Gait during community ambulation and laboratory-based walking in people with mild traumatic brain injury

Abstract

Background

Gait impairments are common in individuals with mild traumatic brain injury (mTBI), presenting in the acute phase and often persisting subtly over time. Despite the prominence of laboratory gait evaluations, a comprehensive understanding of gait deficits post-mTBI necessitates the examination of various gait domains in real-world environments. Assessing gait during the community ambulation task (CAT) may capture real-world challenges and influence interventions or rehabilitation in individuals with mTBI.

Research Question

The aim of the study was to compare gait performance across independent gait domains in individuals with and without mTBI using wearable sensors during both the CAT and laboratory tasks (i.e., 1-minute walk test). Additionally, associations between the CAT and single- and dual-task walking were investigated.

Methods

In this cross-sectional study, 107 participants, including individuals with mTBI (n=52) and healthy controls (n=55), underwent gait assessments during the CAT and 1-minute walk tests (single- and dual-task) using wearable sensors. Four independent gait domains (i.e., gait variability, pace, rhythm, and turning), consisting of thirteen gait variables, were analyzed. Statistical methods included t-tests and partial correlations, adjusted for covariates.

Results

Individuals with mTBI exhibited decreased gait pace and increased gait variability in both the CAT and laboratory tasks. Notably, the CAT task exhibited greater gait differences in terms of gait variability, rhythm, and pace compared to the laboratory tasks. Gait rhythm and pace during the CAT revealed stronger correlation with dual-task walking compared to single-task working.

Significance

The community ambulation walking task showed more abnormalities across gait domains compared to laboratory walking tests, highlighting the potential importance of incorporating real-life, community ambulatory tasks into post-mTBI evaluation of mobility.

1. Introduction

Gait performance serves as an important functional measure in people with mild traumatic brain injury (mTBI). Impaired gait can manifest during the acute phase of mTBI [1–3] and subtle deficiencies may persist for several years after injury [4]. To better understand gait characteristics following mTBI, Stuart et al. employed principal components analysis to derive a conceptual model, using thirteen gait metrics, categorized into four independent domains: gait variability, rhythm, pace and turning [5]. Gait domains are independent from one another, and evidence suggests that specific brain regions relate to different gait domains [6,7]. Due to the heterogeneous nature of mTBI as well as the potentially diffuse white matter damage, it is possible that people with mTBI could display gait impairment across all four gait domains [8,9]. Indeed, this model has been applied to distinguish pathological effects of mTBI on one’s gait performance [4,10]. The ability to identify gait deficits is important as they provide insights towards the compensatory mechanisms adapted from postural instability as well as impaired coordination, cognitive functions, sensory integration and motor activations. Increased gait variability and reduced gait and turn metrics were found to be associated increased fall risks [11,12]. These gait deficits could play an important role as target markers for rehabilitation and in deciding when to safely return to activities, such as sports or military activities. To date, most studies primarily focus on gait speed during single-task gait evaluations (e.g., 1-minute walk test) in people with mTBI [4,13]. This inclination is likely derived from the simplicity of measurement and the intuitive nature of gait speed. While gait speed reflects overall performance and is sensitive to pathology and age [14], it lacks specificity and may not capture subtle underlying deficits [15]. A recent literature review that compared common gait metrics during single-task gait, including gait speed as well as stride length, stride width, and stride time, reported that gait metrics were largely unaffected beyond the acute phases of mTBI [4]. However, comparing gait metrics from the four independent gait domains [5,15,16], it has been previously reported that individuals with chronic mTBI exhibited slower pace and turning during single-task gait, compared to the control groups [10]. In addition to single-task gait, assessment of gait needs to reflect the challenges encountered in daily activities to accurately capture deficits post-mTBI and to comprehend their impact on a patient’s life.

