biosensors

Article
Reliability, Validity, and Identification Ability of a Commercialized
Waist-Attached Inertial Measurement Unit (IMU) Sensor-Based
System in Fall Risk Assessment of Older People
Ke-Jing Li 1 , Nicky Lok-Yi Wong 1 , Man-Ching Law 1,2,3 , Freddy Man-Hin Lam 4 , Hoi-Ching Wong 1,3 ,
Tsz-On Chan 1,3 , Kit-Naam Wong 1,3 , Yong-Ping Zheng 1,2,3 , Qi-Yao Huang 5 , Arnold Yu-Lok Wong 2,4 ,
Timothy Chi-Yui Kwok 6 and Christina Zong-Hao Ma 1,2, *

1 Department of Biomedical Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China;
kejing.li@connect.polyu.hk (K.-J.L.); nickyly.wong@connect.polyu.hk (N.L.-Y.W.); tess.law@polyu.edu.hk (M.-C.L.);
katherinehc.wong@polyu.edu.hk (H.-C.W.); tsz-on-jeffrey.chan@polyu.edu.hk (T.-O.C.);
kn-jessica.wong@polyu.edu.hk (K.-N.W.); yongping.zheng@polyu.edu.hk (Y.-P.Z.)
2 Research Institute for Smart Ageing, The Hong Kong Polytechnic University, Hong Kong SAR, China;
arnold.wong@polyu.edu.hk
3 Jockey Club Smart Ageing Hub, The Hong Kong Polytechnic University, Hong Kong SAR, China
4 Department of Rehabilitation Sciences, The Hong Kong Polytechnic University, Hong Kong SAR, China;
freddy-mh.lam@polyu.edu.hk
5 School of Fashion and Textiles, The Hong Kong Polytechnic University, Hong Kong SAR, China;
qi-yao.huang@polyu.edu.hk
6 Department of Medicine & Therapeutics, The Chinese University of Hong Kong, Hong Kong SAR, China;
tkwok@cuhk.edu.hk
* Correspondence: czh.ma@polyu.edu.hk; Tel.: +852-2766-7671; Fax: +852-2334-2429

Abstract: Falls are a prevalent cause of injury among older people. While some wearable inertial
measurement unit (IMU) sensor-based systems have been widely investigated for fall risk assessment,
their reliability, validity, and identification ability in community-dwelling older people remain un-
Citation: Li, K.-J.; Wong, N.L.-Y.; clear. Therefore, this study evaluated the performance of a commercially available IMU sensor-based
Law, M.-C.; Lam, F.M.-H.; Wong, fall risk assessment system among 20 community-dwelling older recurrent fallers (with a history
H.-C.; Chan, T.-O.; Wong, K.-N.; of ≥2 falls in the past 12 months) and 20 community-dwelling older non-fallers (no history of falls
Zheng, Y.-P.; Huang, Q.-Y.; Wong, in the past 12 months), together with applying the clinical scale of the Mini-Balance Evaluation
A.Y.-L.; et al. Reliability, Validity, and
Systems Test (Mini-BESTest). The results show that the IMU sensor-based system exhibited a sig-
Identification Ability of a
nificant moderate to excellent test–retest reliability (ICC = 0.838, p < 0.001), an acceptable level of
Commercialized Waist-Attached
internal consistency reliability (Spearman’s rho = 0.471, p = 0.002), an acceptable convergent validity
Inertial Measurement Unit (IMU)
(Cronbach’s α = 0.712), and an area under the curve (AUC) value of 0.590 for the IMU sensor-based
Sensor-Based System in Fall Risk
Assessment of Older People. Biosensors
receiver-operating characteristic (ROC) curve. The findings suggest that while the evaluated IMU
2023, 13, 998. https://doi.org/ sensor-based system exhibited good reliability and acceptable validity, it might not be able to fully
10.3390/bios13120998 identify the recurrent fallers and non-fallers in a community-dwelling older population. Further
system optimization is still needed.
Received: 12 September 2023
Revised: 31 October 2023
Keywords: fall risk assessment; wearable system; inertial measurement unit (IMU) sensor; the
Accepted: 8 November 2023
Mini-Balance Evaluation Systems Test (Mini-BESTest); community-dwelling older people
Published: 25 November 2023

Copyright: © 2023 by the authors. 1. Introduction

Licensee MDPI, Basel, Switzerland.
Falls are a major health issue for older people worldwide, and about one-third of
This article is an open access article
people aged older than 65 years fall each year [1]. Given the fact that China has the largest
distributed under the terms and
elderly population in the world, with 264.02 million older people (≥60 years old) [2], fall
conditions of the Creative Commons
incidence and the subsequent injuries [3] in older people have placed a heavy burden on
Attribution (CC BY) license (https://
the health system. As the first step of a fall prevention program [4], fall risk assessment is an
creativecommons.org/licenses/by/
4.0/).
essential element to identify the people with a higher risk of falls and facilitate the provision

Biosensors 2023, 13, 998. https://doi.org/10.3390/bios13120998 https://www.mdpi.com/journal/biosensors

Biosensors 2023, 13, 998 2 of 14

of customized/individualized fall prevention interventions thereafter. However, existing

fall risk assessment strategies vary a lot across different settings and populations [4]. A
reliable and valid fall risk assessment strategy for community-dwelling older people is
urgently needed.
Gait and balance disorders, as one of the most common causes of falls in older peo-
ple [5], are significant predictors of falls [6]. Some clinical scales related to the evaluation of
gait and balance disorders have been widely used as fall risk assessment tools, including
the Berg Balance Scale (BBS), Timed Up and Go (TUG) test, Balance Evaluation Systems
Test (BESTest), and Mini-Balance Evaluation Systems Test (Mini-BESTest). Among these
clinical tests, the Mini-BESTest, a widely used comprehensive balance measurement scale
consisting of 14 items [7], has been recommended as the most accurate tool for identifying
the older people with a history of falls in the past 12 months [8]. Although all the above clin-
ical scales were proven to exhibit good reliability and validity in fall risk assessment [9,10],
they all require experienced professionals to conduct the assessment and are influenced by
the assessors’ clinical experience, which potentially make them unsuitable for large-scale
community-based fall risk screening.
Recently, some instrumented fall risk assessment tools have been developed and
investigated to help provide more accurate, affordable, and easy-to-use services to com-
munities [11,12]. Typical instrumented fall risk assessment tools include wearable sensing
technology, floor-mounted force plates [13], electronic walkways [14], and posturography
using motion capture systems [15]. Wearable sensing technology can be used to collect
the users’ clinically relevant data accurately and provide an objective, continuous health
monitoring service, by placing sensors into wearable objects or directly on the human
body [16–18]. Compared to clinical scales and other instrumented assessment tools, wear-
able sensing technology has the advantages of being portable, unobtrusive, and inexpen-
sive [19], which enables it to address the inevitable increasing need for daily/continuous
fall risk monitoring and large-scale fall risk assessment in the current aging society [20].
The inertial measurement unit (IMU) is one commonly used wearable sensor that can
be utilized in fall risk assessment. By incorporating an accelerometer, a gyroscope, and
sometimes a magnetometer, the IMU can obtain the linear and angular acceleration, velocity,
orientation, and gravitational force data of the target body segments accurately [21,22].
Studies have also suggested that IMU sensor-based systems can enhance the quantifica-
tion/identification of standard clinical scales in fall risk assessment and prediction [23,24],
and can be used in longitudinal monitoring protocols in daily activity [25]. However, so
far, the reliability, validity, and identification ability of such IMU sensor-based systems
in identifying community-dwelling older fallers and older non-fallers has remained in-
conclusive [26]. Therefore, before the implementation of IMU sensor-based technology
in large-scale fall risk assessment programs for community-dwelling older people, the
performance of IMU sensor-based systems in fall risk assessment needsto be carefully
evaluated to ensure the promotion and execution of evidence-based practice.
Up to now, there has been a lack of evaluations of IMU sensor-based systems’ applica-
tion in fall risk assessments for community-dwelling older people. This study bridged this
research gap and evaluated the performance of a commercially available IMU sensor-based
system from the aspects of reliability [27], validity [27], and identification ability with
reference to the Mini-BESTest [28]. The findings of this study provide evidence on the
current status of IMU sensor-based systems in assessing fall risks of community-dwelling
older people, and they are expected to inspire future development and optimization of
such instrumented fall risk assessment tools.

