Learning and generalization tasks predict short-term cognitive outcome in nondemented elderly.

TL;DR: 2 baseline variables, delayed paragraph recall and generalization performance, were predictive of follow-up outcome with sensitivity of 81% and specificity of 91%—better than the classification accuracy based on either of these measures alone.

Abstract: This study examines whether behavioral measures obtained in nondemented elderly can predict cognitive status at 2-year follow-up. Prior studies have established that delayed paragraph recall can help predict short-term risk for decline to mild cognitive impairment and Alzheimer disease. It was examined whether prediction accuracy can be improved by adding a discrimination-and-generalization task that has previously been shown to be disrupted in nondemented elderly with hippocampal atrophy, a risk factor for Alzheimer disease. Fifty nondemented, medically healthy elderly patients received baseline clinical diagnosis and cognitive testing; 2 years later, patients received a follow-up clinical diagnosis of normal, mild cognitive impairment, or probable Alzheimer disease. In all, 2 baseline variables, delayed paragraph recall and generalization performance, were predictive of follow-up outcome with sensitivity of 81% and specificity of 91%-better than the classification accuracy based on either of these measures alone. These preliminary results suggest that these behavioral tasks may be useful tools in predicting short-term cognitive outcome in nondemented elderly.