Objective To construct a predictive tool for the efficacy of intravenous immunoglobulin (IVIG) therapy in children with Kawasaki disease (KD) in Beijing, China. Design This was a cohort study. Data set (including clinical profiles and laboratory findings) of children with KD diagnosed between 1 January 2010 and 31 December 2015 was used to analyse the risk factors and construct a scoring system. Data set of children with KD diagnosed between 1 January 2016 and 1 December 2016 was used to validate this model. Setting Children's Hospital Capital Institute of Pediatrics and Beijing Children's Hospital. Patients 2102 children diagnosed with KD. Interventions No. Main outcome measures Responsiveness to IVIG. Results The predictive tool included C reactive protein >= 90 mg/L (3 points), neutrophil percentage >= 70% (2.5 points), sodium ion concentration < 135 mmol/L (3 points), albumin < 35 g/L (2.5 points) and total bilirubin > 20 mu mol/L (5 points), which generated an area under the the receiver operating characteristic curve of 0.77 (95% CI 0.71 to 0.82) for the internal validation data set, and 0.69 (95% CI 0.58 to 0.81) and 0.63 (95% CI 0.53 to 0.72) for two external validation data sets, respectively. If a total of >= 6 points were considered high-risk for IVIG resistance, sensitivity and specificity were 56% and 79% in the internal verification, and the predictive ability was similar in the external validation. Conclusions The predictive tool is helpful in early screening of high-risk IVIG resistance of KD in the Beijing area. Consequently, it will guide the clinician in selecting appropriate individualised regimens for the initial treatment of this disease, which is important for the prevention of coronary complications.