Improving Diagnostic Accuracy: Comparison of Nomograms and Classification Tree Analyses for Predicting the Diagnosis of Oppositional Defiant Disorder

Unstructured clinical interviews are inaccurate tools for diagnostic decision-making. While structured diagnostic evaluations improve reliability, they are infrequently used in clinical practice. Empirical approaches are a hallmark of evidenced-based assessment and may reduce burdens of structured interviews. We explore two approaches to empirical prediction of diagnosis, the naive nomogram, and classification tree analysis (CTA). To illustrate the clinical utility of each approach, we compared their use in a sample of 6-year-olds (N = 619) to predict structured-interview diagnoses of oppositional defiant disorder (ODD). Findings indicate the accuracy of both approaches in predicting the absence of a disorder and improved detection of ODD using CTA for subgroups of children. Both empirical prediction techniques have applicability to diagnostic decision-making in psychiatry and pediatrics. Impact Statement Application of statistical prediction models to mental health diagnosis in the traditional clinical setting may ultimately reduce time burden on busy clinicians and improve the accuracy of ascribed diagnosis.