- Select the diagnosis group and enter the current CKiD U25 eGFR and urine protein-to-creatinine ratio (UPCR): these three are required.
- Add any of the optional predictors. The tool automatically uses the best-fitting model (lowest Akaike information criterion, AIC) for the data entered, as the original online tool does.
- Read the time by which 10%, 25%, and 50% of children with the same profile are expected to start KRT, and the 2- and 5-year risk.
All computation runs in your browser; no values are stored or transmitted beyond your own device.
When to Use
Use for children and adolescents with stable CKD under pediatric nephrology care, eGFR roughly 15–90 mL/min/1.73 m², to frame prognosis with families and to time dialysis access, transplant evaluation, and preemptive living-donor work-up.
When NOT to rely on it
Not for acute kidney injury, children already on KRT, or adults. Only two diagnosis groups are modeled; responses to specific treatments (for example, for focal segmental glomerulosclerosis, lupus nephritis, or atypical hemolytic uremic syndrome) are not captured. Estimates are percentiles for a population with the same profile, not a deterministic forecast; times beyond 10 years are not estimated.
Pearls & Pitfalls
How the diagnosis group is coded
Hemolytic uremic syndrome (HUS) behaves like non-glomerular disease (congenital anomalies of the kidney and urinary tract, cystic and other non-glomerular causes) and is grouped with it; "glomerular" means glomerular disease other than HUS. The diagnosis group changes the slope of eGFR and the shape of the survival curve, so do not read any single coefficient in isolation.
Use the right eGFR
The models were built on the CKiD under-25 (U25) equations (Pierce et al., 2021), preferably combining creatinine and cystatin C. Adult CKD-EPI or the older bedside Schwartz equation will shift the prediction. Get the input from the CKiD U25 eGFR calculator.
Pitfalls
(1) Blood pressure must be classified with the 2017 American Academy of Pediatrics (AAP) percentiles; use the pediatric BP calculator. (2) In Korean validation, discrimination was excellent but calibration was imperfect; the enriched model with eGFR slope aligned best with observed risk. (3) The models predict the timing of a clinical decision (starting KRT), which also depends on local practice and access.
CKiD Time to KRT
Optional predictors
Next Steps
- Median under about 2 years or 2-year risk high: start KRT preparation now: family education on dialysis modalities, vascular access or peritoneal dialysis catheter planning, and transplant referral with living-donor evaluation to allow a preemptive transplant.
- Modifiable predictors: treat hypertension to AAP targets, reduce proteinuria with renin-angiotensin system blockade, correct metabolic acidosis and anemia, and support nutrition and growth; re-run the model after a year to see the effect of a slower eGFR slope.
- Counseling: share the 25th and 50th percentile times as a range, not a date, and revisit at each annual review.
- Calculate inputs with the CKiD U25 eGFR and pediatric BP calculators.
Evidence
Model structure (Ng et al., 2023)
Time to KRT follows a generalized gamma distribution: log T = β + σ·W. All predictors modify the location β; diagnosis group also modifies the scale σ and the shape κ. eGFR enters as log(eGFR/45) with a change in slope above 45; UPCR as log(UPCR/0.5) with a change in slope above 0.5; the eGFR slope as the annual change in log eGFR.
| Model | Predictors (beyond diagnosis, eGFR, UPCR) | AIC | c-statistic |
|---|---|---|---|
| Elementary | None | 1476.5 | 0.865 |
| Partially enriched 1 | BP, eGFR change | 1470.1 | 0.866 |
| Partially enriched 2 | BP, anemia | 1457.7 | 0.874 |
| Partially enriched 3 | BP, albumin, chloride, CO₂ | 1456.7 | 0.871 |
| Partially enriched 4 | BP, anemia, albumin, chloride, CO₂ | 1446.2 | 0.875 |
| Enriched | All of the above plus eGFR change | 1445.1 | 0.875 |
Derivation: 890 children in CKiD (median age 11.4 years, median eGFR 50 mL/min/1.73 m², 29% started KRT over a median of 5.2 years). External validation of the elementary model in the European ESCAPE cohort: c-statistic 0.854, calibration slope 0.996.
Korean validation (Park et al., 2026)
533 children in KNOW-Ped CKD (median age 10.8 years, median eGFR 57.6 mL/min/1.73 m²); 32.1% started KRT over a median of 4.8 years. C-statistics were 0.911–0.972 across models and calibration slopes exceeded 0.9, but the goodness-of-fit test showed miscalibration; models including the eGFR slope matched observed risk most closely.
Implementation check: this page reproduces both worked examples in Ng et al. (2023): median 3.3 years for a non-glomerular profile with eGFR 30 and UPCR 2 (elementary model), and 2.5, 5.1, and 10.1 years for the published enriched-model glomerular profile.
