Validity, reliability, and psychometric properties of a computerized, cognitive assessment test (Cognivue^®)

Background: Cognitive issues such as Alzheimer's disease and other dementias confer a substantial negative impact. Problems relating to sensitivity, subjectivity, and inherent bias can limit the usefulness of many traditional methods of assessing cognitive impairment.

Aim: To determine cut-off scores for classification of cognitive impairment, and assess Cognivue^® safety and efficacy in a large validation study.

Methods: Adults (age 55-95 years) at risk for age-related cognitive decline or dementia were invited via posters and email to participate in two cohort studies conducted at various outpatient clinics and assisted- and independent-living facilities. In the cut-off score determination study (n = 92), optimization analyses by positive percent agreement (PPA) and negative percent agreement (NPA), and by accuracy and error bias were conducted. In the clinical validation study (n = 401), regression, rank linear regression, and factor analyses were conducted. Participants in the clinical validation study also completed other neuropsychological tests.

Results: For the cut-off score determination study, 92 participants completed St. Louis University Mental Status (SLUMS, reference standard) and Cognivue^® tests. Analyses showed that SLUMS cut-off scores of < 21 (impairment) and > 26 (no impairment) corresponded to Cognivue^® scores of 54.5 (NPA = 0.92; PPA = 0.64) and 78.5 (NPA = 0.5; PPA = 0.79), respectively. Therefore, conservatively, Cognivue^® scores of 55-64 corresponded to impairment, and 74-79 to no impairment. For the clinical validation study, 401 participants completed ≥ 1 testing session, and 358 completed 2 sessions 1-2 wk apart. Cognivue^® classification scores were validated, demonstrating good agreement with SLUMS scores (weighted κ 0.57; 95%CI: 0.50-0.63). Reliability analyses showed similar scores across repeated testing for Cognivue^® (R ² = 0.81; r = 0.90) and SLUMS (R ² = 0.67; r = 0.82). Psychometric validity of Cognivue^® was demonstrated vs. traditional neuropsychological tests. Scores were most closely correlated with measures of verbal processing, manual dexterity/speed, visual contrast sensitivity, visuospatial/executive function, and speed/sequencing.

Conclusion: Cognivue^® scores ≤ 50 avoid misclassification of impairment, and scores ≥ 75 avoid misclassification of unimpairment. The validation study demonstrates good agreement between Cognivue^® and SLUMS; superior reliability; and good psychometric validity.