Assessment of reliability in the clinical evaluation of depressive symptoms among multiple investigators in a multicenter clinical trial

The objective of this work was to determine the severity of depressive symp toms when multiple clinical examiners evaluate a single subject, as prepara tory to their participation as evaluators in a clinical trial. Using the 17 -item Hamilton Depression Rating Scale (HDRS), 37 psychiatrists independent ly assessed the videotape of a patient with symptoms of depression. A new m easure for the detection of multiple examiners not in consensus (DOMENIC) w as used to identify scale items with low reliability and raters with low in ter-rater reliability, from among the remaining raters. Overall inter-rater agreement on the full HDRS was 'excellent' (97%). All raters but one showe d adequate agreement both on individual items and on total scores. Two of t he 17 HDRS symptomatology items had unacceptable levels of inter-rater scor ing variability (<70% agreement). The use of DOMENIC allows for the detecti on of items of low inter-rater reliability and identification of raters tha t deviate from the group's ratings prior to the beginning of a clinical tri al. (C) 2001 Elsevier Science Ireland Ltd. All rights reserved.