A novel model for predicting non-responsiveness to intravenous immunoglobulins in Kawasaki disease: The Singapore experience.

Abstract

We aimed to assess the utility of four published risk-scoring methods in predicting intravenous immunoglobulins (IVIG) non-responsiveness in Kawasaki disease (KD) patients from Singapore and develop a new predictive model. We reviewed the medical records of 215 KD children. The performance of existing scoring methods in identifying non-responsive cases based on sensitivities (SN) and specificities (SP) was evaluated in 122 Singaporean Chinese. From our dataset, a model involving six predictors was built. The following respective SN (%) and SP (%) were obtained: Egami: 26%, 68%; Kobayashi: 21%, 62%; Sano: 13%, 86% and Fukunishi: 46%, 71%. These results indicated that the existing scoring methods performed poorly compared to those reported in their respective original publications, which ranged between 68 and 87%. The new predictive model was derived with an improved SN (80%) and SP (80%). Currently available risk-scoring methods have less applicability in the Singaporean Chinese population. The proposed new risk-scoring predictive model derived based on data from Chinese cohort demonstrated much better SN and SP.