Wearable inertial sensors to measure gait and posture characteristic differences in older adult fallers and non-fallers_ A scoping review

Background: Wearable inertial sensors have grown in popularity as a means of objectively assessing fall risk. This review aimed to identify gait and posture differences among older adult fallers and non-fallers which can be measured with the use of wearable inertial sensors. In addition to describing the number of sensors used to obtain measures, the concurrent anatomical locations, how these measures compare to current forms of clinical fall risk assessment tests and the setting of tests. Methods: Following the development of a rigorous search strategy, MEDLINE, Web of Science, Cochrane, EMBASE, PEDro, and CINAHL were systematically searched for studies involving the use of wearable inertial sensors, to determine gait and postural based differences among fallers or those at high fall risk compared with non-fallers and low fall risk adults aged 60 years and older. Results: Thirty five papers met the inclusion criteria. One hundred and forty nine gait and posture characteristic differences were identified using wearable inertial sensors. There were sensor derived measures which significantly and strongly correlated with traditional clinical tests. The use of a single wearable inertial sensor located at the lower posterior trunk, was most the most effective location and enough to ascertain multiple pertinent fall risk factors. Conclusion: This review identified the capabilities of identifying fall risk factors among older adults with the use of wearable inertial sensors. The lightweight portable nature makes inertial sensors an effective tool to be implemented into clinical fall risk assessment and continuous unsupervised home monitoring, in addition to, outdoor testing.


Introduction
One in three adults over 65 years fall each year [1,2] and this increases to 40% of individuals over 80 years [3,4]. Falls are associated to a considerable increase in morbidity, immobility, mortality and loss of independence [5,6] and can also result in a heavy psychological burden due to a fear of falling [7]. Together these result in diminished mobility, social isolation and reduced quality of life [8,9] posing high economic burdens [10]. Therefore, primary prevention of falls is of paramount importance.
There are a variety of different clinical tests used to quantify fall risk. Examples of these include: the Timed-Up-and-Go test [11], Sit-To-Stand test [12], Dynamic Gait Index [13], Berg Balance Scale [14] and Tinetti Performance Oriented Mobility Assessment [15]. Despite being relatively quick and reliable whilst providing potentially relevant information on fall risk, clinical tests are steeped in a history of subjective observation [16]. Although, some tests such as the Timed-Up-and-Go test is measured by time taken and could be considered objective. Nevertheless, these clinical tests lack construct validity and fail to provide data from activities of daily living in a habitual environment setting, which will undoubtedly be different to a clinician's simulated assessment [17]. There have also been previous suggestions that clinical-based testing may be contaminated by the Hawthorne effect [18]. This is where participants perform differently and at times better than they normally would, due to awareness of being examined. Recent research has suggested clinical assessments may not provide a true reflection of the individual being assessed. Giannouli et al. [17] identified that in older adults without any mobility impairments, mobility capacity related measures had little significance for predicting real-life performance. Additionally, Cofré Lizama et al. [19] reported the ability to perform accurate tracking of mediolateral centre of mass deteriorates with age, meaning that this potentially crucial factor may be missed with current clinical based assessments.
Recently, wireless and wearable technology, such as accelerometers and gyroscopes, have emerged as a potential alternative to clinical/laboratory testing. Such devices have the potential to be a worthy replacement for previously mentioned fall risk assessments and can even be used in conjunction with clinical tests; whilst providing objective data, resulting in more informed decisions regarding fall risk and subsequent fall preventative treatments. A wearable inertial sensor (WIS), is also known as a wearable inertial measurement unit (IMU). They are small, lightweight, inexpensive and does not require the arduous set-up times of traditional motion capture systems. Additionally, it possesses a long battery life and does not have to be confined to a clinic/laboratory. Therefore, WIS can provide continuous real time kinematic data during activities of daily living, as opposed to artificial movements of traditional fall risk assessment tests.
Current literature has identified the use of inertial sensors to detect when falls have occurred [9,20]. Despite potentially reducing time spent on the floor following a fall, the overall health consequences from falling will still remain. Therefore, primary prevention of falls is required and WIS-based technology may possess a pivotal role in achieving this. To successfully prevent falls from occurring whilst using WIS a thorough list of characteristic difference and potential fall risk factors, which have been measured using said WIS, is required. Recent reviews [8] and [21], provided insightful preliminary results regarding feature classification models and fall risk assessment and prediction. However, a rigorous search strategy was not provided, and systematic search results suggest that potentially meaningful papers may have been missed. Additionally, these reviews did not explore the potential clinical relevance of individual sensor derived features and how they correlate with traditional clinical assessments.
Therefore, the aim of this review was to identify, describe and evaluate a thorough list of characteristic differences and potentially fall risk factors related to gait and posture among older adults who are fallers and/or high fall risk, which have been measured with the use of WIS. In addition to, how they compare to current clinical assessments. The study also aimed to describe the number of WIS used, anatomical sensor locations and the setting of tests.

