Slip-Fall Predictors in Community-Dwelling, Ambulatory Stroke Survivors: A Cross-sectional Study

Abstract

Background and Purpose:

Considering the multifactorial nature and the often-grave consequences of falls in people with chronic stroke (PwCS), determining measurements that best predict fall risk is essential for identifying those who are at high risk. We aimed to determine measures from the domains of the International Classification of Functioning, Disability and Health (ICF) that can predict laboratory-induced slip-related fall risk among PwCS.

Methods:

Fifty-six PwCS participated in the experiment in which they were subjected to an unannounced slip of the paretic leg while walking on an overground walkway. Prior to the slip, they were given a battery of tests to assess fall risk factors. Balance was assessed using performance-based tests and instrumented measures. Other fall risk factors assessed were severity of sensorimotor impairment, muscle strength, physical activity level, and psychosocial factors. Logistic regression analysis was performed for all variables. The accuracy of each measure was examined based on its sensitivity and specificity for fall risk prediction.

Results:

Of the 56 participants, 24 (43%) fell upon slipping while 32 (57%) recovered their balance. The multivariate logistic regression analysis model identified dynamic gait stability, hip extensor strength, and the Timed Up and Go (TUG) score as significant laboratory-induced slip-fall predictors with a combined sensitivity of 75%, a specificity of 79.2%, and an overall accuracy of 77.3%.

Discussion and Conclusions:

The results indicate that fall risk measures within the ICF domains—body, structure, and function (dynamic gait stability and hip extensor strength) and activity limitation (TUG)—could provide a sensitive laboratory-induced slip-fall prediction model in PwCS.

INTRODUCTION

Falls among people with chronic stroke (PwCS) are a serious threat, especially among the PwCS who achieve community ambulation and are exposed to external environmental disturbances, such as slips or trips.^1–3 Recovery from slips is more challenging than trips,⁴ and could potentially lead to more slip-falls than trip-falls. Furthermore, compared with trip-falls that commonly lead to wrist fractures, slip-falls can result in severe physical consequences such as hip fractures and further lead to psychosocial, and socioeconomic distress.^5–8 Thus, there is a need to identify PwCS at fall risk, especially slip-fall risk. However, considering the varying severity of stroke-specific impairments and the multifactorial nature of falls, it is challenging to identify measures that best predict slip-fall risk among PwCS.⁹

Several measures have been established as fall risk predictors for PwCS. Common clinical measures include Berg Balance Scale (BBS), Timed Up and Go (TUG), Fugl-Meyer Assessment (FMA), and muscle strength, which assess various intrinsic fall risk factors such as balance, functional mobility, sensorimotor, and strength impairment, respectively.^10–12 While these measures may be used to predict fall risk,^13,14 findings from other studies either were inconclusive or showed no difference in performance of these measures between fallers and nonfallers.^15,16 Moreover, individuals who performed well were still at risk for falling, possibly due to the ceiling effect, especially among high-functioning PwCS.^17–19 Furthermore, these measures assess an individual’s performance on discrete tasks, which may not be representative of real-life fall scenarios. Additionally, these measures do not directly assess recovery responses needed to prevent an impending fall. Thus, it is essential to further determine whether such impairments can identify PwCS at fall risk.

It has been suggested that a comprehensive model that goes beyond physical functioning could better predict fall risk among PwCS.^20,21 Psychosocial factors assessed using the Activities-specific Balance Confidence (ABC) Scale and Community Integration Questionnaire (CIQ) have been identified as fall risk predictors.^20,22 Studies indicate that fall history may induce fear of falling, resulting in self-imposed restriction of physical activity and declining functional capacity, further predisposing individuals to fall risk.^23,24 However, few studies demonstrated a discrepancy between self-efficacy scores and fall risk.^25,26 As self-efficacy refers to beliefs about one’s own ability as well as functional limitations,²⁷ we postulate that there may be a mismatch between one’s own perception of their balance control and their actual gait impairments. Such a mismatch would limit the contribution of fear of falling for fall risk prediction.

