Archives of Physical Medicine and Rehabilitation journal homepage: www.archives-pmr.org Archives of Physical Medicine and Rehabilitation 2014;-:-------
ORIGINAL ARTICLE
Spatial-Temporal Gait Variability Poststroke: Variations in Measurement and Implications for Measuring Change Amanda E. Chisholm, PhD,a,b Shelley Makepeace, BSc,a Elizabeth L. Inness, MSc, BSc, PT,a,b Stephen D. Perry, PhD,a,b,c William E. McIlroy, PhD,a,b,d,e,f Avril Mansfield, PhDa,b,d,e From the aToronto Rehabilitation Institute, Toronto, Ontario; bGraduate Department of Rehabilitation Science, University of Toronto, Toronto, Ontario; cDepartment of Kinesiology and Physical Education, Wilfrid Laurier University, Waterloo, Ontario; dHeart & Stroke Foundation, Center for Stroke Recovery, Toronto, Ontario; eSunnybrook Research Institute, Toronto, Ontario; and fDepartment of Kinesiology, University of Waterloo, Waterloo, Ontario, Canada.
Abstract Objective: To determine the responsiveness to change of spatial-temporal gait parameters among stroke survivors for 3 different variability measures: SD, coefficient of variation (CV), and median absolute deviation (MAD). Design: Retrospective chart review. Setting: Clinical laboratory in a Canadian hospital. Participants: Stroke survivors (NZ74) receiving inpatient rehabilitation. Interventions: Not applicable. Main Outcome Measures: Spatial-temporal gait variability was calculated for step length, step width, stance time, swing time, and double support time. Responsiveness to change was determined by comparing (1) trials without versus trials with a concurrent cognitive task and (2) admission to discharge from rehabilitation. Results: Variability estimators (SD, CV, and MAD) increased with the addition of a cognitive task and decreased from admission to discharge of rehabilitation. However, these changes were not statistically significant when change in gait velocity was included as a covariate. The effect size values were similar for all variability estimators with a trend toward a greater SD response to temporal parameters. The CV displayed a larger response to change for step length than did the SD and MAD. Although gait variability decreased between admission and discharge, the effect size was larger for the condition without the cognitive task than for the condition with the cognitive task. Conclusions: Our results show that gait variability estimators demonstrate a similar responsiveness to a concurrent cognitive task and improved walking ability with recovery from stroke. Future work may focus on evaluating the clinical utility of these measures in relation to informing therapy and response to gait-specific training protocols. Archives of Physical Medicine and Rehabilitation 2014;-:------ª 2014 by the American Congress of Rehabilitation Medicine
Supported by the University of Toronto, Natural Sciences and Engineering Council of Canada (PGS-D), and Canadian Institute of Health Research (grant no. MOP-77772). We acknowledge the support of the Toronto Rehabilitation Institute, which receives funding under the Provincial Rehabilitation Research Program from the Ministry of Health and Long-Term Care in Ontario. No commercial party having a direct financial interest in the results of the research supporting this article has conferred or will confer a benefit on the authors or on any organization with which the authors are associated.