2. Methods
2.1. Participants
This study was conducted at the Jockey Club Smart Aging Hub of the Hong Kong
Polytechnic University. A total of 40 participants, including 20 community-dwelling
older recurrent fallers (≥65 years old and had a history of 2 or more falls in the past
Biosensors 2023, 13, 998 3 of 14

12 months) [29] and 20 community-dwelling older non-fallers (≥65 years old and with no
history of falls in the past 12 months), were recruited through the convenience sampling
via the Jockey Club Smart Aging Hub. This sample size of 40 participants yielded a
statistical power of 0.809, suitable for the case–control study design, with an effect size
of 0.92 based on the published literature [30] and a significant level of 0.05. None of the
participants had previously participated in similar trials or studies involving wearable
systems. Participants were excluded if they had any active medical conditions or active
inflammatory/pathologic changes in lower-extremity joints within six months prior to the
experiment. All data collected were kept confidential and identifiable only through codes
known to the researchers. This study adhered to the principles of the Declaration of Helsinki
and received approval from the Institutional Review Board of the Hong Kong Polytechnic
University (Reference number: HSEARS20211116005). All participants provided written
informed consent before the experiments.

2.2. Equipment
A commercially available 9-axis inertial measurement unit (IMU)-based system (Booguu
Biosensors 2023, 13, x FOR PEER REVIEW 4 of 15
AspireTM Fall Risk Management System, Booguu Company Limited, Hong Kong SAR,
China) was used and evaluated in this study (Figure 1a). This system, as described in
the user manual and on the company website [31], was developed to offer a simple, fast,
hands and rest
and accurate fallonrisk
theassessment
chest, and for
thensit and stand repeatedly
community-dwelling olderatpeople.
the fastest pace forthe
Specifically, 15
seconds, to assessathe
system includes functional
waistbelt capacity
with of the lower
an embedded limbs;
6-axis and (4) Gait (Figure 1d):
accelerometer/gyroscope andwalk
an
at a normal 3-axis
embedded speedmagnetometer
for about 10 meters, to assess
for data an individual’s
detection, walkingfor
a tablet application pa user
ern.operation,
The final
total
and anassessment result was
online database for represented by scores
data processing ranging(Table
and storage from 1).
1 toTo10.conduct
Additionally,
the fallafter
risk
each assessment,
assessments, userstheweresystem generated
instructed separatea 5-min
to complete resultstask
for [32].
each Throughout
of the four thesub-tasks,
assess-
which
ments,were
data also represented
captured by the by scores ranging
waistbelt from 1 to
were uploaded to 10.
theThese
onlineresults werefor
database obtained
further
through a patented categorization algorithm and the company’s own database [33–35]. As
data processing. The whole assessment procedure could be monitored and controlled
specified in the
via the tablet user manual,
application, withtotal scores falling within
a comprehensive report the range ofthe
containing 1 toassessment
3, 4 to 7, and 8 to
results
10 indicated
generated theeach
after high,assessment
medium, and low risks of falls,
(Supplementary respectively.
Material: Fall Risk Assessment Report).

Figure 1. (a) The validated IMU sensor-based system in this study. (b–d) Illustrations of the system
Figure 1. (a) The validated IMU sensor-based system in this study. (b–d) Illustrations of the system
a achment during assessment on one participant as an example: (b) stand still with eyes open (sub-
attachment during assessment on one participant as an example: (b) stand still with eyes open
task 1); stand still with eyes closed (sub-task 2); (c) sit and stand repeatedly with eyes open (sub-
(sub-task 1);(d)
task 3); and stand
walkstill with eyes
at normal closed
speed with(sub-task
eyes open2);(sub-task
(c) sit and
4). stand repeatedly with eyes open
(sub-task 3); and (d) walk at normal speed with eyes open (sub-task 4).
2.3. Experimental Protocol
To apply the system for fall risk assessment, the waistbelt was placed tightly against
Experiments
the end wereback
user’s lower conducted
using an in elastic
an openstrap.
roomTheat the Jockey Club
5-minute Smart Aging
task included fourHub,
sub-
the Hong Kong Polytechnic University, where participants had a spacious and
tasks [32]: (1) Postural stability I (Figure 1b): stand still for 30 s in a 28 cm × 28 cm spacequiet
environment
with eyes open,in which to undergo
to assess the healththecondition
assessments without
of the sensoryinterruptions. All participants
system and nervous system
received guidanceconditions;
under eyes-open and protection fromstability
(2) Postural the same experienced
II (Figure researcher
1b): stand to skeep
still for 30 in a
consistency in data collection and interpretation. The entire experiment was completed
within a one-hour timeframe on the same day for each participant. The experiment
consisted of three parts: demographic data collection, instrumented fall risk assessment,
and clinical fall risk assessment. The demographic data collection was completed by
documenting participants’ self-reported age, gender, medication history, and fall history,
Biosensors 2023, 13, 998 4 of 14

28 cm × 28 cm space with eyes closed, to assess the health condition of the sensory system
and nervous system under eyes-closed conditions; (3) Dynamic movement (Figure 1c):
cross hands and rest on the chest, and thensit and stand repeatedly at the fastest pace for
15 seconds, to assess the functional capacity of the lower limbs; and (4) Gait (Figure 1d):
walk at a normal speed for about 10 meters, to assess an individual’s walking pattern. The
final total assessment result was represented by scores ranging from 1 to 10. Additionally,
after each assessment, the system generated separate results for each of the four sub-tasks,
which were also represented by scores ranging from 1 to 10. These results were obtained
through a patented categorization algorithm and the company’s own database [33–35]. As
specified in the user manual, total scores falling within the range of 1 to 3, 4 to 7, and 8 to
10 indicated the high, medium, and low risks of falls, respectively.

Table 1. Technical specifications of the IMU sensor-based system.

Items Specifications
Dimensions 72.6 mm × 40.5 mm × 12.6 mm
Weight 32 g
Processor ARM® Cortex™-M4F 32-bit processor
Accelerometer/Gyroscope Bosch ® 6-axis accelerometer/gyroscope
Magnetometer Bosch ® 3-axis magnetometer
Inertial sensor platform Bosch ® 9-axis sensor fusion
Battery Rechargeable lithium ion/polymer battery

2.3. Experimental Protocol

Experiments were conducted in an open room at the Jockey Club Smart Aging Hub,
the Hong Kong Polytechnic University, where participants had a spacious and quiet en-
vironment in which to undergo the assessments without interruptions. All participants
received guidance and protection from the same experienced researcher to keep consis-
tency in data collection and interpretation. The entire experiment was completed within
a one-hour timeframe on the same day for each participant. The experiment consisted of
three parts: demographic data collection, instrumented fall risk assessment, and clinical
fall risk assessment. The demographic data collection was completed by documenting par-
ticipants’ self-reported age, gender, medication history, and fall history, followed by onsite
measurements of weight and height. The instrumented fall risk assessment was completed
by conducting three assessments using the IMU sensor-based system. The clinical fall risk
assessment was completed by conducting one assessment using the Mini-BESTest. The
sequence of the three IMU sensor-based assessments and the Mini-BESTest was randomized
for each participant to minimize potential order effects. A five-minute rest was provided
between each assessment to mitigate fatigue and minimize possible learning effects.

