Current diabetes management strategy consists of a combination of regular visits to the doctor, self-administered medication or insulin shots, and tests to ascertain diabetes progression through daily glucometer readings or periodic lab tests. It is well established that a modest amount of weight loss through dietary changes and increased physical activity sharply reduce the patient’s chances of developing diabetes (National Institute of Health, 2008). The current system of diabetes management assumes patients are self-motivated and disciplined with diet control, exercise, and medication adherence. However diabetic patients struggle with issues related to motivation and the ability to achieve behavioral change and proactive management.
Our solution is a mobile application based Diabetes Personal Coach that is powered by applying Machine Learning (ML) on a continuous data stream of blood glucose levels, physical activity, and dietary habits across patients. Recent technological advancements in Continuous Glucose Monitors (CGM) and persuasive technologies (activity monitors such as Fitbit, Garmin Vivomove, and Apple Watch) now enable us to monitor real-time blood glucose and activity levels. Our solution overlays a patient’s blood glucose levels with the patient’s activity and dietary data to identify different patient contexts. ML will be used to determine motivation and ability at the user level to prompt the user with appropriate behavior change triggers. For demonstrative purposes, consider the following three scenarios: First, when motivation and ability are high, the person will work hard to induce lifestyle changes. Our approach would use ML to optimize actionable triggers for such patients.