Objective: To assess interrater reliability and validity of NINCDS-ADRDA (National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer's Disease and Related Disorders Association) criteria for Alzheimer's disease (AD).
Design: A multisite reliability and validity study in which clinicians from each site diagnosed 60 case summaries yielding a preconsensus estimate of reliability and validity. A consensus conference was conducted for each disagreement, leading to a postconsensus estimate of validity. The criterion standard was a diagnosis of AD by autopsy.
Setting: Three academic medical centers.
Subjects: A convenience sample of 60 detailed case summaries, 40 with AD and 20 with other dementing disorders.
Main outcome measures: The kappa coefficient, sensitivity, and specificity.
Results: The kappa coefficient for preconsensus agreement on a diagnosis of probable or possible AD vs non-AD was 0.51; the sensitivity of a diagnosis of probable or possible AD for a pathological diagnosis of AD was 0.81, and the specificity was 0.73. The postconsensus sensitivity was 0.83, and the specificity was 0.84.
Conclusions: The results support the reliability and validity of NINCDS-ADRDA criteria and show that the consensus process may improve diagnostic accuracy. The cases are reviewed with a focus on the sources of diagnostic disagreements and errors and possible changes that might improve the accuracy of the criteria.