2.4. Data and Statistical Analysis

Overall, the instrumented fall risk assessment results were represented by the total
and sub-tasks score generated from the IMU sensor-based system, ranging from 0 to 10 [32].
The clinical fall risk assessment results were represented by the scores generated from
the Mini-BESTest, ranging from 0 to 28 [36]. Statistical analysis was conducted using the
Statistical Package for Social Sciences (SPSS, version 26.0, IBM Corporation, Armonk, NY,
USA). The detailed data and statistical analysis can be divided into four parts as follows:

2.4.1. Reliability Evaluation of the IMU Sensor-Based System

The test–retest reliability of the IMU sensor-based system was assessed using the
total and sub-task scores obtained from the three IMU sensor-based assessments. A
single-measurement, absolute-agreement, 2-way random-effects model of the intra-class
coefficient (ICC) [37] was used to examine the test–retest reliability of the IMU sensor-
based system.
Biosensors 2023, 13, 998 5 of 14

The internal consistency reliability of the IMU sensor-based system was assessed
using the averaged total and sub-task scores obtained from the three IMU sensor-based
assessments. Cronbach’s coefficient α [38] was used to examine the internal consistency
reliability of the IMU sensor-based system.

2.4.2. Validity Evaluation of the IMU Sensor-Based System

The convergent validity of the IMU sensor-based system was assessed by calculating
the correlation between the averaged total scores obtained from the three IMU sensor-
based assessments and the scores obtained from the Mini-BESTest. Spearman’s rank
correlation coefficient was used to examine the convergent validity of the IMU sensor-
based system [39,40].

2.4.3. Identification Ability Evaluation of the IMU Sensor-Based System

The fall risk identification ability of the IMU sensor-based system was assessed by
using the receiver-operating characteristic (ROC) of the averaged total scores obtained from
the three IMU sensor-based assessments [41]. The sensitivity and specificity of the IMU
sensor-based system in fall risk identification were further analyzed based on the cut-off
scores and the participants’ groups (recurrent fallers and non-fallers). Three cut-off scores
were used for analysis: one based on the ROC result of this study, and the other two based
on the system’s recommended thresholds for identifying fall risks in older people (3.5 and
7.5 points [32]).

2.4.4. Identification Ability Evaluation of the Mini-BESTest

Similar to the IMU sensor-based system, the fall risk identification ability of the Mini-
BESTest was also assessed in this study, by analyzing the receiver-operating characteristic
(ROC) of the scores obtained from the Mini-BESTest [41]. The sensitivity and specificity
of the Mini-BESTest in fall risk identification were analyzed based on the cut-off scores
and the participants’ groups (recurrent fallers and non-fallers). Three cut-off scores were
used for analysis: one based on the ROC result of this study, and the other two based on
recommendations from two previous studies for identifying fall risks in older people using
the Mini-BESTest (16 points [8] and 19.5 points [42]).

3. Results
3.1. Demographic Data
As shown in Table 2, a total of forty community-dwelling older people (age: 70.4 ± 5.9 years;
gender: 18 males and 22 females; height: 160.1 ± 9.6 cm; weight: 60.4 ± 11.3 kg; BMI:
23.5 ± 3.7 kg/m2 ) participated in this study, including twenty recurrent fallers and twenty
non-fallers. No significant difference was observed between the recurrent faller and non-
faller groups in age (p = 0.605), height (p = 0.351), weight (p = 0.631), or BMI (p = 0.083). No
falls or any other adverse events occurred during the experiments.

Table 2. Participants’ demographic data (mean ± SD).

Recurrent All
Females Males Non-Fallers
Fallers Participants
(n = 22) (n = 18) (n = 20)
(n = 20) (n = 40)
Age (year) 70.6 ± 6.8 70.1 ± 4.7 70.8 ± 6.7 70.0 ± 5.2 70.4 ± 5.9
Height (cm) 154.2 ± 7.5 167.3 ± 6.4 158.6 ± 10.5 161.5 ± 8.5 160.1 ± 9.6
Weight (kg) 54.2 ± 8.8 68.0 ± 9.4 61.3 ± 10.3 59.6 ± 12.4 60.4 ± 11.3
BMI (kg/m2 ) 22.9 ± 4.0 24.3 ± 3.2 24.3 ± 3.0 22.8 ± 4.2 23.5 ± 3.7
Total score of the IMU sensor-based system 6.7 ± 2.0 6.9 ± 2.4 6.4 ± 2.5 7.2 ± 1.7 6.8 ± 2.1
Total score of the Mini-BESTest 22.8 ± 3.6 25.0 ± 3.0 23.1 ± 3.5 24.5 ± 3.4 23.8 ± 3.5
Biosensors 2023, 13, 998 6 of 14

3.2. Reliability of the IMU Sensor-Based System

Regarding the test–retest reliability, the results are presented in Table 3. Specifi-
cally, good to excellent test–retest reliability was observed in the total scores (ICC = 0.838,
p < 0.001), moderate to good test–retest reliability was observed in the first (ICC = 0.717,
p < 0.001) and the second (ICC = 0.698, p < 0.001) sub-task scores, poor to good test–retest
reliability was observed in the third sub-task scores (ICC = 0.653, p < 0.001), and good
to excellent test–retest reliability was observed in the fourth sub-task scores (ICC = 0.843,
p < 0.001).

Table 3. The intraclass correlation coefficient (ICC) values of the total and sub-task scores generated
by the IMU sensor-based system during three assessments.

95% Confidence Interval

Score ICC Value p-Value
Lower Bound Upper Bound
Total 0.838 0.745 0.904 <0.001
1st task (postural stability with eyes open) 0.717 0.576 0.827 <0.001
2nd task (postural stability with eyes closed) 0.698 0.553 0.814 <0.001
3rd task (dynamic movement and gait I) 0.653 0.495 0.783 <0.001
4th task (dynamic movement and gait II) 0.843 0.753 0.907 <0.001

Regarding the internal consistency reliability of the IMU sensor-based system, an

acceptable level of internal consistency reliability was observed (containing four sub-tasks,
Cronbach’s α = 0.712).

3.3. Validity of the IMU Sensor-Based System

As shown in Figure 2, there were also significant moderate positive correlations [43]
between the scores of the Mini-BESTest and those of the IMU sensor-based system in all
participants (Spearman’s rho = 0.471, p = 0.002, Figure 2a) and in the recurrent faller group
Biosensors 2023, 13, x FOR PEER REVIEW
(Spearman’s rho = 0.536, p = 0.015, Figure 2b). No significant correlation was observed 7 of in
15
the non-faller group (Spearman’s rho = 0.327, p = 0.107, Figure 2c).

Figure
Figure 2. 2. Correlation
Correlation betweenthe
between thescores
scoresofofMini-BESTest
Mini-BESTestand
andthethescores
scoresofofthe
theIMU
IMUsensor-based
sensor-based
system in (a) all participants, (b) recurrent fallers only, and (c) non-fallers only (* p < 0.05).
system in (a) all participants, (b) recurrent fallers only, and (c) non-fallers only (* p < 0.05).