Due to the limitations of aforementioned measures, studies have used objective measures for fall risk prediction.^28,29 For example, the Sensory Organization Test (SOT) is a form of computerized posturography that quantifies postural sway under various sensory conditions.^30,31 SOT composite scores are known to predict fall risk in PwCS.³² However, as most falls occur during walking, assessing walking stability may provide better fall prediction than posturography during quiet standing. Over the years, a conceptual framework has been developed that humans must simultaneously control their center-of-mass (COM) position and velocity (COM motion state) to prevent a fall. This gave rise to the biomechanical measure of dynamic gait stability, which measures the distance from a person’s instantaneous COM motion state to their limits of stability.³³ Although dynamic gait stability is a significant fall risk predictor in healthy older adults,³⁴ it is unclear whether it can potentially identify slip-fall risk among PwCS.

Although several studies have examined fall risk predictors under one of the International Classification of Functioning, Disability and Health (ICF) domains, limited studies included predictors across all domains.^10,35 Additionally, previous studies consider retrospective fall history to classify fallers and nonfallers, which could induce inaccuracies in fall risk prediction due to recall bias or misunderstanding a fall event. Secondly, falls can include both slips and trips and results may be biased depending on the fall type prevalent in the study. To overcome these barriers, researchers have now successfully induced realistic slip and trip falls in laboratory settings. Previous literature indicates that real-life like perturbations can be simulated in safe, controlled laboratory environment, thus providing an opportunity to validate causal-effect relationship of falls and offer a diagnostic and prognostic tool to detect and prevent real-life falls albeit in a safe and controlled laboratory setting.^34,36 Considering the impact of slip-falls, this study focused on examining fall risk predictors in response to realistic, laboratory-induced slip-falls. Further, to conduct a stochastic study to systematically and comprehensively examine slip-related fall risk factors in PwCS, there would be a need to collect enough number of real-life slip-falls in community-dwelling ambulatory individuals with stroke, resulting in a longitudinal study design. However, longitudinal studies take substantial amount of time and are prone to significant attrition of sample size. Further, falls collected in real life may not be accurately and sensitively reported due to the lack of understanding of a fall event, recall bias, and inability to differentiate or recall fall type such as a slip or trip. Thus, the purpose of this study was to examine fall risk predictors across the ICF domains to obtain a model that can best predict fall risk in PwCS in a safe laboratory environment.

METHODS

Sixty-one community-dwelling PwCS (stroke onset >6 months) capable of independently walking 7 m (length of the instrumented walkway) with no complaints of musculoskeletal or cardiovascular conditions, systemic disorders, or any other neurological conditions were included in this study. Participants with low bone density (T score < −2), cognitive deficits (score of ≤26/30 on the Montreal Cognitive Assessment Scale), or speech deficits (aphasia score of ≥71/100 on the Mississippi Aphasia Screening Test) were excluded. Following the screening, clinical measures were performed, and previous/retrospective 1-year fall history was recorded. This study was approved by the Institutional Review Board of the University of Illinois at Chicago, and written informed consent was obtained from all participants prior to their enrollment.

Slip Simulation and Fall Reproduction

Data were collected by 2 trained research assistants who were responsible for conducting the laboratory slip test sessions and performing the assessments described in the sections that follow. Slip simulation and fall reproduction was induced by a low-friction, moveable platform using a computer-controlled release mechanism on a 7-m instrumented walkway. Participants first had to walk 25 to 35 unperturbed baseline trials to become familiar with the laboratory walking environment before receiving a single, novel slip perturbation. Previous evidence suggests laboratory-induced slip fall incidences are similar for both paretic and nonparetic limbs.³⁷ However, as stability is lower during paretic slips than nonparetic slips,³⁷ the former could potentially be more harmful. Thus, a single novel paretic slip was induced to assess and validate various fall risk measures. Upon touchdown of the paretic foot, the platform unlocked to induce a slip and would slide a maximum slip distance of 45 cm at a velocity of 1.96 m/s.

