Prediction of Kawasaki disease coronary artery lesions in the Japan Environment and Children's Study.

Objective: This study aimed to validate a composite of previously identified risk factors for predicting coronary artery lesions in Japanese children with Kawasaki Disease (KD). Materials and Methods: Children diagnosed with KD from the Japan Environment and Children's Study (JECS) cohort were included. We assessed the discriminative ability of the model using the area under the curve (AUC) and evaluated its calibration using calibration plots. Based on clinical insights, we derived our original model, which consisted of variables established in previous score models or studies. Using the bootstrap method, internal validation was assessed with the AUC and calibration slope. Results: Our analysis included 959 children with KD, derived from 104,062 pediatric records contained in the JECS. Our model comprised the following variables: sex, hematocrit, platelet count, C-reactive protein, and serum sodium levels. The mean and 95% confidence intervals of the AUCs for the Harada, Egami, and Kobayashi models were 0.66 (0.59-0.72), 0.59 (0.52-0.66), and 0.64 (0.56-0.72), respectively. Our score model with bootstrap analyses produced an AUC of 0.80 (0.68-0.91) and a calibration slope of 1.01. Conclusions: We built and evaluated our score model using indicators that were easily accessible with a moderate-to-high AUC and stable calibration results, indicating the high reliability of the model for clinical application. Further studies are required to validate our model in other cohorts.