Evaluating dual-task gait conditions involving a secondary cognitive task that diverts attention from walking is thought to replicate real-life walking demands [17]. After an mTBI, individuals reportedly experience impaired dual-task gait predominantly during the acute phase [4] and often display deficits for several months post-injury across gait domains [18–21]. In dual-task scenarios, individuals can prioritize either the motor or cognitive aspect, or both tasks might exhibit deficits simultaneously (mutual interference). Extensive exploration of dual-task gait where mobility tasks are combined with simultaneous cognitive activities reveals the diagnostic potential of dual-task gait in individuals with mTBI [4,21–24]. However, the benefit of assessing dual-task walking remains uncertain as some literature showed no difference in dual task cost [10]. Besides adding cognitive load towards walking, assessment of non-linear movement and ambulatory turns for navigating complex environments is typically not included in most gait and functional clinical tests utilized after mTBI [25]. Walking with turns while moving through varied environments are crucial components of everyday mobility, which often involving non-linear paths. Advancements in wearable inertial sensor technology, such as accelerometers, gyroscopes, and magnetometers now allow for the evaluation of various objective gait measures in natural and unconstrained settings [26,27] that occur in everyday mobility. Continuous assessment of ambulatory tasks during natural movements might offer more precise indicators of mobility issues post-mTBI compared to assessments conducted in laboratories or subjective evaluations in clinical settings.

The aim of this cross-sectional study was to compare gait performance, using wearable sensors, between individuals post-mTBI and healthy controls across various gait assessments during both a prescribed community ambulatory and laboratory gait tasks which may lend themselves to clinical testing, and to determine whether single- or dual-task walking associates with community ambulation. We hypothesized that a prescribed community ambulatory task could better identify gait impairment across all four gait domains, represented by slower pace, rhythm, and turning, as well as increased variability, compared to the more common tasks of single and dual task walking. Additionally, we expected that dual-task gait would exhibit a greater number of gait variables significantly associated with the prescribed community ambulatory task compared to the single-task gait assessment.

2. Methods

2.1. Participants

Participants were enlisted as part of a larger study (ClinicalTrials.gov identifier: NCT03892291) [28]. The data was gathered from a total of 107 individuals across three locations: Courage Kenny Research Center in Minneapolis, MN, Oregon Health & Science University in Portland, OR, and University of Utah in Salt Lake City, UT (Table 1). For inclusion, individuals with mTBI met specific criteria: 1) received a diagnosis of mTBI following Veterans Affairs/Department of Defense (VA/DoD) guidelines [29], 2) aged between 18 and 50 years, and 3) were beyond the acute phase (>3 weeks post-mTBI) but within 3 years from their most recent mTBI while still reporting symptoms. Symptoms related to mTBI that individuals experienced were evaluated using the Neurobehavioral Symptom Inventory (NSI) [30]. Healthy control participants were either free from any history of mTBI or had surpassed 7 years since their last mTBI without any lingering symptoms. Exclusion criteria for both groups encompassed: 1) past history of other injuries, medical or neurological conditions affecting balance (such as lower extremity injury, recent surgery, or stroke), 2) existing substance abuse disorder, 3) pregnancy, and 4) inability to communicate proficiently in English. The study received approval from the Institutional Review Board at all participating sites. Before taking part in the study, participants provided written informed consent.

Table 1.

Demographics for the mTBI and healthy control groups

	mTBI	Controls	Total	Group difference p value	Site difference p value
N	52	55	107	-	-
Site
Courage Kenny Research Ctr.	23 (44.2%)	24 (43.7%)	47 (43.9%)	0.903	-
Oregon Health & Science University	10 (19.2%)	9 (16.3%)	19 (17.8%)
University of Utah	19 (36.6%)	22 (40%)	41 (38.3%)
Age (years)
Mean (SD)	31.8 (9.4)	30.9 (9.5)	31.3 (9.4)	0.646	<0.001
Gender
Male	10 (19.2%)	15 (27.3%)	25 (23.4%)	0.326	0.757
Female	42 (80.8%)	40 (72.7%)	82 (76.6%)
BMI (kg/m 2 )
Mean (SD)	24.5 (5.9)	24.8 (4.1)	24.6 (5.0)	0.750	0.016
NSI score
Mean (SD)	27.7 (15.2)	5.2 (4.3)	16.1 (15.8)	<0.001	0.727
Days since injury
Mean (SD)	313.1 (261.4)	-	313.1 (261.4)	-	<0.001
Number of previous mTBIs
1	29 (55.8%)	-	29 (55.8%)	-	-
2	19 (36.5%)	-	19 (36.5%)	-	-
3 or more	4 (7.7%)	-	4 (7.7%)	-	-

Significance denoted by bolded p-value (p <0.05).

