Abstract: TH-OR69
A Prediction Equation for Incident CKD Using Routinely Collected Data: The Kidney Disease Risk Equation (KDRE)
Session Information
- Preventing Progression and Reassessing Race in GFR Estimation
November 04, 2021 | Location: Simulive, Virtual Only
Abstract Time: 04:30 PM - 06:00 PM
Category: CKD (Non-Dialysis)
- 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Sood, Manish M., Ottawa Hospital Research Institute, Ottawa, Ontario, Canada
- Dixon, Stephanie, Institute for clinical and evaluative sciences, London, Ontario, Canada
- Rhodes, Emily, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada
Background
The identification of individuals at risk for incident CKD (eGFR < 60 ml/min, stage 3a) is an important first step for disease surveillance, monitoring, education and allocation of key therapies to reduce CKD progression. Despite recommendations, albuminuria measurements in appropriate individuals remains poor. As such, we set out to develop and validate a prediction equation for new onset CKD with and without an albumin creatine ratio (ACR).
Methods
Population-level administrative data cohort of 1,109,905 adults (>66 years old) from Ontario, Canada April 1, 2008 and December 31, 2017 with a minimum of 2 eGFR measures (one for baseline > 70 ml/min, one for outcome) were included. Prediction equations stratifying individuals with (n=191,690) and without (n=998,825) ACR were derived, internally validated by bootstrapping and externally validated in 122,144 (22,809 ACR, 99,335 non-ACR) individuals in Manitoba, Canada. The study outcome was a single eGFR measure < 60 mls/min/1.73 m2 with up to 10 years follow-up. In additional analyses, we examined two eGFR measures < 60 mls/min and a single eGFR < 45 ml/min as study outcomes.
Results
Among individuals (54.5% women, mean age 64 SD 7, mean baseline eGFR 82 SD 8, median ACR 1 IQR 1-3), an eGFR < 60 ml/min occurred in 37.2% during the follow-up. The final model including up to 6 variables (age, sex, baseline eGFR, hemoglobin, time from hypertension and diabetes mellitus diagnosis) yielded a 5-year c-statistics of 0.77 (no ACR) and 0.78 (with ACR) with excellent calibration. Model performance was similar in additional analyses and in an external validation.
Conclusion
An equation incorporating readily available and routinely collected administrative data variables can accurately predict the onset of CKD with or without ACR.