Participants wore athletic shoes, orthoses (if any), and a full body safety harness throughout the experiment. The harness was attached to a ceiling mount with a load cell that recorded the weight supported by the harness. Participants were told to walk at their self-selected speed, and baseline walking trials were collected to ensure they consistently landed on the moveable plate with their paretic limb. Following baseline walking trials, participants were informed that a slip may occur at any time and they should try to recover their balance and continue walking. At regular intervals throughout the experiment, participants’ vitals (blood pressure, heart rate, and oxygen saturation) were monitored. Further, fatigue was assessed using a Visual Analogue Fatigue Scale (VAFS), which assessed fatigue on a scale of 0 to 10 points, with 0 being worst and 10 being normal. As fatigue can influence balance in PwCS,³⁸ and thereby affect perturbation response, fatigue was also assessed before the experiment started and after the baseline walking trials (and before the slip trial). Before the experiment started, the mean VAFS score was 8.10 (1.40) on 10. The mean VAFS score did not change significantly after baseline walking and 7.25 (1.67) on 10. A forward slip under the paretic limb was induced with no warning or practice, thereby simulating realistic slips while walking. Slip outcomes were classified as either a fall or recovery based on the force data recorded by the load cell. A fall was identified when the load cell force exceeded 30% of the participant’s body weight and was verified via visual inspection of all video records.³⁹ Following the slip trial, the participants were subjected to the NASA task load index, which is a tool for measuring and conducting a subjective mental workload assessment while performing a task (in this case, a slip trial).⁴⁰ It rates performance on 6 dimensions, that is, mental demand, physical demand, temporal demand, effort required, performance, and frustration level.⁴⁰

OUTCOME MEASURES

Body Structure and Function Domain

Dynamic Gait Stability Assessment

The dynamic gait stability assessment is an instrumented measure that considers both COM position and its instantaneous velocity.^33,34 The regular walking trial at a self-selected speed immediately prior to the unannounced slip was analyzed to obtain dynamic gait stability. Helen Hayes marker sets with 30 full-body reflective markers were attached to anatomical landmarks to assess stability. Marker coordinates were recorded using an 8-camera motion capture system (Motion Analysis Corporation, Santa Rosa, California). Marker displacement data were low-pass filtered at optimal cutoff frequencies ranging from 4.5 to 9 Hz using a fourth-order Butterworth filter. Force plate data, harness load cell data, and trigger-release onset signals were collected at 600 Hz using a 64-channel, 16-bit analog-to-digital converter. The ground reaction force and motion data were time synchronized using a trigger signal during data collection. The vertical ground reaction force data were used to determine the gait events (step touchdown and lift-off) and verified from foot kinematic data collected via 3D motion analysis.

The motion analysis data were used to compute the absolute COM position using a 12-segment body representation model. The instantaneous COM velocity was calculated by taking the time derivative of the COM position trajectory. The COM position was expressed relative to the heel of the base of support (BOS), normalized by the participant’s foot length, whereas velocity was expressed relative to the BOS, normalized by a dimensionless fraction of $\sqrt{g\times h}$ (g = gravitational acceleration, h = participant’s height). Thus, dynamic gait stability was measured as the shortest distance between the instantaneous COM motion state and previously predicted computational limits of the feasible stability region.⁴¹ Stability values more than 0 (more positive) indicate decreased likelihood of backward balance loss whereas stability values less than 0 (more negative) indicate greater likelihood of backward balance loss.⁴²

Sensory Organization Test

We assessed postural sway under progressively challenging sensory conditions (eyes open, eyes closed, platform moving with eyes open, platform moving with eyes closed, surrounding moving with eyes open, both surrounding and platform moving with eyes open) of the SOT (Balance Master, NeuroCom International, Inc, Clackamas, Oregon). Two 20-second trials were performed for each condition. A composite score ranging from 0 (higher sway) to 100 (no sway) was calculated as the weighted average for all conditions.^43,44

Fugl-Meyer Assessment Scale for Lower Extremity

The Fugl-Meyer Assessment Scale for Lower Extremity (FMA-LE) is a stroke-specific, performance-based index that assesses sensorimotor impairment based on ratings of reflexes, movement, coordination, joint range of motion, and joint pain. Each item was scored on a 3-point scale, with scores ranging from 0 to 34.^45,46

Strength Assessment

Maximum voluntary isometric force-generating capacity was assessed for paretic lower limb muscles (hip flexors and extensors, knee flexors and extensors, ankle plantarflexors, and dorsiflexors) using an isokinetic dynamometer (Biodex System 3 Pro, Biodex Corp, Shirley, New York).⁴⁷ Participants performed 3 trials for each movement with rest periods between trials. They were instructed to exert their maximum effort and were verbally encouraged throughout the assessment.

