|
Abstract:
|
Diabetes is responsible for significant morbidity and mortality across the developed world. Improving health outcomes for patients with diabetes requires continuing care including both medical treatments like insulin and lifestyle considerations like healthy eating and physical activity. Mobile technology offers an effective and scalable platform to assist patients in managing their illness. Examples include: (i) tools to inform timing and dosing of insulin treatments; (ii) a channel for immediate and anytime communication between the patient and their healthcare provider; and (iii) a delivery system for teaching, training, and social support. Thus, there is tremendous potential to use mobile technology to improve patient outcomes. We develop a general framework for estimating precise, i.e., deeply tailored, treatment strategies that map individual patient information collected using a mobile device to recommended interventions. This framework combines dynamical systems models with online reinforcement learning algorithms. We demonstrate our framework by constructing an optimal insulin dosing strategy for a population of patients with type I diabetes in North Carolina.
|