Abbreviations: Ctr.: Center, SD: Standard deviation, BMI: Body mass index, NSI: Neurobehavioral Symptom Inventory

2.2. Procedures

Participants completed three gait performance assessment tasks in a randomized order including a single- (ST) and dual-task (DT) one-minute walk test (1MW) and a Community Ambulatory Task (CAT). The 1MW test consisted of walking at a comfortable pace between two marked lines on the ground spaced 6 meters apart, involving 180-degree turns. The cognitive overlay for the 1MW dual-task involved memorization and recall of an 8-digit alpha-numeric grid coordinate administered verbally. This task was introduced in the context of a geocaching activity to provide a cognitive load that is ecologically appropriate [31,32]. Regarding the CAT, participants walked at their own chosen pace while following directions based on landmarks within a building (e.g., “walking down the hallway toward the green doors”) [33]. These instructions were given verbally by a researcher who walked several steps behind the participant to ensure that participants followed the instructions and to avoid influencing their pace. The CAT was approximately 400-m long, spanning between two floors. Each walking route took 6–7 minutes to complete and included various elements typically encountered during everyday walking, such as walking in straight and non-straight paths, turning at various angles, navigating through hallways, going up and down stairs and ramps, reading signage, and being aware of pedestrian traffic (Fig. 1).

Fig. 1.

Simplified schematic of the CAT with provided instructions and approximate distances (left), detailed schematics of the CAT according to the floor plan from OHSU. Abbreviations: CAT: Community Ambulatory Task, OHSU: Oregon Health and Science University.

During all assessments, participants wore five motion-tracking sensors positioned on the forehead (head), sternum (trunk), pelvis (lumbar spine), and both feet (with a sampling rate of 128 Hz; APDM Inc., Portland, OR, USA). Thirteen gait variables were computed using MobilityLab (ML, APDM, Inc., Portland, OR, USA) [34,35] to assess all gait domains (variability, rhythm, pace, and turning) providing an in-depth assessment of gait performance [5]. Gait variables included gait speed (m/s), stride time (s), stride length (m), single support time (s), double support time (s), foot strike angle (degrees), turn duration (s), turn velocity (degrees/s), and standard deviation (SD) of stride time, stride length, foot strike angle, single support time, and double support time.

2.3. Statistical analyses

The demographic details and assessments for both the mTBI and control groups are presented as mean (SD) where applicable. The Shapiro-Wilk test was used to check the normality of the data distribution. For normally distributed data, either Student’s or Welch’s t-tests was employed, depending on the equality of variances, to assess the differences between the mTBI and control groups in each assessment. For non-normally distributed data, Mann-Whitney U test was used instead. Since the comparison was exploratory, multiple comparison corrections were not applied. To explore if there was a pattern of group differences among gait performance assessments, Cohen’s d effect sizes (Cohen’s d_s) were calculated. The effect size was considered weak when the value was less than 0.5, moderate when the value is between 0.5 and 0.8, and strong when the value is larger than 0.8 [36]. Pairwise partial correlations between the CAT and 1MW ST, and the CAT and 1MW DT were measured using Pearson or Spearman correlation coefficients, depending on the data distribution. The partial correlation analysis was adjusted for age, gender, BMI and NSI score to account for covariates. Statistical significance was set at p < 0.05 for all tests.

3. Results

People with mTBI had more deficits including gait variability, rhythm, and pace compared to healthy participants during the CAT than during laboratory gait tests

During the CAT, mTBI patients had a significantly slower gait speed and shorter stride length with moderate effect sizes, shallower foot strike angle and shorter double support time compared to healthy controls (Table 2). Similarly, patients with mTBI had significantly greater stride time SD, single support time SD and double support time SD compared to healthy controls (Table 2). The variability of the stride time showed moderate effect size (Fig. 2A).

Table 2.

