It is well established that individuals with stroke are at increased risk of falls,1 with significant physical2 and psychosocial3 consequences that can contribute to decreased independence, activity, and participation.4 Fall rates are reported to be as high as 22% in the acute care setting,5 47% within inpatient rehabilitation,6 and up to 73% after discharge from hospital.7 Although there are numerous factors that have been linked to falls, a critical factor is the ability to execute successful balance-recovery reactions in response to instability.8 The focus of this study was to explore the characteristics of balance-recovery reactions after stroke, in light of their important link to mobility and fall risk.
A critical response in recovering from loss of balance is the ability to take a rapid, reactive step.8,9 Despite the importance of reactive stepping and its known link to falls in the elderly,10,11 research within the stroke population is only just emerging. It has been reported that 71% of ambulatory individuals with stroke have impaired reactive stepping performance at time of discharge from inpatient rehabilitation; this impairment is not clearly identified by commonly used clinical measures.12 Impaired performance is characterized by the need for assistance, an inability to initiate a step freely with either the nonparetic or paretic limb, decreased foot clearance, multiple-step responses, or a lack of attempt to step.12–15 Importantly, recent studies have confirmed that features of impaired reactive stepping performance are associated with falls after stroke in inpatient rehabilitation13 and predictive of falls upon return to the community.16 This would suggest that the assessment of reactive stepping after stroke is an important focus for clinical attention.
The failure to routinely assess reactive balance control within current clinical practice is of concern.17 In response, we implemented a standardized lean-and-release methodology as a measure of reactive balance control within an inpatient stroke rehabilitation setting, and demonstrated potential for clinical uptake.18,19 Measurement technology (ie, force plates) was added to reveal underlying balance control issues, which can be masked by observation-based methods.20,21 The capacity to reveal the temporal characteristics of the response may be of particular importance. Markedly delayed stepping responses have been observed poststroke.14,15 Specifically, delays in early “time to foot off” (TFO) phases, which occur before observable limb movement, are known to influence the overall success or failure of the stepping response14 and are associated with increased fall rates within inpatient rehabilitation.16 The potential to modify the temporal properties of reactive stepping, through task-specific balance “perturbation” training, has also been demonstrated.14 Collectively, this suggests that it may be important to both measure and target this early phase of reactive stepping within clinical rehabilitation settings.
Surprisingly, little information is available about paretic limb timing within reactive stepping responses; only one pilot study has a reference to a paretic mean TFO value.15 In this study, TFO was slower in trials where the individual with stroke stepped with the paretic versus nonparetic limb.15 Studies of perturbation-evoked feet-in-place responses have also reported slower, more variable paretic limb muscle onset latency,22–25 which is associated with less paretic lower limb motor recovery.22 Therefore, it is plausible that paretic limb TFO would be slower than nonparetic limb TFO after stroke. However, unique stroke-related impairments may need to be considered. Individuals with stroke tend to bear more weight on their nonparetic limb26 and the capacity to load the paretic limb is less than the nonparetic when weight shifting27; these impairments may differentially influence the time required to unload the nonparetic and paretic limbs to initiate a reactive step.
The objectives of this study were, therefore, to (1) determine whether there were differences in reactive stepping TFO within the paretic versus nonparetic lower limb; (2) investigate the determinants of TFO within the paretic and nonparetic limb; and (3) investigate the influence of TFO on the reactive stepping performance of those with stroke. We hypothesized that (i) paretic TFO would be slower than nonparetic TFO; (ii) TFO would be negatively associated with paretic lower limb motor recovery, positively associated with the load borne under the paretic or nonparetic stepping limb, and negatively associated with capacity to maximally load the paretic and nonparetic stance limbs during weight shift; and (iii) slower TFO would be associated with an increased likelihood of the individual to “fall” in response to a postural perturbation.
Standardized assessment of reactive balance control is routinely conducted at the participating inpatient stroke institution. As a result, this study was able to be conducted as a retrospective chart review and was approved by the research ethics board of the Toronto Rehabilitation Institute–University Health Network.
