Unpacking the Complexity of Diabetes Care through Investigating Disease Control and Therapeutic Inertia among Patients with Type 2 Diabetes

Abstract

Type 2 diabetes (T2D) is a prevalent chronic condition requiring complex management and routine assessment of patient’s diabetes treatment. Therapeutic inertia occurs when patients do not receive timely, diabetes treatment modifications despite sub-optimal disease management. The long-lasting challenges with diabetes management in the United States (U.S.) called for further research. Innovative methodologies were needed to understand how the multifactorial factors predicted therapeutic inertia in T2D, along with investigations of health outcomes succeeding therapeutic inertia exposure. This dissertation aimed to examine factors associated with disease control in diabetes (glycemic levels), predict patients’ risk of therapeutic inertia, and investigate subsequent health outcomes after therapeutic inertia exposure. This dissertation completed three Aims using a retrospective, observational approach to study real-world data sources (patient-reported surveys and electronic health record (EHR) data). Aim 1 analyzed the National Health and Nutrition Examination Survey (NHANES) to examine glycemic levels among a nationally representative sample of people with diabetes stratified by antihyperglycemic medications and contextual factors. Aims 2 and 3 analyzed EHR data from the OneFlorida Clinical Research Consortium (linked data from 1,240 practices/clinics across Florida). Aim 2 applied machine learning methods to predict therapeutic inertia among patients with T2D and estimated the therapeutic inertia prevalence. Aim 3 assessed diabetes-related complications, glycemic levels, and healthcare utilization after exposure to therapeutic inertia in T2D. For results, meeting guideline-based glycemic levels (52% met A1C<7%) was associated with antihyperglycemic medication use and contextual factors (e.g., gender, language spoken, food security, healthiness of diet, healthcare use, insurance coverage). Machine learning predicted patients’ risk of therapeutic inertia (prevalence=40%; C-statistic=0.84). Some top predictors of therapeutic inertia risk included patient (age, prior ambulatory visits, metformin prescription), provider (provider specialty), and healthcare system (health insurance) factors. Therapeutic inertia exposure was significantly associated with lower composite scores for diabetes-related complications and severity, higher glycemic levels, and lower rates of ambulatory visits. This dissertation generated real-world evidence for disease control and therapeutic inertia across the T2D trajectory. The public health significance was to provide scientific knowledge for optimizing diabetes care and mitigating therapeutic inertia in T2D, ultimately to improve the lives of patients with diabetes in the U.S.