Activities-Specific Balance Confidence Scale

The ABC Scale rates a participant’s confidence level from 0% to 100% in 16 functional activities of daily living. The self-rated scores for each activity are averaged to yield the mean ABC score. Higher scores represent higher balance confidence.^48,49

Activity Limitation Domain

Berg Balance Scale

The BBS is a measure composed of 14 functional tasks. Each task was scored on a scale from zero to 4 based on participants’ ability to meet task-specific goals for a maximum possible score of 56.^50,51

Timed Up and Go

The TUG assesses functional mobility. Participants were instructed to stand up from a chair, walk for 3 m, turn around, return to the chair, and sit back down “as quickly and as safely as possible.” The score was the time taken in seconds to complete the test.^52,53

Participation Restriction Domain

Physical Activity Scale for the Elderly

The Physical Activity Scale for the Elderly (PASE) measures self-reported physical activity and determines the frequency and duration spent on leisure, household, and occupational activities over a 7-day period. Higher PASE scores represent greater physical activity, with scores ranging from 0 to 400 or more.^54,55

Community Integration Questionnaire

The CIQ consists of 15 items regarding participants’ activities from a home, social, and community integration perspective. The basis for scoring is primarily the frequency of activity performance with secondary weightage given to whether activities were performed alone, with someone’s help, or completely by someone else. Scoring ranges between 0 and 29 points.^56,57

Contextual Factors

The ICF consists of contextual factors (environmental and personal). This study included fall history as a personal contextual factor. A previous fall experience can affect a person’s participation level, thereby affecting activities and, in turn, body structure and function.⁵⁸ One-year fall history was recorded using a self-reported questionnaire. A fall was defined as an event where the participant unintentionally came to rest on the ground or another lower level.⁵⁹

Data Analysis

Data for all variables were summarized using descriptive statistics (means ± standard deviations) and compared using t tests between slip-induced fallers and nonfallers. A Spearman correlation was performed to examine the relationship between fall history and laboratory-induced slip-fall outcome. Univariate logistic regression was performed to identify predictors of laboratory-induced slip-falls. Sensitivity, specificity, and overall accuracy were calculated for each variable. Subsequently, variables with a significance level of less than 0.3 were included for further analysis in the multivariable regression model.^34,60 A backward stepwise logistic regression was performed to determine a model that could best predict laboratory-induced slip-falls. Odds ratio was also calculated. The area under the curve (AUC) was found using the receiver operating curve (ROC) for variables as well as for the best predictive model. The Youden index (sensitivity + specificity − 1) was calculated based on the ROC, and the maximum value of the index enables the selection of an optimal cutoff score.⁶¹ Thus, cutoff scores were calculated for all variables in the best predicted model based on the Youden index. Statistical significance was established at an α level of 0.05 throughout the study. All statistical analyses were performed using SPSS 25.0 software (IBM Corp, Armonk, New York).

RESULTS

Of the 61 participants, 3 had missing stability data due to missing markers for motion marking; for 2 participants, a slip was not induced appropriately due to technical issues. Of the 56 included participants, 42% (24) fell and 57% (32) successfully recovered from the laboratory-induced slip. There was no effect of age, height, and weight of study participants on the perturbation outcome. Further, there was no correlation between fall history and laboratory-induced slip-fall outcome (r = 0.102, P = 0.417). Scores on the TUG (P = 0.01) and hip extensor strength (P = 0.04) were significant measures indicating that slip-induced fallers walked slower (performed TUG slower) and had less strength compared with nonfallers. Dynamic gait stability was also significantly different between slip-induced fallers and nonfallers (P = .01) (see Table 1).

Table 1.

Sample Demographics Organized by Slip Outcome (Fall vs Recovery During the Laboratory Test) With Their Respective Means, Standard Deviations, and Statistical Significance Based on t-Test Results

