases, and potential misuse. The scheme’s risk management framework should include AI, with regular risk assessments, risk mitigation strategies, and contingency plans.

Legal implications of AI decisions                                                                        

Decisions made by AI systems can have legal implications so the scheme’s legal advisory team should analyse implications of AI-based decisions and create pathways for decision review, including how to override decisions if necessary.

Image generated for this article using Midjourney as a generative AI program.

AI Technological Integration and Innovation

Integration with existing systems                                                                         

Integrating AI into existing IT systems can be challenging and costly – the biggest hurdle may be the shortage of skilled staff. Schemes can plan phased and scalable AI integrations, starting with pilot projects to test compatibility and effectiveness while ensuring IT infrastructure support.

Innovation and continuous improvement

AI technology is rapidly evolving. Adoption requires a culture of innovation within the organisation, investing in learning and R&D to stay abreast of change, challenges, and opportunities.

Stakeholder Engagement and Workforce Empowerment

Stakeholder engagement and transparency                                                      

Lack of understanding and trust in AI among stakeholders, including customers and employees, is understandable. Chatbots, for example, have pros and cons and are not always popular. Who wouldn’t prefer to speak to a human? Schemes will need to implement transparent communication policies regarding AI use, benefits, and limitations while engaging stakeholders through education and open dialogue.

Skill development and training                                                                             

Successful implementation of AI requires a workforce skilled in new technologies. Investment in employee training and development programmes and partnerships with academic institutions or specialised training providers may pay off.

Customer-centric AI solutions                                                                               

AI solutions must align with customer needs and enhance their experience. Involve customers in developing and refining AI applications, with regular feedback gathering.

Image generated for this article using Midjourney as a generative AI program.

Where to From Here?

Since the November 2022 public launch of ChatGPT the pace of change has been frantic; the following twelve will validate or refute Gartner’s predictions.

AI is already in our systems, though often invisible to users, but the turning point in overt, general adoption in healthcare work of various kinds is coming when Microsoft, Google, Meta, and other “Big Tech” vendors whose products we use every day embed these technologies in their systems.

But, healthcare has unique considerations, including the complexity of delivery and funding systems, the sensitivity and privacy of health data, and unique requirements for reliability and equity, so caution is needed.

AI in healthcare is a key focus for Insight. We want to enable and support the ethical, compliant, safe, and beneficial use of AI technologies in health care in SA. Expect more from us on this topic soon.