We characterized dynamic balance measured by the distance walked on 4-, 8-, and 12-cm-wide beams under single and dual task conditions in individuals with and without a neurological condition. We also determined if dynamic balance would predict prospective falls over 12 months. We found that age, disease, and beam width affect dynamic balance as measured by distance walked on narrow, low-lying beams and that beam walking distance predicts future falls.
Characterization of dynamic balance as measured by beam walking distance — In agreement with our study hypothesis, age, disease, and beam width all affected dynamic balance as measured by the distance walked on low-lying beams (Table 2, Fig. 1). Beam walking distance was unaffected on the 12-cm-wide beam in the five age groups of healthy individuals (3.76/4.00m). However, the distance walked on the 8-cm-wide beam decreased by 0.34m already in the 20-year-old group. This reduction was ~ 3x greater, 1.1m, in the 60-year-old group. In contrast to these large reductions, the distance walked during tandem gait over a 4-cm-wide tape on the floor was unaffected. Thus, beam walking vs. tandem-walking on the floor represents a different and challenging balance task. Reductions in distance walked on the beams suggest the presence of sub-clinical impairments in the abilities that control walking balance. These impairments seem to remain undetected by standard balance tests such as tandem-walking on a tape glued to the floor (4.0/4.0m) or by the frequently used SPPB (11.6/12.0, Tables 1, 2). The additional and large reductions in beam walking distance on the narrowest, 4-cm-wide beam point to a floor effect: this condition is too difficult for even healthy individuals age 20–60. The potentially greater sensitivity of beam walking vs. tandem-walking to detect subtle impairments in walking balance could be related to a reduction not only in the base of support (i.e., distance between the two feet) but to the reduction also in the contact area at the interface between the feet and the board. Such a mechanical constraint can strongly but transiently augment instability as the center of mass pivots over the stance leg. Instability increases during beam-walking but less so during tape-walking because even during normal gait, the path of the center of mass travels outside the medial border of the supporting foot [27] especially in old adults [28]. Thus, the path of the center of mass passes close to the beam edge, making older individuals ‘feel’ that they could lose their balance and they step off the beam (‘fall’). This sense of imminent balance loss reduces beam-walking distance. Beam walking thus increases the difficulty of postural control and its sensitivity to sub-clinical motor impairments. It ensures a fall-specific, sharp end-point in the form of an actual loss of balance even in older individuals who self-report to be ‘healthy’ [2, 10]. In total, these data seem to suggest that an optimal beam width probably lies around 6-8cm that could avoid floor and ceiling effects for a new balance test to be established. Our data imply that beam walking could complement or even replace certain ‘functional tests’ currently in use to measure walking balance based on walking speed without an actual loss of balance [23, 24].
Neurological diseases strongly affected beam walking distance and much more so than it affected SPPB (9.1/12 points) (Table 2, Fig. 1). Some studies actually suggest that a score of 9 on the SPPB is a ‘high performance’. Contrasting with these SPPB scores, beam walking distances decreased sharply by 0.8m on the 8 vs. 12cm beam and by additional 1.6m on the 4 vs. 8cm beam. The 0.53m distance walked by patients on the 4-cm-wide beam suggests a floor effect: the task was extremely difficult. An interesting observation was that patients walked numerically identical distances, i.e., 2.9m, on the 12-cm-wide beam and the 4-cm-wide tape glued to the floor. These data suggest that walking on a wide beam may not provide additional benefits over tandem walking but could provide additional insights into walking balance over ‘functional tests’ (SPPB, walking speed) in patients we examined in the present study. However, additional studies are needed because a recent study reported that tandem-walking failed to identify ~ 25% of patients with vestibular disorders [29]. Because dysfunctional walking balance is a precursor to falls in neurological patients, an accurate identification of fall-risk factors remains a priority in this population and beam walking might be an effective adjuvant to ‘functional tests’ currently in use in such patients [25].
Beam walking with a cognitive dual task did not significantly reduce the distance walked on the three beams (Table 2). These data are unexpected and contrast with a previous study that reported strong effects of cognitive dual-tasking on beam-walking distance [11]. One would expect that when the motor task is difficult and demands attention, adding a secondary cognitive task would strongly reduce motor performance. In that study older individuals were ~ 6 years older than our participants in the 6th decade and they walked significantly slower when dual-tasking on the beams. Because in the present study the number of errors (1.0-1.5) while dual-tasking did not differ between age groups and beam widths, individuals perhaps prioritized the motor element of motor-cognitive dual-tasking. Our patients walked on the tape ~ 1m shorter distance (2.9 m) than age-similar healthy individuals (3.9 m). The tape-walking performance was already so low that dual-tasking had little potential to reduce it further (p > 0.05; reduction of 0.4 m, Table 2). Additional data are needed to confirm the effects of cognitive dual-tasking on walking balance. This is because adding a secondary cognitive task to the Timed-Up-and-Go test did not increase the accuracy of fall prediction [30]. Therefore, the role of cognitive dual-tasking in walking balance remains unclear.
