Does Falls Efficacy Influence the Relationship Between Forward and Backward Walking Speed After Stroke?

Abstract

Objective

Forward walking speed (FWS) is known to be an important predictor of mobility, falls, and falls-related efficacy poststroke. However, backward walking speed (BWS) is emerging as an assessment tool to reveal mobility deficits in people poststroke that may not be apparent with FWS alone. Since backward walking is more challenging than forward walking, falls efficacy may play a role in the relationship between one’s preferred FWS and BWS. We tested the hypothesis that people with lower falls efficacy would have a stronger positive relationship between FWS and BWS than those with higher falls efficacy.

Methods

Forty-five individuals (12.9 ± 5.6 months poststroke) participated in this observational study. We assessed FWS with the 10-meter walk test and BWS with the 3-meter backward walk test. The modified Falls-Efficacy Scale (mFES) quantified falls efficacy. A moderated regression analysis examined the hypothesis.

Results

FWS was positively associated with BWS (R² = 0.26). The addition of the interaction term FWS × mFES explained 7.6% additional variance in BWS. As hypothesized, analysis of the interaction revealed that people with lower falls efficacy (mFES ≤ 6.6) had a significantly positive relationship between their preferred FWS and BWS, whereas people with higher falls efficacy (mFES > 6.6) had no relationship between their walking speed in the 2 directions.

Conclusions

FWS is positively related to BWS poststroke, but this relationship is influenced by one’s perceived falls efficacy. Our results suggest that BWS can be predicted from FWS in people with lower falls efficacy, but as falls efficacy increases, BWS becomes a separate and unassociated construct from FWS.

Impact

This study provides unique evidence that the degree of falls efficacy significantly influences the relationship between FWS and BWS poststroke. Physical therapists should examine both FWS and BWS in people with higher falls efficacy, but further investigation is warranted for those with lower falls efficacy.

Introduction

Stroke is the number one cause of long-term disability in the United States.¹ Nearly 7.2 million Americans were reported to be living with stroke between 2011 and 2014.² Approximately 73% of people poststroke have limited mobility³ and fall within the first 6 months,⁴ leading to additional fear of falling⁵ and subsequent limitation of activities,⁶ community participation,⁷ and reduced quality of life.⁸ Fear of falling is defined as “low perceived falls efficacy or confidence at avoiding falls.”⁹ Fall-related efficacy is a valid and important variable to measure and address in poststroke rehabilitation secondary to its relationship to physical activity, mobility, and falls.^5,10,11 Both forward walking speed (FWS) and fear of falling are reported to be strong predictors of falls in individuals poststroke.^12,13 FWS is the most frequently measured clinical outcome in hemiparetic gait assessment^14,15; however, a paradigm shift has recently occurred in gait rehabilitation and assessment, whereby backward walking speed (BWS) has emerged as a stronger predictor of falls in the elderly^16,17 and health care providers are being encouraged to assess BWS in older adults^16,17 as well as in people poststroke.¹⁸

Backward walking is an essential walking adaptability task,¹⁹ especially when one needs to negotiate tight spaces like backing up to open a door, completing household chores such as sweeping or vacuuming, stepping back in a restroom to sit down on a toilet seat, or while turning.¹⁷ Many times one also needs to take a step back to prevent a fall as a stepping strategy. In fact, falls occur not only while walking forward, but also while taking steps backward in such tight spaces or when avoiding a sudden obstacle. Moreover, in a cohort of 62 elderly individuals, BWS accurately identified fallers, more so than FWS.¹⁷ Specifically, BWS of less than 0.6 m/s identified 100% of the fallers whereas a FWS of less than 1.0 m/s identified only 83% of elderly fallers.¹⁷ In a more recent study by Carter et al, a cutoff of 3.5 seconds to complete the 3-meter Backward Walk Test (3MBWT) was concluded to be better than the 8.0 or 13.5-second cutoff to complete the forward Timed “Up and Go” test in the prediction of falls in community-dwelling elderly individuals.¹⁶ Spatiotemporal characteristics of backward gait are also known to decline more rapidly than forward-gait characteristics with increasing age.¹⁷ Thus, as backward walking is a crucial aspect of mobility in everyday life, assessment of backward walking in the prediction of falls and mobility deficits has become increasingly important in older adults^16,17 and in people with neurological disorders such as stroke¹⁸ and Parkinson disease.²⁰

