The validity and reliability of 6 sets of clinical criteria to classify Alzheimer's disease and vascular dementia in cases confirmed post-mortem: added value of a decision tree approach.

Data from 204 participants from the Oxford Project to Investigate Memory and Ageing, who were diagnosed post-mortem using the histopathological criteria of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD), were used to assess the validity of the clinical criteria for Alzheimer's disease (AD) of the 'National Institute of Neurological and Communicative Disorders and Stroke/the Alzheimer's Disease and Related Disorders Association' (NINCDS/ADRDA). Cases who had been diagnosed as NINCDS/ADRDA 'probable AD' in life were usually confirmed at autopsy, but half of the NINCDS/ADRDA 'negative' cases were not (low specificity). It was hypothesized that the overall clinical impression may have taken precedence over the use of the actual criteria. We therefore investigated the validity and reliability of the clinical criteria using a computerized 'dementia diagnosis system' for each of 6 sets of criteria [Diagnostic and Statistical Manual of Mental Disorders (DSM-IV), NINCDS/ADRDA and three sets of criteria specifically for vascular dementia (VaD): NINCDS-AIREN, State of California Alzheimer's Disease Diagnostic and Treatment Centers (ADDTC), and Vascular Cognitive Impairment (VCI)] to classify a subset (n = 96) of the cases confirmed post-mortem. The use of the computerized system significantly (p = 0.01) increased the specificity (81%, similar to sensitivity) of the NINCDS/ADRDA diagnoses, which were shown to have 'moderate' inter-rater reliability. The DSM-IV criteria had good validity for AD when compared with post-mortem confirmation and showed 'substantial' inter-rater reliability. The ADDTC and VCI criteria for VaD had good specificity (88%) and sensitivity (75%), but only for one rater. The DSM-IV and NINCDS-AIREN criteria for VaD showed poor validity and inter-rater reliability. We conclude that the forced use of decision trees through a computerized system enhances the accuracy of the clinical diagnoses of dementia.