A random survival forest model accurately predicted kidney function deterioration in patients with stage 4 cardiovascular-kidney-metabolic syndrome, achieving an area under the curve of 0.930.
Cohort (n=23,014)
No
A random survival forest model using 8 routine clinical variables (including NT-proBNP and LVEF) accurately predicts the risk of kidney function deterioration in patients with stage 4 CKM syndrome.
Cardiovascular-kidney-metabolic (CKM) syndrome is a progressive disease that can affect multiple vital organs. The specific factors contributing to the deterioration of kidney function in stage 4 CKM patients remain unclear, and no relevant clinical prediction model has been established. A retrospective analysis was conducted on eleven years of inpatient data. Stage 4 CKM Patients who fulfilled the diagnostic criteria and did not meet the exclusion criteria were enrolled. The outcome was kidney function progression, defined as a sustained decline in estimated glomerular filtration rate of ≥ 40% from baseline or initiation of kidney replacement therapy. Based on clinical indicators, predictive models were constructed using Cox regression, LASSO-Cox regression and random survival forest algorithms. Calibration curves, receiver operating characteristic curves, and decision curve analysis were employed to validate the models. The SHapley Additive exPlanations method was used to interpret the final model. Based on the model, a web-based risk calculator was constructed for clinical practice. A total of 23,014 subjects with stage 4 CKM were included and randomly divided into two cohorts at a ratio of 2:1 to the development cohort and the validation cohort. During follow-up, 1,772 outcomes (11.6%) occurred in the training cohort and 942 (12.3%) in the validation cohort. Key predictors of kidney function deterioration in stage 4 CKM patients included N-terminal pro-B-type natriuretic peptide, serum albumin, left ventricular ejection fraction, hemoglobin, age, blood urea nitrogen, and urinary protein level. Among the three models, the random survival forest model demonstrated the best discrimination and calibration in the validation set (area under the curve = 0.930, 95% confidence interval: 0.925–0.934 in 12 months). The random survival forest model was highly accurate in identifying the special patients with an elevated risk of kidney function deterioration in stage 4 CKM syndrome patients. Our model can be utilized for prospective monitoring and support personalized management strategies for this high-risk population. Not applicable.
Wang et al. (Mon,) conducted a cohort in Stage 4 cardiovascular-kidney-metabolic (CKM) syndrome (n=23,014). Random survival forest model vs. Cox regression and LASSO-Cox regression models was evaluated on Kidney function progression (sustained decline in eGFR ≥40% from baseline or initiation of kidney replacement therapy) (95% CI 0.925-0.934). A random survival forest model accurately predicted kidney function deterioration in patients with stage 4 cardiovascular-kidney-metabolic syndrome, achieving an area under the curve of 0.930.