T1 Assist: NHS Hack Day

Imagine getting over a hundred people to listen to your (and dozens of others) health problems and then they volunteer a weekend of their own time to help solve it. That's a NHS Hack Day

Graphic Credit: Ayesha Garrett at London Lime

Last weekend I went along to the Cardiff NHS Hack Day. Organised once again by Dr Anne Marie Cunningham and a small team of NHS and digital health advocates, their energy and enthusiasm, together with the NHS Hack Day profile, drew around a hundred people to a damp Cardiff for a weekend of digital health exploration.

There's sixty seconds to pitch the problem you'd like to see worked on, around an hour of coffee and chat to help people form groups and decide which ones get pursued over the weekend.And then just over 24 hours to see what can be done.

So what did I ask for?

Well, diabetes, particularly Type 1 diabetes, does not run like clockwork and yet our NHS clinical support structure assumes it does. No two days or two weeks are the same, and yet the system assumes best care is served up by seeing your clinical team every three months (as a paediatric patient). Like clockwork.

So what if we could re-purpose all of that glucose and insulin pump data some patients have for their own use - and for sharing 1-1 with their clinicians in clinic - to empower clinicians to choose when and how to contact and see their patients, offering support and advice when you most need it, rather than seeing them the week before the wheels fall off?


What did we do?

After getting over the fact that there were at least half-a-dozen other pitches that I'd gladly volunteered to assist and getting more coffee, a multi-talented team started to form and we got to work scribbling out the problem, putting together an outline of version 0.1 of a possible solution.

We started with Janki's Nightscout data, which contains CGM, finger prick, carbohydrate and insulin pump data:

Janki and her 640g had a busy weekend in Cardiff (including a new pizza restaurant)

A huge thank-you to two other Nightscout parents who volunteered their anonymous data for the Hack Day build :)

Now armed with three very different T1 datasets, the team were able to collate multiple Nightscout data-streams onto a single AWS site and then build our own API to handle queries of that multi-patient data.

One of the team also converted the Nightscout sensor and blood glucose data stream into a common data structure standard (known as HL7 FHIR), which is a huge step towards seamless interoperability with other digital health systems, such as Electronic Patient Record (EPR) systems held in GP surgeries and hospitals. For a data-type that's not as digitally 'mature' as say, medical imaging data - where one standard (DICOM) is universally used - settling on a universal format would be a big step. It would be great if regulators could start pushing for that even harder than they are.

The team produced a very visual dashboard, designed for clinical teams to use on a daily or weekly basis, prioritising patients based on a live-steam of their recent data. An iPhone app complements the detail served up by the dashboard, with the ability to choose an individual patient and select actions such as messaging them or their carer.

Prototype T1 Assist Dashboard for clinical teams

Oh, and we gave all this a name: T1 Assist.

To Sam, Yesh and the three Mark's, who gave up their weekend to help start these wheels in motion, a huge Thank-You. 

And of course none of this would be possible without the #WeAreNotWaiting Community and all of those who have put Nightscout together.

Next Steps?

The conversation is ongoing to see how and where we can take this next.

Tidepool and others, such as Glooko give clinicians access and insight to T1 data, but are primarily designed to facilitate 1-1 interactions. What we're trying to explore is whether the technology is there to deeper change and search for a better way to help keep those with diabetes healthy within the limited healthcare resources available.

Technology-wise, it's clearly feasible.

Our next key steps is to explore the problem with clinical teams and make sure we're addressing both the clinical and political challenges.There's also a human design element required to ensure all patients feel supported between what could be very variable clinic visit schedule.

I think T1 Assist can help diabetes clinical teams reconnect with both their patients and their patients data, driving support when most needed but also offering the opportunity for positive reinforcement and encouragement when things are going well - and, dare I say it, allowing the clinician to ask "How did you do that!?" to start a truly two-way conversation within a team that fully includes the patient.

Thanks for reading and feel free to leave any comments below or on the Facebook page, where you can also message me directly.

Graphic Credit: Ayesha Garrett at London Lime