Backward walking is generally a novel task even for healthy men and women. It is more challenging than forward walking because it requires greater muscle activity²¹ and occurs with a greater metabolic cost.²² Backward walking demands additional reliance on proprioception because of a lack of peripheral visual feedback of people’s footfalls as well as of the ground conditions behind them.²³ Unlike forward walking with eyes closed, backward walking is able to elicit this increased reliance on proprioception even with one’s eyes open. Similar to walking forward with obstacle negotiation, backward walking involves increased cognitive demand and increased prefrontal activation.²⁴ While it may be argued that forward walking with eyes closed or obstacle crossing present similar challenges to mobility as backward walking, in a recent study with 61 individuals poststroke, backward walking emerged as the most difficult task followed by obstacle negotiation during forward walking and forward walking alone.²⁵ Additionally, backward walking along with the cognitive challenge of serial-7 subtractions proved to be more challenging than forward obstacle crossing with similar cognitive tasks in this study.²⁵ Hawkins et al suggest that backward walking can reveal kinematic and spatiotemporal walking deficits that may not be apparent while assessing forward gait in community ambulators poststroke.¹⁸

People poststroke with higher perceived falls efficacy generally walk faster than those with lower perceived falls efficacy in the forward direction poststroke.^26,27 However, even those individuals who walk faster than 0.8 m/s and demonstrate the capacity for unlimited community ambulation poststroke take slower, more cautious steps while walking backward as compared to healthy individuals.¹⁸ During backward walking, these community ambulators poststroke tend to reduce their gait speed, take shorter steps with increased cadence, and increase their double support time, but have similar forward gait characteristics when compared to healthy controls.¹⁸ Similarly, as compared to elderly nonfallers, elderly fallers demonstrated increased double support time, shorter step length, and a wider base of support during backward walking assessment, but not during forward walking assessment.¹⁷ Thus, backward walking assessment elicits a cautious gait pattern and stability threat even in those individuals poststroke who walk faster than 0.8 m/s,¹⁸ and probably demonstrate high falls efficacy.²⁶ A few individuals poststroke with high falls efficacy may still be at risk of falling,²⁸ which may not be apparent with their forward gait speed. It is plausible that assessment of BWS may reveal mobility deficits in these individuals with higher falls efficacy. While it may be hypothesized that lower falls efficacy will lead to a slower BWS, it unknown if BWS increases linearly with increased falls efficacy poststroke. Preferred FWS is highly correlated with preferred BWS in healthy young adults.²⁹ It is not known if BWS can be predicted from preferred FWS in people poststroke and how this relationship may differ in accordance with one’s falls efficacy. Thus, it is essential to analyze the effect of this psychological construct on the relationship between FWS and BWS.

The purpose of this study was to examine if the relationship between preferred FWS and BWS differs according to the degree of falls efficacy in people poststroke. We hypothesized that FWS will be positively associated with BWS; however, this relationship will be moderated by one’s falls efficacy. Lower falls efficacy will lead to stronger positive relationship between FWS and BWS, whereas for people with higher falls efficacy, the relationship between FWS and BWS will not be as strong.

Methods

Participants

Data for this cross-sectional study were drawn from baseline measurements from 2 larger clinical trials of walking interventions aimed to aid recovery of locomotor deficits poststroke. Individuals with chronic poststroke hemiparesis (6 months to 5 years poststroke) were recruited from the community between January 2016 and October 2019. Individuals were included if they were age 18 to 85 years and could walk at least 3 meters with no more than maximum one-person assist. We excluded those who had other neurological conditions or presented with serious cardiac conditions or orthopedic diagnoses. The institutional review board at the University of Florida approved this study and all participants signed a written informed consent to participate. This cross-sectional study follows the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines.

Procedure

To characterize the stroke-related severity of motor impairments, we administered the Fugl-Meyer Assessment of Motor Recovery for the Lower Extremity (FMA-LE). This scale measures the ability to move the lower extremity selectively, outside abnormal synergistic patterns and possesses good validity and reliability.^30,31 We recorded the use of assistive devices in daily life through self-report. Other participant characteristics such as stroke onset date, stroke type, lesion location, and side of hemiparesis were obtained from medical records.

