Lateral Perturbation-Induced and Voluntary Stepping in Fallers and Nonfallers After Stroke

Abstract

Objective

A loss of balance poststroke from externally induced perturbations or during voluntary movements is often recovered by stepping. The purpose of this study was to characterize stepping behavior during lateral induced waist-pull perturbations and voluntary steps in community-dwelling fallers and nonfallers with chronic stroke.

Methods

This study used a cohort design. Thirty participants >6 months poststroke were exposed to 24 externally triggered lateral waist-pull perturbations and 20 voluntary steps. Balance tolerance limit (BTL) (transition from single to multiple steps) and first step type were determined for the waist-pull perturbations. Step parameters of initiation time, velocity, first step length, and clearance were calculated at and above BTL and for the voluntary steps. Hip abductor/adductor torque, foot cutaneous sensation, and self-reported falls that occurred 6 months prior were evaluated.

Results

Twelve participants were classified retrospectively as fallers and 18 as nonfallers. Fallers had a reduced BTL and took more medial first steps than nonfallers. Above BTL, no between-group differences were found in medial steps. At BTL, the nonparetic step clearance was reduced in fallers. Above BTL, fallers took longer to initiate a paretic and nonparetic step and had a reduced nonparetic step length and clearance compared with nonfallers. There was a between-group difference in step initiation time for voluntary stepping with the paretic leg (P < .05). Fallers had a reduced paretic abductor torque and impaired paretic foot cutaneous sensation.

Conclusion

A high fall rate poststroke necessitates effective fall prevention strategies. Given that more differences were found during perturbation-induced stepping between fallers and nonfallers, further research assessing perturbation-induced training on reducing falls is needed.

Impact

Falls assessments should include both externally induced perturbations along with voluntary movements in determining the fall risk.

Falls, a common secondary complication after a stroke,¹ can result in a fracture,² increased fear of falling,³ and more sedentary behavior.³ The combined residual sensorimotor deficits following a stroke^4,5 and reduced participation and activity limitations after a fall can lead to poor balance function and increased fall risk in stroke.⁶ Balance function is well characterized when a small perturbation disrupts standing balance and the feet remain in place after a stroke.⁷ Muscle activation patterns in stroke are characterized by poor coordination among the lower extremity muscles as well as prolonged time to initiate the muscle activity during internal and external perturbations to standing balance that require postural adjustments or protective stepping.^8,9 However, perturbations that involve a change in the base-of-support configuration, such as protective stepping to preserve balance stability, are more likely to produce a fall.⁹ Understanding the balance recovery strategies used after stroke are central to the development of interventions that target fall prevention in this population.

Protective stepping is a common balance recovery strategy used when balance is disrupted by an internal perturbation or due to external forces that act on the body. The number of falls from either type of perturbation is equivalent after a stroke.¹⁰ Steps evoked during voluntary movements after a stroke are initiated and executed more slowly with a reduced and delayed onset compared with healthy controls.^11–13 Stepping induced from external perturbations to balance are initiated more slowly, with a reduced step length and reduced foot clearance in those with stroke compared with healthy people.^12,14–18 Individuals after stroke are also more likely to initiate a step at a lower perturbation magnitude than age-matched controls.⁹ More importantly, the first step is initiated more frequently with the nonparetic than the paretic leg.^9,18,19 The limited use of the paretic limb could potentially increase risk of falling, especially if the nonparetic limb is obstructed and unable to initiate the first step.²⁰ Therefore, stepping in response to both types of perturbations is impaired and of equal importance. The step characteristics may be useful in distinguishing those at a greater risk of falling from nonfallers.

While the majority of previous studies disrupted standing balance in the anterior-posterior direction, there is less information about balance recovery in the medio-lateral (M-L) direction after a stroke. The unilateral sensorimotor deficits²¹ in this population increase the risk of falling in the frontal plane.⁹ The unilateral deficits from the stroke increase fall risk through unequal weight-bearing^22,23 and reduced strength in paretic muscles,^9,24 which are important for M-L stability. To further address the unilateral deficits that may impair M-L balance, the primary aim of this study was to identify the stepping characteristics that distinguished fallers from nonfallers during lateral externally induced waist-pull perturbations and rapid voluntary steps in community-dwelling individuals with chronic hemiparetic stroke.

