IBM Watson's computer housing case at the company's New York City's offices. The supercomputer, best known for winning on "Jeopardy," is now taking on health care. (Andrew Spear for The Washington Post.)

LAS VEGAS -- One of the most agonizing aspects of having diabetes is the uncertainty. While you can try to manage your blood-sugar levels as best as you can through food, sleep, activity, insulin and constant monitoring, there's always a fear that it might plummet unexpectedly, leading to a medical emergency.

What if there was a way to predict when this might happen?

In what could lead to one of the most important breakthroughs for diabetes management in years, Medtronic and IBM have teamed up to work on an app that aims to give patients an early warning that you may be headed in this direction. Powered by Watson, the supercomputer that won on "Jeopardy," the app analyzes data from Medtronic's insulin pumps and continuous glucose monitoring devices to look for patterns that suggest that something isn't quite right.

A screen shot of the glycemic alert app. (Photo courtesy IBM)
A screenshot of the glycemic alert app. (IBM)

In an early research project involving 600 patient cases, the team was able to predict near-term hypoglycemic events — in which the glucose in the blood is too low and can lead to seizures, coma or even death — up to three hours in advance of the symptoms.

That's plenty of time, Medtronic's Annette Bruls pointed out, to enable a person with diabetes to take action to head off the event.

"It's a very complex equation all the things that people need to manage their glucose levels," Bruls, president of diabetes services and solutions at Medtronic said in an interview at the Consumer Electronics Show. "What we are doing here is providing predictions, insights and in the future actionable recommendations to patients."

The algorithms are based on de-identified data from 150 million patient days collected through Medtronic wearable devices, which consist of electrodes that are inserted under the skin to measure glucose levels in tissue fluid and that beam the information in real-time to a monitoring device, such as a smartphone.

(Photo courtesy IBM)
(IBM)

Deborah DiSanzo, General Manager of IBM Watson Health, described the app is a kind of personal medical assistant for diabetes patients.

"You can imagine you are a patient with diabetes, it would help you tremendously to know when I get up in the morning and I walk my glucose level is at this level, when I eat oatmeal it goes to this level, when I eat ice cream to this level," DiSanzo said.

In the future, she said, Watson might suggest that you go for a walk at certain times of day or could help you choose between blueberries as a snack or something else.

The first iteration of the app is scheduled to come out in the summer of 2016, but DiSanzo said it will constantly be updated.

"Watson is a cognitive computer, so the more data it gets, the more it will learn," she said.

Diabetes, a chronic disease that affects the body's ability to produce or manage insulin, is one of the world's most devastating diseases. There are more than 400 million people diagnosed with the disease and consequences of poorly managed diabetes are severe — heart failure, blindness, renal failure. Each year an estimated 1.5 million people globally die from diabetes.

(Medtronic)
(Medtronic)

Read more:

This diet study upends everything we thought we knew about ‘healthy’ food

Cutting sugar from kids’ diets appears to have a beneficial effect in just 10 days

Scientists (sort of) settle debate on low-carb vs. low-fat diets

Doritos, deconstructed (mesmerizing photos of the 34 processed ingredients in your favorite snack)

Hot topic: Could regularly eating spicy foods help you live longer?

Beware the rule-following co-worker, Harvard study warns

For more health news, you can sign up for our weekly newsletter here.