Abstract: PO2341
Developing an Electronic Health Record (EHR)-Based Model for Delineating Advanced CKD Cohort in Veterans Affairs (VA) System
Session Information
- Reassessing Race in Predicting Progression
November 04, 2021 | Location: On-Demand, Virtual Only
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Purvis, Zachary Pitt, VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Chamarthi, Gajapathiraju, University of Florida, Gainesville, Florida, United States
- Shell, Popy, VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Fu, Devin S., VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Orozco, Tatiana, VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Shukla, Ashutosh M., VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
Background
Late diagnosis of chronic kidney disease(CKD) and non-referral to nephrology are important limiting concerns for pre-end stage kidney disease(ESKD) nephrology care, including dialysis modality education. Querying existing electronic database with longitudinal patient-level data can improve recognition of advanced, stage 4/5 CKD and facilitate evidence-based pre-ESKD care. Using a mixed approach of electronic query followed by manual chart review, we report the development of an Electronic Health Records (EHR)-based model that allows identification and quantification of advanced CKD burden in a regional Veterans Healthcare System(VHS).
Methods
We identified all Veteran enrollees at a large regional VHS using VA Informatics & Computing Infrastructure data set. Among these, we identified all Veterans with an eGFR below 30ml/min or an existing ICD-10 diagnostic code for stage 4/5 CKD within last 12 months. We applied diagnostic and procedure codes for dialysis, ESKD, and acute kidney injury(AKI) in an iterative approach to improve the accuracy of identifying non-dialysis advanced CKD cohort.
Results
Of 148,164 active enrollees within VHS, our initial model of using a single eGFR <30 ml/min identified 3,813(2.57%) Veteran enrollees with advanced CKD. Manual review of a select cohort(n=787) showed 63.3% error rate, with high rates of ESKD and AKI being major confounders. Successive iterations involved exclusions of ESKD and AKI codes and incorporation of a second latest eGFR >90 days before latest eGFR. The final EHR-based advanced CKD model included 1,329(0.89%) Veteran with the residual error of 14.4% on manual chart review without the possibility of further automated exclusions. Of these, 872 were found to have definite advanced CKD and 457 were classified as probable advanced CKD based on whether both or only one of the latest two eGFRs more than 90 days apart were below 30 ml/min with CKD.
Conclusion
An EHR-based model to identify advanced CKD can be successfully developed for a regional VHS with over 85% accuracy. Further testing is needed to determine its wider applicability across additional VHA sites, and if validated, this model can be applied across the VHA electronic data to identify the burden of advanced CKD for needs assessment and clinical care among Veterans.
Funding
- Veterans Affairs Support