Abstract
Background Individuals with stroke are at increased risk for falls soon after hospital discharge. The ability to react to a balance perturbation, specifically with a rapid step, is critical to maintain balance and prevent falls.
Objective The purpose of the study was to determine the prevalence of impaired reactive stepping responses in an ambulatory group of patients with stroke who were preparing for discharge from inpatient rehabilitation and the relationship to patient performance on commonly used clinical measures of balance, mobility, and lower limb impairment.
Design This study was a retrospective analysis of patient admissions over a 3-year period.
Methods Charts were reviewed for patients who, at time of discharge, had completed a perturbation-evoked reactive stepping assessment.
Results Ninety-nine (71%) of 139 patients had impaired stepping reactions characterized by the need for assistance, an inability to step with either lower limb, or the need for multiple-step responses. There was a statistically significant difference in clinical scores between those with and without impaired stepping, but groups were characterized by considerable variation in clinical profiles. For example, Berg Balance Scale scores ranged from 25 to 55 versus 20 to 56 and gait speeds ranged from 0.17 to 1.43 versus 0.26 to 1.55 m/s for patients who demonstrated a failed step versus a successful step, respectively.
Limitations Not all patients who attended stroke rehabilitation received a reactive stepping assessment at discharge.
Conclusions Impaired reactive stepping is a prevalent problem for ambulatory patients with stroke preparing for discharge, possibly increasing their risk of falling when faced with the challenges of community ambulation. Specific tests that target the capacity to perform perturbation-evoked stepping reactions may be important to identify those at risk for falls and to direct appropriate intervention strategies.
The risk of falls for survivors of stroke is high,1,2 with significant physical3 and psychosocial4 consequences that contribute to decreased independence, activity, and participation.5 Falls are common upon return to home after stroke. Fall rates as high as 73% have been reported,2 and the majority of falls occur within the first 2 months following discharge from rehabilitation.6 It is possible, therefore, that we are not optimally identifying those at most risk, nor are we preparing these individuals during hospitalization for the challenges they will encounter in their everyday living environment.1
While challenges in the community may expose fall risk, the key factor that ultimately determines whether an individual will fall is their ability to recover from a loss of balance, specifically using a rapid stepping response.7,8 Reactive stepping responses are not only “last resorts” to large-magnitude perturbations but also the preferred response to small-magnitude perturbations8 commonly observed in real-life situations.9 Numerous age-related changes in reactive stepping responses have been observed. Elderly people are more likely to demonstrate a failed capacity to recover from instability than younger adults10,11 and are more likely to take multiple steps to restore balance,12,13 a consequence of ongoing instability after the initial step.14,15 Furthermore, such multi-step responses have been found to be predictive of falls in daily life among older adults.15,16
Despite the importance of reactive stepping and despite the fact that early work by Harburn and colleagues17 proposed using a weight-drop cable pull as a clinical method to test, there exists little research in this area with individuals with stroke. Findings from our previous pilot work suggest this is an important area of further study. Across patient cases, individuals in subacute stages of stroke demonstrate impaired anticipatory postural adjustments, delays in timing, an inability or unwillingness to initiate a step with the paretic limb, and the use of multi-step responses or the need for assistance to regain stability.18,19 Importantly, features of reactive stepping have been associated with falls after stroke in inpatient rehabilitation20 and, more recently, predictive of falls upon return to the community.21 Given the clinical attention directed toward balance and mobility retraining and falls prevention within rehabilitation, there is no doubt that greater insight into the magnitude of this clinical problem for patients with stroke is warranted.
Reactive balance control is less frequently assessed in clinical practice than other aspects of balance, possibly influenced by outcome measures that are most commonly used in clinical settings.22 A potential limitation of many clinical measures of balance, mobility, or limb control is their focus on volitional limb control and self-governed speed of movement that is fundamentally different from the control and speed required for reactive stepping.23–25 Therefore, it also is important to establish the association between patient performance on measures of reactive balance control and typical clinical balance, mobility, and limb impairment measures.
Research conducted in the early phases of recovery is important to inform clinicians and guide interventions to potentially achieve important outcomes prior to discharge home. The present study affords a unique opportunity to examine reactive balance control performance in the early stages of stroke recovery with a focus on the point of discharge from inpatient rehabilitation to the community.
