How AI technology is the answer to relieving pressure on A&E
Professor Matthew Cooke discusses urgent care pressures, Health Navigator, and why he joined our advisory board
Almost as soon as the clocks changed and we said goodbye to the summer, we were seeing headlines that A&E waiting times were the worst on record. Just 83.6% of patients attending all types of A&E departments, which includes dental A&E and urgent care centres, in England were admitted, transferred or discharged within four hours. That’s the lowest level since the emergency care access standard was introduced in 2004. We know that such delays are associated with worse clinical outcomes.
Present strategies to reduce emergency attendances and admissions have generally failed, so we need a new approach. One of the answers, I believe, lies in Artificial Intelligence and tools such as Health Navigator – an AI-guided health coaching service which is proven to reduce emergency hospital admissions.
Health Navigator uses its AI technology to identify the patients that use the greatest proportion of A&E services and who would benefit from early intervention through its personalised health coaching service.
I have a particular interest in the use of AI within healthcare and, after meeting with Health Navigator CEO Joachim Werr at a conference last year, I was invited to join its Advisory Board which launched this month.
Health Navigator is unique, not only for the service it delivers, but for the fact that it has very robust evidence to demonstrate the efficacy of its services. It has undergone four years of intense R&D in the UK culminating in what appears to be a very successful randomised controlled study commissioned by seven Acute hospital trusts and CCGs, of which NHS Vale of York CCG and York teaching hospital is one. The trial has so far demonstrated a reduction in A&E attendances by 36%, unplanned hospital admissions by 30%, and planned admissions by 25%. The study has recruited 1,800 patients and is ongoing for another year and a half aiming to recruit a total of 3,000 patients.
I’m an academic by background – I love evidence – and there’s very few operational interventions that have undergone this level of analysis. Usually the approach is to see if it works somewhere and then if it does, try it somewhere else.
Part of my role on the Health Navigator Advisory Board is ensuring the evaluations are done effectively and in a way that is the most informative.
The opportunities are, in my opinion, just obvious. If 1% of people take up some 50% of all unplanned bed days in hospital, it just makes sense to focus your intervention on those people to free up the beds. If we free up the beds then the patient flow increases and those headline figures of people waiting in A&Es and people with delayed discharge will improve. The knock-on effect of a relatively small number of people being treated differently should be massive and the figures from the Health Navigator randomised controlled trial support this.
The next question will be ‘how far ahead can we predict that these people are going to benefit from the intervention?’. At the moment they have to reach a certain level of use of the NHS to be identified. This group of patients has a very high turnover, so traditional methods of looking at last year’s high users will fail to cause improvement as 80% will no longer be high users after a year. Early identification is essential. But can they be identified, say, five years before this usage peaks? The earlier you intervene, the less specific you can be, so if you go five years out then you could have five or maybe 10 times as many people and only a small portion of them would benefit. Can the AI technology be pushed further to effectively identify these targeted groups at a much earlier stage – now that is what really interests me.
Professor Matthew Cooke