A predictive model and nomogram for coronary artery injury in Kawasaki disease based on laboratory indicators: a retrospective study

Background To explore the changes in various laboratory parameters in children with Kawasaki disease, analyze their correlations with complete and incomplete Kawasaki disease as well as coronary artery lesions and non-coronary artery lesions, and establish prediction models and nomograms. Objective Incomplete Kawasaki disease (IKD) is prone to misdiagnosis, and coronary artery lesion (CAL) represents its severe complication. This study aimed to explore the predictive value of routine laboratory indicators for typical/incomplete KD and CAL, and to construct reliable predictive models and nomograms, thereby providing quantitative tools for early screening and risk stratification in primary clinical practice Methods A total of 131 children with confirmed Kawasaki disease (KD) from January 2023 to June 2025 were retrospectively enrolled, who were assigned to the TKD group ( n = 95) and IKD group ( n = 36), as well as the coronary artery lesion (CAL) group ( n = 39) and non-CAL (NCAL) group ( n = 92). Univariate and multivariate logistic regression analyses were applied to screen out independent influencing factors. The performance of the predictive models was evaluated using receiver operating characteristic (ROC) curve, calibration curve and decision curve analysis (DCA). Bootstrap resampling was adopted for internal validation of the models, and visual nomograms were constructed accordingly. Result Total protein (TP) was the only independent factor for differentiating typical from IKD; hypoalbuminemia, hyponatremia, and elevated lactate dehydrogenase (LDH) were identified as independent risk factors for KD complicated with CAL, with hypoalbuminemia being the strongest predictor (OR = 0.783, P = 0.001). The area under the curve (AUC) of the predictive model for TKD was 0.762, and that of the CAL predictive model was 0.790. Both models showed good calibration and positive clinical net benefit, and the corresponding nomograms enabled rapid individualized quantitative risk prediction. Conclusion This study confirmed the clinical value of routine laboratory indicators in phenotypic differentiation of KD and risk prediction of CAL. The constructed predictive models and visual nomograms feature convenient detection and simple operation, which can provide practical references for the early precise diagnosis and treatment of KD and the prevention and control of CAL risk, and are particularly suitable for primary medical institutions with limited diagnostic resources.