Background: Clinical investigators from Seattle, Honolulu, Tokyo, and Hiroshima participated in two standardization exercises in which data were collected on independent assessments. Exercises were conducted to evaluate the interobserver agreement on clinical diagnoses of dementia and dementia subtypes in a cross-national study of dementia prevalence and incidence rates in the United States and Japan.
Method: Fifteen clinicians from four participating sites assessed the diagnosis of 85 patients based on standardized summaries of clinical and diagnostic test data on each patient. Diagnostic guidelines and conventions were adopted on the basis of group consensus during standardization exercises.
Results: Using DSM-III-R criteria, generally good levels of agreement for all dementia diagnostic categories occurred in both years. For most measures of diagnostic agreement, improvements were observed between the 1995 and 1996 standardization sessions. Interrater agreement was highest for discrimination between dementia and nondementia (1996 overall kappa, K = .90). The kappa values for dementia subtypes in 1996 ranged from .5 to .85, and for all sites combined the value was .67. For dementia subtypes, percent agreement was highest for vascular dementia and Alzheimer's disease, but was less reliable for other types of dementia.
Conclusions: Clinicians from different cultures and medical traditions can reliably use the DSM-III-R criteria to classify dementia cases in cross-national research. The interrater agreement on dementia and its subtypes improved after clear-cut guidelines for interpretation of diagnostic criteria were developed and followed.