Assessments for the mTBI and healthy control groups

Gait Domain	Assessments	mTBI (n = 52)	Controls (n = 55)	Group difference (p-value)	Effect size
CAT
Variability	Stride time SD	0.25 (0.03)	0.23 (0.04)	0.003	0.609
	Single support time SD	1.27 (0.19)	1.18 (0.20)	0.015	0.479
	Double support time SD	1.79 (0.42)	1.62 (0.37)	0.035	0.204
	Foot strike angle SD	2.05 (0.31)	2.11 (0.37)	0.334	0.187
	Stride length SD	0.41 (0.06)	0.41 (0.07)	0.815	0.023
Rhythm	Stride time	1.09 (0.07)	1.06 (0.07)	0.114	0.312
	Single support time	39.91 (1.43)	40.37 (1.21)	0.075	0.349
	Double support time	20.38 (2.80)	19.38 (2.37)	0.048	0.389
Pace	Stride length	1.31 (0.10)	1.36 (0.11)	0.006	0.537
	Gait speed	1.22 (0.11)	1.29 (0.14)	0.002	0.627
	Foot strike angle	22.59 (3.08)	24.09 (3.60)	0.022	0.448
Turning	Turn duration	1.65 (0.20)	1.62 (0.19)	0.244	0.113
	Turn velocity	124.37 (18.49)	126.96 (15.48)	0.304	0.099
1MW ST
Variability	Stride time SD	0.15 (0.02)	0.14 (0.02)	0.086	0.166
	Single support time SD	0.84 (0.10)	0.82 (0.09)	0.165	0.271
	Double support time SD	1.02 (0.15)	1.01 (0.11)	0.465	0.143
	Foot strike angle SD	1.35 (0.25)	1.36 (0.24)	0.770	0.028
	Stride length SD	0.21 (0.03)	0.21 (0.04)	0.884	0.014
Rhythm	Stride time	1.06 (0.08)	1.05 (0.07)	0.432	0.153
	Single support time	40.52 (1.61)	40.90 (1.24)	0.185	0.260
	Double support time	18.88 (3.22)	18.12 (2.46)	0.136	0.144
Pace	Stride length	1.36 (0.13)	1.41 (0.10)	0.031	0.209
	Gait speed	1.29 (0.18)	1.35 (0.14)	0.052	0.383
	Foot strike angle	24.89 (3.69)	26.28 (4.02)	0.065	0.360
Turning	Turn duration	1.96 (0.33)	1.83 (0.25)	0.027	0.437
	Turn velocity	213.95 (37.76)	231.51 (41.04)	0.023	0.445
1MW DT
Variability	Stride time SD	0.15 (0.03)	0.14 (0.03)	0.063	0.181
	Single support time SD	0.83 (0.07)	0.78 (0.07)	0.002	0.610
	Double support time SD	1.01 (0.11)	0.97 (0.12)	0.077	0.347
	Foot strike angle SD	1.28 (0.26)	1.31 (0.21)	0.860	0.017
	Stride length SD	0.19 (0.03)	0.20 (0.04)	0.707	0.037
Rhythm	Stride time	1.07 (0.09)	1.05 (0.07)	0.149	0.284
	Single support time	40.43 (1.68)	40.78 (1.26)	0.219	0.242
	Double support time	19.15 (3.29)	18.40 (2.53)	0.195	0.255
Pace	Stride length	1.35 (0.12)	1.40 (0.09)	0.029	0.433
	Gait speed	1.27 (0.17)	1.34 (0.14)	0.024	0.447
	Foot strike angle	24.46 (3.32)	25.80 (3.75)	0.054	0.379
Turning	Turn duration	2.01 (0.32)	1.84 (0.26)	0.004	0.579
	Turn velocity	208.19 (41.64)	229.49 (39.19)	0.006	0.267

Fig. 2.

Effect size calculations for each gait assessment between the mTBI and healthy control groups. A) CAT. B) 1 MW ST. C) 1 MW DT. Statistically significant differences between groups indicated as *p <0.05. Abbreviations: ST: Single- task, DT: Dual-task, 1MW: 1-min Walk Test, CAT: Community Ambulatory Task.

