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

Abstract

The objective of this work was to determine the severity of depressive symptoms when multiple clinical examiners evaluate a single subject, as preparatory to their participation as evaluators in a clinical trial. Using the 17-item Hamilton Depression Rating Scale (HDRS), 37 psychiatrists independently assessed the videotape of a patient with symptoms of depression. A new measure for the detection of multiple examiners not in consensus (DOMENIC) was used to identify scale items with low reliability and raters with low inter-rater reliability, from among the remaining raters. Overall inter-rater agreement on the full HDRS was 'excellent' (97%). All raters but one showed adequate agreement both on individual items and on total scores. Two of the 17 HDRS symptomatology items had unacceptable levels of inter-rater scoring variability (<70% agreement). The use of DOMENIC allows for the detection of items of low inter-rater reliability and identification of raters that deviate from the group's ratings prior to the beginning of a clinical trial.