#2977 PROGRESSION AND REGRESSION OF CHRONIC KIDNEY DISEASE IN THE UNITED STATES USING ELECTRONIC MEDICAL RECORDS

Abstract

Abstract Background and Aims Chronic kidney disease (CKD) is generally considered to be progressive and may lead to end-stage kidney disease (ESKD), or death. We aim to analyze a large US database and capture CKD progression, regression, ESKD and death over a 5-year follow-up, in a real-world setting. Method A retrospective cohort study was conducted using data from IBM® Explorys electronic medical records (>53 million patients) from 1 Jan 2010 to 31 Dec 2019. We created two main cohorts of patients with CKD: Cohort 1 and Cohort 2. Cohort 1 was based on diagnostic codes only (ICD 9/10, n = 810,987 unique patients), and Cohort 2 on KDIGO screening and staging criteria only (eGFR, UACR/UPCR values, n = 953,396). Within these cohorts, we created individual sub-cohorts for each CKD stage (1-5), though patient categorization was not mutually exclusive (one patient could contribute to multiple main and sub-cohorts). Patients in Cohort 1 were required to have an ICD code specific to a stage, and patients in Cohort 2 were required to have two consecutive laboratory test results within 3–18 months to determine KDIGO CKD stage. Index date was the date of diagnosis or second abnormal test with at least 12 months lookback period. Progression and regression were assessed for Cohort 1 using change in ICD codes and for Cohort 2, two test results (3-18 months apart) were required with a change in KDIGO eGFR stage and at least one of the values showing an increase or decrease by ≥25% in eGFR from baseline. We performed individual follow-up for progression and regression from each CKD stage to a progression or a regression event, death, loss to follow-up, or maximum 5 years, whichever came first. For ESKD, we followed up from each CKD stage to ESKD, death, loss to follow-up, or maximum 5 years. For death, we followed up from each CKD stage to death, loss to follow-up, or a maximum of 5 years. Results Most patients in Cohort 1 (ICD) and 2 (KDIGO) had stage 3 CKD at index. Mean age was 70y in Cohort 1 and 71y in Cohort 2, with 49% and 40% males, respectively. Both cohorts were predominantly Caucasian. The prevalence of cardiovascular diseases increased with CKD stage. The proportion of patients from Cohort 2 who had at least one CKD diagnosis increased from 34% in stage 1 to 86% and 83% in stage 4 and 5, respectively. CKD regression, ESKD and death increased progressively from stage 1–5 in both cohorts, except for stage 5 in Cohort 2. Progression in Cohort 1 decreased with stage and was variable in Cohort 2. The proportion of patients experiencing at least one event corresponding to CKD regression, progression, and ESKD was higher at all stages in Cohort 1 vs. Cohort 2, particularly ESKD which was markedly higher for stage 5 (91.4% vs. 13.6%). Proportion of death was similar in both cohorts and was the highest in stages 4–5. Conclusion Substantial differences in kidney outcomes were noted between CKD classifications derived from ICD codes vs. KDIGO criteria, except all-cause mortality which was comparable across both cohorts and increased notably with stage. Changes in kidney function, both progression, regression, and ESKD occurred more frequently in Cohort 1. Our analysis suggests ICD diagnosis and KDIGO criteria might not be used interchangeably in the context of utilizing real-world data sources.