El análisis de curvas ROC en estudios epidemiológicos de psicopatología infantil: aplicación al cuestionario CBCL

Resum

ROC analysis was applied in order to study diagnostic accuracy of the Child Behavior Checklist (CBCL) and to obtain the optimal cut-off in a sample of 196 pediatric and psychiatric patients 6 to 17 years old. The group of origin, the diagnosis in the Diagnostic lnterview for Children and Adolescents-Revised (DICA-R) and the clinician's diagnosis were used as external validators. The results indicate that the discriminant power of the CBCL for the presence or absence of psychopathology depends on the external validator used. The best results were obtained when it was considered the group of origin and the DlCA-R diagnoses. As screening test, a cut-off between 50 and 54 gave the best sensitivity.