To address the aims of this study, we assessed FWS with the 10-meter walk test (10MWT).³² Participants walked over a 14-meter walkway for 2 trials at their self-selected speed and time was recorded for the middle 10 meters. The 3MBWT quantified BWS.¹⁶ The 3MBWT has been validated for its accuracy to detect falls in elderly individuals.¹⁶ In this study, participants traversed a walkway of 5 meters backward at their comfortable speed for 1 practice trial followed by 2 test trials. Time was recorded for the middle 3 meters. The longer walkways allowed for acceleration and deceleration speeds as participants initiated and terminated their gait. Time was recorded with a stopwatch. For both tests, direction-specific gait speed (m/s) was averaged across the 2 trials. Participants used an assistive device or a brace during the walking tasks if that was customary for their everyday ambulation. The side on which the assistance device was held (typically the nonparetic side) remained constant during both forward and backward walking tasks. For participants who required one-person assistance to successfully accomplish the walking tasks without falling, a physical therapist provided manual contact guard assistance. The contact guard assistance ranged from intermittent touch to continuous manual contact, as deemed necessary to assist with balance during walking, but did not alter the participant’s natural gait speed. For all other participants, the physical therapist provided supervision during the walking tasks, as necessary. Additionally, the physical therapist refrained from providing any verbal encouragement or guidance during both walking tasks to ensure that these did not alter the participants’ natural gait speed.

The modified Falls-Efficacy Scale (mFES) was administered to assess falls efficacy in performing basic and instrumented activities of daily living without falling. This 14-item questionnaire assesses the stem question “How confident are you that you can do the following activities without falling?” on a scale of 0 (not confident at all) to 10 (completely confident). The mFES has demonstrated excellent test-retest reliability of 0.97 in people poststroke.³³ An average of the 14 items was calculated as the final score in which higher values depicted higher falls efficacy.

Statistical Analysis

To assess the role of falls efficacy as a moderator of the relationship between FWS and BWS, we conducted a hierarchical moderated regression. We tested the normality of the saved standardized residuals with the Kolmogorov-Smirnov and Shapiro-Wilk tests to determine if general linear model assumptions were met.³⁴ We examined the bivariate correlations between the independent variables as well as their variance inflation factors to test for the problem of multicollinearity. As expected, the interaction term of mFES × FWS was highly correlated with FWS (r² = 0.83, Tab. 1) and the variance inflation factors of FWS in the final model was 4.³⁵ Iacobucci et al recommend mean centering of individual predictors when one wants to test the effect of an interaction on a dependent variable, above and beyond individual predictors, without the interference of nonessential or micromulticollinearity (that is the strong correlation between FWS and mFES × FWS).³⁶ Accordingly, we centered the FWS and mFES variables using the Aiken and West centering approach,³⁷ and computed the product term between FWS and mFES using the centered variables. Iacobucci et al advise that mean centering helps to reduce the micromulticollinearity without affecting the overall model fit (the R² or the macromulticollinearity).³⁶ After ensuring normality of residuals and no multicollinearity issues between variables post-centering, we progressed with the analysis, with a P less than or equal to .05 significance level.

Table 1.

Bivariate Relationship Between Falls Efficacy, Forward and Backward Walking Gait Speed in All Participants^a

Assessments	mFES	FWS
FWS	0.617b
BWS	0.461b	0.513b
FWS X mFES	0.826b	0.476b

^a BWS = Backward Walking Speed; FMA-LE = Fugl-Meyer Assessment of Motor Recovery after Stroke—Lower Extremity; FWS = Forward Walking Speed.; mFES = Modified Falls-Efficacy Scale.

^b P less than .001.

Block 1 of the moderated regression model consisted of FWS alone predicting the variance explained in BWS. We then added the moderator mFES to block 2 to assess if the moderator itself explained any additional variance in BWS, above and beyond FWS. Lastly, to test the effect of mFES on the relationship between FWS and BWS, the interaction or the product term of mFES and FWS was added in block 3. Additional significance was determined by significant F change and R² change in the 3 blocks. All analyses was conducted using Statistical Product and Service Solutions (SPSS), version 25, New York.

