The use of questionnaires in child psychiatry research: measuring their performance and choosing an optimal cut-off

Abstract

This paper introduces the basic principles of Receiver Operating Characteristics (ROC) analysis, a technique to evaluate and compare the discriminant power of questionnaires employed in psychiatric research. ROC curves may assist in the selection of an optimal cut-off point, which has been shown to depend upon the prevalence rate, the consequences of correct and incorrect classifications, and the distributions of scale scores among normal individuals and cases. Then, ROC analysis is applied to the Child Behavior Checklist on a sample of 564 6-11-year old French children. The performances of the social competence scales of this instrument are poorer than those of the behavior problem scales. The findings suggest also that the cut-offs should be varied within this same age group and be higher for screening in community surveys. Further applications of ROC analysis are discussed.