Background: The purpose of this study was to determine a score on the Obsessive Compulsive Scale (OCS) from the Child Behavior Checklist (CBCL) to screen for obsessive compulsivedisorder (OCD) in children and to rigorously test the specificity and sensitivity of a single cutpoint. Methods: A receiver operating characteristic (ROC) curve analysis was a pplied to data from 61 patients with clinically determined OCD, 64 clinical controls and 73 general population controls to determine the best sum score on the CBCL-OCS to predict confirmed OCD in children. Using the ROC-determined cutoff, this score was applied to a national sample of CBCL data from 2460 singleton children ages 4-18 and to 20,016 children ages 7-18 from three large general population twin samples to determine the estimated prevalence in the general population. Results: Using a CBCL-OCS score of 5 demonstrated an area under the curve (AUC) of .88 with high sensitivity (92%) and moderate specificity (67%) compared to clinical controls. Compared to the general population controls, the AUC was .96 with high sensitivity (92%) and specificity (89%). In the twin samples, the number of participants with CBCL-OCS scores above this cutpoint was 2.3-7.1%. Conclusions: These findings suggest that the OCS of the CBCL may provide a highly effective way to screen for childhood OCD, and that the prevalence of childhood OCD may have been underestimated, thus prompting the need for further research into screening children for this condition. © 2005 Association for Child Psychology and Psychiatry.