Rannsókn á áreiðanleika DIS greiningarviðtalsins

Abstract

The basic design of this study was a test-retest administration of the NIMH Diagnostic Interview Schedule (Version Ill-a), where two highly experienced lay interviewers a) each administered respectively 33 and 32 interviews and, b) six weeks later readrninistered the DIS to half of their own and half of each other's patients. The study resulted in a total of 104 interviews that were administered twice to 52 subjects. The subjects were 20 males and 32 females, all born in 1931. Three levels of analysis are presented. Reliability was calculated for diagnoses, symptoms and for individual items and chapters of the interview. An Item Consistency Index was used for calculating item consistency across the two administrations. The results suggest, that generally, compared to other similar studies, the interview has adequate reliability, but also show that this reliability drops when the interview is readrninistered by a different interviewer. The analysis of the data also shows that some chapters of the interview are less reliable than others, thus lowering the total reliability of the instrument and thereby diminishing the reliability and validity of certain diagnoses. The results are finally discussed in the light of that the study was conducted in a community sample, with low base-rates, which raises some important questions concerning the use of clinically validated diagnostic interviews in the normal population.