The kinematics and strategies of recovery steps during lateral losses of balance in standing at different perturbation magnitudes in older adults with varying history of falls

An increased susceptibility to falling sideways is a major cause of hip fracture and functional disability [1‐3]. The impaired ability to recover from an unexpected lateral loss of balance is especially relevant to falls among older adults. When balance is lost unexpectedly to the right or left, arm lift [4] and trunk movements [5] are used to decelerate the center of mass (CoM) motion and preserve balance at minor perturbation magnitudes. A lateral recovery step is used at higher perturbation magnitudes, and the foot is placed more to the right or left to preserve balance during locomotion [6‐9] and during standing [10‐12] to adequately regulate the relationship between the body’s CoM and the base of support (BoS) provided by the feet to prevent a fall. The spatiotemporal characteristics of the recovery steps (e.g., direction, timing, and length) need to match the requirements for controlling balance. Many past studies of the ability to recover from an unexpected loss of balance have focused on postural strategies and kinematics in the anterior–posterior direction, and explored age-related differences. The lateral direction kinematics in older adults with a varying history of falls and the strategies they use to recover from balance loss at increasing perturbation magnitudes has been less investigated.

In the current study, we aimed to investigate the differences in balance recovery to a lateral balance loss between non-fallers (NFs), recurrent fallers (RFs), i.e., high-risk fallers, and older adults who had fallen only once (OFs), which could indicate either a slightly impaired balance control system or an isolated event that does not truly represent their balance ability. We compared the kinematic characteristics of the recovery stepping at the single step- and multiple step-thresholds levels. These step threshold levels were shown previously to be independent predictors of a future fall [13‐18]. In addition, we compared the recovery step kinematic and stepping strategies used by NFs, OFs, and RFs in single-step and multiple-step reactions that were observed at increasing magnitudes of perturbations. These investigations may help in understanding mechanisms of balance recovery and future fall risk.

We first hypothesized that compared with NFs, step threshold levels will be reduced in OFs, and even more so in RFs and that RFs will have a higher probability of taking steps to recover their balance (higher probability of stepping reactions). Additionally, we hypothesized that these step threshold levels will be negatively associated with fear of falls and positively associated lower extremity function. Our second hypothesis was that the kinematics of the first recovery step during single-step and multiple-step threshold trials will be impaired in RFs, especially step initiation duration, as well as their ability to control the CoM motion. Thus, the total time to fully recover balance during the single- and multiple-step threshold trials will be delayed compared with OFs and NFs. Third, we hypothesized that RFs will use a different first step recovery strategy, i.e., less crossover strategy compared with NFs and OFs, since crossover stepping requires a more complex swing trajectory to move one foot across the other foot while circumventing the stance leg, and this may result in a delay in step initiation and balance recovery in RFs. Fourth, we hypothesized that the kinematics in single-step and multiple-step reactions along all perturbation magnitudes will be impaired in RFs.

As Table 1 shows, NFs had a significantly lower FES score than OFs (19 vs. 21, p = 0.009), and significantly higher overall function, basic lower extremity function, and advanced lower extremity function compared with RFs (67 vs. 57, p = 0.005; 81 vs. 70, p = 0.004; and 60 vs. 50, p = 0.008, respectively).

A total of 1999 fix of support and change of support balance recovery trials were observed (1268 for NFs, 445 for OFs, and 286 for RFs). Eighty-three participants reached their own single-step threshold level, 65 reached their own multiple-step threshold level (i.e., 15 subjects who completed the protocol did not reach their multiple-step responses, one participant asked to stop the experiment before reaching this threshold, and with two participants, we encountered technical problems at the end of the study protocol; thus, no multiple-step threshold trials were observed). The group comparison of kinematics of multiple-step events were made for these 65 participants.

