Lessons learned from the multi-agency advisory service – helping developers of AI and data-driven tech to navigate the regulatory pathway

Clíodhna Ní Ghuidhir and Rebecca Boffa outline the common challenges that developers of artificial intelligence (AI) tech face when navigating their way to the health and care market, and link to their series of articles, which provide further advice on what can be done to avoid these.

Clíodhna Ní Ghuidhir and Rebecca Boffa, principal and senior scientific advisers at NICE

AI and data-driven tech have the potential to revolutionise health and care delivery. But developers of these technologies tell us that navigating the regulatory system can be complex and challenging. For the last 8 months, we’ve been trying to solve this tricky issue. The multi-agency advisory service (MAAS) aims to create a single source of information and guidance for both developers and adopters of these technologies. It’s an exciting project that has the potential to improve access to the best health and social care innovations, making a real difference to people’s lives. We’re proud to be involved.

We’ve recently been mapping the regulatory and market access pathway: from product conceptualisation, all the way through to market entry and post market regulation. This is a first step in determining where there are opportunities for streamlining and how we can best help developers and adopters to understand what needs to be done, why, and when. This process has also identified areas where multiple regulatory bodies have a remit. We know that when a number of different organisations are involved in stages of the pathway, it can be particularly difficult to navigate. The MAAS will streamline the process by providing a single resource on how to meet the different requirements across the pathway.

We’ve also been involved in a number of other projects across the health innovation landscape, such as the AI in Health and Care Award. As part of the award process, we’ve been advising on the evaluation of new technologies using NICE’s Medtech Early Technical Assessment (META) Tool. We worked alongside developers to help them think through their evidence generation plans and how to tackle their most critical gaps. Through this work, we’ve seen first-hand the challenges that developers face on their journey to market.

One thing that has really struck us is the importance of having a clear, purposeful intended use and value claim when conceptualising and developing an AI product. Developers often put forward broad claims around the value their products can bring. This is understandable as they are keenly aware of their product’s potential. For example, they may have developed a triage tool that could be used in multiple settings to reduce admin costs and prevent morbidity and mortality. The challenge is that evidencing even one intended use through the regulatory process and one value claim through the evaluation process is resource demanding and time consuming. As a result, developers that have multiple value claims often risk spreading their resources too thinly.

We’ve also learnt how important it is for developers to plan ahead when navigating the regulatory pathway. We want the MAAS to make it easier for developers to understand all relevant regulatory requirements early on. This should include those that apply to their products now, and those that may apply as they develop their product further. Often, companies have a good feel for the evidence requirements and approvals required for the current state of a product. But they do not have a plan for navigating the regulatory pathway if and when they make any changes. As one of the unique benefits of AI is the speed at which the algorithms can evolve over time, understanding the regulatory requirements associated with these changes is critical.

These are just a couple of the challenges we’ve identified while working on development of the MAAS. We believe it’s important to share these learnings so that the AI development community can benefit. In fact, in the coming month we will be presenting this to applicants for Round 3 of the AI Award. If you’d like more information, see our series of articles on the NHSX website, covering initial steps when developing a product and how to generate the right evidence. These outline in more detail the common traps developers fall into when developing their products and offers guidance on what can be done to avoid them.

We look forward to sharing more as we continue to develop the MAAS.

Blog comments

There are no comments

Leave comment