Methods

Thirty community-dwelling adults with hemiparesis from a stroke participated in the study. Participants included in the study were >6 months post-stroke, ≥50 years of age, could walk 10 m with/without an assistive device, could stand unsupported for 5 minutes, and had no medical or neurological condition that significantly impacted their ability to walk beyond the effects of the stroke. All participants gave informed consent to participate in the study, and the University of Maryland, Baltimore Institutional Review Board approved the study protocol.

All participants attended 1 laboratory session and performed (1) externally induced steps generated by a lateral waist-pull perturbation device controlled through a motor-driven system,²⁵ and (2) visually cued lateral voluntary steps initiated under a choice reaction-time condition.²⁶ The participants received a total of 24 lateral waist-pull perturbations and 3 trials at 4 perturbation magnitudes toward the paretic and nonparetic side. The instructions to the participants before the lateral waist-pull perturbation were to “respond naturally and prevent yourself from falling.” The perturbation magnitudes had a fixed acceleration of 720 cm/s², and the velocity (v) and displacement (d) were as follows: Level 1 v = 18.0 cm/s, d = 8.6 cm; Level 2 v = 27.0 cm/s, d = 12.1 cm; Level 3 v = 36.0 cm/s, d = 15.7 cm; and Level 4 v = 45 cm/s, d = 19.3 cm. The lateral waist-pull system used previously in studies of older adults^27,28 and stroke¹² was set at perturbation magnitudes that are known to induce stepping.²⁸ There is a unique biomechanical feature of the lateral waist-pull perturbation to standing balance, as the body’s center of mass was passively moved in the direction of the perturbation, the leg opposite to the direction of the waist-pull was passively unloaded.²⁹ The balance tolerance limit (BTL) magnitude was determined for the external waist-pull perturbation. BTL was defined as the perturbation magnitude level whereby the induced balance recovery transitioned from a single step to a multiple-step pattern. At BTL, the first step was classified into 1 of 3 step types: (1) a lateral step was a passively loaded leg that stepped sideways in the direction of the waist-pull perturbation; (2) a crossover step was a passively unloaded leg that stepped toward and beyond the loaded leg, either in front or behind the body; or (3) a medial-step was an unloaded leg that stepped toward but not beyond the loaded leg, which always resulted in a lateral step with the opposite limb.²⁹ The lateral voluntary step under a choice reaction condition was chosen because the onset times of the muscles that control ML stability of a rapid voluntary step were a predictor of falls in older adults.³⁰ For the lateral voluntary step, a light cue panel was positioned directly in front of the participant at eye level. Participants were instructed to take a lateral step in the same direction as the light cue. For example, if the light came on the right side of the panel, they would initiate a side step with the right leg. The direction of the light cue, randomly presented, indicated when to initiate the step and the leg used to step. The instructions given were “to step as fast as possible when you see the light,” with no instructions given on the number of steps that they were to take. The purpose was to ensure that the participants would attempt to step as fast as possible without limiting their recovery to a single step. The lateral voluntary steps consisted of 10 trials (5 paretic, 5 nonparetic). The direction of each trial of the waist-pull perturbation and voluntary steps was randomized to reduce the anticipation and learning effects. All participants first performed the waist-pull perturbation trials, followed by the voluntary steps.

Participants wore comfortable shoes and stood in a self-selected position on 2 adjacent force platforms (Advanced Mechanical Technology Inc., Watertown, MA) during the stepping trials. For each trial, the feet were placed in the same position on the force platforms by tracing the outline of the feet onto a paper taped over the platforms to ensure the same position at the start of each trial. A harness fitted on each participant was used for safety and did not provide support unless the participant was unable to recover their balance. The sampling rate for the ground reaction forces was 600 Hz. The ground reaction forces were monitored visually, preceding the start of each trial to encourage symmetrical weight-bearing. A reflective marker was affixed on the lateral malleoli and recorded for 7 seconds per trial at a sampling rate of 120 Hz using a 10-camera motion analysis system (Vicon, Oxford, UK).