As hypothesized, the levels of single-step and multiple-step thresholds were significantly lower in OFs compared to NFs (Fig. 1). Lower single-step and multiple-step thresholds were found earlier in older adults compared with young’s [22] and in stroke patients compared with age matched healthy controls [28]. A recovery step at a lower perturbation magnitude in OFs may suggest a lower balance recovery function, i.e., less ability to control the CoM motion over the BoS, or may be due to a past single fall event which resulted in a greater fear of falling (FES-I) that does not actually represent true balance abilities. A lower FES-I between OFs and NFs, a similar self-reported lower-extremity function between OFs and NFs (Table 1), as well as similar step recovery kinematics and step-recovery strategies (Tables 2 and 3), support the interpretation that due to a past single fall event, OFs have a more “responsive” over-reactive balance control system than NFs, and this does not truly represent their balance control abilities. Psychological factors related to postural threat are modulating cortical activity associated with postural reactions to unpredictable perturbations [29], it was found earlier that postural threat alters spinal reflex excitability reflected by soleus H-reflex attenuation [30]. This finding suggests that when postural control is threatened in older adults that have higher fear of fall such as in OFs, there is a shift to a more cortical involvement and a more responsive reaction.

Surprisingly, and in contrast to our hypothesis, the RFs single-step and multiple-step threshold levels were similar to that of NFs (Fig. 1). The question is why these thresholds in RFs were remarkably similar. First, the RFs in our study showed a low probability of stepping, especially at low perturbation magnitudes, suggesting less willingness to use the change BoS strategies, i.e., rapid recovery stepping responses at low perturbations, i.e., 1–8 (Fig. 2). Recovery stepping responses require rapidly lifting of the leg off the ground and then swinging it to the target location. This is also dependent on neuromotor mechanisms related to the build-up of muscle force and power to move the leg. Since RFs also reported lower low-extremity function (Table 1), it seems that they tried to avoid the use of a recovery step as much as they could. But when a recovery step was unavoidable at higher perturbation magnitudes, RFs in our study exhibited a delayed recovery step initiation duration in their single-step threshold trial compared with NFs (of about 100 ms), i.e., the lowest perturbation magnitude to evoke a recovery stepping response (Table 2A). This is further supported by the linear mixed-effects models of the single-step trials of all perturbation magnitudes that revealed that the step initiation duration of RFs was delayed compared with NFs and OFs (Table 3A). This, accompanied by larger recovery step length and a slower recovery step duration in RFs, compares with NFs and OFs (Tables 2A and 3A). In regard to the step recovery strategies, our findings show that the RFs used a more significant LLSS strategy compared with OFs and NFs (Fig. 3). It should be noted that in the LLSS strategy, there is a need to first unload the loaded leg, and then to swing the loaded leg sideways. Thus, we found that it took longer for RFs to lift the leg and initiate stepping (about 100 ms). Maki and Mcllroy [31] found that when the LLSS strategy was performed, the step initiation duration was delayed about 200 ms compared with the unloaded leg strategies (i.e., ULSS, COS, Leg Abduction). However, they found that when the LLSS strategy was used, the swing phase duration was shorter and step length was smaller [31]. In contrast, we found that the step length tended to be longer in RFs compared with both NFs and OFs. We assume that this resulted from the weaker ability of RFs to control the moving CoM over the BoS, i.e., a larger CoM displacement, resulting in a longer recovery step in the first recovery stepping (Tables 2A and 3A). The differences in recovery step length between Maki and Mcllroy [31] and our findings were due to a different population that was studied, i.e., younger people in Maki and Mcllroy’s study [31] versus older adults with a varying fall history in our study. Also, in Maki and Mcllroy [31], the participants were instructed to “try not to step”; thus, their younger participants tried to avoid stepping by decelerating the moving CoM over the BoS (a 200 ms delay in step initiation), and their recovery step length was shorter. We however, instructed our older adults to react naturally; thus, the larger first recovery step length was due to their lesser ability to decelerate the CoM, and a longer first recovery step was needed in RFs to catch the moving CoM.

Interestingly, the COS strategy was rarely seen at a low perturbation magnitude. At higher perturbation magnitudes, however, all three groups predominantly used the unloaded leg strategies, especially the COS strategy, which accounted for 49.3–63.4% of their balance strategies (Fig. 3). This was accompanied by a faster step initiation duration in the RFs in the multiple-step threshold trial versus their single-step trials (see Table 2A vs. B and Table 3A vs. B). The COS strategy was performed due the fact that higher perturbations magnitudes induced a faster CoM displacement to load the standing leg and unload the swing leg, allowing a more rapid foot-lift of the unloaded leg to “crossover” the loaded leg. The drawback of this strategy is that this was accompanied with a longer step recovery time (Table 2A vs. Table 2B), probably due to a more complex swing path to move the foot across the standing loaded leg (either in front of or behind it) with a larger eCoM path displacement in the first recovery step (Table 3A), and longer step recovery duration.