Setting and Participants
Assessments were administered within an onsite clinic that integrates technological and clinical measures to assess balance and gait. The reactive balance control assessment (summarized later) is one component of the larger assessment and administered to individuals within stroke inpatient rehabilitation at the discretion of the primary physical therapists, as part of their routine practice. Individuals considered for assessment were required to 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, comprehend and follow instructions, and have the capacity to stand unsupported and walk at least 5 m without physical assistance, with or without an assistive device. Information was extracted from the clinic database for individuals who completed a discharge assessment between October 2009 and September 2012. Of the 180 individuals who received a reactive balance control assessment at time of discharge, 75 were excluded: 23 did not have an identified paretic and nonparetic limb (ie, bilateral impairments or “no” affected side), 13 had musculoskeletal issues (eg, previous hip or knee arthroplasty), 2 had concurrent neurological diagnoses other than stroke, 1 did not initiate any stepping responses during the assessment, 13 had targeted reactive balance control training during the course of their therapies, and 23 did not have trials for both paretic and nonparetic steps. Therefore, a final sample of 105 individuals with stroke was included in subsequent analyses.
Participant characteristics extracted from the database included age, sex, time poststroke, side of paresis, level of functional disability (Functional Independence Measure28—total, motor, and cognitive score), and functional balance (Berg Balance scores).29 Lower limb sensory impairment was also extracted from the physical therapists' assessments, as a binary variable (yes/no). Sensory assessments were not standardized across therapists and, therefore, these data were not included as a predictor variable.
Response Variable: Assessment of Time to Foot off During Reactive Stepping
Reactive stepping was evaluated using a “lean-and-release” balance perturbation method.30 The lean-and-release assessment simulates a forward fall; the individual leans forward on a horizontal cable, attached at the level of his/her chest, which is released at an unpredictable time, eliciting a reactive stepping response. In addition to physical therapist supervision, the individual is attached to an overhead safety harness system to allow for unrestricted movements but safety should balance recovery fail. Participants are assessed under 2 conditions: up to 5 trials each of “usual response” and “encouraged-use.” In usual response conditions, the individuals were instructed to “respond however you naturally would to recover your balance.” In encouraged-use trials, the preferred stepping limb (the limb most frequently used in the usual response condition) was blocked by the physical therapist (placing their hand or foot approximately 5 cm in front of the shin), to force stepping with the opposite limb. Individuals were instructed to “respond however you would naturally to recover your balance knowing that I have blocked this limb.” Individuals were assessed in usual, flat footwear, and ankle-foot orthoses if prescribed. Individuals stood with 1 foot on each of 2 forceplates (Advanced Medical Technology Inc, Watertown, MA) in a standardized foot position (heel centers 0.17 m apart, 14 cm between the long axes of the feet30). A load cell placed in series with the horizontal cable measures the force placed on the cable when leaning, to ensure consistency of perturbation amplitude. Participants were encouraged to lean forward from the ankles such that 8% to 10% of their body weight was consistently supported. Previous research has determined that cable load values of this amplitude consistently elicit stepping responses among this patient cohort.18 Lesser lean angles were allowed at the physical therapist's discretion, according to patient ability or preference; however, trials with cable load of less than 3% body weight were excluded from analyses. The load cell was also used to detect perturbation onset (ie, time when force recorded was <1 N).
Time to foot off was measured as the time between perturbation onset and when the vertical force recorded under the stepping limb was less than 1% body weight. Load cell and forceplate data were sampled at 256 Hz. The assessment was video-recorded and reviewed to confirm performance including the initial stepping limb (paretic or nonparetic) and occurrence of a perturbation-evoked “fall” (ie, need for assistance by the supervising physical therapist or harness).
Lower limb impairment was determined from the Chedoke-McMaster Stroke Assessment leg (CMSA-Leg) and foot (CMSA-Foot) stage of motor recovery scores.31 The CMSA is a commonly used clinical measure of motor impairment with established intrarater and interrater reliability and concurrent validity, when used with individuals within stroke rehabilitation.32 The CMSA assigns a score between 1 and 7, with higher CMSA scores indicating greater motor recovery or less limb impairment.