Methods
The review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A scoping review was chosen due to the statistical heterogeneity of the studies within this field, deeming a standard systematic review with meta-analysis to be inappropriate. A protocol was developed following consultation with topic and methodological experts (https://ore.exeter. ac.uk/repository/handle/10871/26862).

Search strategy and eligibility criteria
The search strategy did not have data limits, ensuring a thorough search of the literature was performed. Table 1 highlights the search strategy developed and was applied in the following databases: MED-LINE, Web of Science, Cochrane, EMBASE, PEDro, and CINAHL.
Additionally, to ensure all relevant papers had been retrieved, a forward and backward search of the literature was also performed.
Studies were included within the review providing they successfully met the following selection criteria. All studies must include the use of WIS to identify gait and posture characteristics or fall risk factors, by comparing fallers and/or high fall risk older adults to non-fallers and low fall risk older adults aged over 60 years, who may be community-dwelling, in residential care or hospital patients. Additionally, if a study involved participants younger than 60 years they were still included, if the mean age minus one standard deviation was over 60 years [22]. No restrictions were placed on how measurements featuring WIS were obtained. This could be as participants performed a fall risk assessment task (clinical based) or any other activity away from the clinic/laboratory (e.g., continuous home walking or activities of daily living). All study designs were eligible for inclusion. No language restrictions were imposed on the inclusion criteria. Papers were excluded if they failed to meet any of the aforementioned inclusion criteria. Initial database searches were conducted on 7 th January 2017, with an updated search performed on 20 th April 2018.

Study selection
Two reviewers independently screened all titles and abstracts of retrieved articles from the electronic database searches. Both reviewers then repeated the process whilst screening the full texts of the remaining papers. Any discrepancies were resolved via discussion when necessary. Data were extracted from included studies by one reviewer and checked by a second. The data management process was performed using EndNote X8.0.2 (Clarivate Analytics, USA).

Data extraction
Data were extracted on population characteristics (gender, age, body mass, and clinical condition when applicable), study characteristics (sample size, study duration and data collection setting, study design [prospective or retrospective]), methods used to measure characteristic differences or fall risk factors (quantity and type of inertial sensors used and their anatomical location(s)) and the type of activity performed when measurements were taken (clinical assessment; static or dynamic, or activities of daily living). A characteristic difference was extracted when a statistically significant difference between fallers/ high fall risk versus non-fallers/low fall risk older adults were identified (P < 0.05). How these variables compared to current clinical assessments were extracted from studies who conducted correlation analysis between sensor derived and clinical assessment measures.

Data synthesis
Following data extraction a single reviewer independently synthesised data. All data were initially tabulated and subsequently reported descriptively [23] to answer the key questions of the review.

Risk of bias assessment
In order to determine the quality of the studies within the review a modified version of the Downs and Black Quality Index was used [24]. The original index was modified as it also evaluates clinical trials, these items were subsequently not included. The modified index was comprised of twenty questions related to information reporting (1-9), external validity (10 and 11), internal validity (12)(13)(14)(15) and selection bias (16)(17)(18)(19)(20). Questions were answered with '1′ (yes) and '0′ (no). A score of 0-6 suggested a high risk of bias, 6-13 equals a moderate risk of bias and 14-20 proposed a low risk of bias.

