Why I’m Building Software for Diabetes Management
Diabetes devices have evolved, but the software is in the Dark Ages.
When I was diagnosed with type 1 diabetes almost nine years ago, I remember thinking “Ok, I’m an engineer, I’ll just turn this thing in to an algorithm, and math the shit out of it.”
I’m sure I wasn’t the first engineer to think that. More than most conditions, diabetes lends itself to modeling and data analysis. Your goal is to keep your blood sugar, which can be simplified as a single number, within a certain range. Your blood sugar naturally rises. Insulin lowers your blood sugar. Eating raises your blood sugar. Exercise increases the effectiveness of insulin.
Simple, I thought.
At that time, I had three tools at my disposal:
1. “Long acting” 24-hour insulin, which you take once per day to control your blood sugar between meals.
2. “Short acting” 2 to 3 hour insulin, which you use to lower your blood sugar quickly if it’s too high, and to keep the food you eat from raising your blood sugar.
3. A blood glucose meter which tells you what your current level is at any moment from a blood sample from your finger. I could use this to check my status six to ten times per day. Any more than that was both incredibly inconvenient and not covered by insurance.
From here, all I had to do was define the factors in the equation with values that worked, and solve for X. Right?
Not so fast
It turns out, in my haste to maintain a positive attitude and turn my shiny new chronic, incurable illness in to little more than an inconvenient math problem, I had failed to acknowledge something important and obvious — the human body is goddamned complicated. I soon realized was that I was trying to do something akin to maintaining a jet engine, blindfolded, with nothing but a sledgehammer and a screw driver. And naturally, I sucked.
There were the typical human failings: Shit, did I forget to take my 24 hour insulin? How many units did I just take? And how long ago was it? I’d really like to test right now, but I’m on a crowded subway so I guess I’ll have to wait until I get where I’m going. Then there were the failings of my body and the tools themselves: My 24 hour insulin I took 12 hours ago is suddenly way too much insulin if I decide to go for a run. My meter tells me I’m currently at 120 mg/dL, but it doesn’t tell me whether it’s going up or down. If you get sick, your body releases tons of sugar in to the blood for what I’m sure is some great biological reason but to me is a huge pain in the ass. Your insulin sensitivity, i.e. how effective 1 unit of insulin is at lowering your blood sugar or covering food, doesn’t just change over time, it changes over the course of each day. Once I take insulin, I can’t un-take it if my plans change or my food is different than I expected.
Two new tools made life easier
Luckily, technological advances made things better. After 5 years, I got a digital insulin pump system which eliminated my need for two different types of insulin. The pump delivers a trickle of insulin constantly and allows me to take larger doses when I eat. I can also turn off the pump when I want to go running or change the amount of insulin I’m taking based on the time of day. Finally, I can know exactly when and how much insulin I’ve taken throughout the day and week. Hooray for digital! Hooray for data! Hooray for the convenience of push button insulin doses! Hooray for no more syringes!
Then, a year or two after that, things got much better when I got a continuous glucose monitor (CGM). This is by far the most useful device I own. It reads my blood glucose levels automatically every 5 minutes. Now, instead of knowing my blood glucose 6–10 times a day, I now know my blood glucose level 280 times a day, with trend information. So what does that mean for me?
At a glance, no matter where I am, I can always know where I’m at and what direction I’m headed. I can see whether my insulin has finished working and then take more. Or I can see that in 20 minutes, if I don’t eat something, my blood sugar is headed for a low.
The next huge opportunity to improve treatment
When I first imagined writing an algorithm to manage my diabetes, I wanted data streams like the ones that my pump and my CGM provide. When combined, this data could show exactly how my body is reacting to specific events. Can you remember what you had for lunch two Thursdays ago? I can’t, but my CGM remembers that I had a low around midnight. And my pump remembers exactly when and how much insulin I was taking leading up to it.
Analyzing this data in aggregate, over time could help me and my doctor make changes to and evaluate the effectiveness of my self care regimen. Do I have high blood sugar in the mornings or lows during the weekend? Can I improve how I’m dosing myself for breakfast? What if my doctor could see all of my data and give me recommendations on the fly?
Insights trapped by fragmented, inaccessible, and frankly crappy software
Right now, all of this awesome data is critically underutilized. My doctor only looks at data from my CGM because the CGM manufacturer’s software doesn’t talk to the insulin pump software. The software for my insulin pump isn’t even cloud enabled at all and doesn’t talk to my CGM. Also, the pump software is Windows only. Yes, I’m serious, this is still an affliction suffered by some of the software in the industry. Oh, and my doctor hates using it, because it sucks. Great.
But, let’s go back to the CGM data that she does use. She has to manually sync the data to a desktop computer via a USB cable (and requires software to be pre-installed on each exam room desktop to do so), despite the fact that the CGM is Bluetooth enabled and continuously sending the data to my phone, which uploads it to the cloud over 4G and wifi real time. For other patients, she has to use completely different software, with a drastically different UI and report format, because they wear devices from a different manufacturer. Many patients are still shuttling printouts of tables and graphs back and forth to visits because at home they use desktop software from manufacturers that is not cloud enabled, or their doctor has no way to sync the data from their device in the office.
Connecting the dots: doctors, patients, devices, and data
There’s no reason we have to accept this. We can and we must have better software that unifies data streams, makes uploading data effortless and gives us a better view of our health. Our doctors should be able to easily get access to our data, quickly find the information they need to make recommendations, and stay connected with us.
My partners and I, all former entrepreneurs and Samsung engineers, are founding Tower Health to solve these problems. Our goal is to build a cloud based data analysis platform that leverages data from our devices, offers insights and recommendations, tracks progress toward goals, and makes it easy for us to stay in contact with our doctors, who will have a complete and continuous clinical view of our treatment regimen.
No more USB cables for bluetooth enabled devices. No more printouts. No more trying to remember specific events. No more sugar coating our numbers. Have a question for your doctor? Want to discuss something between visits? She’ll have access to your data to provide important context. Want to follow progress toward a goal for overall insulin usage or bolus/basal ratio? Want to know which days of the week are your worst? We’ll let you know automatically.
If you’re interested, visit http://towerhealth.co and give us your email, we’ll be recruiting doctors and patients for a pilot program in early March. If you have feedback, and/or are a doctor or patient interested in making your voice heard by doing a user interview, email us any time at team@towerhealth.co. And follow us on twitter at http://twitter.com/towerhealthco. Let’s put our data to work.