The
The sub-taskscore
sub-task scoreanalysis
analysis ofof the
the IMU
IMUsensor-based
sensor-basedsystem
systemis shown in Figure
is shown 3. For
in Figure 3.
all participants, there were significant though weak positive correlations
For all participants, there were significant though weak positive correlations between the between the
scores
scores of of
thethe Mini-BESTest
Mini-BESTest andand those
those ofofthe
thefirst
firstsub-task
sub-task(Spearman’s
(Spearman’srho rho==0.338,
0.338,pp== 0.033,
0.033,
Figure 3a), the second sub-task (Spearman’s rho = 0.323, p = 0.042, Figure
Figure 3a), the second sub-task (Spearman’s rho = 0.323, p = 0.042, Figure 3b), the third 3b), the third
subtask (Spearman’s rho = 0.332, p = 0.037, Figure 3c), and the fourth sub-task
subtask (Spearman’s rho = 0.332, p = 0.037, Figure 3c), and the fourth sub-task (Spearman’s (Spearman’s
rho
rho = 0.384,
= 0.384, p=p =0.014,
0.014,Figure
Figure3d)
3d)ofofthe
theIMU
IMUsensor-based
sensor-basedsystem.
system.
 The sub-task score analysis of the IMU sensor-based system is shown in Figure 3. For
all participants, there were significant though weak positive correlations between the
scores of the Mini-BESTest and those of the first sub-task (Spearman’s rho = 0.338, p = 0.033
Figure 3a), the second sub-task (Spearman’s rho = 0.323, p = 0.042, Figure 3b), the third
Biosensors 2023, 13, 998 subtask (Spearman’s rho = 0.332, p = 0.037, Figure 3c), and the fourth sub-task7(Spearman’s
of 14
rho = 0.384, p = 0.014, Figure 3d) of the IMU sensor-based system.

Figure3.3.Correlation
Figure Correlation between
between thethe scores
scores of Mini-BESTest
of Mini-BESTest and
and the the sub-task
sub-task scores
scores of of the
the IMU IMU sensor
sensor-
basedsystem
based systemforfor
(a)(a) sub-task
sub-task 1; sub-task
1; (b) (b) sub-task
2; (c) 2; (c) sub-task
sub-task 3; and 3;
(d)and (d) sub-task
sub-task 4 (* p < 0.05).
4 (* p < 0.05).

3.4. Fall Risk Identification Ability of the IMU Sensor-Based System

3.4. Fall Risk Identification Ability of the IMU Sensor-Based System
As shown in Figure 4a, the area under the curve (AUC) of the IMU sensor-based system
was 0.590 (with 95% confidence interval = 0.411–0.769, p = 0.330, Figure 4a). According to
the IMU sensor-based system’s ROC curve, the best sensitivity and specificity values were
achieved with a cut-off score of 3.5/10, which aligned with one of the recommended cut-off
scores by the company. Using this cut-off score, the IMU sensor-based system demonstrated
a sensitivity of 30% in correctly identifying recurrent fallers and a specificity of 95% in
correctly identifying non-fallers.
When the cut-off score of 7.5/10 [32] was used, the IMU sensor-based system demon-
strated a sensitivity of 50% in correctly identifying recurrent fallers and a specificity of 55%
in correctly identifying non-fallers.

3.5. Fall Risk Identification Ability of the Mini-BESTest

As shown in Figure 4b, the area under the curve (AUC) of the Mini-BESTest was
0.668 (with 95% confidence interval = 0.494–0.841, p = 0.070, Figure 4b). According to
the Mini-BESTest’s ROC curve, the best sensitivity and specificity values were achieved
with a cut-off score of 24.5/28. Using this cut-off score, the Mini-BESTest demonstrated
a sensitivity of 65% in correctly identifying recurrent fallers and a specificity of 65% in
correctly identifying non-fallers.
When the cut-off scores of 16/28 [8] and 19.5/28 [42] were used, the Mini-BESTest
demonstrated sensitivities of 5% and 5% and specificities of 95% and 85%, respectively.
 system was 0.590 (with 95% confidence interval = 0.411–0.769, p = 0.330, Figure 4
According to the IMU sensor-based system’s ROC curve, the best sensitivity a
specificity values were achieved with a cut-off score of 3.5/10, which aligned with one
the recommended cut-off scores by the company. Using this cut-off score, the IMU sens
Biosensors 2023, 13, 998 based system demonstrated a sensitivity of 30% in correctly identifying recurrent
8 of 14 fall
and a specificity of 95% in correctly identifying non-fallers.

Figure 4. The
Figure ROC
4. The ROCcurves
curvesofof the (a)IMU
the (a) IMU sensor-based
sensor-based system
system and (b)and (b) Mini-BESTest.
Mini-BESTest. Do ed li
Dotted lines
represent random
represent classifiers.
random classifiers.

4. Discussion
When the cut-off score of 7.5/10 [32] was used, the IMU sensor-based syste
The performance of a commercialized waist-attached inertial measurement unit (IMU)
demonstrated
sensor-baseda system
sensitivity of 50%
in fall risk in correctly
assessment identifying older
of community-dwelling recurrent
peoplefallers
was and
specificity of 55%
evaluated instudy.
in this correctly identifying
Generally, non-fallers.
the findings of this study indicate that: (1) the IMU
sensor-based system exhibited good reliability, acceptable validity, and poor identification
3.5. Fall Riskand
ability; Identification Ability ofexhibited
(2) the Mini-BESTest the Mini-BESTest
inconclusive identification ability in the fall
risk assessments of community-dwelling older people. More details are provided in the
As shown sections.
subsequent in Figure 4b, the area under the curve (AUC) of the Mini-BESTest was 0.6
(with 95% confidence interval = 0.494–0.841, p = 0.070, Figure 4b). According to the Mi
4.1. Good Reliability of the IMU Sensor-Based System
BESTest’s ROC curve, the best sensitivity and specificity values were achieved with a c
The significant moderate to excellent test–retest reliability and the acceptable level of
off score of 24.5/28. Using this cut-off score, the Mini-BESTest demonstrated a sensitiv
internal consistency reliability of the IMU sensor-based system generally indicated that
of 65%
suchina system
correctly identifying
possesses recurrent
good reliability fallers
in fall and a specificity
risk assessment of 65% in correc
of community-dwelling
identifying non-fallers.
older people. This aligns with previous research that corroborated the reliable performance
When the cut-off
of IMU sensors scoresaof
in detecting 16/28
range [8] and 19.5/28
of balance-related [42] were signals
biomechanical used, [26,44,45],
the Mini-BEST
demonstrated sensitivities of 5% and 5% and specificities of 95% and 85%, respectively
as well as in fall risk assessments of patients with multiple sclerosis [46] and stroke [47].
It is also noteworthy that the good test–retest reliability of the average total scores of
the IMU sensor-based system (ICC = 0.838) is comparable to that of the commonly used
4. Discussion
clinical balance tests, including the BBS, the BESTest, the Mini-BESTest, and the Brief-
The
BESTestperformance
(0.886 ≤ ICC ≤of0.945)
a commercialized waist-a
in fall risk assessments ached
of older inertial
people measurement
[9]. These results u
lend support to the notion that IMU sensors can yield considerably reliable results for fall
(IMU) sensor-based system in fall risk assessment of community-dwelling older peop
risk assessment of community-dwelling older people, especially when compared with the
was evaluated in this
existing, widely study.
applied Generally,
clinical the findings of this study indicate that: (1) the IM
balance assessments.
sensor-based system exhibited good reliability, acceptable validity, and po
identification ability; and (2) the Mini-BESTest exhibited inconclusive identification abil
4.2. Acceptable Validity of the IMU Sensor-Based System

in the fall risk assessments

The significant of community-dwelling
moderate positive correlation between theolder
averagedpeople. More
total scores details a
of the
IMU sensor-based system and the scores of the Mini-BESTest suggests that the IMU sensor-
provided
based in the subsequent
system sections.
exhibited acceptable convergent validity in fall risk assessment of community-
dwelling older people. It is also encouraging to note the significant though weak positive
4.1. Good Reliability
correlations of thetheIMU
between fourSensor-Based
sub-task scoresSystem
of the IMU sensor-based system and the
scores of the Mini-BESTest. The Mini-BESTest is a widely utilized comprehensive balance
The significant moderate to excellent test–retest reliability and the acceptable level
internal consistency reliability of the IMU sensor-based system generally indicated th
Biosensors 2023, 13, 998 9 of 14