This study aimed (1) to characterize the prevalence of residual impairment to reactive stepping among patients being discharged from inpatient stroke rehabilitation and (2) to determine if commonly used clinical measures of balance (Berg Balance Scale [BBS]),26,27 walking capacity (gait speed), and lower limb impairment (Chedoke-McMaster Stroke Assessment [CMSA] Impairment Inventory)28 could differentiate between patients with varying abilities of reactive balance control.
Method
This study was a retrospective chart review.
Setting and Participants
The Balance, Mobility and Falls Clinic of the Toronto Rehabilitation Institute—University Health Network provides assessments of balance and gait using both technological and clinical measures as part of routine care. Assessments are administered at the discretion of the treating physical therapist. All patients are considered for assessment, but must be medically stable; have no musculoskeletal or other condition that could be exacerbated by the balance perturbation; have the cognitive-communicative ability to consent to the assessment and comprehend and follow instructions; and have the capacity to stand unsupported and walk at least 5 m without physical assistance, with or without a gait aid. Information was extracted from the clinic database20,29 for patients assessed between October 2009 and September 2012. A total of 437 patients were discharged from stroke inpatient rehabilitation to the community during that time frame. Out of the 180 patients identified as having completed a discharge assessment of reactive stepping, 41 were excluded. Seventeen patients did not complete both test conditions (outlined below). Seventeen patients had participated in enhanced balance retraining. Three patients received nonstandardized instructions that could influence their responses. For 4 patients, there were technical difficulties that prevented observation and coding of video-recorded responses. Therefore, a final sample of 139 patients was included in subsequent analyses.
Protocol for Reactive Stepping Assessment
Reactive stepping was evaluated using a “lean-and-release” balance perturbation method (Fig. 1). Participants were instructed to lean forward from their ankles (and reinstructed if improper form was used), and body weight was supported by a cable attached at chest height. At a varied and unexpected time, the cable was released, forcing participants to elicit a stepping reaction to regain stability.19,30,31 Participants wore a safety harness attached to an overhead support, and a physical therapist provided supervision to ensure safety should balance recovery fail.
The “lean and release” balance perturbation method. The patient wears a safety harness that is attached to an overhead support structure and leans forward on cable connected to the wall. The cable is released unexpectedly inducing a forward fall.
The participants were assessed in their usual flat footwear and ankle-foot orthoses (if prescribed). They stood in a standardized foot position (heel centers 0.17 m apart, 14° between the long axes of the feet32) with one foot on each of 2 forceplates (Advanced Medical Technology Inc, Watertown, Massachusetts). Perturbations were delivered under 3 conditions and in the following order: 5 trials of unconstrained conditions, 1 trial of a dual-task condition, and 5 trials of encouraged-use conditions. The secondary task of the dual-task condition is nonstandardized; therefore, these data were not included in this study.
Preperturbation cable load was monitored using a load cell (A-Tech Instruments Ltd, Scarborough, Ontario, Canada) mounted in series with the cable. Tension on the load cell was achieved through forward leaning, and cable load was used to determine the magnitude of the perturbation.31,33 The load on the cable was expressed as a percentage of body weight averaged over 1 second prior to the perturbation. There was no significant difference in patient cable load between the unconstrained and encouraged-use conditions (P=.82); therefore, cable load values represent the average of the 2 conditions. Mean cable load across participants was 8.5% body weight (SD=2.9%). A lean of 11% body weight corresponds to a whole-body lean angle of approximately 9 degrees from vertical; this perturbation is of sufficient magnitude to consistently elicit a stepping response in healthy young adults with no balance impairment.31 In unconstrained conditions, participants were instructed to respond however they would naturally respond to recover balance. In encouraged-use conditions, the preferred stepping limb (the limb used most frequently in unconstrained conditions) was blocked to force stepping with the opposite limb. A physical therapist placed his or her hand approximately 5 cm in front of the patient's shin. Participants were instructed to respond however they would naturally respond to recover balance knowing the preferred stepping limb had been blocked. All tests were video-recorded and reviewed to code responses.