During the 1MW ST, mTBI patients had a significantly slower turn velocity and smaller stride length compared to healthy controls (Table 2). Similarly, patients with mTBI had a significantly longer turn duration compared to healthy controls. All gait measures during 1MW ST showed small effect sizes (Fig. 2B). During the 1MW DT, mTBI patients had a significantly slower turn velocity gait speed, and shorter stride length compared to healthy controls (Table 2). Similarly, patients with mTBI had a significantly longer single support time SD and turn duration, with moderate effect sizes, compared to healthy controls (Fig. 2C; Table 2).

The CAT is significantly associated with the 1MW ST and 1MW DT in most gait variables

For the CAT and 1MW ST and DT, gait variables across all four gait domains showed significant correlation, although more gait variables were significantly correlated between the CAT and 1MW DT (Table 3). The strongest correlations were in gait rhythm and pace, specifically in stride time, double support time, gait speed, foot strike angle, single support time, and stride length (Table 3).

Table 3.

Partial correlation coefficients between CAT versus 1MW ST, and CAT versus 1MW DT in gait variables

Gait Domain	Gait variables	CAT - 1MW ST		CAT - 1MW DT
		Coef.	p value	Coef.	p value
Variability	Stride time SD	0.146	0.150	0.137	0.179
	Single support time SD	0.105	0.303	0.214	0.035
	Double support time SD	0.037	0.720	0.080	0.433
	Foot strike angle SD	0.221	<0.001	0.290	0.004
	Stride length SD	0.030	0.772	0.111	0.278
Rhythm	Stride time	0.704	<0.001	0.773	<0.001
	Single support time	0.716	<0.001	0.786	<0.001
	Double support time	0.714	<0.001	0.787	<0.001
Pace	Stride length	0.667	<0.001	0.698	<0.001
	Gait speed	0.547	<0.001	0.613	<0.001
	Foot strike angle	0.834	<0.001	0.853	<0.001
Turning	Turn duration	0.289	0.004	0.175	0.085
	Turn velocity	0.382	<0.001	0.362	<0.001

Significance denoted by bolded p-value (p <0.05).

Adjusted for age, gender, Body Mass Index, Neurobehavioral Symptom Inventory score.

Abbreviations: ST: Single-task, DT: Dual-task, 1MW: 1-min Walk Test, CAT: Community Ambulatory Task, Coef: Correlation Coefficient

4. Discussion

We utilized wearable sensors to examine gait performance during both community ambulatory (CAT) and laboratory gait tasks (ST and DT walking) in people with and without mTBI. In support of our hypothesis, individuals with mTBI demonstrated deficits across the four gait domains including increased gait variability along with smaller and slower steps and turns during both CAT and 1MW tests compared with healthy individuals. Specifically, there were more significant differences between those with mTBI and healthy controls during the CAT in terms of gait variability, rhythm and pace. Additionally, the CAT exhibited more strongly correlated gait variables with the 1MW DT compared to the 1MW ST, especially in terms of gait rhythm and pace. These results support the potential use of an essentially unconstrained community-based ambulatory assessment that offers a nuanced insight into mobility impairments in mTBI when contrasted with gait evaluations conducted in a laboratory setting. The measurement of these gait characteristics has the potential to inform personalized intervention approaches.

