Risk-based referral model to nephrologist specialist care in Stockholm

For most patients, clinical management of the early stages of chronic kidney disease is performed in primary care settings. The Kidney Disease: Improving Global Outcomes (KDIGO) 2024 guidelines recommend using a 5-year kidney failure risk equation (KFRE) of 3-5% to guide nephrologist referrals. Here, we aimed to assess the impact of adopting a risk-based referral model compared with traditional referral criteria.

We conducted an observational retrospective study of adults with an estimated glomerular filtration rate (eGFR) <60 ml/min/1.73 m2 (Lund-Malmö equation) from the SCREAM project, a healthcare utilization cohort from Stockholm, Sweden. We evaluated the performance of the non-North American four-variable KFRE and recalibrated it to better fit our setting. KFRE thresholds were compared with traditional models-the clinical Swedish criteria and the classic KDIGO 2012 criteria-both of which are mainly based on age, eGFR and albuminuria thresholds. Sensitivity, specificity, positive and negative predictive values, reclassification matrices, net reclassification improvement and decision curve analyses were used to assess performance and clinical utility.

The study included 887 388 observations from 192 964 individuals. At inclusion, 49% were men, median age was 76 years and median eGFR was 54 ml/min/1.73 m2. During follow-up, 2624 (1.4%) progressed to KRT. The KFRE demonstrated good prediction performance, which further improved after recalibration. Both the non-North American- and SCREAM-recalibrated KFRE provided higher sensitivity and specificity than Swedish and classical KDIGO criteria. KFRE-based referral models yielded better net reclassification improvement, demonstrating superior performance in decision curve analyses. Higher thresholds (15% for the non-North American-recalibrated KFRE, 9% for the SCREAM-recalibrated KFRE) than the KDIGO recommended ones provided the best combined sensitivity and specificity. Compared with traditional referral models, implementation of a risk-based referral would decrease the number of unnecessary referrals by 23% and 25%, respectively.

In a large northern European healthcare system, transitioning to a risk-based referral model would result in an important reduction in unnecessary referrals while maintaining a low rate of missed cases, optimizing resource utilization.