Results
Database searches returned 7252 citations following the removal of duplications. One hundred and sixty six papers underwent full text scrutiny against the inclusion/exclusion criteria. Following this, 35 papers remained and were included within the review (Fig. 1).
The 35 included papers reported 23 studies featuring a comparison of fallers to non-fallers and twelve where subjects were classified as 'high fall risk' compared with 'low fall risk'. There were 2685 participants included although this excluded one study [25] where there was a lack of clarity over sample size, with sample sizes ranging from 12 to 260. Six studies failed to provide sufficient detail on test group characteristics [2,[26][27][28][29][30]. 70 % (1823/2608) of participants were females, although gender was not reported in three studies [25,30,31].
Four different settings were used: a laboratory, home environment, local community centre and a hospital (Tables 2A and 2B). Two studies included mixed environments (laboratory and a home) [32,33]. Of the seven studies set in a home environment, two were short supervised sessions [28,34], whereas five studies involved continuous monitoring over two days to fourteen weeks [27,32,33,35,36].
There were 149 gait and posture differences identified from the 35 studies, as statistically significant differences were observed between groups whilst using WIS. Of this, 127 were dynamic measures (Table 3) whilst 22 were static measures ( Table 4).
Identified characteristics and potential fall risk factors were categorised as temporal, spatial, linear acceleration, angular velocity, position and angle or energy variables. Some characteristics were statistically significant in multiple studies, but were counted only once within the aforementioned totals in the present review. Whereas, many other variables were identified (or measured) in only one study with some variation. Two of the 35 studies failed to reach statistical significance for any WIS based measures between fallers and non-fallers [46,48]. In the event of some studies witnessing a repetition of statistically significant results for the same variables in different conditions, such as: standing with eyes open or eyes closed and on a hard or foam surface [2,28,34,38], during different phases of a clinical assessment [49,52,57], different phases of turning courses [31] and also at different times during the day [28], they have each been recorded once. Tables 3 and 4, contain a full listing of all observed gait and posture differences measured using WIS and the corresponding studies they were identified in. Table 5 highlights the number of measures ascertained from WIS at identified anatomical locations. The lower trunk consisting of the third to fifth lumbar vertebrae, second lumbar vertebrae, sacrum and 'lower back' had the most identified sensor derived measures, with 102 of 172 measures (59.3%). The third to fifth lumbar   [57]. The sternum and sacrum were also desirable locations for various parameters, respectively (20/172 11.6%). If multiple studies had the same location for one measure this was recorded once. However, it should be noted that some measures were determined using multiple anatomical locations, for example, walking speed was measured with a sensor located at the' lower back' [25,33], third to fifth lumbar vertebrae [27,29,31,42,45,49,59], second lumbar vertebrae [57], tenth thoracic vertebrae [58], sacrum [43,50,56] and head [51]. The study by Mohler et al. [34] was omitted from this analysis as five separate inertial sensors were simultaneously used for each measure they reported.
Five studies presented correlation analysis on WIS-based measures to traditional clinical assessment with mixed results [2,25,38,43,56]. Maximum entropy, mediolateral harmonic ratio and the eight step harmonic ratio variation were weakly correlated to walking speed. The root mean square (RMS) of the acceleration signal during quiet standing on a mat with eyes open yielded strong statistically significant correlations with the Timed-Up-and-go test but inversely, albeit strongly correlated with the Berg Balance Scale [2]. The RMS of the angular velocity signal and RMS of the acceleration signal were also strongly correlated with the Berg Balance Scale [38]. Whilst walking speed and step length were found to have strong positive correlations to the Tinetti scale [56]. Four studies did assess the test-retest reliability of WIS-based measures, finding moderate and high reliability [35,41,43,50].
The average score from the risk of bias assessment was 13.2 out of  twenty, suggesting a very low moderate risk of bias. There were 21 (60%) studies which had a moderate risk of bias and fourteen (40%) which had a low risk of bias. Zero studies were determined to have a high risk of bias (Table 6).

Discussion
The aim of the review was to identify characteristic differences and fall risk factors related to gait and posture, which have been measured using WIS in older adults identified as fallers and/or high fall risk. In addition to, how the measures correlated with current clinical tests, the number of sensors used, the concurrent anatomical sensor location(s) and the setting of where measurements took place, have all been identified. Through the use of a systematic methodology, the review was able to effectively explore the aforementioned aims, enabling a rigorous assessment of current literature and areas for future research.
There were 149 potential static (whilst standing still) and dynamic (whilst in motion) gait and posture differences identified that were categorised as temporal, spatial, linear acceleration, angular velocity, position and angle, and energy parameters [8]. Of the 149 differences, 127 were dynamic based measures, whilst 22 were static based measures. Wearable inertial sensor based measurements largely concurred with clinical fall risk assessments [60,61], through correlation analysis, despite being limited to five studies. Therefore, there is a clear scope for WIS to provide a good quantitative alternative to traditional fall risk observations, in the measurement of gait and postural stability for assessing fall risk among older adults [2].

