Development of a nomogram for predicting hyperkalemia in advanced chronic kidney disease

Abstract

Abstract Hyperkalemia (HyperK) is a potentially life-threatening complication in advanced chronic kidney disease (CKD), yet its prediction in real-world outpatient settings remains challenging. In a retrospective cohort study including 395 patients with CKD stages 4–5, all with baseline serum potassium levels within the normal range, followed for up to 2.2 years. Clinical, biochemical, and pharmacological variables were obtained from electronic health records, and logistic regression with LASSO selection was applied to identify independent predictors of HyperK. Sex-stratified nomograms were developed to facilitate individualized risk estimation, and model performance was assessed using AUROC, calibration plots, and internal validation with 1,000 bootstrap resamples. During follow-up, 303 patients (76%) developed HyperK. Independent predictors included higher serum creatinine, calcium, and age, while higher sodium levels, hemoglobin, obesity, and thiazide use were associated with lower risk. In adjusted models, men had a 49% lower risk of HyperK (OR 0.51, 95% CI 0.28–0.92). Sex-specific nomograms demonstrated good discrimination, with AUROC of 0.78 in men and 0.81 in women, and calibration analyses confirmed adequate model fit. Importantly, both models showed a high negative predictive value (>95%), supporting their use in safely identifying low-risk patients who may require less intensive monitoring. Secondary analyses showed that higher phosphate was independently associated with mortality (OR 1.74, 95% CI 1.04–2.93), while increased creatinine predicted the need for kidney replacement therapy (OR 1.29, 95% CI 1.08–1.56). These findings provide validated, sex-specific nomograms that enable individualized risk prediction of HyperK in advanced CKD, supporting personalized management in outpatient nephrology care