Background: The Clinical Global Impression scale (CGI) is frequently used in medical care and clinical research because of its face validity and practicability. This study proposes to improve the reliability of the Clinical Global Impression (CGI) scale in depressive disorders by the use of a semi-standardized interview, a new response format, and a Delphi procedure.
Methods: Thirty patients hospitalised for a major depressive episode were filmed at T1 (first week in hospital) and at T2 (2 weeks later) during a 5' specific interview. The Hamilton Depressive Rating Scale and the Symptom Check List were also rated. Eleven psychiatrists rated these videos using either the usual CGI response format or an improved response format, with or without a Delphi procedure.
Results: The new response format slightly improved (but not significantly) the interrater agreement, the Delphi procedure did not. The best results were obtained when ratings by 4 independent raters were averaged. In this situation, intraclass correlation coefficients were about 0.9.
Conclusion: The Clinical Global Impression is a useful approach in psychiatry since it apprehends patients in their entirety. This study shows that it is possible to quantify such impressions with a high level of interrater agreement.