Outcome-Driven Personalized Treatment Design for Managing Diabetes

摘要

Diabetes affects 422 million people globally, costing over $825 billion per year. In the United States, about 30.3 million live with the illness. Current diabetes management focuses on close monitoring of a patient's blood glucose level, while the clinician experiments with dosing strategy based on clinical guidelines and his or her own experience. In this work, we propose a model for designing a personalized treatment plan tailored specifically to the patient's unique dose-effect characteristics. Such a plan is more effective and efficient-for both treatment outcome and treatment cost-than current trial-and-error approaches. Our approach incorporates two key mathematical innovations. First, we develop a predictive dose-effect model that uses fluid dynamics, a compartmental model of partial differential equations, constrained least-square optimization, and statistical smoothing. The model leverages a patient's routine self-monitoring of blood glucose and prescribed medication to establish a direct relationship between drug dosage and drug effect. This answers a fundamental century-long puzzle on how to predict dose effect without using invasive procedures to measure drug concentration in the body. Second, a multiobjective mixed-integer programming model incorporates this personalized dose-effect knowledge along with clinical constraints and produces optimized plans that provide better glycemic control while using less drug. This is an added benefit because diabetes is costly to treat as it progresses and requires continuous intervention. Implemented at Grady Memorial Hospital, our system reduces the hospital cost by $39,500 per patient for pregnancy cases where a mother suffers from gestational diabetes. This is a decrease of more than fourfold in the overall hospital costs for such cases. For type 2 diabetes, which accounts for about 90%-95% of all diagnosed cases of diabetes in adults, our approach leads to improved blood glucose control using less medication, resulting in about 39% savings ($40,880 per patient) in medical costs for these patients. Our mathematical model is the first that (1) characterizes personalized dose response for oral antidiabetic drugs; and (2) optimizes outcome and dosing strategy through mathematical programming.