The objective of this study was to determine predictive factors in children with Kawasaki disease (KD) with which we could distinguish the patients with KD who are either at very low risk or at very high risk for coronary artery inflammation (i.e., either patients who do not need intravenous immunoglobulin treatment or patients in whom more aggressive or even experimental therapies should be considered). Prospectively collected demographic, clinical, and laboratory data on 344 patients treated for KD were correlated with the patients' echocardiographic findings. The parameters studied were age, sex, duration of the fever, erythrocyte sedimentation rate, hemoglobin, white blood cell count, platelet count, and serum albumin. These were examined both in bivariable comparisons and in multiple logistic regression models. Low serum albumin, age <1 year, and the duration of the fever prior to treatment were risk factors for coronary arteritis. In the multivariable models, their combined predictive value for coronary lesions was poor, especially when identifying the patients at a low risk for coronary artery lesions (CALs). In fact, 44 of 98 patients with CALs were falsely classified to the low-risk group. Ten of 14 patients younger than 1 year of age, who also had low serum albumin (<30 g/L), had echocardiographically verified CALs, and 7 (50%) had a definite coronary artery aneurysm. We could not distinguish a group at such a low risk that these patients could be left untreated. Young patients with low albumin run a very high risk for CALs.