Can AI in healthcare meet the expectations?
The NHS recently pledged £36 million for new AI projects to help revolutionise healthcare but the extent to which this will occur is still hotly debated.
Artificial Intelligence (AI) is becoming increasingly prevalent in business and wider society and is beginning to now be applied to healthcare. These technologies have the potential to transform many aspects of patient care and current ways of working within the NHS.
With more than 15 years of clinical experience, Dr Rui Providência, Clinical Informatics Lead of the sites at Barts Health NHS Trust, has a breadth of practical insight into traditional cardiology – but today, Hospital Times speaks to him about AI and how this can benefit patients and the health service, both now and in the future.
How can AI be used in practice?
AI in healthcare can be used for a variety of applications from diagnosis and management of disease to administrative processes. Specifically, at UCL Hospital, London, it is now being used for planning and bed management. Further, nationwide to improve stock management and reduce wastage. Given my background in cardiovascular disease, I am most interested in how AI has transformed this space and has supported a new era of care in cardiology. AI can now be used in conjunction with insertable cardiac monitors (ICMs) to remotely monitor abnormal heart rhythms and accurately help detect atrial fibrillation (AF) and sinus pauses (Pauses) automatically, and earlier in the condition.
What is the benefit of using AI rather than human expertise?
ICMs monitor the heart rhythm and if an abnormality, such as an AF or Pauses, is detected a clinical team would be prompted via website alerts to conduct a review. Due to the way all ICMs are built and the fact that these are subcutaneous devices, there will be artefacts that can affect the system, and these come through as false alerts. Even the most accurate ICM, such as LINQ II, is prone to identifying false alerts.
Working in one of the largest centres in Europe for ICMs, these artefacts, or false-positives, would lead to thousands of tracings per month requiring checking by our expert team, which would be time consuming, and also extremely frustrating for patients. The introduction of AI has almost eliminated the likelihood of detecting false alerts while ensuring only true heart abnormalities are alerted. This has not only meant that clinicians are saving time in reviewing results, roughly 160 hours per 100 patients, but this technology also allows for a much higher level of accuracy, ensuring almost no true alerts are missed.
Is AI accurate and can we trust it?
All AI tools require testing and validation of their performance before they make it to clinical practice. Even after being in use their performance needs to be audited, and refinements can be introduced if required. Using the LINQ II ICM device as an example, the integration of AI has meant that we can be more confident in our decision making, knowing that the sensitivity of the device has not been compromised, with 99.3% of true AF alerts and 100% of Pause alerts being retained. While this is already high, the technology will only become more accurate.
Of course, when selecting a cardiac device to use in clinical practice it is important to choose a reliable technology – which for me, needs to be based on quality clinical evidence. With the ICM system we use, the AI learned from a huge amount of ICM specific tracings that were classified by physicians, by physiologists, by experts in the fields, and that classification helped the AI learn how to interpret those tracings and give accurate results we can trust.
How does AI fit into the wider digitalisation strategy for the NHS?
The NHS is already under significant pressure and conditions are predicted to worsen, so it is imperative to protect the system as much as possible. Since the pandemic, NHS England has pledged to level up its digital maturity and capabilities, and AI is a key tool on this journey. It is expected that AI will improve efficiency, and will lead to better care and improved patient safety.
While the benefits of AI are clear, what challenges do you see for wider adoption on the NHS?
AI has been a ‘buzz word’ for many years in consumer use, however, it is still relatively new in a healthcare setting. With this in mind, the biggest challenge will be convincing HCPs who are hesitant to use the technology, and there is a level of education that is needed to address this. While this will take time, as soon as AI starts to become more common place, it will become integral to all processes. The controlled introduction of AI in healthcare in the UK, with still only a small number of case-uses (e.g. for dealing with AF and Pauses in ICMs) is just the precipice of the potential for AI and I’m excited to see the reality of the full capabilities and the evolution of this technology. The next 10 years will represent a gigantic leap in medicine.