Dimensional measures of psychopathology. The probability of being classified with a psychiatric disorder using empirically derived symptom scales

Objectives: We sought to develop a ranking scheme that assigns a probability of having one of four psychiatric disorders to children based on their scores on a symptom scale. We then estimated the impact of each scale symptom on the prevalence of the disorder in the population. Method: Logistic regressions were specified for ADHD, ODD, depressive, and conduct disorders using all the individual symptoms in the pertinent scale as predictors. Individual fitted values from the regression function then served as a probability scale measure. We combined the prevalence and influence of each scale symptom to calculate its overall impact on the prevalence of the disorder. Results: Probability distributions had a wide range of values and discriminated between cases and non-cases. Those having a disorder were consistently associated with higher probabilities in the scale. The estimated probability corresponds to the empiric prevalence of the diagnosis in a group of persons sharing the same estimated probabilities. Symptoms varied on their impact on the prevalence of the disorder. Conclusions: We recommend the estimated probability of the disorder based on the empirically defined scales as dimensional measures that complement prevalence of the disorder. Different symptoms are identified as targets for screening when selection is based on their impact on the prevalence of the disorder than when selection is based on the strength of the association with the disorder. We recommend using a common nosology with different classification schemes; the categorical definition of the disorder, the probability of having the disorder, and the impact of each symptom in the prevalence. Different measures serve different purposes. Accepted: 5 April 2002 Correspondence to Maritza Rubio-Stipec