Dynamic tests
Brauer et al. [62] suggests that dynamic tests provide a better prediction of falls, because most falls occur whilst an individual is in motion. Therefore, it is of no surprise that 91% of the studies within the review used a form of dynamic testing (e.g. walking). The primary  dynamic gait difference between fallers and non-fallers identified through WIS coincide with observations from a clinician's fall risk assessment. During an assessment of gait, risk factors include slower walking speed and shorter step and stride lengths, which were consistently found to be characteristic differences in a number of studies. A clinical test of gait interprets fall risk based upon the duration it takes to complete a walking based task, such as the Timed-Up-and-Go test, with a longer time to completion indicating greater fall risk. It is conceivable Lower step/stride frequency [27,42,56,57] Lower stride regularity [42] Slower step/stride speed [28,33,49,57] Slower stride velocity [28] Less single support [28] Greater double support [28,34] Greater swing time variation [28] Greater stride velocity variation [28] Less step time variation [28] Slower maximum walking speed [25,59] Greater step time variability [29,56,59] Longer turn duration [32,57] Slower speed of turning [32] Lower turning cadence [57] Greater mean walk bout duration [34] Lower mean sitting bout duration [34] Greater Timed-Up-and-Go duration [49,57] Longer sit-to-stand duration [52] Longer stand-to-sit transition duration [52] Slower stand-bend time [57] Slower bend-sit time [57] Slower stand-sit time [57] Spatial Shorter stride/step length [27,28,29,42,51,56,58]
that a slower gait with shorter strides and steps is a conscious compensatory mechanism, as older adults adopt a more cautious gait pattern stemming from a heightened fear of falling and decreased muscular strength [28]. This is corroborated by Asai et al. [59], who reported individuals with a fear of falling walked slower than those without a fear of falling. However, as these temporal risk factors may be a conscious alteration made by older adults due to fear of falling, individuals who do not have the same fear may not exhibit such an altered gait pattern. Indeed, Mohler et al. [34] failed to identify any differences in walking speed, stride time and stride length among groups of fallers and non-fallers who were non-frail, pre-frail and frail. Nevertheless, these findings do coincide with previous non-WIS research, where slower walking speeds have been linked to increased fall risk [63]. However, a non-linear relationship between gait speed and falls has been reported [64], which suggests that a slower walking speed equates to greater fall risk when walking indoors, and when walking outdoors a faster gait speed resembles a higher risk of falls. Despite this, all studies included within the review were conducted indoors, including the seven studies conducted in the home environment. Consequently, there is a lack of information on faller and non-faller characteristics and subsequently potential fall risk outdoors. The portable nature of WIS allow for outdoor testing to provide a more complete assessment of fall risk and thus, should be incorporated into future research to establish whether fall risk factors associated with outdoor activities can be identified. Additionally, outdoor unsupervised testing will reduce the possibility of the Hawthorne effect, where subjects are consciously aware of being tested.
A consistent finding was that fallers and high fall risk older adults walked with a less smooth and less stable gait pattern, but the location of the sensor varied and there were inconsistencies as to the optimal location (upper or lower trunk). The majority of studies positioned an inertial sensor at the lower trunk, but it was the upper trunk that yielded the highest specificity for predicting future falls in a prospective study design [45]. Vertical, anteroposterior and mediolateral acceleration root mean squares (RMS) were found to be lower among fallers and high fall risk older adults, although findings were mixed on centre of mass acceleration. The acceleration signal during the walking phase is a measure of balance during gait and increased acceleration variability during steady state gait is indicative of reduced balance [65]. However, it is suggested that a lower acceleration RMS during walking represents higher gait instability [47,66,67], as individuals are walking more rigidly [57]. This coincides with Van Schooten et al. [27] who reported a higher logarithmic rate of divergence which suggests a lower local stability (greater movement from a point of equilibrium) as a prospective fall risk factor from a home based study. This perhaps confirms older adults who are fallers walk rigidly causing greater instability during gait. These WIS-based findings not only coincide with clinical observations but assist in giving clinicians further pertinent information regarding an individual's fall risk which may otherwise be missed.

