Labor has developed a 21-item Fall Risk Index (FRI) 2 for early detection of the risk of falls in elderly people. It has been found that fallers are older, more disabled, and more depressed and had higher scores on the FRI than nonfallers 3 and that FRI score was closely associated with aging and sex differences in community-dwelling elderly persons in Japan. 4 Although the significance of the FRI has been demonstrated in cross-sectional settings, the cutoff score for predicting falls in elderly people has not been established in longitudinal studies.A cohort study was conducted to evaluate the predictive value of the FRI in community settings. The study population consisted of 1,053 community-dwelling elderly subjects aged 65 and older (443 male, 610 female, mean age 74.6 AE 6.7) living in T town, Kochi, Japan. Participants were assessed using the FRI in June 2004 during their annual geriatric functional assessment, and the question ''Do you often fall?'' was used to detect fallers. Subjects who answered yes to the question were considered to be fallers, and those who answered no, not at all, were considered to be nonfallers. The FRI consists of 21 questionnaire-based items identified through a search of international and Japanese research articles that surveyed the history of falls within a year on a yes/no basis. 2,4,5 The 21 items included in the FRI encompassed physical, cognitive, emotional, and social aspects of functioning and environmental factors. It included history of stumbling, ability to climb stairs, slow walking speed, ability to cross the road within the green signal time interval, ability to walk 1 km without rest, ability to stand on one leg for 5 seconds, use of a cane, ability to squeeze a towel, dizziness, hunched back, knee pain, visual disturbance, hearing disturbance, forgetfulness, fear of falling, use of five or more prescription medications, darkness in the home, presence of obstacles inside the house, presence of barriers on the floor, daily stair use at home, and steep slopes around the house (Appendix A).Response rate of the eligible population was 94.7% at baseline and 74.4% at 1-year follow-up. Statistical analysis was performed using SPSS (SPSS, Inc., Chicago, IL) and a receiver operating characteristic (ROC) curve was used to determine the cutoff point on the FRI for predicting falls.Of 1,053 subjects, 186 (17.7%) were fallers at baseline and 219 (20.8%) at 1-year follow-up. Figure 1 shows the ROC curve of FRI to detect fallers at 1-year follow-up. Area under the ROC (AUC) of FRI was 79.4% (95% confidence interval (CI) 5 76.1-82.6%). Using a cutoff point of 9 out of 10 on the FRI, sensitivity was 67.7%, specificity was 76.4%, positive predictive value was 42.9%, and negative predictive value was 89.8% for fallers at 1-year follow-up.A past history of falls was found to be a strong predictor of falls in community-dwelling elderly people, 5,6 and subjects who reply yes to the question ''Do you often fall?'' LETTERS TO THE EDITOR 2369 JAGS DECEMBER 2009 should be immediately adv...