re-thinking mobile diabetic applications

During my last two semesters in grad school, I researched and began designing a mobile phone application for use by diabetics for tracking and monitoring their progress in controlling the disease. I envisioned an application that would offer customized coaching and real-time feedback combined with flexible data input methods. This solution was designed to address deficiencies in diabetic treatment resulting from the one-size-fits-all approach to diet and exercise guidelines offered by nutritionists/diabetic counselors, and empower patients to learn and implement what works best for them individually. This project was born out of my frustration with the process as a gestational diabetic and a survey of available tools and technologies that felt inadequate. You can read more about the solution I envisioned in my portfolio.

One of the big challenges of the solution I designed was the intensive data processing required. Users would be gathering data constantly throughout the day (blood sugar readings, food eaten, insulin taken, exercise completed, journal notes, etc.). The analysis of this data would require access to massive external databases (such as food guides with carbohydrate quantities), as well as complex algorithms for analyzing and offering relevant suggestions based on the data. For example, let’s say that the user scanned a bar code on a box of Raisin Bran at 9am to document what they had for breakfast. An hour later, the application reminds them to test their blood sugar. The result is very high. The user wants to know why. The app would need to be able to analyze the ingredients, explain why they are problematic, and offer some suggestions as how to alter the outcome of eating this (add protein such as nuts, use a higher-fat milk, etc) or suggest not eatingĀ  it. It would also need to flag this item for future reference so that if the user did adopt the suggestion of adding protein when consuming this, that they could later see if the tip reduced the resulting blood sugar.

As you can see, the analysis needed to make this app successful almost nears impossibility… or at least a degree of artificial intelligence. Which is likely why all the current diabetic technology tools merely collect the data and display graphs and charts, but do not analyze the data.

Anyhow, today I watched a program about the Open Source Mobile Data Kit on the University of Washington’s TV channel. The Open Data Kit is a suite of tools designed for Android phones that allow for remote collection of data with an array of data types including video, photos, and barcodes in addition to the normal text based data. Server-side tools allow for analysis of the data with results pushed back to the phones for field use. The Open Data Kit is being used in a number of current and planned deployments by field workers in Africa and other places for help with large-scale health initiatives such as HIV treatment and prevention.

This got me to thinking that perhaps the mobile diabetic application I envisioned could run on this platform. But instead of having all data analyzed by smart computer systems, perhaps data could go to a group of diabetic experts who have access to all the data in useful formats and could in turn offer custom guidance to the patient/user of the diabetic system.

In the Raisin Bran example above, when a user checks a box on their mobile phone interface to Analyze why this food might have caused such a blood sugar spike, a signal is sent to the data center and becomes part of a queue (like a support call to a call center). When this request comes up in the queue, the analyst/diabetic expert has access to all this patient’s data and can offer advice based on this individual’s specific situation/history. The analyst can enter the advice in the system, and beam it back down to the user’s phone via the application. For higher severity issues related to insulin dosages or other emergencies, the request can go to the front of the queue, and the patient can receive a phone call back to talk through the situation.

Though this system might not provide info as fast as an intelligent computer, it is still a huge improvement over the current healthcare system where one only gets a short appointment with a diabetic counselor on a weekly or monthly basis, when most of the teachable opportunities have faded from memory. This system would also have greatly reduced development costs and time to market, so patients could benefit sooner.

Now I’m wondering about the legal aspects/liability of offering custom healthcare advice through mobile technology. More research to do!