After failure of diet and lifestyle intervention, step-wise addition of glucose-lowering medications is the usual course of therapy. Yet, responses to diabetes medications vary substantially among individuals, suggesting a smarter and more individualized approach to medication selection is urgently needed. Electronic Medical Records (EMR) provide complementary information to Randomized Controlled Trail (RCT) data. They contain valuable medical information including patient and diagnostic profiles, prescriptions, health care utilization, and lab results, often over many years, that allow one to analyze and understand medication treatment patterns and effectiveness in real world settings. Given the large size of many EMR, they permit study of heterogeneous treatment effects in complex cohorts. It is estimated Veterans Health Administration (VHA) databases contain information on about 6 million veterans. To fully take advantage of EMR, developing and implementing tools that can turn acquired information into knowledge toward supporting diabetes decision making for healthcare providers is crucial, but this technology is currently limited. Our objective is to study real-world glucose lowering treatments patterns and its effectiveness, then to adapt and implement recent artificial intelligence algorithms to “learn from” EMR databases, finally provide data-driven treatment option predictions given individual patient profiles (i.e. precision medicine).
Menu