To interpret the interaction, we used both the simple slopes approach³⁷ and the Johnson-Neyman approach with the CAHOST Microsoft 2013 Excel worksheet created in 2017 by Carden et al.³⁸ The simple slopes approach tests the significance of the interaction at 3 levels of the moderator: average, above 1 SD, and below 1 SD, essentially categorizing a continuous variable into 3 levels. Alternatively, the Johnson-Neyman technique retains the properties of a continuous variable and provides a range for which the moderator has a significant interaction effect on the dependent variable. Carden et al argue that the Johnson-Neyman technique is a more robust technique to explore the effect of an interaction in a moderated regression when compared to the simple slopes approach, especially for continuous variables.³⁸

Role of the Funding Source

The funders played no role in the design, conduct, or reporting of this study.

Results

Forty-five adults (17 females), age 58.9 ± 10.01 years, and 12.9 ± 5.6 months poststroke, participated in this study. Table 2 illustrates the participants’ demographic and clinical characteristics. To ensure safety against falling during the 10MWT, 3 participants required continuous contact guard assistance and 8 required intermittent, light-touch contact guard assistance. During the 3MBWT, 7 participants required continuous contact guard assistance, and 7 required intermittent, light-touch contact guard assistance to assist with balance and to prevent a fall. All other participants were provided supervision only during both the 10MWT and 3MBWT.

Table 2.

Participant Characteristics^a,b

Characteristic	Mean (SD) (Range)
Mean age, y	58.9 (10.01) (31–83)
Sex	17 F/28 M
Mean time since stroke, mo	12.9 (5.6) (6–24)
Stroke type	Ischemic: 36 (80%)
	Hemorrhagic: 7 (15.5%)
	Ischemic and hemorrhagic: 1 (2.2%)
	Unknown: 1 (2.2%)
Lesion location	Cortical: 21 (46.7%)
	Subcortical: 7 (15.5%)
	Mixed: 7 (15.5%)
	White matter:c 5 (11.1%)
	Brainstem: 4 (8.9%)
	Unknown: 1 (2.2%)
Mean FMA-LE (range, 0–34)	23.24 (5.2) (11–32)
Affected side (R/L)	19R/26 L
Assistive devices	None: 18 (40%)
	Cane (any type)d: 25 (56%)
	Walker (any type)e: 2 (4%)

^a F = female; FMA-LE = Fugl-Meyer Assessment of Motor Recovery after Stroke—Lower Extremity; L = left; M = male; R = right.

^b n = 45.

^c White matter: corona radiata or internal capsule lesions.

^d Cane types: single-point cane, quad cane, hemiwalker.

^e Walker types: 2-wheeled walker, 4-wheeled walker.

Table 3 represents the results from the hierarchical moderated regression analysis. Block 1 revealed FWS alone significantly explained 26% of the variance in BWS (β = .51, P < .001). Block 2 demonstrated the addition of mFES did not contribute to any additional variance in BWS, above and beyond FWS. However, as seen in block 3, the interaction term between FWS and mFES explained an additional 7.6% (β = −.293, P = .03) of the variance in BWS, above and beyond the individual variables of FWS and mFES. This suggests that the moderator mFES suppresses the relationship between FWS and BWS, weakening it as falls efficacy increases. The final model explains 37.3% of the variance in BWS (P = .031).

Table 3.

Models of Backward Walking Gait Speed Poststroke^a

DV: BW Gait Speed
Model Statistics				Predictor Statistics
Blocks	R 2	F Change	P	Predictors	β	95% CI for B	P
Block 1
FWS	0.26	15.33	<.001	FW gait speedc	0.51	0.13 to 0.39	<.001
Block 2
mFES (+FWS)	0.30	2.02	.16	FW gait speedd	0.37	0.02 to 0.035	.03
				mFES	0.23	−0.01 to 0.04	.16
Block 3d
mFES × FWS + (FWS, mFES)	0.37	4.97	.03	FWSd	0.34	0.01 to 0.33	.04
				mFES	0.16	−0.01 to 0.03	.33
				mFES × FWSd	−0.29	−0.13 to −0.01)	.03

^a β = standardized regression coefficients; 95% CI for B = 95% CI for unstandardized regression coefficients; BWS = Backward Walking Speed; DV = dependent variable; FWS = Forward Walking Speed; mFES = modified Falls-Efficacy Scale.

