Trends in Medication Utilization and Glycemic Control Among Type 2 Diabetes Using a Common Data Model Based on Electronic Health Records From 2000 to 2019

Analyzing the treatment patterns of type 2 diabetes (T2DM) in real practice helps to understand the flow of diabetes management and establish further management plans. Observational Health Data Sciences and Informatics (OHDSI) is an international collaboration created an international data network (Observational Medical Outcomes Partnership Common Data Model, OMOP-CDM). This study was aim to analyze treatment patterns of T2DM using the OMOP-CDM based on electronic health record (EHR) data and to assess whether CDM analysis was feasible to diabetes research. This is a retrospective, observational study using the EHR data of Jeonbuk National University Hospital (JNUH) transformed into OMOP-CDM. The data consisted of medical records of patients visits from January 2000 to December 2019. ATLAS ver. 2.7.6, an OHDSI’s open-source software is publicly available, was used for analysis. The 20 year old EHR data of a JNUH contain about 1.5 million patients. The proportion of adult patients treated for T2DM increased from 1,867 (1.6%) in 2000 to 9,972 (5.1%) in 2019. Sulfonylurea (SU) was the most prescribed drug (73%) followed by metformin (55%) in 2000. On the other hand, in 2019, metformin was the most prescribed (64%), and DPP-4 inhibitor prescription increased rapidly up to 55%, while the SU prescription rate decreased to 36%. The rate of insulin treatment ranged from 16% to 24%, which is higher than national surveyed based on health insurance data. Over time, monotherapy decreased while dual, triple, and quadruple combinations steadily increased. Dual combination was the most common with metformin and DPP-4 inhibitor, triple combination was the most with metformin, SU, and DPP-4 inhibitor in 2019. In analysis of annual HbA1c trends, the proportion of patients with HbA1c of 7% or lower increased (from 32.8% 2000 to 50.2% in 2019). Proportion of patients with HbA1c of 9% or more decreased from 30% to 12%. However, it was found that about half of T2DM patients still had HbA1c values above the target range. In addition, the number of patients who visited our emergency room for severe hypoglycemia did not decrease. Present study revealed that CDM analysis was feasible for diabetes research. Medication utilization patterns have changed significantly over the past 20 years with a shift towards newer drugs. Despite these changes and clinical efforts, improvement in glycemic control is still a challenge and hypoglycemic is still a problem to overcome.