Stepping limb load was determined by the percentage of total body weight (% BW) under the paretic/nonparetic stepping limb under 2 conditions: (i) usual stance, measured as the % BW on the limbs during quiet standing with eyes open, averaged over 30 seconds; and (i) preperturbation, measured as the % BW on the limbs, averaged over 1 second immediately before the onset of the balance perturbation during the lean-and-release test.
The capacity to maximally load the paretic/non-paretic lower limb was determined by the % BW able to be borne under the respective limb, averaged over the duration of the trial. The individual stood on forceplates, as outlined earlier, and was instructed to weight shift to the paretic side/nonparetic side and bear as much weight as possible on that limb, maintaining this position for up to 20 seconds.
Amplitude of the perturbation (cable load) was also included as a covariate of TFO, given that the magnitude of the perturbation could vary across individuals with varying functional abilities. Cable load was expressed as the % BW (averaged over 1 second before the perturbation) associated with the preperturbation lean angle.
All statistical analyses were performed using SAS 9.2 (SAS Institute Inc, Cary, NC). Descriptive statistics were used to characterize the sample. The mean values of TFO across trials were determined for the paretic and nonparetic limbs of each individual with stroke, calculated with the first perturbation trial removed as it is known to have different characteristics than subsequent trials.33,34 Encouraged-use trials, where the individual initiated a step with the blocked limb, were also excluded. Paired t tests were used to determine differences in mean paretic versus nonparetic TFO within individuals (α = 0.05). Multivariate regression analyses were used to establish associations of predictor variables with nonparetic and paretic TFO. A stepwise method of regression analyses was then performed with variables entered in the model at a significance level of P ≤ 0.15, to determine the most predictive variables explaining TFO. Correlational and variance inflation factor analyses were calculated to determine possible influence of multicollinearity. CMSA-Leg and CMSA-Foot scores were significantly correlated (r = 0.63; P < 0.0001) with variance inflation factors of 1.8 and 1.9, respectively. Regression analyses were repeated independently with significantly correlated variables removed. There was no impact on statistical inference with both or either CMSA-Leg/Foot scores included; therefore, the final model used only CMSA-Foot scores. Logistic regression was used to determine whether paretic and nonparetic mean TFO was associated with increased likelihood for the individual with stroke to “fall” during the assessment (ie, need for the physical therapist or harness assistance within any trial).
Participant profile is displayed in Table 1 for the 105 individuals with stroke who completed a reactive stepping assessment, at time of discharge from inpatient rehabilitation. Participant limb load values varied across quiet stance and preperturbation conditions. Mean paretic limb load was significantly greater during quiet stance (47.3; standard deviation [SD], 7.5% BW) as compared with preperturbation (44.8; SD, 10.0% BW) when stepping with the paretic limb (mean difference 2.5% BW; SD, 9.7; 95% confidence interval [CI], 0.6-4.4; P = 0.009) and significantly less during quiet stance as compared with preperturbation (50.1; SD, 9.2% BW) when stepping with the nonparetic limb (mean difference −2.8% BW; SD, 9.3; 95% CI, −4.6 to −0.9; P = 0.003). Preperturbation paretic limb load was also significantly less when stepping with the paretic limb versus the nonparetic limb (mean difference −5.3% BW; SD, 6.3; 95% CI, −6 to −4; P < 0.0001).
Reactive TFO in the Paretic and Nonparetic Lower Limb
The participants' mean TFO values are displayed in Figure 1. There was no significant difference between mean paretic and nonparetic limb TFO (mean paretic 351 ms; mean nonparetic 365 ms; mean difference −14 ms; P = 0.20).
