Diabetes - time for open data and disruptive technology
alexblandford.tumblr.com
alexblandford.tumblr.com
And yes, that pump will be imperfect, but it probably will do a better job than you can do yourself, let alone than you will do yourself.
I know that, traditionally, patients used glucose level measurements, stared at the numbers, guessed at what caused them and then tried to correct for similar future events, but AFAIK, that has never been very effective. It did give you some sense of being in control, but that was more an illusion than reality. e.g: your blood sugar is high. Is that because you drank a beer last night, because you went to bed late, because you did not sleep well, because you are stressed about something, because you had a fever two days before, because you took the bus to the train station?
However, except from the truly bloody obvious, that didn't help much maintaining blood sugar levels, unless you lived like a robot. Normal daily schedules simply vary too much to allow you to determine the true cause of atypical blood sugar level variations.
IMO, the time of stand-alone glucose meters has passed, and I do not see how having users process logs will help improve their lives. Reading individual measurements definitely will, but that is so that you can react at short notice, and adjust insulin intake.
I've been a diabetic for 14 years and that's exactly how I feel. The only way I can have perfect glucose levels is to live like a robot: wake up every day at the same time, eat exactly the same measured amount of food, check my glucose level every time my alarm beeps and do nothing out of the ordinary.
I'm trying to exercise more and the diabetes is the biggest barrier, I'm like a roller coaster right now.
I don't want to stare at a graph and do even more calculations, I want those calculations done by a machine, which can infuse insuline constantly and use more or less depending on my glucose trend.
There are new technologies out there, like this CGM [1] (continuous glucose monitor) which is very small and wireless (minilink minimed) which IIRC can work with an insulin pump (there is one that's very similar, wireless too, with a reservoir for three days of insulin)
[1] http://www.diabeteshealth.com/media/images/article_images/50...
The problem? Price. Most insurances won't cover this, the initial purchase is not cheap and the maintenance cost a lot too.
I hear that a lot, but I think this may be a cached judgment and not be accounting for recent decreases in price and increases in CGM sensor quality. If you haven't asked your insurance recently, ask again.
Public insurance won't cover CGMs right now and are hesitant to give out pumps. I checked yesterday again, a Dexcom G4 [1] is around 1.5K, and I would need around 200€ each month on sensors.
I'll probably end buying one out of pocket, but it's not something most people can afford!
[1] http://farm9.staticflickr.com/8457/8066666840_465937321d_o.j...
There are a number of biological factors that can make a difference, but for type 1 diabetics, the point should be that you know the ratio of carbohydrates eaten to insulin to take. This can vary, but good data helps you get the ratio right. It can also identify if you have high glucose points that match your circadian rhythm, or if you're becoming hypoglycaemic at a certain point regularly and just haven't mentally joined the dots.
Now, yes, you could do all of this through a log book on paper. I could sit down for a half hour each month and look for trends. But, I take readings on a device with digitised output between 4 and 8 times per day. I don't want to have to go analogue with that, I essentially want the blood glucose equivalent of google analytics. I can account for unusual variations (cycled for 10 miles/got blind drunk/etc.) much better if I have a dataset that demonstrates what the benchmark should be.
That makes sense, as companies are almost giving glucose meters away (in my N=3 sample, at least one owns more than one pump) to gain/keep market share for their 'razors', and as they improve quality of life so much for a relatively modest sum of money.
Also, suppose that you find the perfect formula for determining insulin intake. How are you going to apply it without an automated device?
This does not align with the practical reality of diabetics or the biological reality of what pumps and sensors do. The situation is much better now with continuous sensors, which display a graph with a sample every 5 minutes with alarms for high, low, and changing too fast.
But you cannot take the human out of the decision process because the most important information needed to decide how much insulin to give is what (ie how much carbohydrate) you're eating, because the effect of insulin is delayed too much. Trying to make decisions from glucose measurements alone is completely inadequate. People are trying to do this anyways, by using dual pumps giving both insulin and glucagon (glucagon has the opposite effect as insulin). This hasn't left the research stages and I don't think it'll work very well.
Basically, we're seeing little movement because open data has huge perceived risks and minute upside.
+ Open Data = Risk: The reason big device companies don't make open APIs for data is that it opens them up to risk. If some 3rd party creates software that uses device data and a patient ends up in a hospital, or worse, who will the family sue? The 2 person startup experimenting with health software, or the multi-billion dollar device company that didn't police every app made with their data? It's slightly irrational, but from the perspective of device company execs, opening data creates huge potential problems and risks with almost no upside. Also, the data privacy laws in the US and Europe scare/confuse many.
+ Open data attracts almost no customers: Unfortunately, there is no real money making/saving opportunity with data. A lot of promising concepts, but until there is a business model, there won't be a huge amount of focus on this area. Once there is, it will explode.
+ People won't pay for better products: Most people with diabetes have multiple meters, most of which they receive for free. The device companies work on a razors and blades model so are incented to keep costs down and use retrograde technology. When companies try to make higher quality products, relatively few people buy them.
So the best selling market today is a product that was originally designed to be a disposable meter for use in the 3rd world. It's super cheap and does the job while some of the awesome products you mentioned flounder.
Diabetes, and healthcare generally is a hugely complex market. The web of stakeholders blows the minds of most people who enter it. It will change eventually, but it's extremely difficult to bring a startup mentality to the industry because it usually requires thinking about physical manufacturing AND opening your company up to regulation AND operating in a Byzantine payment system. Possible, but improbable. If you're working on something in this area, we'd love to talk.
Re: the iPhone meter - we designed it to have its own battery and display, so even if your iPhone dies, the meter has enough charge to last 10 or so days, and a separate power cord to charge it for more. There was no way we'd tether a person's ability to test to a failing phone battery:)
My goal is to create software that uses Bayesian algorithms to tell me exactly what I need to do (medications, diet, exercise) to keep my glucose measurements nice and flat.
http://www.hanselman.com/blog/TheSadStateOfDiabetesTechnolog...
The main thing that works is to maintain a consistent diet, timing of meals, exercise, and insulin dosage. I consulted with a dietician to plan my meals. It was simple and helped a log.
Great article.
And I'd like my insurance to cover it, but that's an entirely different story.
I think with diabetes, I certainly want the data to be adaptive beyond "constant diet, timing of meals" etc.
@j_s - Yes, I think he makes some fantastic points.
@inetsee - Yes, lots more variables than BG readings, but getting that right is a good start.
@axsar - Thanks, will have a look, although it still relies on proprietary data cable.
Not that it's right, but it's possible.
This way way, the process stays the same independent of what meter I use, the data is easily exportable and I don't need to use Accu-Check 360°, which is dire.
[1] https://play.google.com/store/apps/details?id=com.wonggordon...