Measures
Features of reactive stepping extracted from the database were: level of independence following the perturbation (ie, no assistance versus reliance on the harness or physical therapist to prevent a fall), multi-step responses (≥3 steps), and limb used for the initial step. Previous research linking multi-stepping with falls in elderly people has defined multi-step responses as responses involving more than 1 step.15,16 Given that it may be a natural response in a forward lean to use a follow-up second step to re-establish stability and base of support, we used a more conservative definition of multi-step responses. We defined multi-step responses as responses involving 3 or more steps. We examined only the first trial response of both the unconstrained and encouraged-use conditions because this test situation is most similar to that adopted in clinical settings34,35 and may have better ecological validity, representing the unpracticed response triggered by a fall in everyday life.36,37 Preperturbation limb load was calculated as the percentage of body weight borne under each limb, as measured by the forceplates.
The specific clinical measures extracted from the database were: measures of functional balance (BBS),26,27 walking capacity (gait speed), and lower limb impairment (CMSA Impairment Inventory).28 Preferred gait speed was measured using a pressure-sensitive mat (GAITRite, CIR Systems Inc, Clifton, New Jersey).38,39 Participants walked over the 4.6-m mat 3 times, wearing regular footwear. If the participant was tested with and without a walking aid, the average gait speed was chosen from the condition that yielded the fastest pace.
Participant characteristics extracted from the database included sex, age, affected hemisphere, time since onset, inpatient rehabilitation length of stay, functional mobility status (items of the Clinical Outcome Variables Scale denoting walking independence, use of aids, and endurance),40 and patient balance self-efficacy (Activities-specific Balance Confidence [ABC] Scale).41 The ABC was administered verbally only to those who had sufficient cognitive and communicative ability or English language comprehension to complete the questionnaire.
Data Analysis
All statistical analyses were performed with SAS 9.3 (SAS Institute Inc, Cary, North Carolina). Descriptive statistics were used to characterize the patient sample. Frequency values were used to describe the prevalence of patient trials exhibiting impaired stepping features. Based on their exhibited stepping reactions, participants were categorized into 3 groups: (1) failed step group: participants who demonstrated a failed capacity to step either by requiring assistance in unconstrained or encouraged-use conditions or attempting to step with the blocked limb during encouraged-use conditions; (2) multi-step group: participants who did not require assistance but who required multiple steps to regain stability; and (3) successful step group: participants who recovered balance in both conditions with 2 or fewer steps and without assistance.
A one-way analysis of variance was conducted to determine mean differences between groups on clinical measures; the Tukey test was used for pair-wise comparisons. The exact McNemar test was used to detect differences in the frequency of impaired stepping reactions (failed or multi-step responses) between unconstrained and encouraged-use conditions. Paired t tests were used to determine differences in cable load between unconstrained and encouraged-use conditions. The Fisher exact test was used to detect proportional differences in impaired stepping reactions between those who initiated a step with their affected versus unaffected lower limbs. For all statistical analyses, α=.05.
Results
At discharge, participants were at a high functional mobility level. Mean walking speed was 0.81 m/s (SD=0.32). Eighty-nine percent (n=124) of the participants were able to walk independently on indoor surfaces (with or without a mobility aid or ankle-foot orthosis). Approximately half of the sample (n=73) could walk distances of greater than 500 m (see Tab. 1 for full clinical profile). As described above, the entire cohort was subdivided into 3 groups based on reactive stepping ability. There were no statistically significant differences between groups in age (F2,136=1.47, P=.23), time since onset of stroke (F2,136=1.85, P=.16), inpatient rehabilitation length of stay (F2,136=2.72, P=.07), or balance self-efficacy (F2,100=2.01, P=.14) for those with the cognitive-communicative ability to complete this questionnaire. In addition, there were no statistically significant differences between groups in the preperturbation limb load (F2,135=2.13, P=.12) or in the amplitude of perturbation applied to the individuals as measured by the preperturbation cable load (F2,135=0.03, P=.97).
