Consensus criteria for the diagnosis of vascular dementia (VaD) are gradually being replaced with data-based criteria. We report the inter-rater reliability of a new set of empirically-derived criteria for vascular cognitive impairment (VCI). Stratified sampling, with optimal allocation, was employed to randomly select 36 patients from the Queen Elizabeth II Health Science Centre's Memory Disability Clinic. Chart reviews were conducted independently by 4 physicians. Each physician classified the patients as having either: no cognitive impairment, VCI or Alzheimer's disease (AD). VCI was further classified both clinically (VCI without dementia, VaD or AD with a vascular component) and radiographically (infarcts, white matter changes, single strategic stroke). The intraclass correlation coefficient (ICC) for the diagnosis by physicians of VCI or otherwise was based on a repeated-measures analysis of variance with raters as the independent variable. A significant coefficient of reliability (average ICC = 0.88, 95% CI = 0.80-0.93) was obtained (H(o): rho </= 0.80, p = 0.03). Where differences in diagnosis occurred, the discrepancies most commonly resulted within the subtypes of VCI (9 cases) or between the diagnoses of AD and VCI (9 cases). Instances of diagnostic incongruity were typically due to the disagreement of a single rater (10 cases). This study demonstrates a high degree of reliability of criteria for VCI by physicians in a memory clinic, and can also be understood as an aspect of construct validation of those criteria. In the absence of a readily available biological marker for VCI, clinical criteria are necessary and can be reliably employed.
Copyright 2000 S. Karger AG, Basel