Peak isokinetic joint torques of the nonparetic and paretic side were measured in 5 trials at 30°/s using the Biodex System Pro4 (Biodex Medical Systems, New York, NY) for hip abduction and adduction. The tests were performed in side-lying. The participants completed clinical assessments of balance and mobility, balance confidence, motor recovery, and cutaneous sensation. The Community Balance and Mobility, a validated measure used in stroke,³¹ assessed high-level balance and mobility required by individuals living in the community.³² The activity-specific balance confidence assessed balance confidence and, when used in community-dwelling individuals after stroke,³³ remained lower 1 year following a stroke compared with healthy controls.³⁴ Motor recovery was evaluated with the subscales of the leg and foot of the Chedoke-McMaster Stroke Assessment (CMSA) Impairment Inventory³⁵ and graded on the stages of motor recovery ranging from 1 to 7, with 7 classified as normal and 1 as being flaccid. The cutaneous sensation of the plantar aspect of the foot was measured with a series of Semmes-Weinstein monofilaments ranging from 1.65 to 6.65. The lowest value reflected normal cutaneous sensation.³⁶ Adequate somatosensation is a key factor for maintaining upright stance³⁷ and withstanding perturbations to standing balance.³⁸ Each participant reported the number of falls in the 6 months before enrollment. A fall was defined as an event that resulted in a person coming to rest inadvertently on the ground or floor or lower level.³⁹

Data Analysis

Participants who reported 1 or more falls in the prior 6 months were classified as a faller, and those who did not report a fall were classified as a nonfaller. For the waist-pull perturbation-induced steps, the first step type (lateral, crossover, medial) and leg used (paretic vs nonparetic) were reported as a frequency, calculated as a percent of all trials. The first step characteristics of step initiation time, step length, step clearance, and step velocity were determined at BTL and 1 level above BTL for the waist-pull perturbation-induced steps and for all voluntary step trials. The perturbation-induced step and voluntary step parameters were calculated with customized MATLAB programs. Independent variables included (1) step initiation time, defined as the interval of time between the instant of the waist-pull perturbation or light-cue onset and first step liftoff; (2) first step length, defined as the maximal foot displacement at first ground contact in the combined anterior/posterior and medial/lateral direction⁴⁰; (3) first step clearance, defined as the maximum vertical displacement; and (4) step velocity, defined as the step length divided by the step duration. The first step length and step clearance were expressed as a percentage of body height. Peak isokinetic torque was defined as a deficit ratio relative to the nonparetic leg (paretic peak torque/nonparetic torque).

Statistical Analyses

SPSS for Windows v22.0 (IBM Company, Chicago, IL) was used for data analyses. The between-group differences (faller and nonfaller) were analyzed with parametric and non-parametric independent t tests and chi-square test (χ²) for demographics and clinical outcome measures. For the induced waist-pull perturbation, the between-group difference in BTL was compared using a Mann-Whitney U test. Between-group differences (faller, nonfaller) in the frequency of nonparetic and paretic leg first step responses and step type (medial, lateral, and crossover) during the induced waist-pull perturbation were tested with a χ² at BTL. The χ² method used contingency tables to compare the categorical variables.⁴¹ A multivariate analysis of variance (MANOVA) was used since the Pearson’s correlation analyses across all participants showed that the step characteristics of step length, step clearance, and step velocity were related parameters, but not redundant, with values ranging from 0.47 to 0.86. A MANOVA was performed for step initiation time, step length, step clearance, and step velocity, with group (faller and nonfaller) as the independent variable. A MANOVA was carried out for the paretic leg and nonparetic leg separately at BTL and above BTL and for the voluntary step. Additional analyses were also performed with step type as a covariate in the MANOVA since there was a significant difference in step type between fallers and nonfallers during the induced waist-pull perturbation. The step type as a covariate was coded from the smallest step to the largest step based on the spatiotemporal parameters. For example, the medial step had the smallest spatiotemporal characteristics due to the constraints of space when moved towards the stance leg and was coded as 1. Thus, based on this principle, lateral steps were coded as 2 and crossover steps were coded as 3. The values presented are means and SDs, and P values of 0.05 or less indicate statistical significance.

Role of the Funding Source

The contents of this publication do not necessarily represent the policy of the funders of this study (National Institute on Disability, Independent Living, and Rehabilitation Research; Administration for Community Living; Department of Health and Human Services). Endorsement by the federal government should not be assumed. The funders played no role in the design, conduct, or reporting of this study.

Results

Twelve participants were classified as fallers and 18 as nonfallers. Six of the fallers had more than 1 fall, while the remaining 6 reported 1 fall. The demographics and clinical outcomes measures of the faller and nonfaller groups are presented in Table 1. The faller group had significantly lower scores on the community balance and mobility (P < .05) and CMSA (combined subscale of the foot and leg) (P < .05) compared with the nonfaller group. Fallers had significant reductions in the paretic hip abductor torque deficit (P < .01) and cutaneous sensation of the plantar surface of the paretic foot (P < .05) compared with nonfallers.