Clinical Profile of Patients by Category of Reactive Stepping Abilitya
Prevalence of Impaired Stepping Responses
The frequency of participants exhibiting a failed, multi-step, or successful stepping reaction across both unconstrained and encouraged-use conditions is displayed in Table 2. At time of discharge, only 40 out of 139 participants (29%) were able to exhibit a successful step in both unconstrained and encouraged-use conditions (successful step group) or, conversely, 99 out of 139 participants (71%) had impaired stepping reactions. In terms of participant performance across both the unconstrained and encouraged-use conditions, 59 out of 139 participants (42%) exhibited a failed step (failed step group), and 40 out of 139 participants (29%) exhibited a multi-step reaction to regain stability (multi-step group).
Frequency of Participants (N=139) Who Exhibited Impaired Reactive Stepping Performance Across Unconstrained and Encouraged-Use Conditionsa
The frequency of failed steps and multi-step reactions did not differ between unconstrained and encouraged-use conditions (P=.74 and P=.32, respectively). In unconstrained conditions, 62 out of 126 participants (49%) initiated a step with their affected lower limb (13 participants with bilateral or unspecified impairments were not included in this statistic). The frequency of failed or multi-step reactions was not significantly different for participants who initiated a step with their affected limb versus their unaffected limb (P=.44 and P=.31, respectively). In encouraged-use conditions, 30 out of 139 participants (22%) attempted to initiate a step with the blocked limb; 15 out of 30 participants had their unaffected limb blocked.
Post hoc analysis revealed that participants who initiated a step with their paretic limb in the unconstrained conditions bore significantly less weight (P=.036) under their stepping limb (mean paretic stepping limb load=43.4% of total body weight [SD=9.1%]) than those who initiated a step with their nonparetic limb (nonparetic stepping limb load=47.3% of total body weight [SD=11.5%], respectively). There was no significant difference in CMSA leg scores between paretic and nonparetic steppers (5.1 [SD=1.0] and 5.0 [SD=1.1], respectively; P=.40). Chedoke-McMaster Stroke Assessment foot scores tended to be higher in paretic steppers (4.8 [SD=1.3]) than in nonparetic steppers (4.4 [SD=1.3]); this difference approached statistical significance (P=.08).
Relationship to Clinical Measures
There were significant differences in BBS scores (F2,134=14.67, P<.0001), walking speed (F2,135=10.33, P<.0001), and CMSA leg (F2,128=10.32, P<.0001) and foot (F2,128=11.62, P<.0001) scores across patient groups. Pair-wise comparisons revealed significant differences (P<.05) between those in the failed step and successful step groups for BBS (mean difference=6.6/56; 95% confidence interval [95% CI]=3.3, 9.8), walking velocity (mean difference=0.28 m/s; 95% CI=0.13, 0.43), and CMSA leg (mean difference=0.9 out of 7; 95% CI=0.4, 1.4) and foot (mean difference=1.2 out of 7; 95% CI=0.6, 1.7) scores. Significant differences also were evident between those in the failed step and multi-step groups for BBS (mean difference=5.8 out of 56; 95% CI=2.6, 9.1) and CMSA leg scores (mean difference=0.6 out of 7; 95% CI=0.1, 1.0). Significant differences between those in the multi-step and successful step groups were evident for mean walking speed (mean difference=0.16 m/s; 95% CI=0.003, 0.33). Despite these significant statistical differences between groups, there was a wide range of clinical scores across groups of participants with varying levels of ability (Fig. 2). For example, participants in the failed step group demonstrated BBS scores ranging from 25 to 55 out of 56, walking speed values from 0.17 to 1.43 m/s, and CMSA leg and foot scores from 2 to 7 and 2 to 6 (both out of 7), respectively.
Scatterplot of individual participant (A) Berg Balance Scale (BBS) scores and (B) walking speed values by subgroup of reactive stepping ability. Bars represent mean clinical scores within each subgroup. Asterisks represent significant group mean differences.
Discussion
To our knowledge, this is the first study to examine the prevalence of impaired reactive stepping in a large cohort of patients with stroke. The results of this study confirm our early pilot work18 and demonstrate that, despite having attained a high level of functional mobility, the majority of ambulatory patients discharged from inpatient rehabilitation are unable to successfully use reactive stepping to recover balance following an induced forward fall. Indeed, 71% of patients demonstrated the need for assistance, a failed capacity to evoke a step freely with either limb, or the need for multi-step reactions to regain stability. Such balance control issues could put these individuals at significant falls risk when faced with the daily challenges of community mobility and, therefore, are worthy of more focused clinical attention.