measurement scale comprising 14 items [7], and it includes four balance-related subscales:
anticipatory, reactive postural control, sensory orientation, and dynamic change [48]. The
observed acceptable convergent validity indicates that the IMU sensor-based system aligns
with the Mini-BESTest, suggesting that the IMU sensor-based system could potentially
serve as an efficient adjunct or supplement to the Mini-BESTest in fall risk assessment of
older people.
When we consider assessments besides the Mini-BESTest, the IMU sensor-based
technology has already been proven to be effective in enhancing other clinical balance
assessment scales in previous studies, such as the Timed Up and Go (TUG) test and the
Berg Balance Scale (BBS), by providing automated kinematic data assessments [23,24,49,50].
Given that most clinical scales used for fall risk assessments of older people need to be
applied by professional experts, which requires substantial health resources and manpower,
it has been challenging to provide widespread, long-term, cross-institutional fall-risk
screening services for older people in the community. In light of the positive outcomes of
this study, the IMU-related technology may serve as a surrogate or supplement of clinical
scales in large-scale fall risk assessment [51–53] of community-dwelling older people in
future clinical practice.

4.3. Poor Identification Ability of the IMU Sensor-Based System

This study observed an AUC value (AUC value = 0.590, p = 0.330) in the IMU sensor-
based system ROC curve that was not meaningfully greater than chance. This suggests that
the IMU sensor-based system may not be able to effectively distinguish between recurrent
fallers and non-fallers among community-dwelling older people. The identification ability
of an IMU sensor-based system is determined by multiple factors, and such negative results
could be explained by the following aspects.
The placement of the IMU sensor and the detected parameters need careful consider-
ation and evaluation. While most studies place the IMU sensor close to the center of the
body for body orientation detection [20], other commonly used strategies include placing
the IMU sensor on a lower limb [54], the sternum [55], or the head [56] for the detection of
gait parameters, trunk acceleration, and head stabilization, respectively. It can generally
be concluded that placing IMU sensors on different body parts could be a good means of
comprehensive parameter detection and analysis. For example, Liu et al. [51] used a single
IMU attached to different body parts for the detection of different parameters, suggesting
that the combination of different parameters performed better in fall risk identification
compared to individual parameters. However, there is currently a lack of guidelines on
IMU sensor placement and parameter detection specifically tailored to fall risk assess-
ment, especially for community-dwelling older people. Future protocol developments and
investigations are, therefore, necessary.
The classification algorithms and scoring criteria need optimization and further investi-
gation. Given that a fall is a multifactorial event, different fall risk classification algorithms
and scoring criteria are needed for assessing different parameters and for use in older
people with different health conditions. Previous studies have demonstrated that the
integration of machine learning models [50] and deep learning approaches [57,58] with
IMU sensor-based technology can improve fall risk identification accuracy. However, these
studies exhibit considerable variation in terms of the population, the utilized algorithms,
and the scoring criteria. Further investigations are needed, not only on the improvement of
fall risk classification accuracy but also on the establishment of correspondence between
the scoring and its clinical implications.
The sensor type and number also require further evaluation and optimization. For
example, Qiu et al. [59] reported that five IMUs worn on the lower back and lower limbs
could accurately assess fall risk in older people. The force sensor, another type of commonly
used wearable motion sensor, has been widely used together with IMU sensors to develop
fall risk assessment devices/systems [54,60]. Such positive findings indicate that integrating
multiple IMUs with other types of wearable motion sensors could be helpful for future
Biosensors 2023, 13, 998 10 of 14

studies to provide more comprehensive assessments and address the current unsatisfactory
results of using the single IMU sensor-based system, as identified in this study.
Unfortunately, the details of the IMU sensor-based system, including the raw data and
parameters detected, background algorithms, scoring criteria, and other technical details
(e.g., sampling frequency, ranges), were not disclosed by the company. While the user
manual recommended tightening the IMU sensor at the lower back, the reason for this
placement was not provided. Disclosing more technical details would not only enable the
investigation and explanation of the current insignificant findings but also facilitate the
optimization of such commercial IMU sensor-based systems. This would ultimately lead
to more accurate and validated assessment results in identifying the risk of falls among
community-dwelling older people in the future.

4.4. Inconclusive Identification Ability of the Mini-BESTest

This study observed an AUC value of 0.668 (p = 0.070) with marginal significance in the
Mini-BESTest ROC curve. Given that ROC curves with an AUC ≤ 0.75 are not considered
clinically useful/significant, the ability of the Mini-BESTest to distinguish between recurrent
fallers and non-fallers among community-dwelling older people remains uncertain. This
result aligns with a previous study, which also reported that the Mini-BESTest had poor
fall risk identification ability (AUC = 0.54, 58 older people, mean age = 78 years) in older
adults with self-reported balance problems [61]. It should be noted that as a commonly
used clinical scale, the Mini-BESTest has been recommended as a stable and reliable tool
for fall risk detection in older people [61] and has been shown to be more accurate than
the BESTest, TUG, and BBS tests in identifying older people’s fall history [8]. Previous
studies investigating the Mini-BESTest’s accuracy in identifying older people with a history
of falls in the past 12 months, have found positive results and reported the cut-off scores
to be 16/28 (AUC = 0.84, 200 older people, mean age = 70) [8] and 19.5/20 (AUC = 0.76,
122 older people, mean age = 76) [42], respectively. Despite these inconsistent findings
above, one study involving 264 older people found that Mini-BESTest performed well in fall
risk prediction for older adults, with different cut-off scores corresponding to different age
groups [62], which suggested that recruiting more participants, preferably with a different
grouping method and more specific subject inclusion/exclusion criteria, could be potential
promising ways to conduct future in-depth investigations on the Mini-BESTest’s fall risk
identification ability in older people.

4.5. Limitations
There are several limitations of this study. Firstly, this study had a relatively small
sample size, and the participants were categorized as recurrent fallers or non-fallers based
on their self-reported fall history without multi-labeling. During the data collection phase,
it was observed that some participants categorized as “recurrent fallers” were highly active
and frequently engaged in outdoor activities such as hiking or street running, which in-
creased their exposure and likelihood of falling. Conversely, some participants categorized
as “non-fallers” exhibited poor balance ability (as evidenced by the Mini-BESTest score)
and tended to live a sedentary life, leading to less exposure and likelihood of falling. Given
that the fall risk increases for both the highly inactive and the highly active older people as
compared to those who are moderately active [63], the current self-reported data collection
strategy cannot fully reflect the participants’ true statuses and is susceptible to memory
bias. Thus, implementing regular prospective follow-up interviews and standardized
questionnaires, such as the International Physical Activity Questionnaire modified for the
elderly (IPAQ-E) [64], to collect fall events and physical activity levels could provide a more
reliable reference for the grouping method. Furthermore, considering that factors such as
age, gender [65], and co-morbidities/diseases such as poliomyelitis [66] and diabetes [67]
can significantly affect the balance performance and the occurrence of falls, future studies
with larger sample sizes and more diverse populations should be conducted to address
this limitation.
Biosensors 2023, 13, 998 11 of 14

Secondly, there might be some learning effects that existed across different trials
in the current experimental protocol. Despite incorporating a 5-minute recovery time
between each trial to mitigate potential fatigue and interactions, the learning effect may
still have existed. To achieve a more accurate evaluation of the test–retest reliability of
the system, conducting reliability assessments on different days would offer a superior
assessment protocol.
Thirdly, this study has evaluated the performance of the IMU sensor-based system
and compared it only to the Mini-BESTest. It should be noted that different outcomes
may arise when comparing the system with other commonly employed clinical balance
assessments. While the Mini-BESTest has been suggested to be the most accurate tool in fall
risk identification for older people among most of the clinical balance scales [8], researchers
also highlighted the absence of a universally recommended tool (or “gold standard”) for
fall risk assessments [68]. By undertaking comparisons between the IMU sensor-based
system and other clinical scales, a more comprehensive understanding of the evaluation
results and their interpretations can be attained.
Finally, this study adopted a retrospective study design without prospective follow-
ups. Incorporating a prospective study design in fall risk assessment studies would not
only support unambiguous data collection concerning fall events but also enable the investi-
gation of the system’s fall prediction abilities. Furthermore, future research should address
the long-term user compliance and solicit feedback regarding the system’s acceptability.
This research should involve both the end users, such as older people and individuals with
disabilities, and the assessors, including clinicians and rehabilitation therapists. Conduct-
ing such studies will further support the community-based implementation and integration
of the system into clinical practice.

