Background: Comparative effectiveness studies using electronic health records (EHRs) often define chronic kidney disease (CKD) outcomes using laboratory-based definitions. Recommendations for defining CKD require multiple longitudinal measurements. The implications of applying these definitions to EHR data are poorly understood.
Methods: Using a cohort of 36,025 individuals with diabetes from three health systems who initiated a new glucose-lowering medication, we compared concordance and CKD incidence rates and rate ratios for glucose-lowering medication classes for laboratory-based CKD definitions requiring one versus multiple abnormal tests.
Results & conclusion: CKD incidence rates differed more than twofold based on the definition. Reassuring for CER, incidence rate ratios for glucose-lowering medication classes on risk of developing CKD were relatively unaffected by CKD definition.
Keywords: chronic kidney disease; comparative effectiveness research; diabetes mellitus; electronic health records.