Risk prediction for hemodialysis initiation: The role of Kidney Failure Risk Equation and personalized timely vascular access referral

Introduction: Optimal timing for arteriovenous fistula creation in chronic kidney disease is challenging. Late referral results in hemodialysis (HD) initiation via a central venous catheter (CVC), whereas early referral may lead to unnecessary procedures. We evaluate the predictive value of estimated glomerular filtration rate (eGFR) and the 2-Year Kidney Failure Risk Equation (KFRE) in forecasting HD initiation. Methods: Included adults referred for vascular mapping with ⩾3 months of nephrology follow-up who selected HD as their preferred modality. The index date was the date of modality selection. We assessed HD initiation timing, vascular access (VA) type, mortality, and the predictive performance of eGFR and KFRE. Results: Included 179 patients, of whom 43.6% ( n = 78) and 58.7% ( n = 105) initiated HD within 12 and 24 months, respectively, with most (59.9%) starting via a CVC (59.9%). Within 24 months, 6.7% ( n = 12) died before HD initiation. Higher urinary protein-to-creatinine ratio, higher KFRE, lower eGFR, male sex, and heart failure with reduced ejection fraction were associated with increased HD risk. Predictive thresholds for HD initiation within 12 months were eGFR 32.8% (sensitivity: 75.6%, specificity: 68.3%, p ⩽ 0.001). For 24 months, thresholds were eGFR 31.0% (sensitivity: 71.4%, specificity: 71.6%, p ⩽ 0.001). Patients exceeding the KFRE threshold had a higher risk of HD initiation than those below the eGFR threshold. A KFRE >40.0% increased specificity, while lower thresholds (>30%) improved sensitivity. Combining eGFR with KFRE enhance specificity but reduced sensitivity. Conclusion: KFRE demonstrates superior predictive performance compared to eGFR for VA planning. Balancing different thresholds values and integrating KFRE with eGFR, can refine HD initiation risk assessment and VA referral, preventing unnecessary surgery and minimizing HD initiation via CVC.