Using EHR Data to Evaluate the Burden of Diabetes Mellitus in a National Network of Children’s Hospital Health Systems

Abstract: Diabetes mellitus is a group of disorders characterized by hyperglycemia resulting from defects in insulin production, insulin action, or both. In children and adolescents 0-17 years-old, pediatric diabetes mellitus (PDM) is one of the more common chronic diseases. Mounting evidence suggests that rates of both Type 1 and 2 diabetes mellitus in children and adolescents have been increasing over the past 30 years. Nonetheless, there has been limited research on how these rates differ by sociodemographics (e.g., race/ethnicity, geography) and clinical characteristics (e.g., body mass index) across diverse regions of the US. To help fill this gap, the proposed project will use electronic health record (EHR) data to assess the prevalence and incidence of PDM, overall and by diabetes type and patient sociodemographic and clinical characteristics. Data will come from PEDSnet, a national pediatric clinical research network that has transformed EHR data to a common data model for over 6.5 million children. PEDSnet includes 8 pediatric medical centers that provide care to children in all 50 states; however, the 11 states with the greatest concentration of children are: CO, DE, FL, IL, IN, KY, MO, NJ, OH, PA, and WA.

In addition to participating in the DiCAYA consortium, we propose: (Aim 1)—to evaluate and improve the quality of the EHR data that will be used for identifying patients with PDM; (Aim 2)—to implement an EHR-based computable phenotype methodology for each type of PDM to support accurate, efficient, and timely surveillance; and, (Aim 3) to compute prevalence and incidence rates of PDM, overall and by diabetes type and patient sociodemographic and clinical characteristics. Aim 1 (data quality) will take advantage of PEDSnet’s well-established data quality program that evaluates both structural and semantic data quality and works with institutional data contributors to remediate data quality problems. Aim 2 (Computable Phenotyping) will implement validated algorithms for identifying children and adolescents with PDM. And, the denominator population for Aim 3 (RateComputations) will be patients who reside in one of the 62 counties for which PEDSnet has representative data and who have >1 contacts with a PEDSnet institution during the observation period. We plan to harmonize our methods with the rest of the DiCAYA consortium to enable standardized assessments of disease rates. Our team has extensive experience working in consortia, such as PCORnet and OHDSI, that share and execute each other’s data science methods. Our attention to evaluating and improving EHR data quality, constructing and testing pediatric EHR-based computable phenotypes, and use of a national network of major pediatric medical centers are key strengths of this proposal.