In contrast to earlier studies,18,29 this group did not demonstrate a preference to use the nonparetic limb for the initial step. This finding could be attributed to methodological differences. Our previous study exploring determinants of limb preference measured paretic and nonparetic limb use across multiple step trials,29 whereas this study considered only the first, most novel step. However, in support of earlier study observations,29 results obtained from this larger cohort similarly showed that patients who initiated a step with their paretic lower limb tended to have better distal lower limb motor recovery (higher motor CMSA foot, but not leg, scores) and bear less weight on their paretic limb prior to perturbation than those who stepped with their nonparetic limb. The latter finding could have been due to an unwillingness or inability to load the paretic limb; however, given the higher motor recovery scores of this subgroup, the latter finding may reflect a compensatory strategy to facilitate the necessary rapid response.
It is noteworthy, however, that patients were profoundly and equally challenged when stepping in both unconstrained and encouraged-use conditions and when initiating a step with the affected and unaffected limb. This finding may reflect the unique challenges of those with stroke; difficulties in speed and precision of lower limb control may limit the patient's ability to step with the affected limb, whereas challenges in loading the affected limb may limit the patient's ability to successfully execute a step with the unaffected limb. It also may be important to differentiate between the capacity to initiate a step and the capacity to execute a step of appropriate length, time, and precision to successfully regain stability. Future research is warranted in order to better understand the spatiotemporal characteristics and underlying control issues of reactive stepping that may differentially influence the success or failure when stepping with the affected and unaffected limbs.
When participants were placed in conditions that constrained use of their preferred stepping limb, irrespective of whether it was the affected or unaffected limb, 22% of the participants initiated a step with their preferred, but blocked, stepping limb. The failure to freely evoke a balance reaction with either limb could put these individuals at obvious falls risk. The current study was limited to anterior perturbations where the selection of either limb in unconstrained conditions is a possible solution to the balance control challenge. However, when faced with unpredictable, multidirectional balance perturbations of daily life, the need to be able to step with either limb is essential. This finding reinforces the need to assess (and train) stepping reactions of both the nonparetic and paretic limbs and supports the use of encouraged-use paradigms of reactive stepping to provide valuable and additional clinical insights to patient performance.
The BBS is the most commonly used clinical measure in stroke rehabilitation for balance42 and fall risk.43 Gait speed also often is used clinically as an overall measure of walking capacity and preparedness for safe community mobility.44,45 Arguably, safe community mobility also encompasses the ability to successfully respond to the countless perturbations to balance (eg, sudden stops, turns, bumps, slips, and trips) that occur in daily activities. It is noteworthy that neither of these measures could clearly discriminate between patients who, despite being poised for discharge to the community, had impaired stepping reactions that could put them at risk in this environment. The wide range of clinical scores for those with failed stepping reactions when balance was perturbed suggests that there are challenges in using these measures to predict performance at the level of the individual patient. Clinical measures that assess balance and mobility through voluntary movement may not appropriately challenge the individual with the timing or stability requirements necessary for successful perturbation-evoked balance responses; they may provide misleading information about patients' balance abilities in these situations.
A strength of this study was its ability to characterize performance in the “typical” patient. This characterization was made possible by the implementation of a lean-and-release method in routine clinical practice, a safe, standardized protocol to measure capacity for reactive stepping within the subacute stages of stroke. Not all patients admitted to inpatient rehabilitation, however, received a discharge assessment. Patients who do not receive this reactive balance control assessment tend to be at lower levels of functional mobility with greater lower limb impairment than those assessed (E.L. Inness, unpublished data, 2014). It is possible, therefore, that the prevalence of impaired reactive stepping would be greater than the present results suggest if all patients were included in the present study. However, given that the majority of patients after stroke regain the capacity to walk46 and walking is the most common activity in which falls occur upon discharge to the community,1 we believe the present sample is representative of, and provides insights into, the magnitude of reactive balance control impairment within, the ambulatory patient with stroke poised for discharge. The lean-and-release method is limited to a forward fall; it is not intended to mimic all possible “real-world” falls that may occur in various directions or environmental conditions. It is, however, intended to reveal the patient's capacity to respond with a reactive step to challenging and temporally unpredictable perturbations to balance. Through use of this method, it has been determined that reactive stepping is associated with falls after stroke in inpatient rehabilitation20 and is predictive of falls upon return to the community.21 These previous studies provide support for the validity of the lean-and-release test and the relevance of the present findings to potential fall risk for the patient with stroke returning to the community.
