Algorithms are boring until your life depends on them. People with Type 1 diabetes use algorithms all day long. They perform mental calculations to manage their blood-sugar levels, which are measured by hand with a finger prick, or with a wearable sensor that fits under the skin. If they stray too far from baseline, the consequences can range from exhaustion and depression, to coma and death.
Usually when humans are doing something for which a computer would be much superior, such as performing calculations, a company will jump in and automate it. Algorithms for managing diabetes are no different. Any competent programmer with access to either long-term or real-time glucose and insulin data could start creating customized apps and tools that make life easier for people with diabetes.
But that's the problem: the data isn't readily accessible. That's why digital health activists like Dana Lewis, who has type 1 diabetes, are hacking into medical devices, to create tools that manufacturers won't.
"Most people don't understand that medical device technology is about ten years behind everything else," Lewis, a Seattle-based data analyst, told me over Skype. "It's very frustrating to not get your data from your body out of the device."
The first time Lewis, who is 27, hacked into her Continuous Glucose Monitor (CGM), a small device that attaches to her body and reads her blood-sugar level every five minutes, all she wanted was to make it louder. The device beeps a warning if her glucose falls or climbs too steeply. She can either increase or decrease her insulin intake with an insulin pump. But Lewis found that she was sleeping through the alarm at night, and the consequences of not waking up in time left her feeling terrible in the morning.
Once she and her husband Scott Leibrand got glucose data off the device, they realized they could do more than sync it to a louder alarm. So they developed a program and started inputting data points for each action she took with her insulin pump—if she was increasing or decreasing the insulin—in response to the data from her CGM.
With those two datasets, Lewis was able to develop a predictive algorithm that mimicked the mental calculations she and other people with Type 1 diabetes are already performing all the time.
Basically, she taught the pump to read her CGM and respond automatically. When the monitor and the pump work together, they become what Lewis has dubbed the Artificial Pancreas System (#OpenAPS). She calls it "closing the loop."
"Diabetes is a disease of patterns"
Lewis compares managing diabetes to driving on the highway. Most of us have a pancreas that does the "driving" for us—that organ is responsible for regulating our insulin levels, but in people with type 1 diabetes, it doesn't work—and we don't have to think about it. People with Type 1 diabetes have to "drive" themselves. In this analogy, the CGM is their speedometer, and the insulin pump is their gas pedal.
The #OpenAPS system is like cruise control. People with diabetes still have to steer, but this little piece of homemade software makes the driving a lot easier, says Lewis.
An artificial pancreas is a risky piece of technology. If it malfunctions, someone might get the wrong dose of insulin, which is a potentially fatal drug. Lewis argues that people with Type 1 diabetes are accustomed to risk, and can judge for themselves whether the payoff outweighs any danger. #OpenAPS users report more peace of mind and better health, with lower average glucose levels, she said. As of May 17, 59 people had adopted her system.
While not everyone is ready to hand over the reins of their insulin pump to a DIY piece of software, there are a great many who want to be able work with their own data, but aren't allowed because device manufacturers won't open up their APIs or make the data available to download. (Lewis blames this on the fact that most of the existing devices were built, tested, and approved by the FDA before users started demanding access.)
Take Nightscout, a program that makes hacked GCM data available in the cloud, so it can be accessed from anywhere. Nightscout won't tell an insulin pump what to do, but with this tool, parents of children with Type 1 diabetes are able to monitor their kids' GCM stream any time, from wherever they are.
A chief advocate for data liberation is Anna McCollister-Slipp, a digital-health entrepreneur with Type 1 diabetes. She takes 14 different medications every day, has many associated complications (kidney, nerve, and eye disease), and keeps six different devices with her at all times. With these, she generates a lot of data.
"Diabetes is a disease of patterns," she said over the phone from her home in Washington, D.C. "Understanding what impacts your blood-glucose is essential, so you can make critical decisions about how to mitigate and treat those factors. All I've wanted for the last five years is to put all this data into a single timeline."
While device manufacturers show the data on their own apps and software platforms, visibility is not the same as accessibility. There's a lot more that patients could do if they were able to work with the data. Case in point: Lewis's artificial pancreas system.
McCollister-Slipp is calling on the companies who make medical devices to open up their APIs. "I think it's incredibly unethical that companies choose to keep the data streams inaccessible when the health and safety of patients is at stake."