Objective: to establish whether changes in a spoken verbal task performance while walking compared with being at rest could predict falls among older adults. Design: prospective cohort study of 12 months' duration. Setting: twenty-seven senior housing facilities. Participants: sample of 187 subjects aged 75-100 (mean age 84.8 ± 5.2). During enrollment, participants were asked to count aloud backward from 50, both at rest and while walking and were divided into two groups according to their

more »

... ing performance. Information on incident falls during the follow-up year was monthly collected. Measurements: the number of enumerated figures while sitting on a chair and while walking, and the first fall that occurred during the follow up year. Results: the number of enumerated figures under dual-task as compared to single task increased among 31.5% of the tested subjects (n = 59) and was associated with lower scores in MMSE (P = 0.034), and higher scores in Geriatric Depression Scale (P = 0.007) and Timed Up & Go (P = 0.005). During the 12 months follow-up, 54 subjects (28.9%) fell. After adjusting for these variables, the increase in counting performance was significantly associated with falls (adjusted OR = 53.3, P<0.0001). Kaplan-Meier distributions of falls differed significantly between subjects who either increased or decreased their counting performance (P<0.0001). Conclusions: faster counting while walking was strongly associated with falls, suggesting that better performance in an additional verbal counting task while walking might represent a new way to predict falls among older adults.