This paper reviews some of the characteristics of the informants as well as some of the attributes of the DICA-R interview that could influence the test–retest reliability in a sample of 109 psychiatric outpatients aged 7–17 years. Different regression models using reliability coefficients constructed from the kappa statistic were obtained. Of those characteristics evaluated in the children, a high level of psychological impairment proved to be significant when it came to predicting the lowest test–retest reliability of the answers; none of the subject-related characteristics were significant in the adolescent patient model. The attributes of the questions that proved to be significant when explaining the lower reliability obtained for the individual question in the children's model were the length of the questions (longest questions), the content (internalising), the presence of time concepts, comparison with the peer group, and the need to exercise judgement; in the adolescents' model, the significant attributes were found to be the internalising content, the presence of time concepts, evaluation concerning the impairment caused by the disorder, and the need to exercise judgement. In the group of children our results are in accordance with the original paper. Similar results were found with adolescents. These findings have implications for the development and revision of new interview schedules.