Future research should explore the feasibility of implementing more specific and standardized assessments of reactive balance control into clinical practice. There were no differences in cable load between groups; however, there was notable variation in cable load among individuals. This contrast suggests that either patients had varying abilities or comfort level in leaning forward or therapists were preselecting perturbation amplitudes relative to the functional ability of the individual patient. Attention to maintaining comparable perturbation amplitude across assessments would be important for evaluation of change within individual patients. Standardizing criteria for choice of perturbation amplitude across varying patient abilities also would be important for the ongoing development of these methods for use in clinical practice. Alternative methods of perturbation could be explored. Further understanding of the most appropriate and clinically meaningful measures also would be important. The present study restricted its analyses to observational measures. However, important features that cannot be detected through observation (eg, time required for step initiation) can influence the overall success or failure of the stepping response,18–20 suggesting that there is potential added clinical value of pairing technological measures with this method. Future research using kinetic, kinematic, and electromyographic analysis are warranted to gain insight into the underlying control issues of reactive stepping after stroke. The relationship between reactive balance control and other measures beyond fall risk, including functional mobility status, balance confidence, physical activity, and participation, should be explored, allowing for the evolution of clinically meaningful measures to ultimately guide more focused treatment.
In conclusion, impaired balance-recovery stepping reactions are a prevalent problem among ambulatory stroke patients preparing for discharge. Impaired balance-recovery stepping reactions may increase risk of falls for members of this population when they are faced with the challenges in the community. This aspect of balance assessment, therefore, is worthy of more focused clinical attention. Specific tests that target the capacity to perform reactive stepping may be important to identify those at risk for falls and to direct appropriate intervention strategies.
The Bottom Line
What do we already know about this topic?
People with stroke are at increased risk for falls after discharge from the hospital. The ability to take a rapid, reactive step in response to a loss of balance is critical to preventing falls.
What new information does this study offer?
The majority of patients who were ambulatory and ready for discharge from inpatient stroke rehabilitation were unable to successfully use reactive stepping to recover balance following a forward fall. Commonly used clinical assessments did not clearly identify those patients who had problems with reactive stepping.
If you're a patient or a caregiver, what might these findings mean for you?
The inability to take a rapid, reactive step in response to a loss of balance might put you at risk for increased falls. Therapists may need to use assessments that specifically target this aspect of balance to better identify your risk for falls and to provide treatment during your rehabilitation.
Footnotes
All authors provided concept/idea/research design, writing, and consultation (including review of manuscript before submission). Ms Inness provided data collection. Ms Inness, Dr Mansfield, Dr Lakhani, and Dr Bayley provided data analysis. Ms Inness and Dr Bayley provided project management. Dr Bayley and Dr McIlroy provided fund procurement and facilities/equipment. Dr Bayley provided participants and institutional liaisons. The authors thank Lou Biasin, BScPT, Karen Brunton, BScPT, and Julia Fraser, MSc, for their assistance in data collection and preparation of the manuscript.
The authors acknowledge the support of Toronto Rehabilitation Institute–University Health Network. Equipment and space have been funded with grants from the Canada Foundation for Innovation, Ontario Innovation Trust, and the Ministry of Research and Innovation. Ms Inness is supported by a Canadian Institutes Health Research Fellowship (Health Professions).
This study was approved by the Toronto Rehabilitation Institute–University Health Network's Research Ethics Board.
- Received December 16, 2013.
- Accepted July 17, 2014.
- © 2014 American Physical Therapy Association
References
Issue highlights