Static tests
Traditionally, measurements of postural kinetics are confined to a laboratory and based upon centre of pressure movement whilst standing on a force plate [37]. This requires expensive equipment which is mounted into a surface rendering it non-portable. Nevertheless, cheaper and more commercially available alternatives such as: Nintendo Wii Balance Boards and Microsoft Kinect have both been suggested to be reliable and valid, but have mixed success regarding their capabilities [68,69]. However, despite being somewhat portable and cheaper they are still largely limited to an indoor environment and may be susceptible to greater noise and inconsistent sampling rates [69]. The most common WIS location was the posterior lower trunk between the third and fifth lumbar vertebrae (Tables 2A and 2B), and this location also yielded the most measures across all studies within this review (Table 5). Collectively the posterior lower trunk area had the most identified measures, and this coincides with Montesinos et al. [21] who also identified the lower trunk as an effective location for sensor derived features. The lower trunk, particularly the third to fifth lumbar vertebrae closely resembles an individual's centre of mass, which has been reported to correlate highly with centre of pressure platforms [26,70], suggesting WIS to be a reliable measure of balance and postural stability [71]. During static testing it was generally determined that fallers exhibit a greater amount of mediolateral and anteroposterior trunk sway than non-fallers [26,28,37,38]. Vertical motion was also identified [38] but findings were inconsistent, as naturally individuals sway forwards and backwards and side to side in an attempt to maintain balance. These findings are in agreement with current literature, as previous research without WIS have also reported similar results regarding trunk sway [72][73][74][75]. These results highlight the importance of measuring postural stability when assessing fall risk, in addition to the efficiency in which WIS can assist clinicians. This is especially so when considering that clinical tests, such as the Romberg Greater lower trunk ML RMS (EC/ECF) [28,38] Greater lower trunk vertical RMS (EO/EC) [38] Greater ML range [37] Greater ankle sway (EO) [34] Greater hip sway (EO) [34] Greater COM sway (EO) [34] Greater ML COM sway (EO) [34] Greater lower trunk RMS (EO -mat) [ Greater lower trunk AP RMS angular velocity (EO) [38] Greater lower trunk ML RMS angular velocity (EO/EC) [38] Position and Angle Greater total displacement [37] Variables Greater maximal displacement [37] Energy Greater AP spectral edge frequency [28] Variables Lower ML spectral edge frequency [28] Greater peak to peak amplitude [40] Greater 25% quartile frequency [40] Greater median frequency scores [40] Greater 75% quartile frequency [40] RMS = Root mean square; AP = Anteroposterior; ML = Mediolateral; EO = Eyes open; EC = Eyes closed; ECF = Eyes closed foam surface.

WIS based measures compared to clinical testing
There were five studies that presented correlation analysis on WIS derived measures to clinical fall risk assessments with varying but largely promising results. Low and weak correlations were observed for maximum entropy [25] and mediolateral harmonic ratios [43] to walking speed. However, there were strong statistically significant correlations between the TUG and Berg Balance Scale tests to the RMS of the acceleration signal and angular velocity signals during bouts of quite standing [2,38]. Additionally, and perhaps even more promisingly, Senden et al. [56] reported all acceleration based measures correlated with the Tinetti scale. This included strong positive correlations for walking speed and step length, two commonly identified gait differences among fallers and non-fallers which have been suggested as fall risk factors. These strong positive correlations provide further scope of the efficacy and ability of WIS to perform and be applied under clinical fall risk assessments. However, despite providing initial promising results, only three of five studies provided strong correlations between WIS measures and clinical fall risk assessments. Therefore, future studies should incorporate correlation analysis to provide further strength to this notion. Interestingly, the Timed-Up-and-Go test (a commonly used clinical assessment) was not able to distinguish between fallers and non-fallers when measured with a stopwatch, but was when using acceleration derived measures [49]. This finding is consistent within the literature as stopwatch derived Timed-Up-and-Go test duration has previously been identified to not always be a sensitive marker of fall risk among older adults [77,78]. These results suggest WIS are a better tool when assessing fall risk, even when stopwatch durations are similar, coupled with their low-cost portable nature make WIS a desirable option for fall risk assessment in a clinical environment, as they also do not require the reactions of an assessor to start and stop a test.
There were six studies which included a prospective study design. These studies suggest gait and postural based differences to be fall risk factors as opposed to just characteristic differences between fallers and non-fallers. One prospective study in particular, Mancini et al. [32], identified that the quality of turning measured as the coefficient of variation for the number of steps per turn, could distinguish and predict future fallers and non-fallers over a six month period, even when both the Tinetti gait and Tinetti balance tests could not. This suggests that WIS capabilities have the potential to go above and beyond that of Questions in brief: (1) Aim clearly described, (2) Outcomes clearly described, (3) Subjects clearly described, (4) Measurement clearly described, (5) Distribution of confounders clearly described, (6) Main findings clearly described, (7) Estimates of random variability, (8) Important adverse events reports, (9) P value reported, current clinical testing. Although, the overall results of the review and the individual studies themselves hold great meaning, more prospective studies are needed within the literature. This is because, many gait and postural based differences between fallers and non-fallers could stem from a fear of falling or even a gait adaption resulting from a fall itself. Therefore, identifying pre-existing fall risk factors will be pertinent to the eventual prevention of future falls. There was no study included within the review deemed to have a high risk of bias. Information reporting and external validity scored generally well. However, the internal validity and selection bias of studies were not as strong. This could be a consequence of questions regarding assessors being blind to testing and randomisation of measures being difficult to perform with WIS. Additionally, no study reported sample size power calculations, potentially reducing the statistical power of available data.
A limitation to the present review, is that only measures that were identified to be statistically significant in at least one paper were highlighted as a gait and posture difference among fallers and nonfallers, even if that measure was not statistically significant in another paper. Although, it must be noted that this rarely occurred, and more often than not, measures were only included within a single study. Whilst the present study being a systematic scoping review rather than a meta-analysis, could be viewed as a limitation, the aim was to undertake a descriptive synthesis. In addition, the heterogeneity of the included studies precluded the possibility of meta-analyses.

