Novel Application Of Accelerometry Data To Enhance Detection Of Falls Risk In Older Adults

Abstract

PURPOSE: The association between physical activity (PA) and falls is complex and bi-directional, with more active persons often demonstrating greater falls risk. The role of PA in preventing or aggravating falls risk may be illuminated by novel, more sensitive measures of PA. We examined the longitudinal association between accelerometer derived PA metrics and falls in control group participants of STURDY (Study to Understand Fall Reduction and Vitamin D in You), a randomized trial of vitamin D supplementation to prevent falls in older adults. METHODS: PA was assessed at baseline, 12-, and 24-months using the Actigraph Link accelerometer, worn 24-hours per day for 7 days on the non-dominant wrist. Falls were reported using a monthly calendar. Minute level activity counts were examined to derive diurnal patterns of PA and number of active minutes/day in 319 participants (mean age 77 (SD=5.4) years, 43% female). Multiple logistic regression models adjusted for age, sex, and gait speed examined the: i) risk of falls by tertiles of daily PA and ii) risk of falls by demographic and functional characteristics, comparing measures of falls per unit time (year) and falls per unit of activity (active minute). RESULTS: In adjusted models, those in the high and low PA tertiles had 78% and 7% greater odds of falling over the next 12 months compared to those in the moderate PA tertile, respectively (p<.05). Moreover, peak daily PA declined 12% between baseline and 24-month follow-up among fallers vs. 8% among non-fallers (p<.05). Differences in daily PA between fallers and non-fallers tended to be greatest between 8am-12pm. Models comparing risk of falls over 24 months of follow-up demonstrated stronger trends in falls risk at higher ages (80 vs >80), among men, and for those with slower gait speeds (<0.8 m/s vs 0.8m/s) when analyzed per unit of activity (falls/active minute) vs. unit of time (falls/year). CONCLUSION: These results demonstrate a J-shaped association between PA and falls, with stronger differences in PA between fallers and non-fallers during the morning hours. Furthermore, the trend towards more robust results for falls/active minute vs. falls/year highlight the complex nature of the association between PA and falls and suggest that novel PA metrics may serve as more sensitive indicators for discerning falls risk.