Our results indicate that gait variables derived from trunk accelerometry provide good discrimination of the risk of falling among older people. The HR values of the upper and lower trunk were lower in fallers than in non-fallers. Physical performance (i.e., the results of the FCS and TUG tests) was also inferior in fallers. The associations between the HR of the upper trunk and the risk of falling were independent, even in the regression model that included physical performance. By contrast the association between the HR of the lower trunk and the risk of falling was weakened in this model, although the HR of the lower trunk was lower in all directions compared with the corresponding HR of the upper trunk. The HR of the upper trunk showed directional characteristics in that the HR in the VT and AP direction, but not in the ML direction, were significantly lower in fallers, and only HR in the VT direction of the upper trunk was a significant predictor of falls. ROC curve analysis confirmed that the HR of the upper trunk showed high specificity for detecting the risk of future falls.
Measurements obtained from accelerometry showed good clinical usefulness and ability to detect the risk of falls among older people in several studies [21, 26]. As an outcome of gait analysis, the HR of trunk acceleration represents the smoothness of trunk movement during gait, and its association with falling was confirmed in a cross-sectional study . Our prospective study extends these earlier results and confirms the usefulness of measuring trunk acceleration to assess the risk of falling. Our results are consistent with earlier reports, as the fallers had lower HRs of the upper and lower trunk than non-fallers . Furthermore, the HR of the upper trunk were significantly associated with the incidence of falls, even after adjusting for confounding variables, and showed high discriminative values (AUC = 0.81). By comparison, the associations between the HR of the lower trunk and falls were weakened in the regression model. Trunk movement during walking plays important roles in filtering and attenuating oscillations to stabilise the head [14–16]. The amplitude of acceleration and signal regularity are filtered from the lower trunk to the upper trunk, and the postural system maintained within the trunk segment contributes to overall stability . In fact, the deterioration in the HR was greater in fallers than in non-fallers, suggesting that the trunk’s role as a filter may be impaired in fallers. This role of the trunk may lead to the differing associations of HRs between the upper and lower trunk with the risk of falls.
Similar associations between gait variables derived from trunk accelerometry, physical performance and risk of falls to those observed in our study were reported in another study . Bautmans et al. analysed lower trunk acceleration during walking and investigated its relationship with physical performance and the risk of falls using the autocorrelation procedure. They suggested that variability of lower trunk acceleration was correlated with physical performance and that the discrimination of the risk of falls, based on trunk variability, decreased after taking into account walking speed. Our results show similarities in that the variables derived from trunk accelerometry and the relationship between lower trunk variables and the risk of falls was attenuated in the regression model including confounding factors, such as walking speed, although the methods used to processed data and measure trunk acceleration differed between the two studies. Physical performance scores on the FCS and TUG tests were also associated with the incidence of falling in bivariate analysis, while regression analysis suggested that the HR may have an advantage over physical performance for assessing the risk of falling. These results suggest that gait analysis should be incorporated into assessments of the risk of falling and that enhanced trunk stability may lead to robust physical function and thus prevent falls.
The HR shows directional differences across studies investigating the HR of trunk acceleration during gait [8–10, 13, 14]. For example, Mentz et al. reported that participants at high risk of falling walked with decreasing HR in all directions of lower trunk acceleration, and in the VT and AP directions of head acceleration . Our results are consistent with these directional characteristics of the HR in relation to falls. Indeed, the HR of the lower trunk in all directions and the HR of the upper trunk in the VT and AP directions were lower in the fallers than in the non-fallers in our study. Our results also revealed that the HR of the upper trunk in the VT direction may represent the risk of falling. The physical characteristics of older people vary greatly, and ranged from very fit to frail in earlier studies, so differences in physical function among older people should be taken into account when determining the risk of falling and when implementing interventions aimed at reducing the incidence of falling . The participants in our study were not physically fit (the mean walking speed was 0.91 m/s), similar to the people at high risk of falling in the earlier studies that used accelerometry [13, 27]. Furthermore, the fallers in our study were generally older and walked more slowly (mean walking speed: 0.63) than the non-fallers. Considering the differences in physical performance between studies, large cohort studies of older people across the spectrum of physical fitness are needed to better understand the association between trunk acceleration and risk of falls.
Our study had several limitations that should be discussed. First, the number of participants was relatively small, as was the number of multiple fallers, which prevented us from examining the associations between multiple falls and gait variables. Multiple falls is associated with a high risk of injury and post-fall syndrome . Thus, further studies are needed to examine the association between multiple falls and trunk acceleration. Second, other factors may confound the risk of falling, particularly cognitive function and trunk alignment. These potential confounders may also interact with aging. Although the MMSE was used to assess general cognitive function, more detailed domains, focusing on executive function and/or memory, should also be assessed. These cognitive functions may affect the outcome of a fall. Fall is a multifactorial outcome and further studies focusing on several factors, including cognitive function and trunk alignment, are needed to investigate the associations between falling and gait variables derived from trunk acceleration. Finally, the incidence of falling in our subjects was relatively low compared with that in other studies [1, 2], while a recent systematic review estimated that the incidence of falls among older people ranged from 14.7% to 34% . These differences may be due to differences between races and/or physical function status of the participants. Clearly, a large cohort study is needed to determine the incidence of falls among older adults.