Conclusion
The review undertook a comprehensive search of the literature and was able to demonstrate that WIS have the capabilities to objectively identify fall risk among older adults, even when clinical tests could not. It was determined that spatiotemporal measures such as: slower walking speeds, shorter step and stride lengths, in addition to, acceleration based metrics including reduced acceleration RMS were most related to fall risk during dynamic tasks. Whereas, increased trunk sway, often measured with the RMS of the acceleration signal was most commonly related to fall risk during static tasks. However, it should be noted that most studies were of a retrospective study design and as such, highlight characteristic differences between fallers and non-fallers which may not necessarily infer fall risk factors. Consequently, more prospective studies are needed. Greater RMS of the acceleration signal appears to be strongly related with faster walking speeds, however, no study directly reported this. The review also included studies (6 total) where a cohort may have had a pathological disorder (e.g., Parkinson's and Alzheimer's), and findings would suggest the same metrics may relate to fall risk regardless of this. This suggests WIS can effectively be used among different cohorts including older healthy and those with neurological ailments. Nevertheless, as such a small sample of pathological disorder studies were identified, it is difficult to make a definitive conclusion and future studies should explore this. A single WIS located at the lower trunk was frequently witnessed and was highly successful in determining several gait and posture differences among older adults. There was a lack of studies outside of a laboratory and during daily life. Therefore, it is currently unknown what the most accurate methods measuring gait metrics are in this new setting. However, regarding WIS anatomical locations it appears to be the same as in laboratory testing, and this also has a very high compliance rate even when multiple WIS are used. Whilst measures of gait during daily life appear to be similar to testing in a laboratory, there was scope for additional metrics to be observed during daily life, such as turning during gait, which provided pertinent information regarding prospective falls even when current clinical assessments could not. A comparison of metrics of gait during continuous daily life monitoring and outdoors, to metrics of gait inside a laboratory was not identified by any of the studies who included a continuous daily life monitoring aspect. Consequently, how these metrics compare in relation to fall risk is currently unknown. Therefore, future studies should consider how metrics of gait compare indoors and outdoors, which would provide greater insight regarding fall risk among older adults. It was determined that WIS are an effective tool to be implemented into current clinical fall risk assessments and perhaps more so in continuous unsupervised habitual monitoring. The lightweight portable nature enables WIS to be used in outdoor testing, which at present, has been largely neglected. This will provide a greater wealth of information regarding fall risk, which is currently unknown. The use of WIS will enable clinicians and healthcare practitioners to make more informed objective decisions regarding fall risk, and its application should be considered in future fall risk assessments.
Ethics approval and right to participate N/A.

Consent for publication
All authors provide consent for publication.

Availability of data and material
This research was a scoping review with a systematic methodology, and therefore, no data is available.

Funding
This report is independent research funded by the Engineering and Physical Sciences Research Council (EPSRC) (EP/K03877X/1). In addition to being supported by the National Institute for Health Research (NIHR) Applied Research Collaboration South West Peninsula (PenARC). The views expressed in this publication are those of the author(s) and not necessarily those of the EPSRC, National Institute for Health Research or the Department of Health and Social Care.

Author contributions
MP conceived the idea for the review which was developed along with VG and AP. MP and VG developed a search strategy. MP undertook database searches and reviewed all articles along with VG, who acted as second reviewer. MP wrote the initial and final drafts of the manuscript. MP, VG and AP altered the draft of the manuscript to form the final version. All authors read and approved the final manuscript.

Declaration of Competing Interests
There are no competing interests.