^b P less than .001.

^c P less than .05.

^d Block 3 explains significantly more variance than block 2 (P = .03).

To further understand the effect of the interaction in this moderated regression analysis, we used the simple slopes approach (Fig. 1) as well as the Johnson-Neyman Technique (Fig. 2). The simple slopes approach depicts that at an average level of falls efficacy (mFES = 6.78), the forward gait speed has a near positive relationship with BWS poststroke (see Fig. 1). At a lower level of falls efficacy (–1 SD, mFES = 4.59), this relationship strengthens to become more positive. However, at a higher level of falls efficacy (+1 SD, mFES = 8.97), the relationship between FWS and BWS disappears. Instead of dividing the continuous variable of mFES into these 3 categories, the Johnson-Neyman plot gives a 95% CI of the slope of FWS predicting BWS and the range of mFES for which the interaction is significant. The interaction between FWS and mFES is significant only for values less than 6.6 on the mFES scale (see Fig. 2). FWS has a significantly positive association with BWS for people with an mFES score of less than 6.6 and has no relationship at all as the mFES scores range higher than 6.6.

Figure 1.

The simple slopes graph for the relationship of FWS with BWS at 3 levels of the moderator mFES (mFES mean = 6.78, mFES mean + 1SD = 8.97, mFES mean-1SD = 4.59). The relationship between forward and BWS is strongly positive at low levels of mFES and almost 0 for higher levels of mFES. BWS = backward walking speed; FWS = forward walking speed; mFES = Modified Falls-Efficacy Scale.

Figure 2.

Johnson-Neyman plot demonstrating that at 95% CI, the relationship between FWS and BWS is significant only when the value of the moderator mFES ranges between 0 and 6.6. BWS = backward walking speed; FWS = forward walking speed; mFES = modified Falls-Efficacy Scale.

We further conducted a subgroup analysis to compare the demographic and clinical characteristics between participants with higher falls efficacy (mFES > 6.6) and lower falls efficacy (mFES ≤ 6.6). As expected, the lower falls efficacy group had a significantly greater number of individuals who used an assistive device, had significantly lower FWS as well as lower BWS than the higher falls efficacy group (Tab. 4). The 2 subgroups did not differ in terms of mean age, time since stroke, stroke severity, sex, or side of hemiparesis.

Table 4.

Differences Between People With Higher and Lower Falls Efficacy^a

Demographic Characteristic	mFES < 6.6 (Mean [SD]) (n = 20)	mFES > 6.6 (Mean [SD]) (n = 25)
Mean age, y	61.55 (10.65)	56.8 (9.15)
Sex	9 F/11 M	8 F/17 M
Mean time since stroke, mo	17.35 (19.32)	21.0 (18.16)
Mean FMA-LE (range, 0–34)	21.75 (6.5)	24.44 (3.63)
Affected side, R/L	9 R/11 L	10 R/15 L
Assistive devicesb	None: 15% Use assistive device:c 85%	None: 60% Use assistive device:c 40%
Functional outcomes
Characteristic	mFES < 6.6 (Mean [SD])	mFES > 6.6 (Mean [SD])	Effect Size
Forward walking speed, m/sd	0.32 (0.29)	0.59 (0.22)	1.07
Backward walking speed, m/se	0.18 (0.15)	0.26 (0.13)	0.57

^a F = female; FMA-LE = Fugl-Meyer Assessment of Motor Recovery after Stroke—Lower Extremity; L = left; M = male; R = right.

^b P = .002, Mann-Whitney U test.

^c Types of assistive devices: single-point cane, quad cane, hemiwalker, rolling walker.

^d P = .001 (independent sample t tests).

^e P = .04 (independent sample t tests).

