Use of the kidney failure risk equation: a regional retrospective primary care cohort study in England
Kidney failure risk scores identified referral candidates, but death was usually more likely than kidney replacement treatment.
*Retrospective cohort study; Level 2b (OCEBM).
Citation
Stewart S, Kalra PA, Kontopantelis E, Blakeman T, Tilston G, Sinha S. Use of the kidney failure risk equation: a regional retrospective primary care cohort study in England. British Journal of General Practice. 2026. doi:10.3399/BJGP.2025.0490
Background
Primary care clinicians use the Kidney Failure Risk Equation to estimate future kidney replacement needs in chronic kidney disease. However, the equation does not account for the competing risk that a patient may die before kidney failure develops.
Patients
Adults with chronic kidney disease stages 3–5 in Greater Manchester primary care records from 2018 to 2023. Patients already receiving dialysis or transplant, and those with dementia or palliative care codes, were excluded; regression analyses also excluded missing body mass, deprivation, or ethnicity data.
Intervention
Kidney Failure Risk Equation score categories: less than 5%, 5%–20%, and greater than 20% predicted 5-year risk.
Control
The less than 5% risk category.
Outcome
All-cause death and kidney replacement treatment.
Follow-up Period
5–6 years.
Results
The study included 109,543 patients; 30,197 had urine albumin testing in 2018 or 2019 and were included in the main adjusted model. On average, 11% had a risk score above 5%, the threshold at which referral may be considered.
| Risk category | Kidney replacement treatment | Death | No event |
|---|---|---|---|
| <5% | 0.1% (95% CI, 0.1–0.2) | 11.9% (95% CI, 11–13) | 87% |
| 5%–20% | 2% (95% CI, 1–3) | 23% (95% CI, 20–27) | 75% |
| >20% | 14% (95% CI, 8–23) | 29% (95% CI, 25–36) | 57% |
In subgroup analyses, Black and Asian patients with scores above 20% had higher kidney replacement probability than death probability.
Limitations
Retrospective record data may contain coding errors. Urine albumin testing was missing in 39.9%, limiting generalizability. The model did not account for timing of events. Kidney failure was identified by treatment codes, excluding conservative kidney care. COVID-19 may have affected testing and care.
Funding
Kidney Research UK; no funder role. Some authors reported industry payments.
Clinical Application
Use the risk equation for referral triage, but discuss competing death risk and prioritize urine albumin testing.
Discussion
Sign in to join the discussion.
In this retrospective primary care cohort, higher KFRE category was associated with RRT and death (adjusted RRT 14% vs death 29% for KFRE >20%). Given the observational design and missing uACR data, how valid and applicable is using KFRE >5% to change nephrology referral practice? In this cohort, patients with KFRE >20% had a relative risk ratio for RRT of about 203 compared with KFRE <5%, but the adjusted probability of RRT in the >20% group was 14%. Which interpretation is most appropriate?