Incidence and circumstances of falls — Table 3 shows that there were 122 individuals with 423 falls. Over 80% of these falls occurred in patients during the 12-month-long follow-up period. Some previous studies reported falls in healthy young individuals age 20 even after excluding sports-related falls [31]. Admittedly, our study has low sample sizes in the age-decades (Table 1), but we did not observe a single fall in the 20-yearl-old participants (n = 19). However, 15%, 35%, 35%, and 50% of individuals reported falling in the 30-, 40-, 50-, and 60-year age-decade, respectively. These data agree with the 30–40% rates reported previously for the corresponding age-brackets in ~ 25,000 community dwelling US adults [32].
There were only 2 individuals with a neurological condition who did not report falling. Our 98% fall rate is twice as high as the 47% rate for those who reported falling 1-2x and the 60% proportion of patients reporting 3-6x recurring falls is also ~ 2-fold greater than the 32% rate reported previously also in individuals with PD, MS, and stroke diagnoses [25]. We had one patient who reported falling six times over 12 months. The age and sex distribution of patients were similar to those in Italy [25]. The smaller sample size, perhaps the higher level of impairment and the lower quality of outpatient care contributed to the high rate of (recurrent) falls in our study.
In age-decades 40–60, most falls occurred outdoors in the morning, which is probably related to why most falls occurred while wearing firm shoes. Most frequently falls occurred due to the knees buckling or weakness. Our older adults and patients had very low grip strength and low strength and muscle mass are related to falls [33] but this association is not always present and requires further confirmation [34, 35]. Our data contrast with reports suggesting that ~ 40–50% of falls occur while walking [13], as we observed that ~ 50% of falls occurred while rising to a higher position from a lower position, stepping up or down, or while turning in standing in individuals with and without a neurological condition (Table 3). Because the amount of physical activity based on the International Physical Activity Questionnaire or IPAQ scores was ~ 1,000 units higher in our older individuals than in some other studies [36], the reason for the low fall incidence during walking, even in our patient group, remains unclear. Indeed, the association between physical activity, sedentary behavior, and falls is complex. Being up on one’s feet and being physically active naturally increases the potential for a fall to occur. However, high levels of chronic physical activity can at the same time improve fitness, which is known to reduce falls risks [37, 38]. Improving some of these environmental risk factors for falls can also reduce risks for and incidence of falls [39].
Correlates of beam walking performance and prediction of future falls — There is a strong effort underway to identify tests that are associated with falls risks and incidence of falls through the age and disease spectrum [39–46]. Of the commonly examined variables such as age, sex, fall history, body fat, education, marital status, or retirement status, beam walking distance was only associated with age which in turn predicted falls over 12 months in healthy adults aged 20–60 years. These associations and predictions were independent of performing beam walking under single or dual task condition. Our findings complement prior data suggesting inconsistent, weak, or even no associations between risk factors for falls and incidence of falls. While many studies suggest that lower extremity muscle strength and power are associated with balance, fall risks, and future falls, there is also evidence suggesting to the contrary with no such associations [42]. A systematic review found that none of the biomechanical markers of challenging walking tasks correlated with fall risk variables and fall prediction was inaccurate without including fall history [40]. Indeed, our data suggest that performance in a difficult walking balance task is associated with age and predicts future falls without fall history. The logistic regression coefficient is reasonably large (coefficient: = -0.38 ± 0.08, p = 0.001) but dual-task condition did not improve prediction accuracy. The coefficient of -0.38 means that odds of fall (i.e., ratio fall / no fall) change 0.68 time (e^-0.38 = 0.68): with each increase in distance walked by 1 meter, the odds decrease by 0.32 (1-0.68). With additional meters walked, the effect is multiplicated: for a difference of, e.g., 4 out of 12m, the odds change 0.68^4 = 0.21 time. That is, the odds decrease by nearly 80%: the odds of fall/no fall would move from 6/6 to 2/10. We interpret these data as clinically meaningful.
Figure 2 shows that dual-task beam walking distance of ~ 8 of 12m maximum (AUC 0.76) was coupled with specificity (0.66) and sensitivity (0.75). These data imply that based on beam walking distance we would miss to identify many of those who would eventually fall and would erroneously identify many individuals as fallers even though they would actually not experience a fall. In patients, sex and education emerged as correlates of beam walking distance, agreeing with a previous report (43). In contrast to this report’s finding, we found no evidence that the fall risk varies among different disease types. We did not examine or find no associative or predictive role in falls several health conditions (vision, depression, arthritis, alcohol) or functional limitations (ability to climb stairs or perform daily functions) [41, 46].
Limitations – The current study did not compare how accurately conventional ‘functional tests’ vs. dynamic balance measured by beam walking distance predicted future falls. This will be reported in a future study. Our data are limited by the homogeneity of fall incidence in patients, i.e., virtually all patients reported falling. This preliminary and exploratory study cannot provide definitive ‘normative data’ for beam walking distances and clear cutoffs of ‘low’, ‘medium’, and ‘high’ levels of dynamic balance by age and sex due to low samples sizes. While gait analysis with wearables can identify age groups and retrospective falls highly accurately by analyzing dynamical systems outcomes with machine learning, such approaches require large sample sizes and sophisticated algorithms and still miss individual cases, leaving room for ‘analog’ solutions such as beam walking [47–49].