Variables	Fall (n = 22) Mean (SD)	Recovery (n = 24) Mean (SD)	P Value
Age, y	57.16 (10.33)	59.12 (10.27)	0.49
Weight, kg	98.88 (19.24)	93.16 (16.35)	0.32
Height, m	1.71 (0.10)	1.73 (0.10)	0.22
TUG, s	17.78 (8.44)	12.79 (4.48)	0.007
Stability	−0.19 (0.05)	−0.12 (0.09)	0.006
BBS (out of 56)	48.41 (5.92)	49.33 (4.82)	0.53
FMA	24.12 (4.18)	26.83 (4.24)	0.08
Fall history, %	14/22	11/24	0.34
ABC, %	76.87 (16.39)	81.69 (14.99)	0.26
CIQ	16.63 (4.45)	16.75 (4.31)	0.92
PASE	95.13 (61.54)	80.04 (45.48)	0.30
SOT (out of 100)	75.20 (5.93)	73.57 (10.39)	0.50
Hip flexors, Nm	27.53 (19.25)	28.75 (13.28)	0.51
Hip extensors, Nm	91.82 (32.11)	120.73 (45.93)	0.02
Knee flexors, Nm	46.31 (21.59)	40.53 (22.33)	0.38
Knee extensors, Nm	54.96 (23.66)	62.30 (26.61)	0.33
Dorsiflexors, Nm	21.81 (13.38)	22.08 (15.99)	0.95
Plantarflexors, Nm	36.29 (26.26)	36.67 (23.88)	0.96

Abbreviations: ABC, Activities-specific Balance Confidence Scale; BBS, Berg Balance Scale; CIQ, Community Integration Questionnaire; FMA, Fugl-Meyer Assessment; PASE, Physical Activity Scale for the Elderly; SOT, Sensory Organization Test; TUG, Timed Up and Go test.

Based on the Youden index, the cutoff score determined for dynamic stability was −0.15 and for TUG was 13 seconds in our study population of PwCS. The cutoff score for hip extensor strength was determined at 106 Nm. Additionally, ROC results show the AUC for TUG was 0.701 and 0.713 for hip extensor strength, whereas the AUC was 0.782 for dynamic gait stability, indicating a higher prediction capacity of dynamic gait stability (see Table 2).

Table 2.

Sensitivity, Specificity, Overall Prediction Accuracy, and Significance From Univariate Regression Analyses With Area Under Curve Derived From Receiver Operating Curve

Variable	Sensitivity	95% CI	Specificity	95% CI	Overall Accuracy	P Value	AUC	95%CI
TUG, s	41.72	27–67	81.81	51–87	64.91	0.01	0.701	0.564–0.838
Stability	58.34	40–70	80.62	56–90	70.93	0.01	0.782	0.657–0.907
SOT (out of 100)	47.61	27–67	65.25	48–82	56.82	0.49	0.476	0.326–0.626
FMA	29.22	37–77	81.84	58–88	59.64	0.10	0.609	0.452–0.766
BBS (out of 56)	8.30	10–27	97.01	80–94	59.67	0.73	0.542	0.387–0.696
Fall history	66.71	47–85	52.23	33–71	57.96	0.33	0.564	0.413–0.716
ABC, %	16.73	22–62	90.91	71–93	59.65	0.26	0.595	0.437–0.754
CIQ	28.65	11–45	73.92	58–88	57.92	0.92	0.536	0.384–0.688
PASE	42.91	22–62	78.34	67–99	61.44	0.46	0.570	0.409–0.731
Hip flexors, Nm	5.32	16–54	96.04	49–87	56.85	0.51	0.499	0.322–0.676
Hip extensors, Nm	60.05	55–89	75.02	30–70	68.27	0.04	0.713	0.556–0.869
Knee flexors, Nm	15.08	10–44	89.37	60–90	58.36	0.37	0.414	0.250–0.578
Knee extensors, Nm	19.02	27–67	78.64	30–70	53.11	0.32	0.549	0.385–0.714
Dorsiflexors, Nm	21.11	6–36	70.02	55–85	57.18	0.94	0.502	0.334–0.669
Plantarflexors, Nm	44.42	25–63	57.91	38–76	57.45	0.95	0.522	0.350–0.694

Abbreviations: ABC, Activities-specific Balance Confidence Scale; AUC, area under curve; BBS, Berg Balance Scale; CI, confidence interval; CIQ, Community Integration Questionnaire; FMA, Fugl-Meyer Assessment; PASE, Physical Activity Scale for the Elderly; SOT, Sensory Organization Test; TUG, Timed Up and Go test.

Dynamic gait stability, TUG, hip extensor strength, FMA, and ABC scores were included for the multivariable analysis. The analysis resulted in a model with TUG, hip extensor strength, and dynamic gait stability as significant predictors of falls with a sensitivity of 75%, a specificity of 79.2%, an overall prediction accuracy of 77.3% and AUC at 0.863 (see Table 3).