The altered gait variables in individuals with mTBI compared with healthy controls in all three walking tasks (CAT, 1MW ST and DT) are consistent with other studies suggesting that mobility deficits occur after mTBI [10,37–40]. 1MW ST showed slower gait pace and turning following mTBI, which agreed with the previous literature [4,10]. However, the effect sizes for these gait measures were low, and among the gait pace measures, only the stride length showed significant difference. The addition of the cognitive task for 1MW DT showed greater reduction of gait pace (i.e., stride length and gait speed), but did not show significant difference in gait rhythm and variability, except for the variability of the single-support time. The lack of differences in gait rhythm and variability might be due to the controlled nature of 1MW test, with straight walk paths and with minimal environmental challenges often seen in real-world walking. In contrast, CAT revealed greater differences in gait pace, rhythm, and variability compared to both 1MW ST and DT, indicating a potential deficiency in dynamic balance control essential for adapting movements to environmental shifts [1]. It is worth noting that gait temporal variability (i.e., variability of the single/double support time and stride time) was larger following mTBI. These differences in gait variability, along with gait pace, have been suggested to be important characteristics of gait impairment following mTBI [5]. Although it is unclear whether the increased variability is associated with walking in the natural environment or with the added cognitive load, our results suggest that the CAT reveals underlying gait deficits not apparent in laboratory gait assessments. Individuals with mTBI demonstrated impaired ability to turn during the laboratory walking that included 180 degree turns, which agreed with other studies that have shown impaired turning in laboratory assessments after mTBI [3,20]. To our surprise, no group difference was seen in turning during the CAT. The lack of differences observed in turn velocity and turn duration during the CAT might arise from the fact that the CAT often included smaller than 180-degree turns, primarily 90-degree turns. Assessments similar to the CAT that include turns and non-linear paths could prove to be an important outcome in clinical evaluations in mTBI, as reported in individuals with a unilateral vestibular loss [33]. Such an objective evaluation of mobility in real-life situations could complement both laboratory and clinic-based assessments reliant on subjective questionnaires, as the latter may not accurately reflect genuine functional capabilities. Therefore, incorporating a community ambulatory task into gait assessments, which mirrors real-world gait, appears effective in distinguishing individuals with gait deficits after mTBI, potentially enhancing the accuracy of assessments.

Our study showed that gait variables, mainly within the domains of gait rhythm and pace, were more strongly associated between the CAT and 1MW DT compared to the CAT and 1MW ST. The closer resemblance of the CAT with 1MW DT might be due to the added cognitive tasks instead of focusing solely on gait. These results suggest that a clinic-based DT walking task could be used when a community ambulatory task is not feasible or available. While the 1MW DT offered fewer group differences than the CAT, it may be easier and faster to assess in clinical settings. Further, some variability gait variables were significantly associated between the CAT with 1MW DT than with 1MW ST, indicating that DT walking may have a closer connection to real-world gait performance. Additionally, an increase in symptoms post-mTBI has been found to be linked to greater variability, slower pace, and turning, particularly evident during dual-task gait conditions [10].

Our use of wearable sensors to document gait deficits during community ambulation in people with mTBI is novel and the findings from this study should be interpreted with caution. First, the application of wearable inertial sensors is not widely adopted in clinical settings, which might limit the generalizability of the findings. Another limitation was the absence of restrictions on participant behaviors throughout the 6–7-minute duration of the CAT, which could potentially influence internal validity. However, considering that the main aim of this study was to evaluate participants in a manner reflective of real-world conditions, we deemed it crucial not to impose constraints aiming to control performance variability among individuals. While it has been suggested that various of turning assessments are comparable across multiple sites [25], further research is required to establish the consistency of gait metrics during the CAT. Additionally, the mTBI group included individuals with diverse mechanisms of injury rather than being restricted to a single cause. Despite the ongoing gaps in understanding the link between injury mechanisms and symptoms, the heterogeneity within the mTBI exposures could limit findings that may be more evident in a more homogeneous group analysis. Furthermore, our study focused solely on gait variables as outcome measures. Subsequent research endeavors should encompass head-trunk coordination, asymmetry, fatigue, or adaptability measures to delve deeper into the sensitivity of these factors in detecting and monitoring post-mTBI deficits. Lastly, multiple comparison corrections were not applied when comparing gait metrics due to the exploratory nature of this study. Hence, more robust studies are required to investigate the feasibility of utilizing identified gait metrics to evaluate gait performances, especially during the CAT, following mTBI.

In conclusion, this is the first study that employed wearable sensors to scrutinize gait performance in individuals with and without mTBI, both in laboratory and community ambulatory settings. The results validated our hypothesis by revealing gait deficits across domains including increased gait variability, reduced pace and turns in individuals with mTBI compared to healthy controls across various walking tasks. Our results and also highlighted the potential benefits of the community ambulatory assessment. Our study underscores the potential significance of quantifying community ambulation as part of comprehensive evaluations, providing a real-world perspective that complements traditional assessments reliant on subjective questionnaires.

Data Availability

Data will be available to qualified investigators through the Federal Interagency Traumatic Brain Injury Research (FITBIR) Informatics System at doi: 10.23718/study/390, through the National Institutes of Health Center for Information Technology: https://fitbir.nih.gov/content/access-data.