5. Conclusions
This study has bridged a research gap and provided insights into the application of
IMU sensor-based systems in fall risk assessments of community-dwelling older people.
Specifically, the study has evaluated the performance of a commercially available IMU
sensor-based system in comparison with the Mini-BESTest. Good reliability, acceptable
validity, but poor identification ability have been identified for the commercialized IMU
sensor-based system in assessing the fall risk among community-dwelling older people.
Despite the relatively poor performance of the evaluated system in identifying fall
risks, the potential of IMU sensor-based technology in facilitating long-term fall risk screen-
ing among community-dwelling older people cannot be denied. This is particularly true
when considering its compact size, low cost, good reliability, acceptable validity, and the
generally accurate analysis of motion parameters. It is recommended that healthcare practi-
tioners can use the IMU sensor-based system as a valuable reference/adjunct/supplement
due to its objective and accurate motion parameter detection and analysis, but we advise
taking a cautious approach when providing diagnoses or clinical suggestions solely based
on the system’s results. Future optimization of the IMU sensor-based systems remains
necessary, not only for the improvement of fall risk identification accuracy but also for the
exploration of the relevant clinical implications for community-dwelling older people.

Supplementary Materials: The following supporting information can be downloaded at: https://
www.mdpi.com/article/10.3390/bios13120998/s1. An example of the report generated by the IMU
sensor-based system. (Note: The original report was in Chinese, and was translated into English
by the authors. The original Chinese-version report is provided in the later pages of the translated
English-version report. In case of any ambiguity, precedence should be given to the Chinese version).
Biosensors 2023, 13, 998 12 of 14

Author Contributions: Conceptualization, M.-C.L. and C.Z.-H.M.; data curation, K.-J.L. and N.L.-Y.W.;
formal analysis, K.-J.L., N.L.-Y.W., M.-C.L., F.M.-H.L., H.-C.W., T.-O.C., K.-N.W., Y.-P.Z., Q.-Y.H.,
A.Y.-L.W., T.C.-Y.K. and C.Z.-H.M.; funding acquisition, Y.-P.Z. and C.Z.-H.M.; investigation, K.-J.L.
and N.L.-Y.W.; methodology, K.-J.L. and C.Z.-H.M.; project administration, M.-C.L., H.-C.W., T.-O.C.,
K.-N.W. and C.Z.-H.M.; resources, M.-C.L., H.-C.W., T.-O.C., K.-N.W., Y.-P.Z. and C.Z.-H.M.; su-
pervision, C.Z.-H.M.; validation, K.-J.L., N.L.-Y.W., M.-C.L., F.M.-H.L., H.-C.W., T.-O.C., K.-N.W.,
Y.-P.Z., Q.-Y.H., A.Y.-L.W., T.C.-Y.K., M.-C.L. and C.Z.-H.M.; visualization, K.-J.L., F.M.-H.L., Q.-Y.H.
and C.Z.-H.M.; writing—original draft, K.-J.L. and N.L.-Y.W.; writing—review and editing, K.-J.L.,
N.L.-Y.W., M.-C.L., F.M.-H.L., H.-C.W., T.-O.C., K.-N.W., Y.-P.Z., Q.-Y.H., A.Y.-L.W., T.C.-Y.K. and
C.Z.-H.M. All authors have read and agreed to the published version of the manuscript.
Funding: This research was partially supported by Hong Kong Research Grants Council (RGC)
through a Direct Allocation Grant—Funding Support to Small Projects (grant number P0036830),
The Hong Kong Polytechnic University (grant numbers P0034491 and G4Y56R006), and Research
Institute for Smart Ageing, The Hong Kong Polytechnic University (grant numbers P0038945).
Institutional Review Board Statement: This study was conducted according to the guidelines of the
Declaration of Helsinki and approved by the Institutional Review Board of the Hong Kong Polytechnic
University (reference number: HSEARS20211116005; date of approval: 26 November 2021).
Informed Consent Statement: Written informed consent has been obtained from the patients to
publish this paper.
Data Availability Statement: Data are contained within the article.
Acknowledgments: We thank all the older volunteers who participated in this study.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Ganz, D.A.; Latham, N.K. Prevention of Falls in Community-Dwelling Older Adults. N. Engl. J. Med. 2020, 382, 734–743.
[CrossRef]
2. Wang, H.; Chen, H. Aging in China: Challenges and opportunities. China CDC Weekly 2022, 4, 601–602.
3. Montero-Odasso, M.; van der Velde, N.; Martin, F.C.; Petrovic, M.; Tan, M.P.; Ryg, J.; Aguilar-Navarro, S.; Alexander, N.B.; Becker,
C.; Blain, H.J.A.; et al. World guidelines for falls prevention and management for older adults: A global initiative. Age Ageing
2022, 51, afac205. [CrossRef]
4. Perell, K.L.; Nelson, A.; Goldman, R.L.; Luther, S.L.; Prieto-Lewis, N.; Rubenstein, L.Z. Fall risk assessment measures: An analytic
review. J. Gerontol. Ser. A Biol. Sci. Med. Sci. 2001, 56, M761–M766. [CrossRef] [PubMed]
5. Cuevas-Trisan, R. Balance problems and fall risks in the elderly. Clin. Geriatr. Med. 2019, 35, 173–183. [CrossRef] [PubMed]
6. Mancini, M.; Horak, F.B. The relevance of clinical balance assessment tools to differentiate balance deficits. Eur. J. Phys. Rehabil.
Med. 2010, 46, 239. [PubMed]
7. Franchignoni, F.; Horak, F.; Godi, M.; Nardone, A.; Giordano, A. Using psychometric techniques to improve the Balance
Evaluation Systems Test: The mini-BESTest. J. Rehabil. Med. 2010, 42, 323–331. [CrossRef] [PubMed]
8. Yingyongyudha, A.; Saengsirisuwan, V.; Panichaporn, W.; Boonsinsukh, R. The Mini-Balance Evaluation Systems Test (Mini-
BESTest) Demonstrates Higher Accuracy in Identifying Older Adult Participants with History of Falls Than Do the BESTest, Berg
Balance Scale, or Timed Up and Go Test. J. Geriatr. Phys. Ther. 2016, 39, 64–70. [CrossRef]
9. Viveiro, L.A.P.; Gomes, G.C.V.; Bacha, J.M.R.; Carvas Junisor, N.; Kallas, M.E.; Reis, M.; Jacob Filho, W.; Pompeu, J.E. Reliability,
Validity, and Ability to Identity Fall Status of the Berg Balance Scale, Balance Evaluation Systems Test (BESTest), Mini-BESTest,
and Brief-BESTest in Older Adults Who Live in Nursing Homes. J. Geriatr. Phys. Ther. 2019, 42, E45–E54. [CrossRef]
10. Aslankhani, M.A.; Farsi, A.; Fathirezaie, Z.; Zamani Sani, S.H.; Aghdasi, M.T. Validity and reliability of the timed up and go and
the anterior functional reach tests in evaluating fall risk in the elderly. Iran. J. Ageing 2015, 10, 16–25.
11. Montesinos, L.; Castaldo, R.; Pecchia, L. Wearable Inertial Sensors for Fall Risk Assessment and Prediction in Older Adults: A
Systematic Review and Meta-Analysis. IEEE Trans. Neural Syst. Rehabil. Eng. 2018, 26, 573–582. [CrossRef]
12. Howcroft, J.; Kofman, J.; Lemaire, E.D. Review of fall risk assessment in geriatric populations using inertial sensors. J. Neuroeng.
Rehabil. 2013, 10, 91. [CrossRef]
13. Piirtola, M.; Era, P.J.G. Force platform measurements as predictors of falls among older people—A review. Gerontology 2006, 52,
1–16. [CrossRef]
14. Menz, H.B.; Latt, M.D.; Tiedemann, A.; San Kwan, M.M.; Lord, S.R. Reliability of the GAITRite® walkway system for the
quantification of temporo-spatial parameters of gait in young and older people. Gait Posture 2004, 20, 20–25. [CrossRef]
15. Battaglia, G.; Giustino, V.; Messina, G.; Faraone, M.; Brusa, J.; Bordonali, A.; Barbagallo, M.; Palma, A.; Dominguez, L.-J. Walking
in natural environments as geriatrician’s recommendation for fall prevention: Preliminary outcomes from the “passiata day”
model. Sustainability 2020, 12, 2684. [CrossRef]
Biosensors 2023, 13, 998 13 of 14

