This article was inspired by an interview with Shahram Ebadollahi, PhD, Vice President, Innovation & Chief Science Officer at IBM Watson Health.
This is part one in a two-part series on cognitive computing in diabetes management.
If data bytes were grains of rice, your body would generate enough during your life to cover Manhattan.[i]
More and more data is being recorded by the devices and sensors we carry around every day – smart phones, activity trackers – each one capturing an abundance of information about our movements, sleep patterns, moods, heartbeat, what and when we eat. “If you combine all that data from everyone generating it, we’ll be looking at 44 zetabytes of data by 2020. One zetabyte is enough to fill the Pacific Ocean, so imagine 44 Pacific Oceans of data. That’s massive,” says Shahram Ebadollahi, PhD, vice president, Innovation & Chief Science Officer at IBM Watson Health.
When combined with medical device data, this information can offer a potential goldmine of insights about our health status, especially for people with diabetes, whose data is vital to their wellbeing. The problem is, most of these data sets are siloed in their respective technologies, unable to be pooled together. Diabetes is a largely self-managed disease. With 90% of diabetes care coming from the person with diabetes, they make over 180 diabetes decisions EACH day[ii]. So relying on memory, regardless of how dedicated one is to routinely journal their daily meals and activities – is impractical.
Here’s where Watson Health “cognitive computing” comes in. According to IBM, “cognitive computing is the simulation of human thought processes in a computerized model. It involves self-learning systems that use data mining, pattern recognition, and natural language processing to mimic the way the human brain works. The goal of cognitive computing is to create automated information technology (IT) systems that are capable of solving problems without requiring human assistance.” With Watson Health’s cognitive computing, we are now able to see health data that was previously hidden, doing more than we thought was possible.
“Cognitive is in its infancy,” Ebadollahi said. “But today’s machines look at data, look at the context of how the user is looking at that data, and it learns to make connections among the data from those examples.”
Real World Applications
To date, we have 125 million days of aggregate data from Medtronic insulin pumps and continuous glucose monitors (CGM) in our CareLink database, which is unique to the industry. That’s enough information to fill 20% of the books in the Library of Congress.
By applying Watson’s cognitive computing data to Medtronic’s device data and other health data, such as activity trackers, digital scales, population health data, and electronic health records, we believe we will deliver personalized diabetes management that can help make sense of all the data available. Utilizing these critical insights, we are developing cognitive computing apps that could act as a personal assistant for a person with diabetes, providing relevant, real-time insights and coaching to help them improve their ability to understand the impact of daily activities on their condition.
For example, in an early research project of retrospective data from Medtronic diabetes devices, Watson Health was able to predict near-term hypoglycemic events (low blood glucose) up to three hours in advance.[iii] Hypoglycemic episodes can lead to confusion, dizziness, shock, and other dangers. And the long term effects of being outside recommended glucose levels is riddled with complications affecting the nerves, heart, eyes, and cardiovascular system, so think about how valuable that information would be for a person with diabetes. Watson does this by accessing data from millions of people to find patterns that may lead to predictions that are more than 80% accurate for a two- to four-hour window.[iv] When you hear the term “predictive health,” this is the type of anticipatory, preventative medicine they are referring to.
In addition to cognitive computing apps, together Medtronic and IBM will look to develop personalized provider and patient care plans, further develop our closed-loop algorithms, and develop near real-time dynamic personalized care plans. By embracing the data-driven nature of diabetes, such programs have the ability to modernize diabetes care management and provide greater freedom to people living with the disease.
If cognitive computing can provide valuable, insightful health information to a person with diabetes, what impact could it have on the healthcare system?
[i] Health Policy Brief: The Relative Contribution of Multiple Determinants to Health Outcomes. Health Affairs. August 21, 2014.
[ii] Digitale, Erin. New research shows how to keep diabetics safer during sleep. Stanford Medicine. May 8, 2014. http://scopeblog.stanford.edu/2014/05/08/new-research-keeps-diabetics-safer-during-sleep/#sthash.esMasb9U.dpufhttp://scopeblog.stanford.edu/2014/05/08/new-research-keeps-diabetics-safer-during-sleep/
[iii] Preliminary research results based on dataset in controlled environment. Data on File.
[iv] Preliminary research results based on dataset in controlled environment. Data on File.
Tags: big data
, cognitive computing
, diabetes devices
, insulin pump