Comparison of Discriminant Validity Indices of Parent, Teacher, and Multi-Informant Reports of Behavioral Problems in Elementary Schoolers

The clinical relevance of using a multi-informant approach to assess behavioral problems in youths is well established. However, empirical studies have consistently shown weak agreement across informants (parents, teachers, youths). Moreover, there is no consensus regarding the most valid strategy for combining informant reports to assess these problems. The aim of this study was to compare the discriminant validity of different strategies for combining parent and teacher reports (mean score, highest score, independent assessments) by testing their sensitivity and specificity in assessing different constructs of behavioral problems (conduct, oppositional and attention/hyperactivity) in youths referred and non-referred to school-based services for these problems. The strategies were tested in terms of their capacity to discriminate between a sample of children (n = 370, 6–9 years old, 40% girls) with emotional or behavioral problems referred for special school services and another sample of children (n = 183, 6–9 years old, 52% girls) never referred for such services. All the children were recruited in the same school boards across the province of Quebec, Canada. High specificity rates were obtained across informants and combination strategies. Low sensitivity rates were obtained for parents, teachers and mean scores, but the use of the highest score yielded higher sensitivity without lowering specificity rates and optimized classification rates across constructs and cutoffs (98th, 93rd percentiles) among both boys and girls. This combination strategy therefore appeared to optimize balance between specificity and sensitivity. These results suggest that, when available, use of parent/teacher highest score should be preferred over use of a single informant score or of a parent-teacher mean score in assessing behavioral problems in youths.