Parent and teacher identification of children at risk of developing internalizing or externalizing mental health problems: a comparison of screening methods.

The accuracy of screening methods for predicting children's internalizing, externalizing, or total behavior problems by 1-year follow-up were compared. Parents and teachers completed two screening measures each: a measure of exposure to multiple family risk factors (the Family Risk Factor Checklist: FRFC) and a simple nomination question of child risk status, which involved informants answering one question concerning whether they believed a child was at higher risk than average of developing a mental health problem in the future. Both parents and teachers were more accurate at identifying children at risk of externalizing than internalizing disorders. For parents, the FRFC showed similar sensitivity to simple nomination as a screen for predicting later total behavior problems, lower sensitivity for internalizing problems, and there was a nonsignificant trend for the FRFC to have higher sensitivity for externalizing problems than simple nomination. The specificity of simple nomination was somewhat higher than that of the FRFC for all three mental health outcomes. For teachers, simple nomination showed higher sensitivity than the FRFC for predicting internalizing, externalizing, and total behavior problems, but the difference was nonsignificant for internalizing problems. The specificity for both teacher screens was similar across mental health outcomes. For a population base rate of externalizing problems of 13%, teacher nominations for children at risk of externalizing problems had a sensitivity of 69%, specificity of 78%, positive predictive value (PPV) of 32%, and negative predictive value (NPV) of 94%. Teacher nominations of at-risk children had greater predictive accuracy than parent nominations. The simple nomination method shows promise as a cost-efficient screen. When screening children who did not yet have behavioral symptoms, both parent and teacher measures resulted in substantial misclassification errors. However, the use of these screening measures at a population level could benefit large numbers of children who subsequently receive selective preventive interventions.