Conditional Probabilities of Child Interview Symptoms in the Diagnosis of Attention Deficit Disorder

Abstract Based upon the standardized interviews of 76 clinic‐referred boys, children's reports of behaviors suggestive of attention deficit disorder (ADD) were examined to determine which symptoms were most efficient predictors of an ADD diagnosis. Unlike previous research that has focused exclusively upon the use of sensitivity and specificity rates, this investigation also employed the conditional probability rates of positive and negative predictive power to establish the diagnostic utility of symptoms. As such, this study attempted to establish the diagnostic efficiency of some of the child‐reported symptoms and describes a new method to examine any criteria set that is to be used for classification purposes. Results indicated that the optimal child‐reported predictors of an ADD diagnosis were not those symptoms typically considered hallmark indicators of the disorder. In addition, no symptom was found to be an efficient exclusion criterion for diagnosis. These findings are discussed in terms of the role of the child as informant for the diagnosis of children's behavior disorders and the need for multisource assessment.