Develop T2D Patient-Centered Treatment Suggestion Rule using EMR data In clinical practice, physicians and health care providers often follow the treatment guidance based on published research and experts' opinions. The American Diabetes Association (ADA) annually publishes updated recommendations for Type 2 Diabetes (T2D) management. Although the standardized diabetes management approach has resulted in substantial improvement in overall diabetes care, different patients often respond to treatments differently (treatment heterogeneity effects). This proposal aims to develop methods to facilitate personalized treatment recommendations using information from electronic medical records (EMR) for T2D. We will first evaluate the real world effectiveness of different treatments when diabetes patients follow the treatment guidance. We will then assess the treatment differences, identify the baseline information that has predictive ability for the treatment differences, and develop treatment recommendation rules for a single individual or subgroups of individuals. As EMR is a type of observational study, the propensity score matching method will be adopted to determine causal relationships. Finally, we will use cross validation and independent data to validate our results. Methods proposed in this research will be implemented in an efficient and user-friendly software package to further facilitate easy patient-centered treatment decision-making. The proposed method uses information from EMR data, which contains comprehensive baseline information for approximately 20 million US patients and ?real world? drug effectiveness. Therefore our method bridges the gap of using ?big and generalizable? data for patient-centered outcomes research.
Type 2 Diabetes (T2D) is one of most costly chronic medical conditions and its care incurs significant burdens on individuals, society, and the health care system. The standardized T2D management approach has resulted in substantial improvement in overall diabetes care, and is associated with a marked relative reduction in morbidity and mortality among patients with T2D. Although there are tremendous advances in diabetes care and risk identification, it still remains the leading cause of preventable blindness and nontraumatic amputations in the USA. Treatment heterogeneity is substantial among T2D patients. Given the rich information stored in electronic medical records, we thus propose to use the information from electronic medical record data to develop personalized treatment algorithms for T2D patients.
