An early risk prediction nomogram for coronary artery lesions in Kawasaki disease: a retrospective cohort study
Xiaoya Cong, Liyuan Shen, Feifei ShenAbstract
Background:
Coronary artery lesions remain a major concern in Kawasaki disease. This study aimed to develop and internally validate a practical nomogram for early identification of children at risk for Coronary artery lesions.
Methods:
We retrospectively analysed 216 children with Kawasaki disease divided into a training cohort (n = 151) and a validation cohort (n = 65). Coronary artery lesions was defined as early coronary dilatation and/or coronary aneurysm. XGBoost and SHapley Additive exPlanations were applied to explore age’s nonlinear effect. Predictors were screened via multivariable logistic regression with elastic-net regression for supplementary assessment. Model performance was assessed using receiver operating characteristic analysis, calibration curves, and decision-curve analysis.
Results:
Overall, 46 children (21.3%) developed Coronary artery lesions. SHapley Additive exPlanations analysis identified a U-shaped age-risk pattern with cut-off points at 17 and 58 months. Male sex, age group, fever duration, systemic immune-inflammation index, C-reactive protein/albumin ratio, and intravenous immunoglobulin resistance were independent Coronary artery lesions risk factors incorporated into the nomogram. The area under the receiver operating characteristic curve reached 0.835 (95% CI 0.804–0.867) in the training cohort and 0.818 (95% CI 0.767–0.869) in the validation cohort. Calibration and decision-curve analysis confirmed good consistency and favourable clinical net benefit.
Conclusions:
This nomogram integrates routine clinical variables and interpretable machine learning to identify nonlinear age-related Coronary artery lesions risk and provide individualised early risk estimation. External validation in large prospective multicentre cohorts is needed before wide clinical application.