Predicting kidney replacement therapy, cardiovascular disease and all-cause mortality in advanced chronic kidney disease among the Chinese population.

The Grams model, designed to predict adverse event risks in advanced chronic kidney disease (CKD) patients, was evaluated in a Chinese cohort of 1,333 patients with eGFR below 30 mL/min/1.73 m2. The model demonstrated moderate to good discrimination across outcomes, performing well in predicting kidney replacement therapy (KRT) but overestimating the risks of cardiovascular disease (CVD) and mortality. Calibration for KRT was accurate, while other outcomes required recalibration to improve alignment with observed data. Although recalibration enhanced calibration, it did not improve the model's discrimination. Importantly, the study identified key predictors, such as albumin levels, that were not included in the original Grams model but significantly improved prognostic accuracy when incorporated. These findings suggest that while the Grams model has moderate applicability to the Chinese CKD population, its predictive performance can be improved by including additional critical variables. Further efforts are needed to improve the Grams model prior to application in the Chinese CKD population, such as the inclusion of novel valuable predictors.