|Susannah Fox (@SusannahFox)|
Peer-to-peer healthcare: Crazy. Crazy. Crazy. Obvious. (We’re getting there.) susannahfox.com/2011/05/25/pee…
I saw the above tweet in my feed this morning, retweeted by Dana Lewis. I thought it sounded like a very interesting idea and aligned well with a lot of the thinking I’ve been doing and the work that Matt and David at http://afrezzadownunder.com/ and Dana at http://diyps.org/ have done.
My first thoughts were why shouldn’t we generate patient trials and wouldn’t that be a great way to accelerate Diabetes care and technological advance?
Taking that a step further, thinking about Diabetes care models, they’ve always been very insular, capturing personal data on a single device and reviewed in isolation. Clinical trials have, in some cases, used CGM to capture data, but more often than not use finger prick testing. And while they may be across a single device, as required by the trial, they are of limited availability.
But how about capturing “Big Data” for Diabetes? An opt-in model where thousands of people volunteer records including Blood Glucose levels and an accurate data set of food (broken down by Fat, Protein and Carbohydrate), exercise, stress, sleep times and other factors? What if we could capture this for a day, a week or a month? Think of the opportunities that this data would represent for providing an insight into how insulins work, how pumps work, the macronutritional impact of varying ratios on overall control! Participant age and sex, and potentially menstruation cycle timing.
In short, there’s a huge amount of insight that could be gained from this kind of data gathering exercise.
Sure, there’s a fair amount of data normalisation to be done with a data set like this, and then analytics to be generated, but by capturing it in the first place, it also creates an enormous body of data that could be easily accessed by those seeking to determine a better picture of living with diabetes by numbers…
But how do we get to this point? Using NightScout with the Libre has led me to what I believe is the answer. NightScout is a distributed network of data sources captured using a common codebase. Each individual has their own database in which all the data relating to their monitoring, careportal entries, etc. It uses a standardised data format for data capture, and more importantly, has thousands of users.
So how would we take this and turn it into something worth data mining?
Taking a step back, first and foremost, there would need to be a decision as to what data we wanted to capture.
As a first pass, for information purposes, I’d want to see glucose monitoring data and insulin dosing, whether MDI or Pump at the very least. I also think that exercise data and food broken down into macros would be pertinent. For each individual flow, you’d need sex, weight, height, age and potentially Diabetes Type as normalisation factors, and therefore a unique (random) identifier.
There’d need to be an “I want to join in” switch, which would enable the NightScout app to write to a second database, or second aggregator app which would collect the feeds from multiple participants and lock them up in the NightScout.Population database. Although unique identifiers would be included, these shouldn’t be trackable back to the original participant.
Then we’d come to the “experiments”. These are really very simple. Initially just a day in the life of the T1 where you commit to record every little detail that can be added into CarePortal. Food, drink, exercise, stress (!), etc.
Then maybe a week of similar data. And then a month.
Perhaps we could have an option to allow some participants to collect data permanently so that we had a wide population data set of years worth of blood glucose data. Whilst that initially appears to be little use without the other inputs, someone may find it useful.
We could ask for participants to opt in to various experiments, with specific rules in place, looking at the effects of various factors on various different tranches of the population. The possibilities are almost endless!
These are all first thoughts and brainstorming around the opportunities that NightScout opens up. A weeks worth of data from NightScout could easily be used by #OpenAPS to fine tune and back test predictor algorithms, in a vast and safe way. Getting specific example feedback could be one of those earlier experiments.
I want to say that the opportunities are endless. I think that might be overplaying the point, but wouldn’t patient contributed and signed off “Big Data” be a massively useful resource in helping the world treat diabetes? And the changes required to an already amazing tool don’t look like they are massively difficult to implement.
Are we up for the challenge? #WeAreNotWaiting