Step-to-step spatial-temporal variability of walking has gathered much attention over recent years and is thought to be an important index of the integrity of the gait control system.1 Changes in variability can arise from internal (eg, aging effects and pathology), external (eg, environmental demands), and methodological (eg, instrumentation) factors.1 Gait variability is clinically relevant for several reasons: (1) to assess fall risk,2-5 (2) to determine the patients’ ability to adapt to changing conditions during gait,6 and
0003-9993/14/$36 - see front matter ª 2014 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.02.014
2
A.E. Chisholm et al
(3) to evaluate the response to therapeutic interventions.7 In addition, identifying indices of walking that may better reveal recovery after neurologic injury remains an important clinical objective. The challenge is that there are a number of different expressions of gait variability (ie, different task conditions, calculations of variability, parameters of walking) and there is little evidence of the unique benefits of each different expression of variability. Gait variability is commonly determined by calculating the SD and/or coefficient of variation (CV). Chau et al1 proposed the median absolute deviation (MAD) to address issues of inflated variability (eg, sensitivity to outliers) observed with SD and CV estimates of spatial-temporal parameters. The MAD is the median of the absolute difference between sample values and their median value.1 Few studies have compared different calculations of variability; however, CV is often not recommended for parameters with a low mean value.8 Because the MAD is less sensitive to outliers,1 it may be less responsive to detect actual change with different task conditions and recovery from rehabilitation. There is a need to understand the unique ability of these indices of step variability to detect change in the control of walking. The overall objective of this work was to advance understanding of the potential clinical utility of different spatial-temporal variability measures for walking assessments poststroke. Dual-task paradigms are often used clinically to evaluate changes in gait performance due to increased cognitive demands, as a reflection of challenges encountered in daily living. Commonly, dual tasking is imposed to increase the challenge to dynamic stability control; however, this effect may also challenge the control of limb motion among individuals with sensorimotor impairment.8 The additional cognitive demand imposed by dual tasking is more likely to expose gait instability by dividing attention between tasks.9 Studies investigating the effects of a concurrent cognitive task on gait performance have demonstrated that healthy adults will decrease gait speed, whereas temporal gait variability remains unchanged compared with single-task walking.10,11 In contrast, individuals with impaired gait will decrease speed and increase temporal variability.10,11 One interpretation of these findings is that the increased step-to-step variability reflects the increased challenge to the control of dynamic stability that results when cognitive resources are directed to a concurrent secondary task.9,10 Evidence suggests that variability is related to impaired gait performance and can be used to measure changes due to functional recovery after stroke rehabilitation.12 Variability measures may provide specific information about changes in performance not captured by averaged measures. Improved gait velocity and temporal patterns observed after rehabilitation suggest a more consistent walking pattern13-15; however, this has not been specifically investigated. Although gait velocity is a clinically important measure of functional recovery after stroke, a relation
List of Abbreviations: ANCOVA ANOVA BBS CMSA COVS CV ES MAD
analysis of covariance analysis of variance Berg Balance Scale Chedoke McMaster Stroke Assessment Clinical Outcomes Variable Scale coefficient of variation effect size median absolute deviation
between velocity and variability may change because of gaitspecific impairments poststroke. This study aimed to evaluate different methods of measuring variability by assessing the responsiveness of each variability estimator (SD, CV, and MAD) to task difficulty (single task and cognitive dual task), and to change over time (admission to discharge) for spatial-temporal gait parameters in stroke survivors. We hypothesized that the SD will detect (1) increased variability with dual versus single task (because of an added cognitive demand) and (2) decreased variability with discharge versus admission (because of improved gait function) more than the CV and MAD because they will be less responsive to change.
Methods Participants A retrospective chart review was performed from a database of individuals (nZ264) who completed a standardized gait assessment on admission and discharge from inpatient rehabilitation as prescribed by their physical therapist. The criteria for inclusion were (1) ischemic or hemorrhagic stroke and (2) independent walking as defined by the ability to walk 10m without a walking aid at admission and discharge. Participants were excluded if they did not complete gait assessments at admission and discharge or the gait assessments were not at least 7 days apart. Participants received standard physical therapy (eg, focused on strength, balance, range of motion, coordination, and walking) for approximately 1 hour each weekday during inpatient rehabilitation. This study was approved by the institution’s research ethics board.
