Abstract
AIM: Community settings often need simple screening, rather than detailed tests, to identify cognitive impairment. This study aimed to develop models to screen older adults with cognitive impairment. METHODS: This study used data from the Integrated Research Initiative for Living Well with Dementia Cohort Study and included 5830 older adults. Individuals were considered cognitively impaired if their Mini-Mental State Examination score was less than 24. Three screening models were developed: the simple model (age, sex, and education), the base model comprising 13 candidate variables available in the questionnaire, and the enhanced model, where grip strength and gait speed were added to the base model. We performed binary logistic regression analysis with stepwise backward elimination (P < 0.1 for retention in the model) to develop each model. Then, we calculated integer scores from coefficients to develop score-based models. The area under the receiver operating characteristic curve (AUC) was used to evaluate discrimination. RESULTS: Participants with cognitive impairment accounted for 4.0% (n = 233) of the total. The score-based simple model comprised three variables (AUC = 0.72, sensitivity: 72%, specificity: 61%). The score-based base model included nine variables (AUC = 0.76, sensitivity: 70%, specificity: 67%). The score-based enhanced model comprised eight variables, including grip strength and gait speed (AUC = 0.79, sensitivity: 73%, specificity: 70%). CONCLUSIONS: This study developed three screening models with acceptable discriminant validity for cognitive impairment. These models comprised simple questionnaire-based items and common physical performance measurements. These models could enable screening of older adults suspected of cognitive impairment without the need to conduct cognitive tests in community settings. Geriatr Gerontol Int 2022; 22: 292-297.