Today many companies and apps try to help people living with diabetes. These companies can follow one of two approaches: one focuses on the pancreas while the other looks at the entire person in addressing diabetes. Here at Quin we strive to learn from the pancreas approach and we pay close attention to the science of the endocrine system, but in the end, we follow the second, human-first approach, for helping people with diabetes.
The driving idea behind the pancreas approach is pretty intuitive. Today the medical community defines type 1 diabetes as a deficiency of insulin-producing cells in the pancreas. Insulin is a hormone needed to turn sugar into energy. When the pancreas doesn’t produce enough of it, the person has a high level of glucose, a type of sugar, in the bloodstream. This is type 1 diabetes in a nutshell: elevated levels of blood glucose caused by the pancreas not producing enough of the insulin hormone. If the pancreas is not producing enough insulin then, with the pancreas-first approach, we need to somehow replace the work it does. The person with diabetes needs to act like a pancreas, or one needs software that knows what the pancreas knows, or one needs to create an artificial pancreas.
Probably the simplest way to follow the pancreas-first approach is for the person with diabetes to remember some basic medical recipes, for example, “take 1 unit of insulin for every 10 grams of sugar you eat”. This is the sort of advice many people receive when they start taking insulin. All they have to do then is count the carbohydrates they eat (or “carb count”) and take the corresponding amount of insulin. A corner piece of lasagna has about 20 grams of carbohydrates, so one would need to take 2U of insulin.
If the pancreas-first approach were enough to help people with diabetes, Quin’s diabetes app would have been a lot easier to build. In user interviews many people with diabetes tell us that medical recipes are not easy to follow. Corner pieces of lasagna are not all created equal: some are larger; some mostly meat; others have more vegetables. If you order a lasagna at your favourite restaurant, do you really know how much carbohydrates it contains? Our users have a lot of experience guesstimating the amount of carbs in their food, but they tell us mistakes are too easy to make. Little carb-counting mistakes can add up to be significant. There are apps out there that try to codify medical recipes and make carb counting more manageable. But they have a tough time giving personalised recommendations. Around 80 to 92 percent of people who take insulin do not achieve stable blood glucose levels over time. The pancreas-first and medical recipe approaches are clearly not working for many people. This is the main trouble with pancreas focused approaches to diabetes. Let’s face it.
Our bodies react to food and to insulin in different ways at different times.
Formulas, rules and algorithms do not work well enough for everybody and across all situations in their life. If you are going through a stressful job interview, your body may react differently to a unit of insulin than if you are vacationing in the Bahamas or even taking a nap. Some of the factors that have been found to influence blood glucose are sleep, stress, weather, altitude, food, walking, running and other exercise, menstruation, travel, among others. If there is a medical formula here to tell people how much insulin to take, it is complicated and unknown. Medical scientists do not currently know how these factors interact to affect insulin and blood glucose.
A better approach zooms out from the pancreas to the entire human being in addressing diabetes. It is the approach we favour at Quin. The fact is that type 1 diabetes is caused by the pancreas not producing enough insulin, and this is not something anybody should deny. At Quin we simply think that the solution is not to replace the work done by the pancreas. Our thinking is that the solution is broader than that. We have learned from the pancreas first approach that many factors control blood glucose and not all are a function of the endocrine system. The human-first approach takes into all the relevant factors we mentioned: both inside and outside this system.
How can we use this variety of factors to help people decide how much insulin to take? The pancreas first approach has shown us that no medical formula, however complex, is likely to help different people facing different circumstances, but not all is lost. At Quin we know that people who take insulin know their bodies best and that they are the best sources of information and guidance about how much insulin they ought to take. Who knows best how lasagna affects their blood glucose level? Well, you know best how it affects your body and I know best how it affects mine.
We gather our users’ trial and error experiences in a collective body of data. The many trial and error experiments they conduct for themselves contain hard-won wisdom. Once again, we know from the pancreas-first approach that it is not feasible to codify their wisdom as a set of rules. But what if there was an app that used machine learning to help people take insulin? Unlike impersonal medical recipes and generic algorithms, these models are deeply personalized for each user’s experience.
In sum, at Quin our goal is to help people decide for themselves how much insulin to take and when to take it. We do it not with generic medical formulas but with machine learning models customized to the lived experience of each user.