Healthy Nuggets #8: More on AI: predicting falls, powering self-monitoring and repeatable lessons

EDIT: Managed to get the numbering wrong, this is #8 rather than #9! ??

The more things change, the more they stay the same.

Almost a decade ago my old boss Gareth Williams was challenging some product managers at Skyscanner to come up with their Maslow hierarchy towards self-actualisation and trancendence in a travel booking product. He might not remember saying this but having written it down, our challenge was to create a product that could:

book your travel before you knew you needed it.

At the time this was Jetsons stuff, yet within a couple of years my pals Richard MacDonald-Keen , Stefan Sabev (AKA "Bot Head"), Konstantin H. , Nikolay Ivanov , Valentin , Stanley Chung , and Laura Wilson had got made a start in building that individual travel relationship thanks to Filip Filipov 's air cover that we were setting the agenda for the industry and not just on a goose chase.

Over 1 million users were happy to interact with primitive AI and if we had been able to prove a link between that relationship and the actual bookings people made, and break away from the industry's "look-to-book" and "session conversion" ratios in isolation, we might still be doing it. We did pretty well along the way though, even a certain Mr Nadella was demonstrating our product as best-in-class when Microsoft launched Cognitive Services back then.

But moving from self-indulgence back to self-actualisation, there seems no better use of this technology, especially given the pandemic and surge in power of Large Language Models, than our own health.

Whilst trancendence may be attempted by hokey services promising to calm you or clear your headspace, there are signs of companies actually trying to turn this around and prevent us from getting ill, and that's the theme for this week's scribble. Self-actualisation of one's health through intelligent use of today's technology and actionable insights.

The theme was in my head this past when this headline mentioning healthcare group Cera popped up: "Cera has developed and launched AI that predicts people falling at home a week before it happens with 83% accuracy."

Having mentioned Bide created by Dr Tom Adler a couple of months back, which uses a gentle intervention technique to effectively calm the elderly and take more care as they move around at night, something to augment that with prediction and data seemed like a good idea.

The story read:

Cera’s carers and nurses deliver 50,000 home visits per day to patients on behalf of the NHS and local authorities across the UK. They are able to input key indicators such as a poor sleep or a change in mood into the Cera SmartCare App on a smartphone.

This information is then used by Cera’s Fall Prediction AI to accurately predict who is most at risk of a fall. It can predict a fall with 83% accuracy up to a week before it happens, enabling carers to intervene to prevent people falling at home.

If you've ever had to care for an elderly relative or friend, you'll know that a fall is often the onset of decline, and the ability to predict it in advance could ultimately save thousands of lives. Cera's data claims that it could detect 10,000 hospital visits caused by falls, per year, if applied across the NHS - which could easily mean tens of thousands of beds freed up for other concerns.

Once again, an app with a simple method of keeping the data collector interested, with a beneficial outcome, looks like it's onto something.

On a slightly different note, but all ultimately related, a survey this past week showed the hunger for useful technology in our NHS, if only it would work properly.

To tackle this, technology products and services must be intrinsically linked to solving staff concerns. For an already stretched workforce, accessibility and time are two pressurising issues; they need to be addressed to make transformation projects successful, satisfy clinicians and bolster appetites for ongoing digital innovation.

Having seen someone delayed out of hospital for half a way because the ward's prescribing computer wasn't working, the need for distributed, flexible software never seemed more important. The good people in the NHS need a focus on user experience as much as any consumer would, yet something seems to stop such software being trialled at any scale, unless a single NHS Trust takes a punt on it. How do we fix that?

And on a more personal note, I mentioned a few weeks ago how Visible is taking an intelligent approach to Long Covid and Chronic Fatigue Syndrome.

Having suffered from (suspected) Long Covid over the past year and ending up with the thick end of a month off sick as a result, this did pique some interest and this past week I started wearing Visible's round-the-clock heart monitor, and tracking my "PacePoints".

The instant thought was, surely wearing an armband often seen in the gym would be a distraction, especially overnight - but after an hour or two and getting into normal life, you don't know it's there. Secondly, after an evening where a late walk had caused a terrible stint of brain fog in the morning, I then shifted to following the alerts Visible routinely gives out when it detects over-exertion.

And what do you know, the next day was a lot better. And the next day showed up patterns of over-exertion whilst sat at the desk after a non-stop 3-hour stint at the screen.

It's early days but the insights already feel useful and therefore it's not one of the many apps I've taken a look at before walking away. The input over and above the sensor is minimal and unobtrusive. It does start to feel like you can take a bit of control whilst building something that might help others.

Good luck to the Visible team with this and if anyone is interested in the product, the link follows. And yes I'll follow the instructions from now on, sorry for the support request ;)

The important lessons in these tales are about focusing on the user, early simplicity and value, and the products that can deliver it.

All the chat about LLMs and AI recently has focused on adoption of the tech, far less about the uses where it might be viable. Of course LLMs will rewrite something you've written, they'll partially replace Google with a narrative answer or list of facts (some of them actually correct) and with the right training, will provide you with plausible responses to well-prompted challenges. They might even be able to write you the code to interpret some data you've collected.

Equally we live in a world saturated in sensor technology, much of which is willingly sitting in our pocket or our hand round the clock. Nothing good comes of those sensors without a crawl, walk, run mentality able to divine early value and present it to the end user in a method that retains their trust and attention.

Like every good product in existence, and nothing new. So it's time to remind ourselves that tech is our servant and products are within our gift, with more important area ready for "disruption" than our own health.

ADVERTISEMENTS (sort of)

One of Waracle's clients, the brilliant folks at Smart Surgical Solutions , dropped an ad on LinkedIn for a CEO to lead them through the next exciting phase of development. Feel free to reply or pass on the ad if you know someone that fits?

Waracle also posted something this past week about our thoughts on, and work in, computer vision. Thanks to Blair Walker for putting this together and please do get in touch if you find it interesting or relevant.

Small print: This newsletter goes out to subscribers and across LinkedIn every Sunday night around 7:30 pm. Feel free to contact me if you've seen or are creating something interesting in digital health. I work for?Waracle, but all opinions and content selections are my own. Anything in which I have a work or personal interest will be declared.

Cover photo by Jan Antonin Kolar on Unsplash.