Determinants of TFO in the Paretic and Nonparetic Lower Limb
Results of multivariate and stepwise regression analyses are displayed in Table 2. Within the final model, the capacity to maximally load the nonparetic limb, the amplitude of the perturbation (cable load), and the capacity to maximally load the paretic limb explained 23.8% of the variance in, and were all negatively associated with, paretic reactive step TFO (F3,88 = 9.18; P < 0.0001). Within the final model, the amplitude of the perturbation (cable load) and the preperturbation load under the nonparetic stepping limb explained 22.7% of the variance in, and were respectively negatively and positively associated with, the nonparetic reactive step TFO (F2,89 = 37.52; P < 0.0001).
TFO and Consequences for Perturbation-Evoked Falls
Sixteen of the 105 individuals (15%) fell during trials where they initiated stepping with the paretic limb; 7 of the 105 individuals (7%) fell during trials who initiated stepping with the nonparetic limb. The likelihood for the individual to fall was independently associated with nonparetic TFO (odds ratio, 1.009; 95% CI, 1.003-1.015; P = 0.003) but not paretic TFO (odds ratio, 1.000; 95% CI, 0.994-1.006; P = 0.95).
This study revealed that, among those within subacute stages of stroke recovery, reactive TFO did not significantly differ when using the paretic versus nonparetic lower limb; however, the determinants of TFO did differ between limbs. Furthermore, the likelihood of a perturbation-evoked fall was associated with TFO of the nonparetic, but not the paretic, limb.
This is the first study to compare paretic and nonparetic reactive TFO within individuals with stroke. Contrary to our hypothesis, TFO was not slower in the paretic versus the nonparetic limb. In partial support of our hypotheses, the study results suggest that unique stroke-related impairments may differentially influence paretic and nonparetic reactive TFO. Greater capacity to weight shift and load the nonparetic limb (and, less so, to load the paretic limb) was associated with faster paretic TFO; this may suggest mediolateral dynamic stability contributes to paretic reactive step timing in early foot-off phases. Postural asymmetry resulting in greater load on the nonparetic lower limb, just before instability, was associated with slower nonparetic TFO. The aforementioned factors may, therefore, be important to measure and target in interventions aimed at improving the temporal characteristics of the response. There is evidence to suggest that improvements in step timing, specifically TFO, can be achieved with task-specific “perturbation-based” balance training14; research is ongoing to determine the efficacy of such targeted reactive balance control training after stroke.35
There have been few studies to date examining reactive step TFO after stroke. TFO mean values reported within this study are somewhat faster than previous studies examining individuals within stroke inpatient rehabilitation. An initial pilot study15 documented ranges between 515 and 891 ms; however, that study relied on a cable pull system for perturbations, which may have led to slower initial accelerations and delays in timing of balance responses. A previous study, with a subgroup of the current cohort,13 demonstrated mean values of 490 ms for fallers and 440 ms for nonfallers. The differences may be accounted for by differences in stage of rehabilitation/recovery (later in the current study) and the fact that the current study excluded the first stepping trial. The latencies identified in the current study were slower, however, than those previously reported for the healthy elderly using similar assessment methodology. In the current study, the mean paretic and nonparetic TFO values for stroke were at, or beyond, the upper confidence limits of values calculated from data reported by Thelen and colleagues,36 in a cohort of healthy elderly (mean TFO 315; SD, 66 ms; 95% CI, 273-357; n = 12). However, these values for the elderly cohort were associated with greater lean angles (15% BW). Therefore, it is not clear whether the delays in step initiation within the current study were a result of unique stroke-specific impairments or differences in perturbation amplitude. It is unlikely that our participants would have been able to achieve lean angles of this amplitude, given the profound challenge to balance recovery previously revealed at lesser lean angles18; therefore, direct comparisons cannot be made.
Regardless of the differences between studies, rapid TFO latencies were evoked within the current testing paradigm for both the paretic and nonparetic limbs. It was most interesting that paretic limb timing was faster on average than the nonparetic limb. Two possible factors may account for such rapid paretic limb responses: (1) induced instability has the capacity to “reflexively” evoke very rapid reactions (in spite of the slowing that may be evident when individuals are asked to move voluntarily),37 or (2) the individuals may use adaptive strategies (ie, preloading the nonparetic limb in anticipation of a step with the paretic limb) to accomplish a more rapid time to unload the limb, despite poor motor control of the paretic in comparison to the nonparetic limb.
