Objective: To develop and provide initial validation of a brief (18 questions, YES/NO) questionnaire completed by a family member, or companion that can be used to distinguish between patients with probable Alzheimer’s disease and healthy controls. The ADCQ takes between 5 to 10 minutes to complete, requires no participation from physicians or other healthcare professionals, and requires no cooperation or participation from the patient.
Design: Internal consistency reliability and test-retest reliability, sensitivity and specificity.
Materials and Methods: After reviewing existing symptom checklists and consulting with AD experts and caregivers, a list of 56 items that could be answered "Yes" or "No" was compiled. The 56-item questionnaire, divided into 6 general categories (memory, confusion and disorientation, geographic disorientation, reasoning/judgment, language), was then administered to the companions of 93 community dwelling healthy and cognitively intact control subjects and the caregivers of 141 AD patients. The AD patients were consecutive referrals to The Memory Clinic at Southwestern Vermont Medical Center who were subsequently diagnosed with probable AD using NINCDS/ADRDA criteria.
Statistical Methods: Demographic variables (age, education, gender) were analyzed using t tests. Interitem reliability was examined using coefficient alpha, and interitem correlation. Test-retest reliability was evaluated using Pearson product moment correlations. Validity was evaluated by first selecting the three best discriminators (symptom items) in each of six categories. This 18-item pool was then analyzed using a number of different statistical techniques in order to identify the algorithms that had optimal validity for classifying individuals as either "AD" or "normal." Analysis included: correlation and regression tree analysis (CART), logistic regression, and stepwise regression. A Receiver Operator Characteristic (ROC) analysis was used in order to identify a cutoff score for the ADCQ.
Results: There were no differences in demographic characteristics between groups (ps> .05). The results indicated that the instrument was reliable (coefficient alpha = .87) and the mean interitem correlation was .54. Test-retest reliability was also good (r = .71). CART techniques had optimal classification rates (mean sensitivity = .88; mean specificity = .93), whereas the logistic regression (mean sensitivity = .85; mean specificity = .87) and stepwise logistic regression (mean sensitivity = .88; mean specificity = .88) had slightly lower classification rates. ROC rates were determined for the total score (0-18 items endorsed). Using a cutoff score of 5/6 produced a sensitivity of .92 and specificity of .86.
Conclusion: The ADCQ appears sensitive to AD and may be useful in helping to make initial distinctions (i.e., screening) between patients experiencing cognitive changes related to the normal aging process and those experiencing cognitive deficits related to dementing disorders such as AD. It has reasonable interitem and test-retest reliability, does not need to be administered by a healthcare professional, can be completed in a brief period of time, and does not require the cooperation of the patient.
Back to S039 Neuropsychology and Aging: Dementia and Beyond
Back to The Eleventh International Congress