A Multistate Model of Cognitive Dynamics in Relation to Frailty in Older Adults

PURPOSE: To analyze cognitive changes in relation to frailty, using a multistate transition model. METHODS: In a prospective cohort study of older Canadians (n = 9266), cognitive states were defined as the errors in the Modified Mini-Mental State Examination score. Frailty states were defined using a Frailty Index based on 40 self-reported deficits. Five-year transition probabilities were modeled by the use of a truncated Poisson distribution with the Poisson mean dependent on the baseline cognitive state, frailty, and other covariates. RESULTS: In multivariable analyses, age, frailty status, and education level were independently associated with cognitive changes, whereas only age and frailty were related to the risk of death. Frail people less often showed cognitive improvement (13.4%; 95% confidence interval [CI], 11.8-15) or stabilization (8.9%; 95% CI, 7.5-10.3) compared with non-frail people, of whom 23.9% improved (95% CI, 22.8-25) and 17% (95% CI, 16.1-17.9) maintained their cognitive status. Frail people were more likely to die (47.4%; 95% CI, 44.8%-50%) versus 22.3% (95% CI, 20.1%-24.5%) of non-frail older adults. CONCLUSION: A multistate transition model can be used to estimate simultaneously the chances of cognitive improvement/stability, decline, and death, and to analyze how these outcomes depend on frailty and other covariates. Ann Epidemiol 2011;21:507-516. (C) 2011 Elsevier Inc. All rights reserved.