Differential Diagnosis of Attention Deficit and Conduct Disorders Using Conditional Probabilities

Attention deficit and conduct disorders require an important yet often difficult differential diagnosis. Prior efforts to determine which symptoms are optimal for making this differential diagnosis have been limited by a reliance on statistics that do not supply the probability of the disorders given a symptom's presence (positive predictive power) or the probability that the disorder is not present given the absence of the symptom (negative predictive power). This investigation examined the utility of these latter statistics in the differential diagnosis of childhood attention deficit and conduct disorders. The data consisted of symptoms from a standardized maternal psychiatric interview collected for a sample of 76 clinic-referred boys. Results indicated that some symptoms are optimal as inclusion criteria, some as exclusion criteria, some as neither, and some as both. Furthermore, some symptoms that have been traditionally associated with the diagnosis of one disorder were actually found to be more useful in the diagnosis of the other disorder.