It is noteworthy that limb load significantly differed when the individual was in “usual stance” posture as compared with “preperturbation” and, further, differed when the individual was stepping with the paretic versus nonparetic limb. As noted earlier, this could suggest that the individual was preplanning and unloading the respective stepping limb to facilitate step initiation. The results also suggest that the amplitude of the perturbation can independently influence TFO within both lower limbs; larger amplitude results in faster responses. Collectively, this suggests that attention to both preperturbation limb load and cable load would be important for future standardization of methods and interpretation of the temporal characteristics of reactive stepping.
A previous study has demonstrated a positive association between TFO and falls within inpatient rehabilitation, but did not differentiate between the paretic and nonparetic lower limbs.13 The results of the current study suggest that nonparetic, but not paretic, TFO may influence falls. Specifically, for every 1-ms increase in nonparetic mean TFO, there is a 1% increase in the odds of the individual with stroke falling in response to evoked postural perturbations. A previous study has documented that individuals in subacute stages of stroke recovery commonly require multiple steps to regain stability.12 It is plausible that if initiation of a paretic step was delayed, a follow-up step with the nonparetic limb may successfully recapture balance. In contrast, if initiation of a nonparetic step was delayed, the follow-up step with the paretic limb may not be able to successfully regain stability. This is speculative, but may highlight the importance of other features of paretic limb reactive stepping, such as step characteristics (length, time, placement) and the capacity to restabilize at step termination, that need to be considered within the context of reactive stepping and the link to falls.
The determinants of paretic and nonparetic reactive step TFO, as per the regression analyses, explained less than 25% of the variance. We acknowledge the possible contribution of other factors, not included in this study, that may provide additional explanation of the variance in the regression models and require further study. It is recognized that TFO is a composite measure that can be further divided into step onset time, anticipatory postural adjustment time, and stepping limb unloading time.15 Future research should explore possible phase-specific delays of the paretic and nonparetic limbs, and their respective determinants, that may contribute to overall delays in TFO.
Another factor that may have contributed to our results is that the first trial was excluded in the our analyses for methodological reasons, but the first trial may be more ecologically valid than subsequent trials, representing the unpracticed response triggered by a fall in everyday life.33,38 The determinants and consequences of delays in step initiation revealed in this study may not generalize to this more novel stepping response. Future research should explore the temporal characteristics of the first trials and, further, determine how best to incorporate this more novel response into standardized methods of measurement. Finally, the lack of association between clinical measures of lower limb impairment and reactive step initiation suggests that commonly used clinical measures did not clearly reveal impairments in reactive stepping performance. This finding supports the need to incorporate alternate methodologies and technological measures to better reveal and quantify underlying dyscontrol associated with reactive stepping performance.
In persons with stroke, unique impairments of dynamic balance control and limb-load asymmetry of the paretic and nonparetic lower limbs, respectively, may differentially influence reactive step time to foot off in response to a balance perturbation. The amplitude of the perturbation influences reactive step time to foot off within both limbs. Delays in nonparetic, but not paretic, time to foot off increase the likelihood of a fall. The results of the current study may have implications for the future development of standardized clinical assessment methodologies and training strategies to evaluate and remediate reactive stepping and reduce fall risk.
The authors acknowledge the support of Toronto Rehabilitation Institute-University Health Network. Equipment has been funded with grants from the Canada Foundation for Innovation, Ontario Innovation Trust and the Ministry of Research and Innovation. The views expressed do not necessarily reflect those of the Ministry. Support personnel were funded by the Canadian Institutes of Health Research Mobility in Aging Team grant (CIHR #MAT-91865). At the time of this study, Elizabeth L. Inness was supported by a Canadian Institutes Health Research Fellowship (Health Professions). Avril Mansfield is supported by a New Investigator Award from the Canadian Institutes of Health Research (MSH-141983).