Table 1.

Characteristics of the Study Participants

Characteristic	Fallers (n = 12)	Nonfallers (n = 18)	P
Age (y)	60.6 (6.2)	62.7 (7.7)	.46
Sex (% female)	63.6%	57.9%	.68
Side of paresis (%)	36.4% Right	27.8% Right	.63
Time post stroke, y (range)	5.4 (0.7–12.1)	10.1 (1.8–48.2)	.16
Community Balance & Mobility Scale	24.8 (10.6)	35.4 (16.4)	< .05a
Activities-Specific Balance Confidence Scale	73.7 (11.2)	77.1 (12.3)	.45
Chedoke McMaster Stroke Assessment—leg + foot	7.0 (2.1)	9.1 (2.7)	< .05a
Plantar cutaneous sensation, median interquartile range	4.31 (3.42–5.91)	3.7 (3.61–4.31)	< .05a
Hip abductor torque deficit (%)	0.50 (0.29)	0.80 (0.19)	< .01a

^a P ≤ .05.

Induced Waist-Pull Perturbation Steps

There was a significant difference in the BTL, where fallers transitioned to a multiple step recovery at a lower BTL than nonfallers (P < .01) (Tab. 2).

Table 2.

Perturbation-Induced BTL and Step Type at and Above BTL for Fallers and Nonfallers^a

	Fallers b	Nonfallers
Perturbation magnitude (% of participants at each magnitude)
Magnitude 1	46.9%	19.8%
Magnitude 2	25.0%	33.0%
Magnitude 3	15.6%	30.8%
Magnitude 4	12.5%	16.4%
Step type, at BTL (% of all trials)
Medial	64.1%	47.3%c
Lateral	25.0%	25.3%
Crossover	10.9%	27.5%c
Step type, above BTL (% of all trials)
Medial	55.9%	60.0%
Lateral	32.2%	11.0%c
Crossover	11.9%	29.0%c

^a BTL = balance tolerance limit.

^b P < .05 fallers significantly different from nonfallers using Mann-Whitney U test.

^c P < .05 between-group differences between fallers and nonfallers using chi-square test.

Induced waist-pull perturbation at BTL

There was a between-group difference in the leg used to initiate the first step. Fallers used the paretic leg as the first step (46.1%) more often than nonfallers (29.7%; P < .05). There was also a significant difference in the step type used to recover balance. Fallers took a crossover step less often (10.9%) and used a medial step more frequently (64.1%) than nonfallers (crossover 27.5%, P < .05; medial 47.3%, P < .05). No significant difference was found between groups in the lateral steps taken (Tab. 2). After adjusting for step type, the reduced nonparetic step clearance (P < .05) in fallers was the only step characteristic that was significantly different from nonfallers (Fig. 1).

Figure 1.

First step characteristics, step onset, step velocity, step length, and clearance of the lateral waist-pull–induced perturbation of fallers (gray bars) and nonfallers (black bars) at balance tolerance limit, expressed as means and standard error. ^*P < .05 significant group differences in fallers versus nonfallers after adjusting for step type.

Induced waist-pull perturbation above BTL

The number of medial steps used by the faller and nonfaller group was not significantly different above BTL (P = .62). Fallers took more lateral steps (P < .01) and fewer crossover steps (P < .05) than nonfallers (Tab. 2). After adjusting for step type, fallers took longer to initiate a step with the paretic (P < .05) and nonparetic leg (P < .05) compared with nonfallers (Fig. 2). When fallers took a step with the nonparetic leg, they had a reduced step length (P < .01) and step clearance (P < .01) compared with nonfallers (Fig. 3).

Figure 2.

Representative single trials of a faller (dashed line) and nonfaller (solid line) of the stepping foot vertical displacement (step clearance) and step initiation time during the lateral waist-pull perturbation (top panel) above BTL of the paretic (A) and nonparetic leg (B) and of a voluntary step (bottom panel) of the paretic (C) and nonparetic leg (D). Arrows represent the onset of the step initiation time for the respective group.

Figure 3.

First step characteristics, step onset, step velocity, step length, and clearance of the lateral waist-pull–induced perturbation above balance tolerance limit of fallers (gray bars) and nonfallers (black bars), expressed as means and standard error. ^*P < .05 significant group differences in fallers vs nonfallers after adjusting for step type.

