220,000 commuters, tourists, and the just plain lost travel on the London tube's Northern Line every day between 5pm and 7pm. Unless it's a Thursday when it's 219,999 as I work from home. Thanks to a wifi connection I can reduce the burden on the creaking underground system by a (very) marginal amount. And my own stress levels by a very unmarginal amount.
If 30p in every £1 of UK public services expenditure went on the Tube, I would at least get a seat in the morning. 30p in the pound or £140bn in 2017 was spent on healthcare in 2017. And it's not enough. With a further £1.8bn sunk into operating costs recently (earmarked for transformational projects - it never got there), the current approach to mend and make do just won't cut it anymore. It's time for some real intelligence to drive the adoption of the artificial variety.
The use cases are already out there: Moorfields Eye Hospital using Google's DeepMind, IBM Watson at Harrow Council, and of course GP at Hand from Babylon Health (and there are some even more exciting things coming from them). This is not only the territory of algorithms which can predict patterns of diseases and chronic conditions, they connect clinicians and patients in ways which allow them to quickly route to the best provision of care, optimising outcomes from the earliest onset of symptoms.
NHS AI is not a path without its risks.
Tech companies may see the precipice the NHS is on and see an easy mark. The cost savings potential of digitising patient data are so considerable that this cost could be recovered in a reasonable timeframe. Not least there is a healthy market for packaged, anonymised patient cohort data among life sciences companies, CROs etc.
Less obvious, and more nefarious, is that AI may actually be the 'middle class' solution to an affluent patient problem. Patient populations on low incomes or from socially disadvantaged backgrounds, tend to present themselves at a point of care less frequently, and usually when symptoms are more severe. They are less likely to see their GP early, and less likely to provide consistent, useful patient-level data into the AI algorithms which would benefit them in the long term. Not a new problem, ask any US-based provider or Federal health agency. But as the UK has not adopted large-scale population health measures outside the occasional epidemic of flu, it's an unaddressed problem with some deep-rooted causes.
And of course there is the economic / labour market issue - surely AI will be eliminating jobs? The Reform think tank thinks not: AI is a decision support, not a replacement for clinical expertise even though it has already proven itself diagnostically in certain oncologies. 30x more accurate is predicting breast cancer? That's a compelling diagnostic success. But with the NHS tens of thousands of recruits short, a situation likely to be compounded by Brexit, AI won't be replacing anyone, just enabling them to do more with less.
55,000 missed operations in 2016-17. That's another bit of data. But it's not all doom and gloom. We have the data, and we see the symptoms of sickness in the NHS.
With AI the symptoms could be the cure.
Written by Kevin A'Court - Head of Healthcare and Life Science Practice at Consulting Point