16. Chen, M.; Wang, H.; Yu, L.; Yeung, E.H.K.; Luo, J.; Tsui, K.-L.; Zhao, Y. A Systematic Review of Wearable Sensor-Based
Technologies for Fall Risk Assessment in Older Adults. Sensors 2022, 22, 6752. [CrossRef] [PubMed]
17. Ma, C.Z.-H.; Lee, W.C.-C. A wearable vibrotactile biofeedback system improves balance control of healthy young adults following
perturbations from quiet stance. Hum. Mov. Sci. 2017, 55, 54–60. [CrossRef]
18. Ma, C.Z.; Zheng, Y.P.; Lee, W.C. Changes in gait and plantar foot loading upon using vibrotactile wearable biofeedback system in
patients with stroke. Top. Stroke Rehabil. 2018, 25, 20–27. [CrossRef] [PubMed]
19. Regterschot, G.R.H.; Ribbers, G.M.; Bussmann, J.B. Wearable movement sensors for rehabilitation: From technology to clinical
practice. Sensors 2021, 21, 4744. [CrossRef] [PubMed]
20. Subramaniam, S.; Faisal, A.I.; Deen, M.J. Wearable sensor systems for fall risk assessment: A review. Front. Digit. Health 2022,
4, 921506. [CrossRef]
21. Ahmad, N.; Ghazilla, R.A.R.; Khairi, N.M.; Kasi, V. Reviews on various inertial measurement unit (IMU) sensor applications. Int.
J. Signal Process. Syst. 2013, 1, 256–262. [CrossRef]
22. Ma, C.Z.-H.; Wong, D.; Lam, W.; Wan, A.; Lee, W. Balance improvement effects of biofeedback systems with state-of-the-art
wearable sensors: A systematic review. Sensors 2016, 16, 434. [CrossRef] [PubMed]
23. Sample, R.B.; Kinney, A.L.; Jackson, K.; Diestelkamp, W.; Bigelow, K.E. Identification of Key Outcome Measures When Using the
Instrumented Timed Up and Go and/or Posturography for Fall Screening. Gait Posture 2017, 57, 168–171. [CrossRef] [PubMed]
24. Greene, B.R.; O’Donovan, A.; Romero-Ortuno, R.; Cogan, L.; Scanaill, C.N.; Kenny, R.A. Quantitative Falls Risk Assessment
Using the Timed Up and Go Test. IEEE Trans. Biomed. Eng. 2010, 57, 2918–2926. [CrossRef]
25. Nouredanesh, M.; Godfrey, A.; Howcroft, J.; Lemaire, E.D.; Tung, J. Fall risk assessment in the wild: A critical examination of
wearable sensor use in free-living conditions. Gait Posture 2021, 85, 178–190. [CrossRef] [PubMed]
26. Baker, N.; Gough, C.; Gordon, S.J. Inertial sensor reliability and validity for static and dynamic balance in healthy adults:
A systematic review. Sensors 2021, 21, 5167. [CrossRef] [PubMed]
27. Salmond, S.S. Evaluating the Reliability and Validity of Measurement Instruments. Orthop. Nurs. 2008, 27, 28–30. [CrossRef]
28. Parikh, R.; Mathai, A.; Parikh, S.; Sekhar, G.C.; Thomas, R. Understanding and using sensitivity, specificity and predictive values.
Indian J. Ophthalmol. 2008, 56, 45. [CrossRef]
29. Kelsey, J.L.J.L.; Procter-Gray, E.E.; Berry, S.D.S.D.; Hannan, M.T.M.T.; Kiel, D.P.D.P.; Lipsitz, L.A.L.A.; Li, W.W. Re-evaluating the
Implications of Recurrent Falls in Older Adults: Location Changes the Inference. J. Am. Geriatr. Soc. (JAGS) 2012, 60, 517–524.
[CrossRef]
30. Diao, Y.; Lou, N.; Liang, S.; Zhang, Y.; Ning, Y.; Li, G.; Zhao, G. A Novel Environment-Adaptive Timed Up and Go Test System
for Fall Risk Assessment With Wearable Inertial Sensors. IEEE Sens. J. 2021, 21, 18287–18297. [CrossRef]
31. Early Prevention of Elderly Falls (booguu.bio). Available online: https://www.booguu.bio/aspire?lang=en (accessed on
11 September 2023).
32. Booguu Aspire (TM) AI Mobile Fall Risk Assessment—YouTube. Available online: https://www.youtube.com/watch?v=kIcZ6
o8kJNk (accessed on 11 September 2023).
33. Cheung, T.K.J. Gait Measurement with 3-Axes Accelerometer/Gyro in Mobile Devices. U.S. Patent 10307086, 20 August 2015.
34. Cheung, J.T.K. Algorithms for Gait Measurement with 3-Axes Accelerometer/Gyro in Mobile Devices. U.S. Patent 10327671,
16 February 2017.
35. Zhang, D.J. Gait Measurement with 3-Axes Accelerometer/Gyro in Mobile Devices. CN Patent 106061384A, 9 August 2019.
(In Chinese)
36. King, L.; Horak, F. On the Mini-BESTest: Scoring and the reporting of total scores. Phys. Ther. 2013, 93, 571–575. [CrossRef]
37. Koo, T.; Li, M. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J. Chiropr. Med.
2016, 15, 155–163. [CrossRef]
38. Taber, K.S. The use of Cronbach’s alpha when developing and reporting research instruments in science education. Res. Sci. Educ.
2018, 48, 1273–1296. [CrossRef]
39. Lane, H.; Brown, T. Convergent validity of two motor skill tests used to assess school-age children. Scand. J. Occup. Ther. 2015, 22,
161–172. [CrossRef] [PubMed]
40. Hajjar, S. Statistical analysis: Internal-consistency reliability and construct validity. Int. J. Quant. Qual. Res. Methods 2018, 6, 27–38.
41. Florkowski, C.M. Sensitivity, specificity, receiver-operating characteristic (ROC) curves and likelihood ratios: Communicating the
performance of diagnostic tests. Clin. Biochem. Rev. 2008, 29 (Suppl. S1), S83.
42. Marques, A.; Almeida, S.; Carvalho, J.; Cruz, J.; Oliveira, A.; Jácome, C. Reliability, validity, and ability to identify fall status of the
balance evaluation systems test, mini–balance evaluation systems test, and brief–balance evaluation systems test in older people
living in the community. Arch. Phys. Med. Rehabil. 2016, 97, 2166–2173.e1. [CrossRef]
43. Akoglu, H. User’s guide to correlation coefficients. Turk. J. Emerg. Med. 2018, 18, 91–93. [CrossRef]
44. Bautmans, I.; Jansen, B.; Van Keymolen, B.; Mets, T. Reliability and clinical correlates of 3D-accelerometry based gait analysis
outcomes according to age and fall-risk. Gait Posture 2011, 33, 366–372. [CrossRef]
45. Pappas, I.P.I.; Keller, T.; Mangold, S.; Popovic, M.R.; Dietz, V.; Morari, M. A reliable gyroscope-based gait-phase detection sensor
embedded in a shoe insole. IEEE Sens. J. 2004, 4, 268–274. [CrossRef]
46. Craig, J.J.; Bruetsch, A.P.; Lynch, S.G.; Horak, F.B.; Huisinga, J.M. Instrumented balance and walking assessments in persons with
multiple sclerosis show strong test-retest reliability. J. Neuroeng. Rehabil. 2017, 14, 43. [CrossRef] [PubMed]
Biosensors 2023, 13, 998 14 of 14

