The BIG Question – How is AI changing our NHS?

Artificial Intelligence is being credited with miracles in healthcare – from predicting diseases to slashing red tape for our doctors and nurses.

With more than six million patients on NHS waiting lists in England and fierce competition for finite resources, AI has the potential to make a real difference, targeting treatment for need, managing data and freeing up medics to spend more time with patients.

We spoke to a range of LJMU experts, who are researching new technologies and tuned in to their hopes and fears for an AI NHS.

“Healthcare is a real test-case for the application of AI in how we do things as a society,” says Sandra Ortega Martorell, Professor of Data Science at LJMU.

“AI is quickly becoming a key tool to better decision-making in the NHS, better systems and, crucially, better outcomes for patients and the public.”

Professor Ortega (pictured) specialises in ways machines can learn tasks and runs a multi-million pound study TARGET to prioritise patients according to their risk of heart and stroke disease.

By integrating data from health records, IoT-enabled medical devices, physical documents, and even building management systems, Sandra and her co-investigator Professor Ivan Olier aim to offer healthcare professionals a ‘single source of truth’, ensuring improved patient treatment and reducing the risk of errors.

“We’re looking at things in a very different way and that must be a positive for the NHS,” says Sandra.

But it’s early days, she says. AI is far from offering the highest levels of service, as anyone using an online chatbot can attest. Chatbots fill gaps in service but research shows they can encourage over-treatment and even reinforce inequality.

Computer scientist Professor Dhiya Al-Jumeily is going a step further, developing ‘Ami’, a robot doctor. Ami can already provide more accurate predictions than human doctors and quicker, and is capable of friendly, near human conversation and express emotions via facial expressions.

“We’re a long way from AIs directly treating the public but we will soon start to see robots working alongside humans – what we call ‘human-in-the-loop’ collaborations,” he explained.

Dhiya (pictured above with 'Ami') sees a much more advanced role for Ami than simply as a ‘digital front door’, especially in a world where training more human doctors is prohibitively expensive. 

“We’re not far from having 10 billion people on Earth and not nearly enough doctors to serve them,” he adds.

On the frontline

So what do people on the NHS frontline make of AI’s impact? Kevin Cairns is a Lecturer in LJMU’s School of Nursing with 20 years on the wards behind him including at Alder Hey Children’s Hospital, where AI system Lyrebird is established and already supports clinicians with documentation.

He sees AI and healthcare as good bedfellows (pardon the pun!)

“Healthcare produces enormous amounts of information - observations, scans, photographs, test results, clinical records and so on. Clinicians then have to bring all of that information together to make decisions, often in very busy and complex environments.

“AI is particularly good at identifying patterns across large amounts of data, so there is considerable potential to use it to support healthcare professionals.”

For example, Kevin’s own research explores how AI aids the monitoring of pressure ulcers, for example by using photographs taken during dressing changes to measure wounds and track changes over time.

“For me, its greatest potential is not in replacing clinicians, but in giving them better information and more time to focus on patients. AI should support clinical judgement, not substitute for it, he says.”

If people want to see a real human doctor, there's no point in arguing that a robot knows better

Dhiya Al-Jumeily, Professor of Computer Science and Artificial Intelligence

Professor Al-Jumeily agrees, saying that public trust in systems, whether human or machine or combined, is imperative, especially when it involves our health.

“We are doing a lot of work to break some of the boundaries between humans and humanoids, and that needs to happen if the technology is to be effective.”

And it’s not just about putting patients at ease but also doctors and nurses. “We’re involving doctors in workshops and conferences with Ami and also introducing the ‘machine role’ in their own medical training. He stresses that at the end of the day, it’s about what people will accept.

“If people want to see a real human doctor, there’s no point arguing that the robot knows better. I can’t see robots either replacing humans or even taking major medical decisions.”

Professor Olier (pictured below) who is Head of the university’s Artificial Intelligence and Digital Technologies Research Institute, agrees that trust is fundamental.

“Healthcare AI cannot simply be accurate; it also needs to be trustworthy. That means developing systems that are transparent about how they reach decisions, robust when faced with new or imperfect data, and fair across different groups of patients. These are particularly important requirements in healthcare, where an AI recommendation can influence decisions that directly affect people’s lives.”

Ivan and Sandra are also leading AI research within the ARISTOTELES project, focusing on understanding how multiple health conditions develop and interact in patients with atrial fibrillation.

“A major part of this work is about moving away from AI systems that simply give clinicians a prediction without adequately explaining why. In healthcare, it is not enough for a model to say that someone is at high risk. Clinicians also need to understand the background, whether the result is clinically plausible, and how confident they should be in the model’s recommendation.”

One of the approaches Ivan and his colleagues are developing is Causal AI for allows for more meaningful interrogation of data. For example, what might happen to a patient’s risk if a risk factor were reduced, if a treatment were introduced, or if one aspect of their clinical profile changed while everything else remained the same?

“This is particularly relevant to the development of personalised medicine and digital twins, where AI models may be used not only to predict what might happen to a patient, but also to explore different possible future scenarios before a clinical decision is made.

“The aim is not to build AI that replaces clinical judgement. It is to develop systems that clinicians can question, challenge and understand.”

Kevin (pictured below) agrees on the supporting role for AI. “It’s a tool that clinicians can exploit to giving them better information and more time to focus on patients. AI should support clinical judgement, not substitute for it.”

He also issues a word of warning; that AI can reinforce biases in the NHS. For example, we know that pulse oximeters can be less accurate in people with darker skin tones. If data containing these types of inaccuracies is subsequently used to train AI systems, there is a risk that existing inequalities are carried forward or even amplified.

“So we must ask not only how well an AI model performs, but where its data came from, how it was collected and whether it properly represents the patients it will ultimately be used with.”

Much of the success of AI will be in its implementation and not just in the capability of the technology, according to Dr Gemma Dale, senior lecturer in AI and workforce management in Liverpool Business School.

Gemma suggests some jobs may disappear altogether and NHS leaders need to implement very careful change management processes, if the workforce – and unions – are not to take against AI.

“New skills will be needed and substantial investment in learning and development, and consultation. There will be human bumps in the road and it’s up to managers to ensure that they do not become a barrier to AI’s adoption,” she says.

We need to demonstrate not only that an AI system works, but why it works

Professor Ivan Olier-Caparroso, Head of the Artificial Intelligence and Digital Technologies Research Institute

LJMU experts concur on the enormous potential of AI and agree that gradual adoption is the way forward if the benefits are to outweigh the risks.

For Ivan, adoption will ultimately depend on trust: “We need to demonstrate not only that an AI system works, but why it works, when it can be trusted and, equally importantly, when it should not be trusted.

Kevin Cairns also believes it will come down to wider understanding: “The most successful healthcare AI will come from collaboration between people who understand the technology and people who understand the patients, workforce and clinical environments in which it will be used.”

ends