Medtronic demos Sugar.IQ app with IBM Watson at Health 2.0; initial launch in coming days, broad launch in late 2016 – September 27, 2016

On the Health 2.0 main stage today, Medtronic’s Annette Brüls showed the final release version of its highly anticipated first-gen app with IBM Watson, Sugar.IQ (press release). It will launch in the coming days to 100 MiniMed Connect users (to gather initial feedback) ahead of a broader rollout “later this year.” Sugar.IQ will be free to Medtronic device users and only on Apple iOS at launch.

The app uses Medtronic CGM and pump data and logged meal information to “find hidden patterns in diabetes data.” Ms. Brüls showed examples of very actionable pattern insights (pictures below), and we learned about some others from the Medtronic team:

• “Having Tuna salad? Be aware that you tend to go low when you eat this meal.”
• “[At 7 AM] Planning your day? I see you tend to go low on Saturdays between 12 PM and 3 PM.”
• “Eating a high carb meal? I see that you tend to go high when you eat more than 45 g of carbs in one meal.”

The app delivers hundreds of these unique personalized insights based on time of day/week, glucose, meals, boluses, and other behaviors (e.g., fingersticks). Each insight shows the corresponding data that actually informed the pattern – persuasive! –initially flowing from the MiniMed Connect app and subsequently from the standalone mobile CGM app, Guardian Connect (under FDA review; launching by April 2017). Users can give Sugar.IQ direct feedback on whether an Insight was helpful, enabling curation of only the most useful suggestions.

The app also has a neat feature called “Glycemic Assist” – the user can tell Sugar.IQ to “follow” a specific food or therapy action to discover its impact on glucose. For instance, follow when I eat burritos, follow when I eat meals with more than 100 g of carbs, follow all correction boluses. The app then collects these events and shares Insights related to them. We love that this is customizable and allows patients to dig into their own personally challenging areas.

Sugar.IQ gives motivational and adherence insights too (e.g., You are spending more time-in-range this month vs. last month), plus sharp-looking CGM data reports (e.g., time-in-range by day). The data visualizations and overall app experience look very strong – clean, intuitive navigation, and a far cry from the underwhelming user experience on Medtronic Paradigm pumps. The Insights actually pop up in a chat-like interface, with a speech bubble subtly flowing out of the IBM Watson logo – smart branding and a nice way to make it less clinical.

Sustained engagement with Sugar.IQ will be the biggest question – will patients continue to use it? Food logging can be hard to keep up, though we’re glad to see Medtronic has built in some predictive capabilities to make it easier to log meals – e.g., showing typical foods eaten at lunch. We also believe that if the insights are strong enough, patients will invest the time in logging their meals. Fortunately, the app will still give pattern insights even if food data isn’t entered, and we wonder if the burden might even decline over time (GPS location? Pictures?) or if gamification or social features might help sustain engagement.

We’re elated to see this partnership moving at a very fast clip (the initial prototype was only shown in January at CES) and we love the focus on pairing cause and effect – especially with food. Even the most avid CGM users can’t easily recognize patterns like those the app identifies, and we see high potential for this to offer more personalized care and guidance.

• IBM and Medtronic are very committed: we learned that the companies have a pretty large team working on this project. We’re glad to see the teams have the right frame of mind – said one member to us: “Once an app becomes predictable, that’s pretty much the end of it.” The team is focused hard on engagement and evolving the app to continually give novel insights.
• There are five generations of Sugar.IQ on the roadmap, with future plans to add the hypoglycemia prediction feature (“you are likely to go low in three hours”; shown in January at CES, but not included in this version), activity data integration, more multivariate analysis, and hyperglycemia prediction. We also assume Medtronic could eventually launch a coaching service that leverages Sugar.IQ plus accountability to a human coach.
• For Medtronic, this launch: (i) gives it impressive patient-facing data analytics that may make its CGM devices more compelling; (ii) shows its commitment to moving beyond hardware to drive better outcomes with software and analytics; (iii) highlights its drive to improve CGM; and (iv) gives it a strong diabetes data play as both Abbott and Dexcom invest further in this area too.
• Since Sugar.IQ only analyzes retrospective data, we assume it is regulated at a low level. It obviously communicates MiniMed Connect near-term (510(k) cleared) and will eventually talk to the standalone mobile CGM, Guardian Connect (PMA).
• Users must use Sugar.IQ alongside the Medtronic apps that communicate directly with the medical devices: MiniMed Connect and Guardian Connect (once approved). This allows Medtronic to de-couple Sugar.IQ from the hardware-communicating apps, enabling it to iterate more quickly from a regulatory perspective. Dexcom has pursued a similar path with its G5 app (for talking to the CGM device) and the Clarity app (for analyzing data).
• In talking to the Medtronic team, the hardest part of developing Sugar.IQ has been coming up with the right insights, curating them, and defining the right frequency – the app aims to show meaningful and actionable insights that come frequently enough, but not too frequently.  “We don’t want to bombard you and we don’t want you left in the lurch without anything.”
• Once Android MiniMed Connect is available (expected by October 2016), we assume Sugar.IQ could be translated over from Apple iOS fairly easily. The app actually talks to CareLink and the Watson Cloud, meaning it is device agnostic and reliant on a data connection – not a particular device environment.

Sugar.IQ Screenshots

--by Adam Brown, Brian Levine, and Kelly Close