Table 3.

Components of the Predictive Models Using Multivariable Logistic Regression

Model	Independent Variables	P Value	Odds Ratio (95% CI)	Cutoff	Sensitivity	Specificity, %	Overall Accuracy, %	AUC	95% CI
1	Stability	0.011	2.64 (1.34–4.28)	−0.156	75.0	79.2	77.3	0.863	0.754–0.971
	TUG, s	0.058	2.35 (1.46–3.88)	13
	Hip extensor strength, Nm	0.084	2.12 (1.58–3.64)	106

Abbreviations: AUC, area under curve; CI, confidence interval; TUG, Timed Up and Go test.

DISCUSSION

Our study findings indicate a model composed of dynamic gait stability, hip extensor strength, and TUG score was a significant predictor of laboratory-induced slip-fall risk in PwCS.

Fall Risk Predictors in the Body Structure and Function Domain

Dynamic gait stability was a predictor of laboratory-induced slip-falls in PwCS. The ability to maintain stability is assessed by the ability to control COM movement relative to BOS.^62–65 Our study findings showed slip-induced fallers had more negative dynamic gait stabilities than nonfallers on laboratory-induced slip-falls. Previous literature also demonstrates that reduced dynamic gait stability is one of the causative factors of falls.⁶⁶ Thus, negative dynamic gait stability not only increases fall risk but may also make recovery from slips more difficult. Our study results provide preliminary evidence that dynamic gait stability could predict slip-fall risk in PwCS.

In our study, only hip extensor muscle strength was a significant predictor of laboratory-induced slip-fall risk in PwCS. Our results are consistent with previous studies indicating that hip extensor strength is an effective measure of determining falls.⁶⁷ Previous literature on laboratory-induced falls has indicated that, in addition to poor stability, reduced limb support or unintended hip descent can result in a fall.^66,68 Unintended hip descent can be caused by inadequate concentric work by the hip extensors.⁶⁸ Thus, weak paretic hip extensor strength can result in descent of the slipping (paretic) limb, thereby further limiting the ability of the paretic limb to weight bear while the nonparetic limb executes a compensatory step to reestablish the BOS, recover stability, and regain limb support to prevent a fall.

SOT was not predictive of laboratory-induced slip-fall risk, most likely due to its limited ability to assess balance in dynamic upright conditions. Given the ambulatory nature of our study population, it can be argued that such perturbations as well as the nature and intensity of these perturbations may be less challenging or insufficient for testing fall-resisting skills and may not be comparable to realistic slip perturbations, thereby limiting its sensitivity in fall risk prediction. Also, most of the studies done using SOT as a fall risk predictor considered retrospective fall history. Hence, it is unclear whether SOT can predict immediate or future fall risk.

The FMA was not able to differentiate slip-induced fallers from nonfallers, probably because FMA rates an individual on discrete motor tasks and does not assess responses required to prevent a fall. Further, the interval nature of the scale may have limited its sensitivity in fall risk prediction. Similarly, the ABC was not able to predict laboratory-induced slip-fall risk. The mean ABC score for the faller group was 76.87%; however in spite of the score, approximately one-half of the study participants fell upon experiencing a slip. Thus, there may be a mismatch between individuals’ self-perception of their own abilities and their functional mobility, balance, and gait impairment in challenging balance conditions.

Fall Risk Predictors in the Activity Limitation Domain

TUG was a significant predictor of laboratory-induced slip-falls in PwCS, probably because tasks performed during TUG such as sit-to-stand, walking, and turning mimic daily activities that increase real-life fall risk. Additionally, our results indicate that slip-induced fallers had a higher TUG score (slow gait speed) compared with nonfallers; slower walking speed is associated with decreased stability against a backward loss of balance.⁶⁹ Thus, participants who took more time to complete the TUG were at higher risk of fall on exposure to the laboratory-induced slip.

