Building a learning health and social care system with artificial intelligence

During a recent DHC meeting of members and key figures from NHSX, we explored challenges in developing sustainable health and social care AI solutions.

Several key themes emerged:

  • Focus on outcomes. Defining improved health and social care outcomes should be the starting point for AI, as with any other technology solution. Our national AI strategy should therefore be framed within that context, not the other way round.

  • Access to live real-world data. The proposals in the draft data strategy to make data available for AI and machine learning development in trusted research environments are welcome, but continuously improving algorithms also need access to real life data. E.g., if we are to improve the flow of patients through a pathway of care, it’s essential to have live feedback to complete that learning loop. There are ways to do this safely and protect patient data, but this is an essential element that needs to be made available by default.

  • Build patient and public confidence. We need to do more to build public confidence and trust in how data is used.

  • Expand functionality. There is a much wider range of potential applications beyond the early areas of imaging, dermatology and screening, all of which focus on image recognition. We can do much more to help patients flow through the system to the right point-of-care, and develop personalised support to help patients manage their conditions more effectively.

  • Patient safety and innovation. Rapid iteration has been the fastest route to transformative AI deployments in other sectors. In healthcare, it’s clearly vital such iterations are safe, but our regulatory and procurement models must support iterative processes rather than act as a stop-start brake.

  • AI awards – some decisions to award or reject proposals have puzzled providers. We need greater clarity about the decision-making process and better feedback to inform future applications and make the most of this investment.

  • Procurement models are often focused on point solutions rather than transforming outcomes. We need a greater emphasis on problem solving and incentives that encourage improved outcomes.

Going forward, the DHC will work to address the areas covered above, and explore how lessons from how other countries address these challenges.