Reliability and clinical concepts underlying global judgments in dementia: implications for clinical research.

The Clinical Global Impression Scale (CGI) is a recognized rating scale for global clinical judgments comprising scores for disease severity, change of disease conditions, and a so-called "efficacy index." In this report, the authors subject the CGI to a methodological analysis. Thirty-seven physicians working in psychogeriatric wards were interviewed on 12 patients each with a DSM-III diagnosis of a dementia syndrome. After the physicians made global judgments on the patients with the CGI, "personal" assessment criteria were elicited. The CGI data were correlated in the statistical analysis with the physicians' assessments of the patients on their personal criteria and on "recognized" assessment criteria obtained from DSM-III-R diagnostic criteria for dementia syndrome. Interrater reliabilities between physicians and nursing staff as well as retest reliabilities for the CGI criteria were also measured. While the CGI-severity reflects primarily the cognitive aspects of dementia, the CGI global assessment of change of the disease condition was poorly correlated with the assessments based on "personal" or the recognized DSM-III-R criteria. This was also indicated by the result that the reliability scores for CGI-severity were high and did not vary greatly, whereas the reliability scores for CGI change showed wide confidence intervals.