Automated Identification of Diabetic Individuals with Renal Complications
The NIDDK has initiated the Kidney Precision Medicine Project (KPMP), which tests the hypothesis that molecular phenotyping of human kidney tissue with multi-omic platforms will identify disease mechanisms and therapeutic targets that improve patient outcomes. Patients enrolling in KPMP will undergo research biopsies, i.e., kidney tissue is being obtained from subjects who would not be biopsied as standard of care. Given the altruism of KPMP participants, we need to be certain the clinical and molecular data sets are valid and accurate to optimize the likelihood of KPMP success. Numerous studies have evaluated the quality, precision and accuracy of molecular phenotypic data derived from genome-scale analytic platforms. Traditionally coordinators have extracted clinical data from medical records; some studies have used double data entry to quantify human error inherent in this task. The process is laborious and by its very nature limited to those data elements absolutely necessary for the project's execution. In contrast, electronic health record (EHR) data can be mined for data elements that are not usually collected. Like omic tools, clinical data extracted from EHRs requires validation. Diabetic kidney disease (DKD) is a major public health burden and a primary focus of the KPMP. We propose to use rule-based algorithms to identify subjects with DKD from an EHR-derived research database and assess the precision and accuracy of key phenotypic variables and outcomes by comparing EHR data to curated kidney disease databases. Since 2001, the Cleveland Clinic has used an EHR. EHR data elements were used to generate a validated, chronic kidney disease (CKD) registry with over 160,000 subjects. The CKD registry includes Chronic Renal Insufficiency Cohort (CRIC) study subjects with curated demographic, clinical, laboratory and medication data. To achieve our goal, we propose to: Specific Aim 1: Establish a research relational database and verify the accuracy and completeness of the data extracted. Specific Aim 2: Quantify the accuracy of a rule-based algorithm to identify subjects in the CKD registry with DKD and determine the precision of the research database-derived data in replicating medication lists and a critical kidney disease outcome variable, eGFR slope, from curated registries.

Curation Flag Information