Identification of children who are at high risk for having varices using noninvasive tests would enable the selection of children for future studies of primary prophylaxis of variceal hemorrhage, but this has been inadequately studied. The objective of the study was to derive a noninvasive clinical prediction rule that is able to identify children with esophageal varices. Fifty-one consecutive children with liver disease or portal hypertension who underwent endoscopy were included in the present retrospective study. At endoscopy, variceal size was graded on a 4-point Likert scale. Results of physical examination, blood tests, and abdominal ultrasound scan (USS) were recorded. Spleen length on USS was expressed as a standard deviation score (z score). A descriptive univariate analysis was performed on variables that were potentially associated with esophageal varices and multivariate logistic regression was then modeled to derive a clinical prediction rule. Esophageal varices were found in 17 of the 51 children (33%). Variables found to differ significantly between children with and without varices included platelet/spleen-length z score ratio (P < 0.001), platelet count (P < 0.001), international normalized ratio (P = 0.001), aspartate aminotransferase/alanine aminotransferase ratio (P = 0.002), and albumin (P = 0.003). Using multivariate logistic regression, a model with platelet count, spleen length z score, and albumin as the independent variables had the best fit. Area under the receiver operating characteristic curve for this clinical prediction rule was 0.93 (95% confidence interval 0.85-0.99), sensitivity 94%, specificity 81%, positive predictive value 0.83, negative predictive value 0.94, positive likelihood ratio 5, and negative likelihood ratio 0.06. This clinical prediction rule is a simple noninvasive measure that may identify children at high risk for esophageal varices. A prospective validation study is in progress.