Clinical assessment At admission and discharge from inpatient rehabilitation, all participants completed a standardized assessment of balance, functional mobility, and sensorimotor impairment with a physical therapist (>10y of experience) using standardized instructions. These assessments included the Berg Balance Scale (BBS) (range, 0e56),16 Clinical Outcomes Variable Scale (COVS) (range, 0e91),17 and Chedoke McMaster Stroke Assessment (CMSA) (range, 1e7).18 The National Institutes of Health Stroke Scale was administered at admission as a gross indicator of stroke severity.19
Gait assessment Participants were asked to walk 6m over a pressure-sensitive mat (GAITRitea) without a gait aid. There were 2 walking conditions: self-selected pace and self-selected pace with a concurrent cognitive task. Instructions for the single-task condition were to “walk at your comfortable speed.” In the cognitive task condition, participants were instructed to count by 1, 2, 3, or 7 starting at a number given at the start of the trial. Difficulty of the cognitive task was varied on the basis of cognitive abilities of the individuals, as determined by the physical therapist. For both conditions, 2 or 3 walks were completed to record at least 18 steps per condition; rest was allowed as needed. Spatial-temporal gait parameters were measured using the GAITRite software.a The pressure mat is 4.27m long and .88m wide and contains a grid of 48288 pressure sensors arranged 1.27cm apart by centers. Means of spatial-temporal parameters measured via the pressure mat have been shown to be reliable among persons with stroke (intraclass correlation coefficient test-retest reliability values, rZ.86e.94 for step length and step time of the paretic and nonparetic limbs).20 www.archives-pmr.org
Measuring change in gait variability
3
Data and statistical analysis
Table 1
Average gait velocity was calculated over the entire condition. Spatial (step length and step width) and temporal (stance time, swing time, and double support time) parameters were calculated for each footfall. Variability estimators were calculated for each condition at admission and discharge. Data from both limbs were included in the calculations. The above spatial-temporal parameters were used in each of the following equations (n observations of a gait variable x): sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 iZ1 ðxi xÞ SDðxÞZ n
ð1Þ
CVðxÞZðSDðxÞ=ðxÞÞ 100
ð2Þ
MADðxÞZmedðjx medðxÞjÞ
ð3Þ
Paired t tests were used to evaluate changes in clinical scores (CMSA, BBS, and COVS) between admission and discharge. Repeated-measures analysis of variance (ANOVA) was used to evaluate the main effect of time (admission and discharge) and condition (single task and dual task), and the interaction effect (conditiontime) for each variability estimator with the 5 spatial-temporal parameters. Gait velocity was used as a covariate in the analysis of covariance (ANCOVA) in a separate analysis. Spatial-temporal data for each estimator and gait velocity were normally distributed as determined by the KruskalWallis test. Effect size (ES) values were also calculated to compare task conditions and assessment times for each estimator. ES is defined as the mean change in scores between tasks or time divided by the SD of the baseline score. A positive ES indicates a greater dual-task score compared with a single task (condition), and a greater discharge score compared with admission (time), while the magnitude represents the responsiveness. Statistical significance was set at P
ORIGINAL ARTICLE
Spatial-Temporal Gait Variability Poststroke: Variations in Measurement and Implications for Measuring Change Amanda E. Chisholm, PhD,a,b Shelley Makepeace, BSc,a Elizabeth L. Inness, MSc, BSc, PT,a,b Stephen D. Perry, PhD,a,b,c William E. McIlroy, PhD,a,b,d,e,f Avril Mansfield, PhDa,b,d,e From the aToronto Rehabilitation Institute, Toronto, Ontario; bGraduate Department of Rehabilitation Science, University of Toronto, Toronto, Ontario; cDepartment of Kinesiology and Physical Education, Wilfrid Laurier University, Waterloo, Ontario; dHeart & Stroke Foundation, Center for Stroke Recovery, Toronto, Ontario; eSunnybrook Research Institute, Toronto, Ontario; and fDepartment of Kinesiology, University of Waterloo, Waterloo, Ontario, Canada.
Abstract Objective: To determine the responsiveness to change of spatial-temporal gait parameters among stroke survivors for 3 different variability measures: SD, coefficient of variation (CV), and median absolute deviation (MAD). Design: Retrospective chart review. Setting: Clinical laboratory in a Canadian hospital. Participants: Stroke survivors (NZ74) receiving inpatient rehabilitation. Interventions: Not applicable. Main Outcome Measures: Spatial-temporal gait variability was calculated for step length, step width, stance time, swing time, and double support time. Responsiveness to change was determined by comparing (1) trials without versus trials with a concurrent cognitive task and (2) admission to discharge from rehabilitation. Results: Variability estimators (SD, CV, and MAD) increased with the addition of a cognitive task and decreased from admission to discharge of rehabilitation. However, these changes were not statistically significant when change in gait velocity was included as a covariate. The effect size values were similar for all variability estimators with a trend toward a greater SD response to temporal parameters. The CV displayed a larger response to change for step length than did the SD and MAD. Although gait variability decreased between admission and discharge, the effect size was larger for the condition without the cognitive task than for the condition with the cognitive task. Conclusions: Our results show that gait variability estimators demonstrate a similar responsiveness to a concurrent cognitive task and improved walking ability with recovery from stroke. Future work may focus on evaluating the clinical utility of these measures in relation to informing therapy and response to gait-specific training protocols. Archives of Physical Medicine and Rehabilitation 2014;-:------ª 2014 by the American Congress of Rehabilitation Medicine
Supported by the University of Toronto, Natural Sciences and Engineering Council of Canada (PGS-D), and Canadian Institute of Health Research (grant no. MOP-77772). We acknowledge the support of the Toronto Rehabilitation Institute, which receives funding under the Provincial Rehabilitation Research Program from the Ministry of Health and Long-Term Care in Ontario. No commercial party having a direct financial interest in the results of the research supporting this article has conferred or will confer a benefit on the authors or on any organization with which the authors are associated.
