Predictive validity of categorically and dimensionally scored measures of disruptive childhood behaviors

Objective: To examine the predictive validity of categorical and dimensional methods of representing variation in disruptive childhood behaviors. Method: A birth cohort comprising 935 New Zealand children was assessed at age 15 years on measures based on DSM-III-R diagnostic criteria for oppositional defiant disorder, conduct disorder, and attention-deficit hyperactivity disorder. These symptom measures were scored in two ways: (1) as cases or noncases using DSM-III-R diagnostic criteria and (2) as dimensional variables in which the severity of disturbance ranged from none to severe. At age 16 years the cohort was reassessed on a series of measures including substance use behaviors, juvenile offending, and school dropout. Results: The analysis compared the efficacy of the categorically and dimensionally scored measures at age 15 years as predictors of outcomes observed at age 16 years. This comparison showed evidence of continuous and generally linear dose-response functions between symptom severity and outcome risks and that dimensionally scored variables were considerably better predictors of outcome than measures based on a diagnostic classification. Conclusions: These findings support the view that disruptive behavior problems have dimensional properties in which the severity of disturbance ranges from none to severe. While DSM-III-R diagnostic criteria may have considerable value and utility as a means of diagnosing in young people behavior disturbances that merit clinical attention, the routine use of these criteria as a means of describing behavioral variability may result in a system of measurement that produces variables that have less than optimal predictive validity.