47. Hou, Y.-R.; Chiu, Y.-L.; Chiang, S.-L.; Chen, H.-Y.; Sung, W.-H. Development of a smartphone-based balance assessment system
for subjects with stroke. Sensors 2019, 20, 88. [CrossRef]
48. Di Carlo, S.; Bravini, E.; Vercelli, S.; Massazza, G.; Ferriero, G. The Mini-BESTest: A review of psychometric properties. Int. J.
Rehabil. Res. 2016, 39, 97–105. [CrossRef]
49. Marschollek, M.; Nemitz, G.; Gietzelt, M. Predicting in-patient falls in a geriatric clinic: A clinical study combining assessment
data and simple sensory gait measurements. Z. Gerontol. Geriatr. 2009, 42, 317–321. [CrossRef] [PubMed]
50. Shahzad, A.; Seunguk, K.; Samgyu, L.; Jeong, A.L.; Kiseon, K. Quantitative Assessment of Balance Impairment for Fall-Risk
Estimation Using Wearable Triaxial Accelerometer. IEEE Sens. J. 2017, 17, 6743–6751. [CrossRef]
51. Liu, J.; Zhang, X.; Lockhart, T.E. Fall Risk Assessments Based on Postural and Dynamic Stability Using Inertial Measurement
Unit. Saf. Health Work. 2012, 3, 192–198. [CrossRef]
52. Moore, J.; Stuart, S.; McMeekin, P.; Walker, R.; Celik, Y.; Pointon, M.; Godfrey, A. Enhancing Free-Living Fall Risk Assessment:
Contextualizing Mobility Based IMU Data. Sensors 2023, 23, 891. [CrossRef]
53. McManus, K.; Greene, B.R.; Ader, L.G.M.; Caulfield, B. Development of Data-Driven Metrics for Balance Impairment and Fall
Risk Assessment in Older Adults. IEEE Trans. Biomed. Eng. 2022, 69, 2324–2332. [CrossRef]
54. Cates, B.; Sim, T.; Heo, H.M.; Kim, B.; Kim, H.; Mun, J.H. A novel detection model and its optimal features to classify falls from
low-and high-acceleration activities of daily life using an insole sensor system. Sensors 2018, 18, 1227. [CrossRef]
55. Lockhart, T.E.; Soangra, R.; Yoon, H.; Wu, T.; Frames, C.W.; Weaver, R.; Roberto, K.A. Prediction of fall risk among community-
dwelling older adults using a wearable system. Sci. Rep. 2021, 11, 20976. [CrossRef]
56. Lin, C.-L.; Chiu, W.-C.; Chu, T.-C.; Ho, Y.-H.; Chen, F.-H.; Hsu, C.-C.; Hsieh, P.-H.; Chen, C.-H.; Lin, C.-C.K.; Sung, P.-S. Innovative
head-mounted system based on inertial sensors and magnetometer for detecting falling movements. Sensors 2020, 20, 5774.
[CrossRef] [PubMed]
57. Kiprijanovska, I.; Gjoreski, H.; Gams, M. Detection of gait abnormalities for fall risk assessment using wrist-worn inertial sensors
and deep learning. Sensors 2020, 20, 5373. [CrossRef] [PubMed]
58. Tunca, C.; Salur, G.; Ersoy, C. Deep learning for fall risk assessment with inertial sensors: Utilizing domain knowledge in
spatio-temporal gait parameters. IEEE J. Biomed. Health Inform. 2019, 24, 1994–2005. [CrossRef] [PubMed]
59. Qiu, H.; Rehman, R.Z.U.; Yu, X.; Xiong, S. Application of wearable inertial sensors and a new test battery for distinguishing
retrospective fallers from non-fallers among community-dwelling older people. Sci. Rep. 2018, 8, 16349. [CrossRef] [PubMed]
60. Lincoln, L.S.; Bamberg, S.J.M. Insole sensor system for real-time detection of biped slip. In Proceedings of the 2010 Annual
International Conference of the IEEE Engineering in Medicine and Biology, Buenos Aires, Argentina, 31 August–4 September 2010;
pp. 1449–1452.
61. Anson, E.; Thompson, E.; Ma, L.; Jeka, J. Reliability and Fall Risk Detection for the BESTest and Mini-BESTest in Older Adults.
J. Geriatr. Phys. Ther. 2019, 42, 81–85. [CrossRef] [PubMed]
62. Magnani, P.E.; Genovez, M.B.; Porto, J.M.; Zanellato, N.F.G.; Alvarenga, I.C.; Freire, R.C.; De Abreu, D.C.C. Use of the BESTest
and the Mini-BESTest for fall risk prediction in community-dwelling older adults between 60 and 102 years of age. J. Geriatr. Phys.
Ther. 2020, 43, 179–184. [CrossRef]
63. Lu, Z.; Lam, F.M.H.; Leung, J.C.S.; Kwok, T.C.Y. The U-Shaped Relationship Between Levels of Bouted Activity and Fall Incidence
in Community-Dwelling Older Adults: A Prospective Cohort Study. J. Gerontol. Ser. A 2020, 75, e145–e151. [CrossRef]
64. Hurtig-Wennlöf, A.; Hagströmer, M.; Olsson, L.A. The International Physical Activity Questionnaire modified for the elderly:
Aspects of validity and feasibility. Public Health Nutr. 2010, 13, 1847–1854. [CrossRef]
65. Minhee, S.; Kim, D.H.; Insook, C.; Ham, O.K. Age and gender differences in fall-related factors affecting community-dwelling
older adults. J. Nurs. Res. 2023, 31, e270.
66. Lo, J.K.; Robinson, L.R. Postpolio syndrome and the late effects of poliomyelitis. Part 1. pathogenesis, biomechanical considera-
tions, diagnosis, and investigations: Postpolio Syndrome, Part 1. Muscle Nerve 2018, 58, 751–759. [CrossRef]
67. Schwartz, A.V.; Hillier, T.A.; Sellmeyer, D.E.; Resnick, H.E.; Gregg, E.; Ensrud, K.E.; Schreiner, P.J.; Margolis, K.L.; Cauley, J.A.;
Nevitt, M.C.; et al. Older women with diabetes have a higher risk of falls: A prospective study. Diabetes Care 2002, 25, 1749–1754.
[CrossRef] [PubMed]
68. Matarese, M.; Ivziku, D.; Bartolozzi, F.; Piredda, M.; De Marinis, M.G. Systematic review of fall risk screening tools for older
patients in acute hospitals. J. Adv. Nurs. 2015, 71, 1198–1209. [CrossRef] [PubMed]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.