Step-to-step spatial-temporal variability of walking has gathered much attention over recent years and is thought to be an important index of the integrity of the gait control system.1 Changes in variability can arise from internal (eg, aging effects and pathology), external (eg, environmental demands), and methodological (eg, instrumentation) factors.1 Gait variability is clinically relevant for several reasons: (1) to assess fall risk,2-5 (2) to determine the patients’ ability to adapt to changing conditions during gait,6 and
0003-9993/14/$36 - see front matter ª 2014 by the American Congress of Rehabilitation Medicine http://dx.doi.org/10.1016/j.apmr.2014.02.014
2
A.E. Chisholm et al
(3) to evaluate the response to therapeutic interventions.7 In addition, identifying indices of walking that may better reveal recovery after neurologic injury remains an important clinical objective. The challenge is that there are a number of different expressions of gait variability (ie, different task conditions, calculations of variability, parameters of walking) and there is little evidence of the unique benefits of each different expression of variability. Gait variability is commonly determined by calculating the SD and/or coefficient of variation (CV). Chau et al1 proposed the median absolute deviation (MAD) to address issues of inflated variability (eg, sensitivity to outliers) observed with SD and CV estimates of spatial-temporal parameters. The MAD is the median of the absolute difference between sample values and their median value.1 Few studies have compared different calculations of variability; however, CV is often not recommended for parameters with a low mean value.8 Because the MAD is less sensitive to outliers,1 it may be less responsive to detect actual change with different task conditions and recovery from rehabilitation. There is a need to understand the unique ability of these indices of step variability to detect change in the control of walking. The overall objective of this work was to advance understanding of the potential clinical utility of different spatial-temporal variability measures for walking assessments poststroke. Dual-task paradigms are often used clinically to evaluate changes in gait performance due to increased cognitive demands, as a reflection of challenges encountered in daily living. Commonly, dual tasking is imposed to increase the challenge to dynamic stability control; however, this effect may also challenge the control of limb motion among individuals with sensorimotor impairment.8 The additional cognitive demand imposed by dual tasking is more likely to expose gait instability by dividing attention between tasks.9 Studies investigating the effects of a concurrent cognitive task on gait performance have demonstrated that healthy adults will decrease gait speed, whereas temporal gait variability remains unchanged compared with single-task walking.10,11 In contrast, individuals with impaired gait will decrease speed and increase temporal variability.10,11 One interpretation of these findings is that the increased step-to-step variability reflects the increased challenge to the control of dynamic stability that results when cognitive resources are directed to a concurrent secondary task.9,10 Evidence suggests that variability is related to impaired gait performance and can be used to measure changes due to functional recovery after stroke rehabilitation.12 Variability measures may provide specific information about changes in performance not captured by averaged measures. Improved gait velocity and temporal patterns observed after rehabilitation suggest a more consistent walking pattern13-15; however, this has not been specifically investigated. Although gait velocity is a clinically important measure of functional recovery after stroke, a relation
List of Abbreviations: ANCOVA ANOVA BBS CMSA COVS CV ES MAD
analysis of covariance analysis of variance Berg Balance Scale Chedoke McMaster Stroke Assessment Clinical Outcomes Variable Scale coefficient of variation effect size median absolute deviation
between velocity and variability may change because of gaitspecific impairments poststroke. This study aimed to evaluate different methods of measuring variability by assessing the responsiveness of each variability estimator (SD, CV, and MAD) to task difficulty (single task and cognitive dual task), and to change over time (admission to discharge) for spatial-temporal gait parameters in stroke survivors. We hypothesized that the SD will detect (1) increased variability with dual versus single task (because of an added cognitive demand) and (2) decreased variability with discharge versus admission (because of improved gait function) more than the CV and MAD because they will be less responsive to change.
