New UK AI targets reduction in hospital-based patient risks

British researchers say other forms of AI are poised to make a dramatic impact throughout the NHS, as well, including AI that diagnoses skin cancer from pictures of moles and eye disorders from retinal scans.

Jeff Rowe | Feb 20, 2019 09:08 am

British medical researchers have announced that an AI system developed at University College London and set to be piloted in NHS hospitals later this year will identify "at risk" patients and doctors to intervene to prevent death or serious illness. 

According to Professor Tony Young, NHS England's national clinical lead for innovation if the pilot, to be carried out in Essex hospitals, is successful, the technology will be rolled out to other areas.

"Instead of people getting sick or dying because they are not picked up in time, this will allow us to step in earlier which will save lives and an enormous amount of money,” Prof. Young, a consultant surgeon at Southend University Hospital, said.
“I think the potential of AI in healthcare like this is as big as the Industrial Revolution was - and signals a completely new paradigm in the way we manage healthcare.”

The technology is the brainchild of Dr Vishal Nangalia, a consultant anesthetist at the Royal Free Hospital in London and NHS England clinical entrepreneur, who used AI to analyze a billion stored blood samples from 20 different UK hospital trusts dating back 12 years.

His Machine Learning Early Warning System study used complex computer algorithms to predict "in-hospital patient deterioration" by cross-referencing blood samples - as well as where, when and by whom they were taken - with patient outcomes.

The technique retrospectively forecast outcomes of patients with kidney problems with up to 95 per cent accuracy, as opposed to traditional methods highlighting serious patient concerns, which picked up as few as 16 per cent of patients who went on to die.

”This gives us the opportunity not only to save lives but to prevent serious illness, making the health service not only safer but more efficient,” Prof. Young explained. "Instead of waiting for people to get worse, we will be able to treat them earlier.”

According to some experts, at least half of adverse events in hospital leading to death, serious illness or injuries could be prevented by using AI, which equates to tens of thousands of patients a year.

The pilot will see the new technology used to predict patients most likely to deteriorate or die from a range of conditions, including heart attacks, kidney injury, infection - and even falls.

The technique sees computers assess blood test results by picking up subtle changes in red and white blood cells, and electrolytes such as sodium and potassium, that suggest a patient is going downhill.