The authors wish to acknowledge the undergraduate cooperative students and the staff of the Balance, Mobility and Falls Clinic at the Toronto Rehabilitation Institute who aided in data collection. The authors wish to acknowledge Julia Fraser, who assisted in the development of the video abstract.
1. Weerdesteyn V, de Niet M, van Duijnhoven HJ, Geurts AC. Falls
in individuals with stroke. J Rehabil Res Dev. 2008;45:1195–1213.
2. Kanis J, Oden A, Johnell O. Acute and long-term increase in fracture risk after hospitalization for stroke. Stroke. 2001;32:702–706.
3. Andersson AG, Kamwendo K, Appelros P. Fear of falling in stroke patients: relationship with previous falls
and functional characteristics. Int J Rehabil Res. 2008;31:261–264.
4. Schmid AA, Yaggi HK, Burrus N, et al. Circumstances and consequences of falls
among people with chronic stroke. J Rehabil Res Dev. 2013;50:1277–1286.
5. Davenport RJ, Dennis MS, Wellwod I, Warlow CP. Complications after acute stroke. Stroke. 1996;27(3):415–420.
6. Mayo NE, Korner-Bitensky N, Kaizer F. Relationship between response time and falls
among stroke patients undergoing physical rehabilitation. Int J Rehabil Res. 1990;13(1):47–55.
7. Forster A, Young J. Incidence and consequences of falls
due to stroke: a systematic inquiry. BMJ. 1995;311:83–86.
8. Maki BE, McIlroy WE. The role of limb movements in maintaining upright stance: the “change-in-support” strategy. Phys Ther. 1997;77:488–507.
9. Maki BE, McIlroy WE. Control of rapid limb movements for balance
recovery: age-related changes and implications for fall prevention. Age Ageing. 2006;35(suppl 2):ii12–ii18.
10. Maki B, Edmonstone M, Perry S, Heung E, Quant S, McIlroy W. Control of rapid limb movements for balance
recovery: do age-related changes predict falling risk? In: Duysens J, Smits-Engelsman BCM, Kingma H, eds. Control of Posture and Gait. Maastricht, the Netherlands: International Society for Postural and Gait Research; 2001:126–129.
11. Hilliard MJ, Martinez KM, Janssen I, et al. Lateral balance
factors predict future falls
in community-living older adults. Arch Phys Med Rehabil. 2008;89:1708–1713.
12. Inness EL, Mansfield A, Lakhani B, Bayley M, McIlroy WE. Impaired reactive stepping among patients ready for discharge from inpatient stroke rehabilitation. Phys Ther. 2014;94(12):1755–1764.
13. Mansfield A, Inness EL, Wong JS, Fraser JE, McIlroy WE. Is impaired control of reactive stepping related to falls
during inpatient stroke rehabilitation? Neurorehabil Neural Rep. 2013;27(6):526–533.
14. Mansfield A, Inness EL, Komar J, et al. Training rapid stepping responses in an individual with stroke. Phys Ther. 2011;91(6):958–969.
15. Lakhani B, Mansfield A, Inness EL, McIlroy WE. Compensatory stepping in individuals with stroke: a pilot study. Physiother Theory Pract. 2011;27(4):299–309.
16. Mansfield A, Wong JS, McIlroy WE, et al. Do measures of reactive balance
control predict falls
in people with stroke returning to the community? Physiotherapy. 2015;101(4):373–380.
17. Sibley KM, Straus SE, Inness EL, Salbach NM, Jaglal SB. Balance
assessment practices and use of standardized balance
measures among Ontario physical therapists. Phys Ther. 2011;91:1583–1591.
18. Inness EL, Mansfield A, Biasin L, Brunton K, Bayley M, McIlroy WE. Clinical implementation of a reactive balance
control assessment in a sub-acute stroke patient population using a “lean-and-release” methodology. Gait Posture. 2014;41(2):529–534.
