Identification of risk for severe psychiatric comorbidity in pediatric epilepsy

Abstract

Objective: This study identified items on the Child Behavior Checklist (CBCL) that predict those children and adolescents with epilepsy at highest risk for multiple psychiatric diagnoses. Methods: Three hundred twenty-eight children, ages 5–18 years, and their parents participated in separate structured psychiatric interviews about the children, which yielded Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition Text Revision (DSM-IV-TR) diagnoses. Parents completed the CBCL. The sample was divided into a younger (≤12 years, n = 214) group and an older (>12–18 years, n = 114) group. This study identified a reduced set of parent-reported CBCL items associated with Multiple Diagnoses versus Single Diagnosis versus No Diagnosis using chi-square tests and stepwise logistic regression. We then performed a generalized logistic regression with Multiple Diagnoses versus Single Diagnosis versus No Diagnosis as the dependent variable and the reduced CBCL set of items as predictors. We calculated the area under the ROC (receiver operating characteristic) curve (AUC) as a measure of diagnostic accuracy for pairwise comparisons. Results: For the younger group, seven items (clingy, cruelty/bullying, perfectionist, nervous, poor school work, inattentive, and sulks) had high diagnostic accuracy (AUC = 0.88), and for the older group, three items (disobedient at school, loner, and lies/cheats) had high accuracy (AUC = 0.91) when comparing children with multiple psychiatric diagnoses to children with no diagnosis. For both age groups, there was less diagnostic accuracy in identifying children with a single versus no diagnosis (AUC = 0.75 [young]; 0.70 [older]). Significance: These findings suggest that responses to these two subsets of parent-reported CBCL items should alert clinicians to children and adolescents with epilepsy at risk for multiple psychiatric diagnoses and in need of a psychiatric referral.