Start with people: why deploying AI in health and care needs to be driven by a user-centric approach
Recently, we hosted an AI roadshow for NHSScotland in one of our AI labs. Over four days we held five interactive sessions to explore the current application of AI across industry and discuss where healthcare might benefit from the deployment of AI solutions.
It was fantastic to see leaders in NHSScotland who were genuinely excited about the possibilities of AI and what it could achieve. Now, this isn’t to say that everyone who arrived believed AI could solve all their problems, indeed some were wary of the benefits of the technology being overhyped. However, without exception, every single person who attended the session highlighted that data was the future of healthcare and that AI can play a critical role in harnessing the power of that data.
People are aware machine learning is driving many of our interactions as consumers, and indeed are starting to see it crop up in their working lives. Where we used to have a hard time convincing people AI wasn’t just Terminator and R2-D2, the challenge is now conveying the more technical complexities of AI to audiences in a comprehensible way.
Proving the value
The increased awareness of AI comes with a greater focus on what we can do in the immediate future with it. Whether you call it a Minimum Viable Product (MVP), Proof of Concept (PoC), Proof of Value (PoV) or pathfinder, there are plenty of small (and large) projects kicking off across health and care utilizing machine learning to drive clinical outcomes. Moorfields Eye Hospital is one that stands out for its promise, an AI system that is touted with a 94% accuracy in identifying over 50 different eye diseases. If it achieves its potential, the system could revolutionize eye care across the world and enable early identification and prevention of diseases that lead to sight loss and blindness, a condition that effects more than 265 million people worldwide. Undoubtedly amazing stuff, and testament to the power that AI can bring to unlocking clinical outcomes. Yet, that project kicked off in 2016 and is still in the development stage. It needs to undergo rigorous clinical trials and get regulatory approval to be used as a medical device.
Operational efficiency with AI
In my recent article on AI in Policing I emphasized the importance of AI in combating criminality, but also highlighted the potential for AI to deliver value in the back office. At risk of sounding like a broken record I believe the same fundamentally applies to health and care as well. From optimized power usage in hospitals, optimizing capacity and demand management, predictive maintenance, reduce variability in drug and clinical supplies, DNA (Did Not Attend) predictive analytics, translation services, and fraud prevention, to name a few, there’s a myriad of possibilities that AI can deliver value without the need for lengthy clinical and regulatory reviews. These opportunities not only provide valuable cost, and / or efficiency savings, but can also then translate into improving patient care and services with the resources that are freed up.
Start with people
So, the message is clear. For AI in healthcare choose achievable projects, identify the opportunities that deliver the most value over a short time-frame and then scale accordingly once the solution has been proven to work. However, we also need to look beyond the immediate future. As I alluded to earlier, there is a growing sense across all industries that AI is now here to stay. Throughout the week we had attendees not just asking “why use AI”, but instead “how do we integrate AI effectively across NHSScotland?” The answer to this is to start with people.
The successful integration of any technology in business needs to be understood and accepted by the end user, whether that’s a patient interacting with a translation service or operational staff using machine learning tools to allocate resources efficiently. We didn’t just choose to host our week-long AI workshop in Glasgow for the sunny weather, it was to be in our Customer Experience Lab and to emphasize the importance of people in the application of AI. Only when the input of the end-user is fully considered in the creation and deployment of AI systems will those systems achieve the full extent of the value they can offer.