Taking the pain out of implementation for AI medical devices

The market launch of your AI product is a huge achievement. It is the point from which you should finally begin to enjoy returns on years of personal and financial investment. Just meeting the demands of the investors and regulators to get to that point can be all-consuming. Having made it through all of that, what will now determine your product’s appeal to potential adopters?

“You might as well spend your time and resources working through the hard stuff of implementation…” - Healthcare provider technical expert

Investing early in a tailored Go-to-Market strategy helps to answer this critical question. Otherwise, determinants of post-market success are all too easy to lose track of in the pre-market gauntlet. One key determinant of the scale and sustainability of returns your AI product can bring is whether your clients have the skill or will to endure the challenges of implementation.

A decade into the most recent AI revolution, there’s little sign of let-up in the pace of technical progress, or support from healthcare policy. Despite all this time and investment, there are relatively few business or clinical successes from clinical AI to date. This isn’t for lack of AI products on market. It’s down to the complexity of the healthcare context and all the interdependent factors which conspire against successful implementation there.

If you find that frustrating, you’re not alone. It’s the same frustration at implementation failure of evidence-based healthcare interventions that founded the field (and academic journal) of “Implementation Science” back in 2006. At its simplest, it requires the identification of stakeholders  to the healthcare niche your product targets and an understanding of their perspectives. These insights can then be applied to anticipate challenges for implementation and mitigate against them by refining your product, the healthcare intervention it sits in and the strategies that you or your clients could deploy. All this helps to minimise the risk of post-market failure.

“I’ve learned...that this closing the loop is what makes the sale...sometimes, we’re handed a package with the implementation science done” - Healthcare provider leadership

So how can you get the balance right? Navigating the path to market access is a resource-hungry necessity if you are to produce a viable product. Having said that, there isn’t much point in producing that viable product if potential clients can’t or won’t implement it.

For this (and so many things we deal with at Hardian) you probably already guessed that the devil is in the detail. The specific factors which influence implementation vary between products and the healthcare niches they target. However, the good news is there are some high-level strategies for AI implementation that are pretty generalisable opportunities to buoy your post-market prospects. This blog draws on stakeholder perspectives from across clinical AI research to highlight 3 exemplar strategies you could deploy to take the pain out of implementation for your customers:

1.       Education

2.       Clinical champions

3.       Local evaluation

Strategy 1 – Education

Thanks to ISO 13485 all compliant vendors of AI medical devices already ensure that “any user training… is available or planned to be available”. Don’t make the mistake of seeing training as just another requirement to satisfy though. Training represents a great opportunity for you and your product to build positive relationships with adopters, whilst allowing you to learn more about their needs and values.

Developing competencies in AI-enabled healthcare is an unmet need for adopters. This is because the training adopters receive from degree programmes and continued professional development are yet to meaningfully deliver. It represents a great opportunity for welcome contributions from topic experts like you. When planning training materials and events, take time to think about the stakeholder groups you need to target, what their interpretation of your product’s value proposition is and what their learning needs are for their product interactions. A recent NHS report on developing healthcare workers' confidence in AI can be a helpful starting point in conceptualising learning needs.

An obvious time to focus your training efforts for clients is during the initial integration. This period can be very demanding for individuals and organisations and any help you offer through actionable materials or effective external support will be well received and remembered. Another training opportunity comes much earlier though. Putting on educational events for clinicians and other adopters before the start of procurement processes, or even market access, allows you to understand and help the community of individuals who will ultimately determine your product’s success. Besides meeting these individuals’ educational needs around clinical AI, you can also arrange benefits for them through credit toward their mandatory continued professional development. Making these connections with adopters and better understanding the systems in which they work has wider benefits too as we’ll see in our second strategy…

Strategy 2 – Clinical champions

Clinical champions are clinicians engaged with frontline care delivery who are influential and well-connected within an adopter organisation or system. They have a highly authentic sense of the true unmet need  that your product should be targeting and the complexity of the sociotechnical systems in which it must thrive. This group of individuals overlaps with Key Opinion Leaders or Clinical Academics/Entrepreneurs, but they’re typically found closer to clinical leadership than research. Identifying these individuals as an outsider can be difficult, that’s where taking time and engaging with the clinical community through education or other meetings can help.

Buy-in from a clinical champion (and potential early-adopter) is a crucial early determinant of success. These individuals may not be experienced with AI, but they are experienced in the informal evaluation of healthcare interventions - so don’t try to pitch them any ‘magic bullets’! These important relationships should be nurtured with a collaborative and transparent approach. Together you can define the local clinical problem you will solve and how your product will do it. To win their engagement this alignment between your value proposition and their pain-points is crucial.

Clinical champions can also help you discover the pain-points of implementing your product, with more patience than you can expect from other potential clients. With their on-the-ground view, they can anticipate and resolve blockers in their organisation much more efficiently and effectively than you can alone. You can use this to anticipate future clients’ needs and ease some of those implementation pains; technical interoperability, navigating local governance structures, care pathway integration and how to monitor performance and return on investment for providers…

Strategy 3 – Local evaluations

The clinical community may not be widely familiar with the use of AI products yet, but potential adopters are not short of propositions from academic and commercial proponents of AI. There’s real difficulty in knowing who to believe, particularly with their awareness of how brittle performance can be in new contexts. Even with gleaming evidence from clinical investigations, that means that adopters (quite rightly) prefer to see products in action on their own population and digital infrastructure.

Rather than fighting this as yet another expense, think of it as an opportunity to secure customer satisfaction and test out your plans for Post Market Clinical Follow-up (PMCF ). This kind of local evaluation may even be possible pre-market, if the silent trial is for research or service appraisal purposes that do not involve actual patient care. That allows you to build demand for your product and lower barriers to implementation for customers even before you get to market.

To make the most of this offering, take time to understand which of their problems your client is hoping to solve with your product and what outcome measures are feasible to align with that. Importantly, the outcome measure you settle on together needs to be recordable for their current practice so you can clearly demonstrate impact in a dimension that matters to them. Designing practical means of measuring the local value of an AI product long term is challenging. Making sure it's done well is a gift to both you and your client. At the procurement stage, it will also deliver locally tailored evidence to convince the most cynical and distracted of internal decision makers.

Final thoughts

Crossing the chasm is always difficult in healthcare innovation, but the ‘AI chasm’ is proving particularly challenging. The risks in achieving success on market are not always obvious, but are nevertheless inherent to the process. By thinking beyond market access into the factors that will determine the implementation of your product, you can identify and control much of that risk.

Juggling these considerations alongside all the demands of regulatory approval is not easy. With experience and a multi-disciplinary team, identifying and mobilising strategies tailored to your product and implementation niche needn’t be painful though. Think through how you might realise one or more of the options above for your product, or get in touch to explore a more tailored approach.

Hardian Health is a clinical digital consultancy focused on leveraging technology into healthcare markets through clinical strategy, scientific validation, regulation, health economics and intellectual property.