The role of health innovation networks in ensuring AI assurance in healthcare

As the health and care industry increasingly embraces artificial intelligence (AI) technologies, it is crucial to ensure that these innovations are implemented with the utmost care and responsibility.  Dr Hatim Abdulhussein, CEO at Health Innovation Kent Surrey Sussex, explores more in this blog.

In the recent report from the Department of Science, Innovation and Technology, ‘Assuring a responsible future for AI‘ the Secretary of State for Science, Innovation and Technology highlighted that ‘AI assurance provides the tools and techniques required to measure, evaluate and communicate the trustworthiness of AI systems’. In TechUK’s ‘Ethics in Action: From White Paper to Workplace’, the five ethical principles are discussed further with a framework for the approach, covering; 

1. Safety, security and robustness 
2. Appropriate transparency and explainability 
3. Fairness 
4. Accountability and governance
5. Contestability and redress 

While in the NHS ‘Understanding healthcare worker confidence in AI’ report, a framework was developed that included aspects of external and internal validation, liability, interpretability and explainability and the important role of understanding human AI interaction, as critical criteria for the adoption of these technologies.  

As the key organisation in England responsible for implementing innovative practice in health and care, health innovation networks have been finding, testing and implementing new technologies and innovations, many of them digital and AI, at scale to address the NHS’s greatest challenges.  Ensuring that innovation is deployed responsibly, ethically and with human-centric design at the core is central to our work and we are well-versed in supporting safe and confident adoption of these technologies. 

Here is what our experience has taught us about implementing digital and AI technologies with confidence: 

Measure and evaluate 

The Health Innovation Network has supported digital health companies to conduct real world testing and evaluation of their technologies, including AI-powered solutions, to understand their impact. This allows providers to make informed decisions about which solutions are safe, effective and appropriate for their local context.  

Healthcare remains the most heavily regulated sector when it comes to the introduction of new products and practices. This needs to be the case to keep people safe.  

Comprehensive testing and evaluation is critical to build confidence and trust in AI systems amongst clinicians and patients. The innovator support offered by teams within the networks signpost companies to the appropriate regulatory bodies and helps them understand the accreditation required. Our relationships with regulatory bodies such as MHRA and NICE allows us to understand the latest guidance and quickly signpost innovators, as well as the thresholds of evidence needed to accelerate adoption, and we continue to inform the development of appropriate frameworks from these regulatory bodies. 

Communicate 

When supporting adoption of these new technologies, the Health Innovation Network acts as an enabler, communicating with commissioners and providers to understand their contexts, challenges and priorities and how the AI-based innovations can address them. Crucially, we highlight both the benefits these solutions can deliver, as well as the safety and ethical considerations. Sharing real-world insights and case studies of how AI solutions are being used effectively and safely in healthcare settings is key to building trust and driving adoption. This support is crucial for effective procurement and implementation. 

One example is Brainomix, an AI-powered clinical decision support tool for stroke that we supported to evaluate and implement, allowing clinicians to access real-time stroke imaging analysis to improve patient outcomes. Historically, CT brain scans are reviewed by a specialist in limited locations but, through Brainomix’s AI technology e-Stroke, high-quality images can be seen within a few minutes of being processed, anywhere and at any time. The technology helps stroke clinicians make swift decisions relating to transfer and treatment, including ensuring access to mechanical thrombectomy (MT), a life-changing treatment which can reduce disability and prevent or limit long-term care needs in patients with the most severe strokes. 

Health Innovation Oxford and Thames Valley, helped set up and coordinate the Thrombectomy Innovation & Transformation (TITaN) stroke network. It also worked with five Integrated Stroke Delivery Networks (ISDNs) and the NHS England South East regional team to roll out and evaluate e-Stroke. 

Through this the number of MTs carried out in the Oxford region increased from 37 in 2019 to 186 in 2022, achieving the Long Term Plan target for the first time 

Be transparent 

The other challenge with these systems is that they are often described as ‘black boxes’ as the decision-making process can lack transparency and explainability. Once they are implemented there are risks of algorithmic bias and a need for ongoing monitoring to maintain confidence.  

To address some of these challenges, we are exploring the potential of ‘platform’ based approaches to AI in healthcare. British start up Newton Tree out of Crawley in Sussex are developing an enterprise AI platform that allows different applications to integrate in a safe and secure way, with tools to monitor model performance and drift, support collaborative AI development and governance, and enable greater transparency and explainability.  

Equally, a number of health innovation networks are supporting the development of the NHS Secure Data Environments to enable ethically and regulatory compliant access to healthcare data. This is a mix of technical and commercial support depending on the local needs and partners, and can put us on a trajectory towards accelerating the early evaluation of AI technologies to make the NHS the home for the opportunity to develop solutions on large scale data sets representative of real world populations. 

Actively protect against bias  

The ‘Developing Healthcare Worker Confidence in AI’ report highlighted the role of the user layer of AI, recognising the role of bias in cognitive decision making, as well as the need for these tools to improve clinicians’ workflow and decision making. More applied research and implementation support is needed in this space. We worry about ethnicity and gender equity in particular with these technologies. Health Innovation Kent Surrey Sussex is working with the Centre for Population Health to understand this more for women, and with Department of Health and Social Care to identify AI Visionaries leading the charge. Later this month, NHS England and Department of Health and Social Care will be hosting a Responsible AI Event focusing on Women’s Health where these Visionaries will be showcased. 

Scale what works 

An example of an innovation currently being used and making an impact is the Brave AI system that helps primary care staff offer support sooner to people at risk of hospital admission by looking for patterns in registered patients’ records. The technology assesses an individual’s risk of unplanned hospital admission in the next year to identify vulnerable patients who may otherwise go under the radar. Integrated neighbourhood teams of nurses, pharmacists, therapists, health coaches, social prescribers and doctors then use the information to reach out to those in need.  

They can then offer to put in place personalised support, such as setting up remote health monitors, offering apps to self-report wellbeing, or linking up with voluntary groups or classes to avoid loneliness. Currently implemented in 30 areas across the south west, Brave AI has been piloted in a care home in Somerset and the findings included a reduction in falls by 35%, reduction in emergency department attendees by 60% and reduced ambulance call outs by 8.7%.  The innovation is now being scaled across GP practices in Gloucestershire, Wiltshire, Somerset, North Somerset, Dorset, Devon and Cornwall. 

There is no doubt that the adoption of AI in healthcare holds huge potential to transform and improve patient outcomes, but it is critical that innovation is deployed responsibly, ethically and with human-centric design at the core. The health innovation networks and our partners across the ecosystem have a crucial role to play in supporting safe and confident adoption of these technologies to realise this potential. 

As the Secretary of State for Science, Innovation and Technology highlights, ‘the UK already has a growing AI assurance market, which could go beyond £6.53 billion by 2035 if action is taken’. It is the social values we hold, no more demonstrable in the NHS and wider health and care system, which can contribute to this growth, enabled by organisations like the health innovation networks that are truly at the forefront of responsible AI implementation.