Abstract
IgA nephropathy (IgAN), the most common primary glomerular disease worldwide, poses challenges in predicting progression at diagnosis-particularly in primary care. This study developed and validated a model to estimate 5-year renal survival and support early risk stratification for personalized management. A total of 1135 patients with biopsy-confirmed IgAN from Hangzhou Hospital of Traditional Chinese Medicine (2014-2017) were retrospectively enrolled and randomly assigned to training and internal validation cohorts in a 7:3 ratio. An external validation cohort comprising 352 patients was obtained from three independent centers (2015-2020). To identify robust prognostic variables, 1000 bootstrap samples were generated from the training set, each subjected to tenfold cross-validation to determine the optimal regularization parameter (λ) for the LASSO-Cox model. Variables with non-zero coefficients were recorded across iterations, and the five most frequently selected were used to construct both the XGBoost survival model and a corresponding nomogram. Model performance was evaluated through discrimination, calibration, and clinical utility using ROC curves, Brier scores, calibration curve and decision curve analysis. The area under the curve (AUC) of the nomogram was 0.951 (95% CI, 0.914-0.988) in the training cohort and 0.927 (95% CI, 0.877-0.978) in the internal validation cohort. In the external validation cohort, the AUC was 0.913 (95% CI, 0.870-0.955). The Brier scores were 0.029 and 0.045 for the internal and external validation cohorts, respectively. DCA further demonstrated the favorable clinical utility of the nomogram. A clinically practical prognostic model incorporating routine clinical and pathological features was developed to estimate 5-year renal survival in patients with IgAN. Specifically designed for primary care settings, the model leverages easily accessible data to enable early identification of high-risk individuals and support personalized long-term management. Its simplicity and applicability in resource-limited environments make it a valuable tool for improving outcomes beyond specialist centers.