Insights from the CERSI-AI National AI Readiness Forum
By Joshua Graham, Postgraduate Student - LLM Law, Technology and Innovation, The University of Strathclyde
The potential for AI technology to benefit the NHS in Scotland is beginning to be seen, as in the case of AI chest x-ray analysis used to detect lung cancer earlier than ever before. But these new technologies also come with new types of risk - how do we identify these novel types of risk, and what needs to be done to mitigate them so that AI can be used safely, ethically and effectively in the NHS?
This question was at the heart of a recent AI Readiness workshop which took place at the Wellcome Trust and was hosted by the CERSI-AI team from the University of Birmingham, a new UK Government-funded centre of excellence tasked with future regulation of AI and Digital Healthcare in order to accelerate innovation that will improve people’s lives. The workshop sought input from attendees on the creation of a new framework or tool that will help adopter organisations, like Scottish health boards and hospitals, assess and increase their level of AI readiness.
I was pleased to attend as part of a forthcoming project on AI Governance in Scottish healthcare, a collaboration between the University of Strathclyde, the Scottish government and the West of Scotland Innovation Hub. The project seeks to examine AI Governance in healthcare, understood as the policies and procedures for making decisions and managing risks relating to the use of AI.
The big goal of our project is to contribute to the national policy framework for the use of AI in health and care, to be published later this year, and to inform the development of best practice in Scotland from a wider system level right down to locally within health boards. It was really beneficial to hear clinicians, academics and industry figures from across the UK speaking about their work in other parts of the country which can inspire what’s being done here in Scotland.
Some takeaways from the event that can help NHS Scotland get AI-ready and unlock the potential of this new technology:
Be proactive
As the saying goes, the best time to start was yesterday but the next best time is now. One speaker highlighted the need to turn AI from ‘hype’ to ‘habit’, to stop seeing it as a panacea or inevitability and to start thinking practically about how to integrate it into the day-to-day of healthcare.
One great example of this comes from Andy Mayne and the Somerset NHS Foundation Trust, who leads a team of Data Scientists tasked with building and test AI models. He spoke about one tool which was designed to assist triage in an emergency department by predicting the likelihood of a patient requiring a hospital stay. The result: staff had on average nearly 3 hours more time to find a bed for patients who would go on to be admitted. The AI Policy that they created to support adoption is comprehensive and the result of 3 years of development. It provides a useful reference for health boards in Scotland, who can draw on the groundwork of others to get a head start in the making of their own AI policies, and shows some of the practical steps we can take right now to become AI ready.
(credit: Andy Mayne, Somerset NHS Foundation Trust)
Stay problem-focussed
Participants in the workshop emphasised that AI technology should only be introduced with a clear result in mind, not adopted haphazardly or ad-hoc. Developing a strong, shared understanding of governance across the AI product lifecycle would ensure that AI use is always responding to a genuine need. For example, the pre-deployment stage of the lifecycle identifies the problem by concentrating on the ‘pain points’ that require an AI product solution; testing, monitoring and evaluation of a product in its deployment stage ensures accountability and safety; post-deployment and decommissioning mean that tools only remain in use if they remain effective and provide value for money.
Clarity around the AI product lifecycle and a laser focus on the problem at hand will foster stronger use-cases for AI tools and will naturally enable other aspects of AI readiness, as in the creation of funding models, procurement processes, collaborative networks and long-term strategy.
(credit: OneLondon, AI Centre for Value Based Healthcare)
Bring people along together
AI represents a fundamental change to the way we understand and experience the world. It is important to ensure that nobody is left behind on the journey towards AI adoption, especially in the acute context of public health and care.
Patient safety stood out amongst participants as a big priority, but other types of risk associated with AI were highlighted, like operational, financial and legal risks, as well as risks to NHS staff. This complex landscape reflects the need to include diverse perspectives and take a multidisciplinary approach when trying to balance the conflicting priorities and methodologies at play, so we can mitigate these risks and enjoy the benefits of AI.
The importance of Patient and Public Involvement and Engagement was stressed, with the explainability and transparency of AI-based decisions being emphasised, and the involvement of trade unions was valued when considering the changes to the working lives of NHS staff that AI will bring. Ethics committees that are baked into the AI product lifecycle and feature representation from patients and staff, as well as other groups whose input is critical, is another method for ensuring wide representation.
Participation is key, and along with Accountability, Non-Discrimination, Empowerment and Legality, constitutes the Scottish Human Rights Commission’s PANEL principles of a human rights based approach. The focus of our project will be on how such an approach can be embedded in the AI governance of Scottish healthcare, and the workshop provided the opportunity to see great examples of these values in action. Fostering collaborations like this will ensure that all four nations can continue to learn from each other, and will contribute to safe, effective and ethical AI use in NHS Scotland.
(credit: CERSI-AI)