Lateral Voluntary Steps

The fallers took a longer time to initiate a lateral voluntary step with the paretic limb than nonfallers (P < .05) (Figs. 2 and 4). There were no significant differences found in the step velocity, step length, or step clearance between fallers and nonfallers.

Figure 4.

First step characteristics, step onset, step velocity, step length, and clearance of the paretic and nonparetic leg of the lateral voluntary step of fallers (gray bars) and nonfallers (black bars) expressed as means and standard error. ^*P < .05 significant between-group differences in fallers and nonfallers.

Discussion

We investigated the spatiotemporal step characteristics of a lateral induced waist-pull perturbation and a choice reaction voluntary step in a group of self-reported fallers and nonfallers after stroke. The faller group transitioned from a single to multiple-step recovery at a lower BTL. However, at BTL, we found no between-group differences in the spatiotemporal characteristics of steps with the paretic leg, even though fallers used the paretic leg as the first recovery step more frequently than nonfallers. It took longer for fallers to initiate a step with the paretic and nonparetic leg above BTL. In contrast to the nonfallers, the faller group took a smaller first step that had a reduced step clearance when they stepped with the nonparetic leg at BTL and above BTL. The differences in the nonparetic step characteristics may indicate an unwillingness or inability to bear weight on the paretic limb for an extended period of time. Moreover, the voluntary step yielded fewer between-group differences, with fallers only taking longer to initiate a step with the paretic leg compared with nonfallers. The common finding for the 2 different forms of stepping is that the fallers are slow to initiate a step compared with nonfallers, which could impair their ability to avert a fall.

Similar to other studies in healthy older adult fallers^28,42 and stroke,⁹ we found that fallers had a lower balance tolerance threshold. The reason for the difference in BTL may be that individuals after stroke have a reduced margin of stability⁴³ and asymmetrical weight-bearing favoring the nonparetic leg.^22,23 Therefore, a relatively minor perturbation could potentially cause unsteadiness, particularly when perturbed towards the paretic side.⁴⁴ It was challenging for fallers to initiate a step rapidly at a larger perturbation, as indicated by the delayed step initiation time of the paretic and nonparetic leg above BTL. Strength deficits in the paretic hip abductor muscles that we found and are described by others^45,46 may limit the capacity to generate fast coordinated actions important for lateral balance stability. These fast coordinated actions are essential for larger perturbation magnitudes, as the hip abductor muscles usually play a prominent role in the control of mediolateral stability during standing, stepping, and gait. Delayed and reduced responses in the muscle activity in the hip abductor muscles are associated with a lower balance threshold in stroke.⁹ Poor neuromuscular coordination would increase imbalance and likely cause the body’s center of mass to move a greater distance and thereby diminish the likelihood of preserving or regaining balance stability through stepping.^47,48

We found that fallers used medial steps more often at BTL. The occurrence of medial steps in response to induced perturbations in stroke has been reported to be approximately one-quarter of all recovery steps.⁴⁴ The amount is less than the number of medial steps we found in our study. The diminished cutaneous sensation of the paretic foot could be a plausible explanation for the higher incidence of medial steps in the present study. Similar findings reported in healthy individuals, in which foot cutaneous sensation was artificially reduced by hypothermic anesthesia, resulted in a transition from a lateral step to a medial or crossover step.⁴⁹ Cutaneous afferent information from the mechanoreceptors of the feet significantly contributes to standing balance control.⁵⁰ Reduced sensory input may necessitate compensatory strategies, such as a medial step, due to the detection errors in the center of pressure in relation to the position and motion of the body’s center of mass, although, above BTL, the increase in medial steps in the nonfaller group resulted in no between-group differences. It may be more challenging at larger perturbation magnitudes to produce fast, coordinated motor responses necessitating changes in stepping strategies. Based on the CMSA of the foot and leg, none of the participants had a score that would indicate the capacity for achieving rapid coordinated movements. Previous studies have demonstrated an association between reduced motor recovery and deficits in functional tasks, such as temporal asymmetry during gait and decreased stability during sit to stand transfers and squats.^22,51,52 It is not clear what benefits are provided by using a medial first step, but it likely reflects the constraints on the control of stepping due to the sensory and motor impairments after a stroke. However, the medial step is the least stable balance recovery response among the different types of stepping, as that likely exacerbates the risk of falling, as indicated for older adults.^29,53

