Much media attention has this week been given to the Government's 50 point ' AI Opportunities Plan'. In his speech announcing the plan, the Prime Minister Sir Keir Starmer outlined the aim for the UK to become an AI " superpower", with AI being "mainlined" into the nation's veins. The hope that AI can provide solutions to the problems facing the healthcare sector and the NHS in particular was clear with the Prime Minister stating, " AI powered scans can help doctors detect disease earlier. AI can cut waiting lists by scheduling better appointments."
A cool £10billion
However, the NHS has historically been woefully slow to appreciate and adopt new technologies and hugely resistant to change. Further, the ability to design, build and deliver IT projects at scale has been demonstrably lacking - an infamous example being the attempt to introduce an electronic patient record system that was quieltly shelved but not before costing the NHS a cool £10 billion.
Realistic about the application
So, AI is undoubtedly a force for change that offers huge potential for transforming healthcare in ways that we can perhaps only currently guess at. Nevertheless, it is important to be very realistic about how it can be best utilised and applied in the short-term, to help solve problems being faced now. It seems obvious that the biggest application will be in helping the NHS to do simple, routine and everyday operational tasks better - more intelligently in fact. Certainly, using AI to streamline appointment scheduling to reduce waiting lists, to enhance the speed and accuracy of diagnosis and to improve communication with and positive engagement from patients are areas that the NHS is already actively exploring. This may not seem as exciting or futuristic as AI robots as carers or AI eradicating cancer but improving menial operational processes could just help to save the NHS in the meantime.
Informed by latest AI studies
Reflecting on recent studies such as Nature's Empathetic AI can't get under the skin, Prof Shafi Ahmed points out that, "it is clear that while large language models (LLMs) show impressive potential in exhibiting empathy, especially in medical contexts, we must remember that this empathy is simulated, not genuine." Prof Ahmed goes onto say, "As we continue to explore and refine these models, it is essential to balance innovation with ethical considerations, ensuring that their use enhances patient care without compromising privacy or emotional well-being."
Empathetic, patient orientated and unbiased?
So, as AI advances across the medical specialities, a cautionary and balanced view needs to be adopted. Tools such as GPT-4 and GPT-4o appear excellent at language processing, however there are serious questions about their ability and capability in the patient-orientated clinical setting. Of particular concern are the AI Biases, this arises when algorithms are trained on biased datasets and leads to skewed predictions or decisions that may disproportionately impact certain groups. This would result in an ineffective product and, potentially, active harm caused by the bias. After all at the end of the day, AI is only as effective as the quality of the data (see Finding authoritative medical information in the digital age), that it is allowed access to.
While digital access to medical information has expanded dramatically, so has the need to distinguish reliable sources from unreliable ones. Even respected institutions such as the NHS may sometimes oversimplify complex evidence or promote interventions based on limited data. There will be a need to keep the application of AI straightforward and allow a progressive evolution. We cannot expect AI to rescue the NHS in the short term.
Feature pic: Shutterstock
