Skip to main content

An App that Predicts Low Blood Sugars Three Hours in Advance

Updated: 8/14/21 6:00 amPublished: 1/25/16

Twitter Summary: @MDT_Diabetes + @IBMhealthcare announce first product of innovative partnership: mobile app that predicts #hypo 3 hours in advance!

During a packed keynote lecture at the recent Consumer Electronics Show (CES) conference, IBM CEO Ms. Ginni Rometty and Medtronic CEO Mr. Omar Ishrak co-announced the first product to come out of their exciting partnership: a predictive hypoglycemia mobile app launching this summer. For those of you on inslun and taking it through a pump, imagine taking a bolus and receiving a predictive notification: “You are likely to go low within the next three hours.” According to early data, those notifications turn out to be 80-90% accurate!

How is it possible? The app uses the IBM Watson supercomputer – the same one that won Jeopardy! – to analyze CGM data, glucose trends, carb data, insulin pump data, and an individual person’s glucose history to predict future low blood sugars within three hours of taking a bolus. The hypoglycemia prediction app will rely on real-time data sent from a Medtronic pump to the MiniMed Connect keychain device to the new mobile app. At launch the app will function exclusively with Medtronic CGM and pumps, and it will likely only support Apple iOS (Android is planned over time). The trick, of course, is to see technology like this broadened to insulin pens as well.

Medtronic and IBM do not yet have FDA clearance for the app for Medtronic pump users, and it is still in its prototype phase. Given the expected summer 2016 launch, however, we assume conversations with the FDA have gone favorably to date. Alternatively, the app may not need FDA clearance, as it’s arguably a very safe entry into real-time diabetes decision support – a super-charged CGM prediction alarm for lows, so to speak. In theory, it will keep patients safer than what they’re doing now, so it could come with little risk. What do you think should be required?

This has a lot of potential for insulin users all over the world. Several questions come to mind: How many false alarms will the app generate? Is Medtronic’s Enlite sensor accurate enough to reliably detect lows? How much extra data entry will the app require? How frequent will it send notifications? Regardless, it’s encouraging to see Medtronic and IBM moving quickly to bring actionable software to people with diabetes, and of course, we have to walk before we can run!

On that note, Mr. Ishrak share a compelling vision of what could come down the line. For instance, Medtronic could leverage other pieces of technology (e.g., activity trackers), and combined with Watson’s capabilities, could generate even better insights for patients: “What has happened to other patients like me when they have taken X boluses of insulin at a glucose level of Y?”

Hypoglycemia remains one of the biggest challenges in diabetes, and while it’s often thought of as a type 1 problem, it’s also a major issue for so many people with type 2 diabetes (and due to the proportion of people with type 2 compared with type 1, the number of people with type 2 who experience hypoglycemia is far greater than that of people with type 1). We’re excited to see this problem being tackled with IBM Watson’s computing expertise and Medtronic’s experience in diabetes. And quickly too – if all goes well, the app will come out just over one year since the partnership was announced in April of 2015.

As Ms. Rometty stated at the end of her keynote, “I believe every decision mankind makes will be better because of Watson cognitive technology.” We hope that’s especially true in diabetes. We salute IBM, and Ms. Rometty's words from PBS' Charlie Rose talk show last year come to mind - that healthcare could well be IBM's moonshot! –AB/AJW

What do you think?