19. Pak P, Jawed H, Tirone C, et al. Incorporating research technology into the clinical assessment of balance
and mobility: perspectives of physiotherapists and people with stroke. Physiother Can. 2015;67(1):1–8.
20. vanAsseldonk EH, Buurke JH, Bloem BR, et al. Disentangling the contribution of the paretic and non-paretic ankle to balance
control in stroke patients. Exp Neurol. 2006;201:441–451.
21. Garland SJ, Willems DA, Ivanova TD, Miller KJ. Recovery of standing balance
and functional mobility after stroke. Arch Phys Med Rehabil. 2003;84:1753–1759.
22. Badke MB, Duncan PW. Patterns of rapid motor responses during postural adjustments when standing in healthy subjects and hemiplegic patients. Phys Ther. 1983;63:13–20.
23. DiFabio RP, Badke MB, Duncan PW. Adapting human postural reflexes following localized cerebrovascular lesion: analysis of bilateral long latency responses. Brain Res. 1986;363:257–264.
24. Ikai T, Kaimkubo T, Takehara I, Nishi M, Miyano S. Dynamic postural control
in patients with hemiparesis. Am J Phys Med Rehab. 2003;82(6):463–439.
25. Marigold DS, Eng JJ, Timothy IJ. Modulation of ankle muscle postural reflexes in stroke: influence of weight-bearing load. Clin Neurophysiol. 2004;115(12):2789–2797.
26. Eng JJ, Chu KS. Reliability and comparison of weight-bearing ability during standing tasks for individuals with chronic stroke. Arch Phys Med Rehabil. 2002;83:1138–1144.
27. Bohannon RW, Larkin PA. Lower extremity weight bearing under various standing conditions in independently ambulatory patients with hemiparesis. Phys Ther. 1985;65:1323–1325.
28. Granger CV, Markello SJ, Graham JE, Deutsch A, Ottenbacher KJ. The uniform data system for medical rehabilitation: report of patients with stroke discharged from comprehensive medical programs in 2000-2007. Am J Phys Med Rehabil. 2009;88:961–972.
29. Berg K, Wood-Dauphinee S, Williams JI. The balance
scale: reliability assessment with elderly residents and patients with an acute stroke. Scand J Rehabil Med. 1995;27:27–36.
30. Hsiao-Wecksler ET. Biomechanical and age-related differences in balance
recovery using the tether-release method. J Electromyogr Kinesiol. 2008;18(2):179–187.
31. McIlroy WE, Maki BE. Preferred placement of the feet during quiet stance: development of a standardized foot placement for balance
testing. Clin Biomech. 1997;12:66–70.
32. Gowland C, Stratford P, Ward M, et al. Measuring physical impairment and disability with the Chedoke-McMaster stroke assessment. Stroke. 1993;24:58–63.
33. McIlroy WE, Maki BE. Adaptive changes to compensatory stepping responses. Gait Posture. 1995;3:43–50.
34. Marigold DS, Patla AE. Strategies for dynamic stability during locomotion on a slippery surface: effects of prior experience and knowledge. J Neurophysiol. 2002;88(1):229–252.
35. Mansfield A, Aqui A, Centen A, et al. Perturbation training to promote safe independent mobility post-stroke: study protocol for a randomized controlled trial. BMC Neurol. 2015;15:87. doi:10.1186/s12883-015-0347-8
36. Thelen DG, Muriuki M, James J, Schultz AB, Ashton-Miller JA, Alexander NB. Muscle activities used by young and old adults when stepping to regain balance
during a forward fall. J Electromyogr Kinesiol. 2000;10:93–101.
37. Martinez KM, Mille M, Zhang Y, Rogers MW. Stepping in persons poststroke: comparison of voluntary and perturbation-induced responses. Arch Phys Med Rehabil. 2013; 94:2425–2432.
38. Barrett RS, Cronin NJ, Lichtwark GA, Mills PM, Carty CP. Adaptive recovery responses to repeated forward loss of balance
in older adults. J Biomech. 2012;45:183–187.