Background: This article presents a computerized method to help predict individuals at risk for developing Alzheimer's disease (AD). This would be a valuable tool for clinicians in developing treatment plans for potential AD patients. Using the initial level and rates of change in visual memory performance, such a method could predict potential AD patients in a fast and inexpensive manner. A longitudinal case-control study of 52 female and 145 male participants was performed in a gerontology research center using premorbid tests of visual memory and neurologic examinations to identify individuals with and without dementia and AD.
Methods: The classification method for each individual starts on the second examination and proceeds to compute that person's risk of AD one examination at a time based on all the follow-up information of the remaining individuals.
Results: By performing a crossvalidation study, the optimal combination of sensitivity and specificity derived from a receiver operating characteristic (ROC) curve showed 65% of the Alzheimer cases and 75% of the noncases were correctly classified for females, while 65 and 60% of cases and noncases, respectively, were correctly classified for males.
Conclusion: Longitudinal measurements of cognition can be useful in detecting the presence of AD.