The BBS was not able to predict laboratory-induced slip-fall risk. Prior studies suggested BBS was a significant predictor of falls during the acute and postacute stroke periods⁷⁰; therefore BBS might have limited sensitivity to predict laboratory-induced slip-fall risk in PwCS. Based on previous studies, which showed a BBS score above 45 to be indicative of reduced fall risk,⁷¹ there should have been comparatively less falls in our study as participants had a mean BBS score of 48.41. However, despite the higher scores, about half of individuals fell on the laboratory-induced slip perturbation, indicating that BBS might not accurately predict laboratory-induced slip-falls in PwCS.^15,20

Fall-Risk Predictors in the Participation Restriction Domain

The PASE and the CIQ were not predictors of laboratory-induced slip-fall risk. The PASE requires recall of physical activity events over a 7-day period, and it is possible that recall bias may have been partially responsible for this lack of relationship. Furthermore, the PASE comprises duration and frequency of activities such as gardening, wall papering, and lawn work that may not be performed by PwCS. In addition to memory or recall bias, study participants may have felt the need to provide socially desirable answers, which might further limit the sensitivity of these measures.

Contextual Factors as Fall Risk Predictors

Fall history could not predict laboratory-induced slip-fall risk, most likely owing to the recall bias associated with correctly remembering previous fall types (slip or trip). Prior studies, which suggested fall history was a predictor of falls, mostly considered recurrent fall history^14,72; however, most study participants reported a single fall history, which may not be sufficient to classify individuals as “fallers.”

Fall Risk Predictors From Multivariate Logistic Regression Analysis

The multivariate logistic regression analysis determined a model composed of dynamic gait stability, hip extensor strength, and TUG to predict laboratory-induced slip-fall risk. This indicates measures from the body structure and function and activity limitation domain of the ICF can be sensitive predictors of laboratory-induced fall risk. Additionally, the cutoff scores determined for these measures are consistent with the previous literature. The cutoff score determined for dynamic stability was −0.15, indicating that PwCS with more negative values are at higher fall risk. Our cutoff score is similar to a previous study in older adults indicating that older adults with a dynamic gait stability cutoff score of less than −0.15 have 69% sensitivity of fall risk.³⁴ Similarly, the cutoff score for TUG in our study was 13 seconds, indicating that PwCS taking more than 13 seconds to complete the test were at an increased fall risk. Similar cutoff scores for TUG were reported in a previous study, demonstrating the sensitivity of TUG in fall prediction was almost 50% when PwCS took more than 14 seconds in the TUG test.¹² Thus, given the range and complex interaction of factors that contribute to falls in PwCS, it is essential to assess all aspects of fall risk. The ICF provides an ideal framework for organizing these various aspects, and our study results show that a comprehensive analysis provides a sensitive fall risk prediction model.

Clinical Implications

The findings of our study indicate measures from the ICF assessing various fall risk factors may provide a sensitive model for laboratory-induced slip-fall risk prediction and can be implemented to identify PwCS at risk for slip-fall. Assessment of dynamic gait stability, TUG, and hip extensor strength should be included for fall risk prediction. The latter two can be assessed clinically. The TUG is an easy to perform and requires minimal space and equipment; likewise, hip extensor strength can be assessed using an instrumented hand-held dynamometer with appropriate stabilization of the lower limb in the clinical settings.⁷³ Further, with the advent of commercially available cost-effective wearable sensors, dynamic stability could be easily computed using a single sensor affixed to the sacrum.⁷⁴ These measures can help identify problems at different ICF levels, and PwCS identified as being at risk for slip-fall could undergo training protocols to prevent falls.

Limitations

However, the study has few limitations. As the study included ambulatory PwCS, these results cannot be generalized to individuals in the acute stroke stage with limited ability to ambulate. Furthermore, although balance is a crucial factor for preventing falls, BBS was not a significant predictor of fall risk. Future studies should consider including the mini-BESTest (a measure assessing reactive balance),⁷⁵ as it could be a sensitive predictor of fall risk. Lastly, the study recorded fall outcomes based on the participant’s performance in a laboratory-induced slip test. While this setting was chosen to ensure safety and recording accuracy, it is difficult to mimic a real-life fall event. Like walking speed, fall-slip risk may vary over the course of an episode of walking, which may be an important consideration for PwCS who walk in the community.⁷⁶

CONCLUSIONS

A model composed of outcomes of dynamic gait stability, TUG, and hip extensor strength can be used by health care professionals in clinical settings to predict slip-fall risk in PwCS and differentiate fallers from nonfallers. Future fall risk prevention paradigms focusing on improving these measures may help reduce fall risk in PwCS.