A Machine Learning–Based Model to Estimate the Risk of Pulmonary Hypertension in Chronic Kidney Disease Patients

Abstract

Pulmonary hypertension (PH) is a common complication in patients with chronic kidney disease (CKD) and is associated with high mortality. Early detection and proper management may improve outcomes in high-risk patients. This study aimed to develop a simple and effective model for screening PH risk in this population. We retrospectively screened 1082 CKD patients. Feature selection was performed using the least absolute shrinkage and selection operator, univariate and multivariate logistic regression (LR). Nomograms were developed for PH risk assessment. The discriminative ability was estimated by the area under the receiver operating characteristic curve (AUROC), and the accuracy was assessed with a Brier score. Models were validated externally by calculating their performance on a validation cohort. Eight machine learning models were developed, and their performance was evaluated. Decision curve analysis and clinical impact curve were used to assess the model's clinical usefulness. A total of 440 patients were included in the analysis, with 308 in the development cohort and 132 in the validation cohort. The final nomogram included five variables as follows: haemoglobin, gamma-glutamyl transferase, triglycerides, coronary heart disease and NT-proBNP. The AUROC of the model was 0.772 (95% CI: 0.731–0.806). External validation confirmed the model's good performance, with an AUROC of 0.782 (95% CI: 0.696–0.854). Among the eight machine learning models, LR showed the best performance. We developed a machine learning model based on clinical and biochemical features to assess PH risk in CKD patients. It enables early detection and risk stratification during follow-up.