Lessons for safe and effective adoption of AI in healthcare

The workshop, held on 17 March 2023, brought together representatives from academia, industry and the health and social care sectors, along with patients and regulators, who discussed the lessons that can be learnt from the AI technologies already used in healthcare.

DM Schedules is one such AI technology currently being piloted, with the potential to benefit patient health, reduce inequalities and save precious time and money for the healthcare system. The technology selects patients at high risk of missing hospital appointments and sends them tailored reminders and support. Hospital appointments are often missed due to inequalities, such as single parents struggling to find childcare, and are associated with worse health outcomes as well as costing the NHS £1.2 billion a year.

The workshop participants’ views on action to address the main barriers to adopting AI in the UK healthcare system include:

A confidence boost: better communication and involvement

Some doctors, nurses, patients and the public lack confidence in the way AI is used in healthcare. This may be due to a lack of time or incentive, concerns about health data privacy, and/or not enough awareness of the limitations of the current standard of care and how this could be improved by AI.

All those who make, use and benefit from AI technologies should be involved in their development from early on. This includes patients and the public who provide their data, the IT teams who install the technology, and doctors and other healthcare professionals who use it. These groups should also have a say in how to best communicate transparent and clear information about the technologies.

A tech-ready NHS: improved digital infrastructure

Many healthcare settings still use paper notes and lack basic digital systems such as computers with up-to-date operating systems. The digital systems that do exist in NHS trusts and other healthcare bodies are often incompatible with one another, limiting the ability to transfer data and scale-up AI technologies. Digital infrastructure in the healthcare system should be improved and standardised so that systems can speak to one another.

Data collection from patients should ensure that datasets, which are used to train AI-based technologies, are diverse and representative of the whole population, to avoid introducing inequalities. Standards should be developed for data collection and use, which will also help with public trust about how health data is used.

Effective regulation: suitable evaluation of AI

When it comes to assessing effectiveness and value for money, AI-based technologies have specific challenges that regulators need to account for. Some AI-based technologies continue to learn and evolve once in use, based on the data they process, so their performance could change over time. In addition, AI technologies often have multiple positive impacts. A technology designed to improve the accuracy of patient referrals from a GP will also benefit staff time at the specialist clinic they are being referred to. This makes assessing their true economic value, and who should pay for the technology, complex.

A system for monitoring, evaluating, and updating the technology after adoption is essential, as is assessing the effects across the whole system. Use of the new AI and Digital Regulations Service, which provides guidance to developers and aims to improve the pathway of these technologies through the regulatory system, is encouraged.

Coordination: calling for strategic direction

For AI to revolutionise healthcare, strategic direction at a national level is needed to encourage coordination across the currently fragmented healthcare system.

The workshop follows a recently published report by the National Engineering Policy Centre, hosted by the Royal Academy of Engineering, Towards autonomous systems in healthcare, that explored the potential benefits and the challenges that would need to be addressed before adopting such technologies, following a series of interviews held with healthcare experts.

Professor Lionel Tarassenko CBE FREng FMedSci, Professor of Electrical Engineering, University of Oxford, and Co-Chair of the workshop, said: “AI has the potential to make healthcare more inclusive, if implemented and managed safely and effectively. There are many possible use cases where AI could help to improve the healthcare system, including for scenarios such as identifying those likely to miss their appointments and using generative AI with Electronic Health Records to spot early signs of undiagnosed disease, including cancer. The workshop highlighted the importance of strategic direction, thorough monitoring, and evaluation, and of clear communication when adopting new technologies to increase public confidence about changes that may affect many of us. We hope that this workshop report will assist efforts to coordinate strategic direction and regulation of AI applied to healthcare, where the UK has several advantages, including NHS data and world-leading research groups and companies.”

Professor Jackie Hunter CBE FMedSci, Board Director, OI Pharma Partners Ltd, and Co-Chair of the workshop, said: “We are only just beginning to realise the vast benefits AI can offer to our struggling healthcare system and the patients who rely on it. To enable AI to reach its full potential to improve efficiency, effectiveness and the health of the public, the solutions summarised in this workshop report are an important starting point.”