To further understand the relationship between mFES, FWS, and BWS in people with varied degrees of stroke severity, we subdivided our participants into high stroke severity (FMA-LE score < 25) and low stroke severity (FMA-LE ≥ 25), according to the median of FMA-LE (25) in this cohort. The Supplementary Table demonstrates the bivariate correlations between mFES, FMA-LE, FWS, and BWS for the 2 subgroups. For the high stroke severity group, all 4 assessments highly correlated with the others. However, for the lower stroke severity group, BWS assessments showed no correlation with mFES, FMA-LE, or FWS.

Discussion

To our knowledge, this is the first study to assess the effect of falls efficacy on the relationship between FWS and BWS. Our study uniquely demonstrates that FWS does not correlate with BWS in people with higher falls efficacy. Congruent to our hypothesis, the strength of the relationship between FWS and BWS varied in accordance with the degree of reported falls efficacy. FWS was more strongly related to BWS for people with a lower falls efficacy score and for people with higher stroke severity. Specifically, the interaction effect of FWS with falls efficacy was significant only for average scores lower than or equal to 6.6 on the mFES.

Our results suggest that as falls efficacy increases, especially after a cutoff score of 6.6 on the mFES, BWS does not correlate with FWS. Similarly, we demonstrate that FWS and BWS share a strong relationship in people with higher stroke severity, but not in people with lower stroke severity, as measured by the FMA-LE (Suppl. Tab.). While FWS is promoted as the sixth vital sign¹⁴ and is an indicator of successful gait rehabilitation,¹⁵ we suggest that for higher-functioning individuals with lower stroke severity or greater falls efficacy, the assessment of BWS may provide unique information about their gait impairments or balance confidence, above and beyond the typical assessment of FWS. People with higher falls efficacy by definition are not apprehensive while performing usual daily activities like forward walking as compared to people with lower falls efficacy. The most commonly used clinical questionnaires to assess perceived falls efficacy or balance confidence in people poststroke such as the mFES³³ used in this study or the activities-specific Balance Confidence scale³⁹ do not include items on efficacy during daily tasks that may require backward stepping. Hence, even those individuals poststroke who are considered to have high falls efficacy may experience fear of falling during BWS assessment. This finding is bolstered by the fact that community ambulators poststroke with an FWS of more than 0.8 m/s demonstrate a slow, cautious gait with reduced gait speed, shorter step lengths, increased cadence, and increased double support time during backward walking assessment, but have forward gait characteristics similar to healthy controls.¹⁸ Thus, BWS assessment may provide a significant challenge and reveal deficits even when FWS approaches a ceiling effect.⁴⁰ The novelty of backward walking, with its additional challenges of reduced visual cues, greater reliance on proprioception,²³ and increased motor activity²² may unmask gait impairments in people who have higher falls efficacy, lower stroke severity, or in other words, may be presumed to be higher functioning. Our findings corroborate those of Hawkins and colleagues who suggested that the assessment of BWS may unmask mobility impairments in community ambulators poststroke that may not be detected during the conventional forward walking assessments.¹⁸

Furthermore, our study indicates that people with lower falls efficacy, especially those who score lower than 6.6 on the mFES scale, walk slower both in the forward and backward directions, suggesting some overlap between these 2 assessments. This may be intuitive, as people who are apprehensive to walk forward will be even more if not equally apprehensive to walk backward poststroke. Our results also suggest that falls efficacy has a strong positive association both with FWS and BWS in people with higher stroke severity. However, for people with an FMA-LE score greater than or equal to 25, falls efficacy affects only FWS and not BWS, suggesting that factors other than FWS and falls efficacy might be associated with BWS in this subgroup. Further investigation in the assessment and predictors of BWS in the 2 cohorts of individuals poststroke is warranted.

While complex forward walking assessments like obstacle negotiation or cognitive dual tasking may also provide additional information about mobility deficits compared to a simple forward walking assessment, standardization of a cognitive task and portability of an obstacle both present as clinical challenges. In contrast, assessment of BWS is simple, quick, and requires only a stopwatch and a tape measure, and hence has the potential to be the most efficient, cost-effective, and portable gait assessment tool for physical therapists around the globe to predict fall risk poststroke. In a recent study conducted by Carter et al, 81% of elderly individuals who scored lower than 0.66 m/s on the 3MBWT reported falling.¹⁶ Similarly, Fritz et al concluded that elderly individuals with BWS slower than 0.4 m/s were 3.2 times more likely to fall than those who walked faster.¹⁷ Although such cutoffs have not been established for people poststroke, in our study participants with higher falls efficacy had an average BWS of 0.26 m/s making them prone to falls, which may not have been apparent with the assessment of FWS alone (see Tab. 4). Future studies need to investigate valid cutoffs of BWS in the prediction of falls for people poststroke, as its assessment may provide physical therapists important insights in addressing the critical issue of increased risk of falls poststroke.