Methods Participants A retrospective chart review was performed from a database of individuals (nZ264) who completed a standardized gait assessment on admission and discharge from inpatient rehabilitation as prescribed by their physical therapist. The criteria for inclusion were (1) ischemic or hemorrhagic stroke and (2) independent walking as defined by the ability to walk 10m without a walking aid at admission and discharge. Participants were excluded if they did not complete gait assessments at admission and discharge or the gait assessments were not at least 7 days apart. Participants received standard physical therapy (eg, focused on strength, balance, range of motion, coordination, and walking) for approximately 1 hour each weekday during inpatient rehabilitation. This study was approved by the institution’s research ethics board.
Clinical assessment At admission and discharge from inpatient rehabilitation, all participants completed a standardized assessment of balance, functional mobility, and sensorimotor impairment with a physical therapist (>10y of experience) using standardized instructions. These assessments included the Berg Balance Scale (BBS) (range, 0e56),16 Clinical Outcomes Variable Scale (COVS) (range, 0e91),17 and Chedoke McMaster Stroke Assessment (CMSA) (range, 1e7).18 The National Institutes of Health Stroke Scale was administered at admission as a gross indicator of stroke severity.19
Gait assessment Participants were asked to walk 6m over a pressure-sensitive mat (GAITRitea) without a gait aid. There were 2 walking conditions: self-selected pace and self-selected pace with a concurrent cognitive task. Instructions for the single-task condition were to “walk at your comfortable speed.” In the cognitive task condition, participants were instructed to count by 1, 2, 3, or 7 starting at a number given at the start of the trial. Difficulty of the cognitive task was varied on the basis of cognitive abilities of the individuals, as determined by the physical therapist. For both conditions, 2 or 3 walks were completed to record at least 18 steps per condition; rest was allowed as needed. Spatial-temporal gait parameters were measured using the GAITRite software.a The pressure mat is 4.27m long and .88m wide and contains a grid of 48288 pressure sensors arranged 1.27cm apart by centers. Means of spatial-temporal parameters measured via the pressure mat have been shown to be reliable among persons with stroke (intraclass correlation coefficient test-retest reliability values, rZ.86e.94 for step length and step time of the paretic and nonparetic limbs).20 www.archives-pmr.org
Measuring change in gait variability
3
Data and statistical analysis
Table 1
Average gait velocity was calculated over the entire condition. Spatial (step length and step width) and temporal (stance time, swing time, and double support time) parameters were calculated for each footfall. Variability estimators were calculated for each condition at admission and discharge. Data from both limbs were included in the calculations. The above spatial-temporal parameters were used in each of the following equations (n observations of a gait variable x): sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 iZ1 ðxi xÞ SDðxÞZ n
ð1Þ
CVðxÞZðSDðxÞ=ðxÞÞ 100
ð2Þ
MADðxÞZmedðjx medðxÞjÞ
ð3Þ
Paired t tests were used to evaluate changes in clinical scores (CMSA, BBS, and COVS) between admission and discharge. Repeated-measures analysis of variance (ANOVA) was used to evaluate the main effect of time (admission and discharge) and condition (single task and dual task), and the interaction effect (conditiontime) for each variability estimator with the 5 spatial-temporal parameters. Gait velocity was used as a covariate in the analysis of covariance (ANCOVA) in a separate analysis. Spatial-temporal data for each estimator and gait velocity were normally distributed as determined by the KruskalWallis test. Effect size (ES) values were also calculated to compare task conditions and assessment times for each estimator. ES is defined as the mean change in scores between tasks or time divided by the SD of the baseline score. A positive ES indicates a greater dual-task score compared with a single task (condition), and a greater discharge score compared with admission (time), while the magnitude represents the responsiveness. Statistical significance was set at P