During voluntary stepping, fallers took longer to initiate a step with the paretic leg. The prolonged time to initiate a movement and reduced execution speed during the voluntary movements are typical after stroke.^54,55 However, the step execution speed was not different between the 2 groups, which may indicate that the initiation impairment reflects the difficulty in developing a motor plan or in motor preparation rather than the actual performance of the step. Similar findings reported in the upper extremity showed planned movements measured by movement-related cortical potentials took longer to develop with the paretic hand compared with the nonparetic hand.^56,57 Thus, training by emphasizing faster movements may help improve balance recovery and reduce the risk of falling. For example, training rapid movements increased movement speed, and the training improved standing balance control during an internal perturbation in stroke.^55,58 Benefits of exercises that emphasize speed are reported in older adults with improved muscle performance and balance control.^59,60 Our findings from the voluntary stepping task may indicate that when given sufficient time to develop a motor plan, fallers can generate a motor output that is appropriate for the necessary response. Because perturbation-induced stepping is initiated faster than a voluntary choice reaction stepping task in stroke,²⁶ induced step training may be an effective method to facilitate faster movements. Alternatively, in those individuals who never took a perturbation-induced step with the paretic limb, practicing fast voluntary stepping may be useful for facilitating a quicker induced response with the paretic limb.

Among the limitations of this study was a relatively small sample. The study results are only generalizable to community-dwelling ambulatory individuals with chronic stroke. Also, the replication of falls in a laboratory setting differs from naturally occurring falls in the real-world environment that are unexpected, although the control of the relationship between the center of mass and the base of support when stepping is a central mechanism of balance control in real-life situations as well as in the laboratory setting. Therefore, it is likely the differences in balance recovery during the externally induced perturbation between fallers and nonfallers existed in their control of the stepping rather than in the mechanism of the perturbation.

Furthermore, a single fall may be a rare or chance occurrence that is not indicative of a serious risk of falling. Thus, some individuals in our study may have erroneously been classified into the incorrect group. Further, self-reported falls may be underreported in our sample due to forgetting about the event, denial that they experienced a fall, or falls are less likely to be remembered unless an injury occurred.⁶¹ However, one-half of our fallers were recurrent fallers, and past falls are associated with future falls in stroke.⁶² Our data support significant differences between self-reported fallers and nonfallers that may put them at risk for falling.

In summary, compared with nonfallers, individuals with chronic stroke who reported a fall in the prior 6 months exhibited greater strength deficits in hip abductor muscles, which are important for the control of mediolateral weight transfer, balance stability, and sidestepping. The fallers also had diminished cutaneous sensation of the paretic foot. These impairments may contribute to the different stepping behaviors related to fall status. However, individuals successfully initiated stepping when perturbed by the lateral waist-pull and, when sufficient time was allowed, initiated voluntary stepping in a choice reaction time condition. Thus, the prolonged time needed to prepare a stepping response to a loss of balance places individuals after a stroke at greater risk for falls, especially when balance is externally disrupted. The risk of falling due to response slowing may decrease with a smaller perturbation magnitude that allows sufficient time to execute effective stepping and with a longer time to prepare for a voluntary step. The current clinical practice in rehabilitation focuses mainly on training balance function involving predictable and known disturbances to balance that occur during instructed voluntary functional tasks. Further research is also warranted to determine whether interventions that emphasize both time-sensitive perturbation-induced and voluntary stepping can improve the capacity to move quickly enough to recover balance and successfully reduce the incidences of falls.

Author Contributions

Concept/idea/research design: V.L. Gray, M. Fujimoto, M.W. Rogers

Writing: V.L. Gray, M. Fujimoto, M.W. Rogers

Data collection: V.L. Gray, M. Fujimoto

Data analysis: V.L. Gray, M. Fujimoto

Project management: V.L. Gray

Consultation (including review of manuscript before submitting): M. Fujimoto

Ethics Approval

This study protocol was approved by the University of Maryland, Baltimore Institutional Review Board. All participants gave informed consent to participate in the study.

Funding

This study was developed under a grant from the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR) (H133P100014, H133F140027). NIDILRR is a center within the Administration for Community Living (ACL), Department of Health and Human Services (HHS). The contents of this publication do not necessarily represent the policy of NIDILRR, ACL, HHS, and you should not assume endorsement by the federal government. This study was also supported by the American Heart Association (14CRP19880025) and The National Institute on Aging Claude D. Pepper Older Americans Independence Center (P30-AG028747).

Disclosures

The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest.