Watch back – #EvaluateDigiHealth: Regulation and evaluation of AI – are we there yet?
The evidence standards framework developed in 2018 by NICE in conjunction with NHS England, Public Health England, MedCity and DigitalHealth.London was seminal in providing clarity to digital health innovators on ‘What good looks like’ for digital health technologies to demonstrate their value in the UK health and care system. This includes evidence of effectiveness relevant to the intended use(s) of the technology and evidence of economic impact relative to the financial risk. At the time, the standards were not applicable to digital health technologies built on AI.
Neelam Patel, Chief Executive Officer, MedCity
Henry Potts, Professor, UCL
Jeanette Kusel, Director NICE Scientific Advice, NICE
Elizabeth Bohm, Head of Data and AI Policy, HRA
We’ve summarised the key themes that were discussed by the panel:
What are NICE and the HRA already doing collectively in this space?
NICE – Jeanette highlighted that NICE published an updated version of Evidence Standards Framework earlier this year and added that there will be a new update for AI coming. The framework covers all digital technologies and includes AI technologies that have fixed algorithms; however, it does not include continuous learning AI. Jeanette explained that NICE recently started pilots for digital products which proved to be a valuable exercise in testing if their methods were fit for purpose, and that the team will continue to fine-tune their methods and use this to evaluate AI technology.
Moreover, she added that NICE is working with the Accelerated Access Collaborative to ensure that the right evidence is gathered during the Accelerated Access Collaborative’s AI in Health and Care Award. She commented that it was great project to help shape evidence collection and hoped these digital solutions would come through to NICE later in their development stage. Moreover, NICE are also collaborating with other regulators such as the MHRA, CQC and HRA – with the aim to streamline the regulatory evaluation pathway for AI and data driven tech.
HRA – Elizabeth commented that HRA now have an established team that oversee AI and data driven technology and are working with other organisations via the multi-agency advice service. She added that the HRA also have a new program that looks at the HRA processes, particularly around applications to the confidentiality advisory group who grant access to non-identifiable confidential patient information – they are working with the group to streamline this process. Moreover, she added that the HRA are also improving a process for medical devices in conjunction with the MHRA, entitled ‘Streamlining data-driven devices’, it is currently in the user research phase. Looking to the future, she commented that the HRA were looking forward to working further with NICE, CQC and the MHRA to ensure that the regulatory pathway is clear for companies.
What impacts have you seen? Have you measured impact of use?
NICE – Jeanette explained that NICE received key feedback from users such as commissioners on the first draft of evidence standards framework and took this into consideration when producing the current version. She added, as NICE develop a new version of the framework for AI, that they will be seeking feedback from developers and commissioners around what information is missing concerning AI products. She added that NICE saw the impact of their work at the Accelerated Access Collaborative’s AI in Health and Care Award and conducted a gap analysis, the results of which were incorporated into clinical development plans of AI products.
HRA – Elizabeth explained that the HRA found via user research, that some companies are unaware of who the HRA are. Therefore, they are planning to reach out to more communities to raise awareness, including students. She added that the HRA are also identifying challenges users have around data access and would like to empower communities by offering them different products, for example, in collaboration with the MHRA, the HRA created a synthetic dataset around COVID, A&E and Cancer.
What does the future hold with respect to the global landscape?
HRA – Elizabeth explained that one of the HRA’s ambitions is to attract the best researchers in the world, and therefore highlighted the importance of connecting their world-class data store with additional data sets. Elizabeth also commented that the main challenges faced by the HRA is maintaining public trust and disparity. She added that the HRA will be producing guidance on data driven technologies and the interrelationship with data holders, to maintain trust and encourage collaborative working.
NICE – Jeanette agreed that collaboration was key, due to the fast-moving nature of the field. She explained that in addition to developing their evidence standards framework specifically for AI, that they are also working collaboratively with the MHRA in updating their regulatory framework for AI and software. She added that NICE are also working with four key partners to collaborate on the following areas:
1) Evidence requirements
2) Algorithm updates
3) Continuous learning algorithms
Globally, NICE are working with other global health technology association bodies, such as CADTH (Canada), MSAC (Australia) and partners in the UK devolved nations, and the MHRA are looking into EU requirements and are also engaging with the FDA. She highlighted the need for alignment of these requirements through collaborative working.
How can users access the multi-agency service?
Jeanette explained that NICE have just completed user research and are now prototyping different solutions. She added that there would be an information platform where users could access useful information such as: the role of regulators, pathways, legislations to meet, best practice guidance at different points of pathway, and how to engage with regulators. A bespoke advice service may also be added, or a signpost to existing routes.