In regard to the multiple-step threshold trials, RFs had a somewhat longer first recovery step length than NFs (p = 0.066, Table 2B and p = 0.067, Table 3B). Although RFs had a longer first recovery step in the multiple-step trials, they were still unable to catch the moving CoM over the BoS, and the total eCoM path displacement was significantly larger (Table 2B and Fig. 3), resulting in a longer time to fully recover their balance (Tables 2B and 3B). Fujimoto et al. [32] also found that older fallers took multiple steps more often compared with older non-fallers, suggesting that this results from a reduced functional stability limit. They suggested that RFs have a pre-planned balance behavior in response to perturbation magnitudes that elicit a stepping response in which they take several small steps to recover balance instead of a one large single step that requires lower limb muscle power generation. Luchies et al. [33], Maki et al. [34], and Fujimoto et al. [35] suggested that during the execution of the first recovery step, the central nervous system (CNS) is already estimating the level of instability in foot contact and selects a more conservative protective stepping response. Luchies et al. [33] suggested that the CNS estimates the level of instability and preplans the correct balance recovery response and the use of extra small steps to recover balance even before the first recovery step is completed. This suggests that to ensure full recovery immediately after the first recovery step response, the CNS is estimating the level of instability and selects a recovery stepping response based on the subject’s past fall experience and physical abilities. It was previously reported by Maki and Mcllroy [31] that when subjects performed repeated trials, the prevalence of multiple steps increased. This further suggests a preplanning motor function. Our results show that due to the inability to control the CoM path displacement, the RFs were more likely to take extra steps in order to avoid falls and to regain more opportunities to make corrective adjustments in their response following a lateral loss of balance (Tables 2B and 3B). Work by Pai and Patton [36] and Pai et al. [37] demonstrated that the occurrence of a step depends on the interaction between the CoM position and its velocity. Following their model, stepping is necessary if there is a sufficiently high velocity of CoM displacement, even if the vertical projection of the CoM is located within the BoS at step initiation, i.e., a concept based on CoM velocity-position limits.

The main limitation of the present study was use of a convenience sample of relatively healthy community dwelling older adult volunteers and of retrospective fall reports. A prospective study that monitors falls prospectively and includes older adults with a different functional status (i.e., healthy, pre-frail, and frail older adults) might resolve this issue. A second limitation is that this sample of RFs was small and limits the interpretation of the results. Thus, it might have been too small to reveal a significant differences in all kinematic parameters of recovery stepping. A third limitation is that the observation of arm and trunk motion was not reported in this report. It was previously reported that these movements helped to preserve balance and decelerate CoM motion during loss of balance. Further analysis might help to identify how these recovery arm and trunk movements are associated to a systematically increase of perturbation magnitudes. A fourth limitation is that the laboratory perturbations in the present study that were based on surface translation during standing may not mimic a real-life situation of balance loss closely enough due to pushes and trips.

This study showed that while older adults who experienced a coincidental single fall event (OFs) showed a lower step threshold, a more “responsive” over-reactive step response related from their fear of falling and not due to impaired balance abilities. The poorer balance recovery function in RF were associated with poorer self-reported lower extremity function and less able to rapidly lift the stepping leg off the ground thus, they tried to avoid stepping and use fix of support strategy to restore balance. When step response was inevitable the LLSS first step strategy was performed by RFs at a low perturbation magnitude, which slowed down step initiation. This resulted in an increase in the CoM displacement, causing longer step duration to complete the stepping vs. NFs. At higher perturbation magnitudes, when multiple steps were needed to preserve balance and prevent falling, all three groups used the unloaded leg strategies (i.e., CoS or ULSS) as their main balance recovery strategies. The use of these strategy resulted a faster step initiation (i.e., faster leg lift-off) however, RFs still showed a tendency to a longer first step length compared with NFs, causing a larger CoM displacement; thus, more time to fully recover from unexpected balance loss (Tables 2B and 3B) suggesting impaired ability to control of the CoM motion over BoS. Our results suggests that the differences in step kinematic and CoM characteristics does not reflect the different types of strategies taken. Rather, it reflects direct impairments in performance within a given step type in RFs. Thus, for older RFs specific balance recovery training to improve the skill of rapidly controlling the CoM motion is required.