Given the limited time available to assess patients,^41,42 individualizing assessment tools according to each patient’s specific impairments would make prudent use of this time and very well may lead to better clinical decision-making skills.⁴³ The results from this study provide evidence that physical therapists may assess BWS preferentially in individuals with higher falls efficacy. In contrast, as falls efficacy decreases, FWS demonstrates a stronger positive association with BWS. Given the significant overlap between the 2 assessments in individuals with lower falls efficacy, assessment of BWS may be redundant with the assessment of FWS. However, this suggestion must be interpreted cautiously as much of the variance in BWS remains unexplained in this cohort. Future studies may examine the importance of BWS and FWS assessment to predict falls, specifically in people with lower falls efficacy poststroke.

Limitations

Although decreased falls efficacy is closely related to increased falls poststroke, we did not assess fall incidence in this cohort. It is possible that the relationship between FWS and BWS may differ in fallers and nonfallers poststroke. Because this was a cross-sectional study, we cannot establish a causal effect of falls efficacy on the relationship between FWS and BWS. However, this study does corroborate results from other studies to emphasize the importance of assessing backward walking in individuals poststroke. Future studies may assess the difference in FWS and BWS in prediction of falls in individuals across the spectrum of disability poststroke. Another limitation of this study is that the mFES considers only activities that require forward walking and does not include questions regarding falls efficacy during activities that may require backward walking. To the best of our knowledge, current falls efficacy–based questionnaires do not inquire about challenges of taking backward steps during activities like turning in narrow spaces or sitting back in a chair. As backward walking is now emerging as a promising intervention and assessment tool, falls-efficacy questionnaires should consider including items that require backward walking. We recognize that the distance to assess forward walking (10MWT) was longer than that to assess backward walking (3MBWT). However, since backward walking is considered more challenging than forward walking, with shorter step lengths and a slower cadence, a shorter backward walking distance is sufficient to balance the efforts required in the 2 walking tests. Currently, the 3MBWT is the only test that has been validated in elderly individuals,¹⁶ and it is under validation for detecting falls in people poststroke.⁴⁴ Future studies may examine the relationship between the 10MWT with different lengths of a backward walking assessment.

Conclusion

The relationship between FWS and backward speed varies greatly according to the degree of falls efficacy in individuals poststroke. Lower falls efficacy leads to a strong positive relationship between FWS and BWS, and this relationship weakens with an increase in falls efficacy. Physical therapists should consider assessing both FWS and BWS in patients with higher balance confidence, whereas the importance of assessing BWS in individuals poststroke with lower confidence or falls efficacy requires further investigation.

Author Contributions

Concept/idea/research design: K. Bansal, D.J. Clark, D.K. Rose

Writing: K. Bansal, D.J. Clark, D.K. Rose

Data collection: K. Bansal, D.K. Rose

Data analysis: K. Bansal, D.J. Clark, D.K. Rose

Project management: E.J. Fox, D.K. Rose

Fund procurement: D.K. Rose

Providing participants: D.K. Rose

Providing facilities/equipment: E.J. Fox, D.K. Rose

Providing institutional liaisons: D.K. Rose

Consultation (including review of manuscript before submitting): D.J. Clark, E.J. Fox

Acknowledgments

We are grateful to Dolores Miller-Sellers for her assistance in reviewing the manuscript.

Funding

This work was supported by the American Heart Association (grant No. 15MCPRP25670037), the VA (Rehabilitation Research and Development grant No. N-2051P), and the Brooks-PHHP Research Collaboration.

Ethics Approval

The institutional review board at the University of Florida approved this study, and all participants signed a written informed consent to participate.

Disclosures

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