How AI can help patients manage diabetes
Artificial intelligence is providing advice that could help more than 30 million Americans suffering from diabetes. That advice is coming from algorithms designed to help people with diabetes effectively control their glucose levels.
Keeping those levels as low as possible can prevent or at least slow some complications associated with the diabetes, according to the American Diabetes Association. These complications can damage the eyes, kidneys and nerves.
Algorithms developed by Mountain View, Calif.-based startup Glooko could help patients with diabetes control it by using data around how foods, exercise, and insulin affect their glucose levels under different conditions.
“In the case of diabetes, being able to track how your behavior affects glycemic control gives you the tools you need to stay out of harm’s way,” said Michelle de Haaff, vice president of Marketing and Customer Success at Glooko.
While tech visionaries like Elon Musk and Stephen Hawking have warned about some uses of artificial intelligence, the Food and Drug Administration cleared Glooko’s mobile app for marketing in the U.S. in December 2012. More recently, Glooko’s Mobile Insulin Dosing System, which allows doctors to set a dosage and pull data from the patient’s glucose monitor, received FDA approval.
AI For Health Care
In developing the algorithms that drive this product, Glooko used machine learning to analyze billions of data points. The company uncovered patterns of how blood glucose levels change in response to factors that include diet, exercise, and insulin injection. It then translated the findings into algorithms that process data from individual patients.
The algorithms are designed for practicality. They might be used, de Haaff explained, to help a patient choose between bacon and eggs or pancakes. Neither is particularly healthy, she noted, but the app in which they are embedded tells the patient which will have the least negative effect on glycemic control.
This app, which runs on iPhone and Android-based smartphones, is part of Glooko’s Unified Diabetes Management System. An office-based component can also be used by care managers, including nurses, diabetes educators and coaches, as well as physicians.
A clinical study of the system showed its effectiveness. After just two months of use, people with diabetes suffered 10.7 percent fewer hyperglycemic events and showed a 3.5 percent drop in blood glucose, according to research published online by the Journal of Diabetes Science and Technology.
In its simplest form, a patient or caregiver poses a question related to diabetes management. Glooko processes the data in a secure and encrypted cloud environment, then sends its answer to the user. Patients may want to know how their blood glucose will likely respond after different types of exercise or eating certain foods. Questions from many users can be processed simultaneously, de Haaff said.
A care manager might use the office-based platform to ask for a trends analysis of when blood glucose levels for an individual patient were high or low during certain days, weeks, or months. Those results might be used to adjust a patient’s overall care plan.
Caregivers can also broaden a request, asking for a population analysis that takes into account data from multiple patients. The resulting analysis might help care givers assess the effectiveness of care plans for their populations of diabetes patients.
More advanced capabilities are on the horizon, as Glooko is developing AI algorithms that use GPS data from smartphones. With this data, Glooko might be able to recognize the patient has entered a restaurant, de Haaff explained. The platform could then automatically examine the menu of a restaurant and suggest the three best meals for glycemic control.
“The thing with using sensors is that we (need to) be helpful and supportive of our customers,” said de Haaff. Such advanced capabilities are still in early development, she noted.
The company is continuing to develop AI algorithms that work in concert with smartphones because of the practical – and psychological – advantages.
“Using something that is familiar – that people like to use – makes a lot of sense,” she said. “When it applies to their daily lives in a practical way, the fear (about AI) will go away at least for the artificial intelligence that is being used in health care.”
With tens of millions of Americans at risk from diabetes complications, the immediate availability of fact-based advice on how to eliminate or reduce some of those complications could go a long way toward establishing the benefit of artificial intelligence.