A new prediction diagnosis model of incomplete Kawasaki disease based on data mining with big data

Kawasaki disease (KD) is an acute, self-limited febrile illness occurring in children. In actual clinical situations, unlike complete Kawasaki disease (CKD), incomplete Kawasaki disease (IKD) lacks typical symptoms and is difficult to distinguish from many febrile illnesses, which poses a challenge to accurate diagnosis and misleading the treatment. Therefore, we investigated the independent risk factors for early prediction of IKD in children. In this research, 809 children suffering from IKD were recruited from the Children's Hospital of Chongqing Medical University from 2007 to 2017, as well as 2427 children were related to febrile diseases, divided into the IKD group and the other related febrile disease group. According to the results of univariate analysis, the study population was divided into three age groups to develop group-specific models that demonstrated more effective performance. Finally, the 0-24 months old group obtained eight independent risk factors: CRP, LDH, UA, TP, ALB, RDA, PLT, and HGB, with the ROC curve showing an AUC of 0.862 in the predictive model and 0.88 in the new dataset. Meanwhile, LDH, UA, ALB, PLT, and MCHC were in the 24-60 months old group, among which AUC was 0.83 in the predictive model and 0.82 in the new dataset; the older group obtained LDH, UA, MCHC, and PLT, with an AUC of 0.7 in the predictive model and 0.8 in the new dataset. Particularly, UA is a new independent risk factor of IKD. These findings offer valuable insights into guiding the personalized diagnosis of IKD in pediatric patients.