sensors

Review
Smart Devices and Wearable Technologies to Detect and
Monitor Mental Health Conditions and Stress:
A Systematic Review
Blake Anthony Hickey 1 , Taryn Chalmers 1 , Phillip Newton 2 , Chin-Teng Lin 3 , David Sibbritt 4 ,
Craig S. McLachlan 5 , Roderick Clifton-Bligh 6 , John Morley 7 and Sara Lal 1, *

1 Neuroscience Research Unit, School of Life Sciences, University of Technology Sydney, Broadway,
Sydney, NSW 2007, Australia; [email protected] (B.A.H.); [email protected] (T.C.)
2 School of Nursing and Midwifery, Western Sydney University, Penrith, NSW 2747, Australia;
[email protected]
3 Australian AI Institute, University of Technology Sydney, Broadway, Sydney, NSW 2007, Australia;
[email protected]
4 School of Public Health, University of Technology Sydney, Broadway, Sydney, NSW 2007, Australia;
[email protected]
5 Centre for Healthy Futures, Torrens University, Sydney, NSW 2009, Australia;
[email protected]
6 Kolling Institute for Medical Research, Royal North Shore Hospital, St Leonards, NSW 2064, Australia;
[email protected]
7 School of Medicine, Western Sydney University, Penrith, NSW 2747, Australia;





 [email protected]
* Correspondence: [email protected]; Tel.: +612-9514-1592
Citation: Hickey, B.A.; Chalmers, T.;
Newton, P.; Lin, C.-T.; Sibbritt, D.;
Abstract: Recently, there has been an increase in the production of devices to monitor mental health
McLachlan, C.S.; Clifton-Bligh, R.;
and stress as means for expediting detection, and subsequent management of these conditions.
Morley, J.; Lal, S. Smart Devices and
The objective of this review is to identify and critically appraise the most recent smart devices
Wearable Technologies to Detect and
and wearable technologies used to identify depression, anxiety, and stress, and the physiological
Monitor Mental Health Conditions
and Stress: A Systematic Review.
process(es) linked to their detection. The MEDLINE, CINAHL, Cochrane Central, and PsycINFO
Sensors 2021, 21, 3461. https:// databases were used to identify studies which utilised smart devices and wearable technologies to
doi.org/10.3390/s21103461 detect or monitor anxiety, depression, or stress. The included articles that assessed stress and anxiety
unanimously used heart rate variability (HRV) parameters for detection of anxiety and stress, with
Academic Editors: Roozbeh Ghaffari the latter better detected by HRV and electroencephalogram (EGG) together. Electrodermal activity
and Ki H. Chon was used in recent studies, with high accuracy for stress detection; however, with questionable
reliability. Depression was found to be largely detected using specific EEG signatures; however,
Received: 27 March 2021 devices detecting depression using EEG are not currently available on the market. This systematic
Accepted: 11 May 2021
review highlights that average heart rate used by many commercially available smart devices is not
Published: 16 May 2021
as accurate in the detection of stress and anxiety compared with heart rate variability, electrodermal
activity, and possibly respiratory rate.
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
Keywords: wearable devices; smart technology; electroencephalogram; heart rate variability; anxi-
published maps and institutional affil-
ety; depression
iations.

1. Introduction
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
Acute stress is a growing, unavoidable issue in contemporary society induced by
This article is an open access article
physical and or emotional stressors, which physiologically can combine to trigger or
distributed under the terms and exacerbate a wide variety of disease states [1,2]. In conjunction with negative emotions
conditions of the Creative Commons such as anxiety and depression, stress can increase cardiovascular disease risk, the leading
Attribution (CC BY) license (https:// cause of mortality worldwide [3]. Anxiety disorders are the leading mental health illness,
creativecommons.org/licenses/by/ with 264 million affected worldwide [4], with depression projected to be the second major
4.0/). cause of disability in the coming decade [5]. Further, the incidence of these mental health

Sensors 2021, 21, 3461. https://doi.org/10.3390/s21103461 https://www.mdpi.com/journal/sensors

Sensors 2021, 21, 3461 2 of 17

issues is increasingly developing in low- and middle-income countries [4]. Therefore,

researchers are trying to create more compact, portable, and accurate technology to monitor
stress and mental health status (depression or anxiety). Such devices will ultimately reduce
morbidity and the economic burden on the health care system, as patients can seek help
earlier or act to reduce symptoms or triggers [6,7].
Vital signs, neural activity (electroencephalogram ((EEG)), heart rate (electrocardio-
gram ((ECG)), skin temperature, and skin conductance response (electrodermal activity)
can provide important information about an individual’s health status [8]. However, the
challenge is how to make this information more readily available outside of the clinical
environment using semi-validated, wearable devices that are tolerated by people and have
regulatory approval for their stated purpose [6]. Within the past decade, the creation of
commercially available smart devices and wearable technologies to monitor health has
grown exponentially [9]. Many smart wearable devices are being developed, including
smart textiles, pedometers, wearable EEG systems, smart watches with photoplethysmog-
raphy, and many other devices that can non-invasively measure several health-related
factors [8].
However, stress and mental illness are in a different paradigm and often difficult to
monitor objectively, with the recent focus on the feasibility of creating technology capable
of detection of mental states. Further, research over the last ten years has demonstrated that
people are reluctant and find it uncomfortable to wear invasive or large intrusive devices
for measuring health status [1,6]. Since smartphones and wearable devices are often carried
on one’s person as an integral part of life in modern society, they are often chosen as the
instruments to detect and monitor stress, anxiety, and depressive symptoms. This review
focuses on both bulky wearables and sensor wearables; devices which are tolerable to the
wearer, portable, and proposed to be capable of detecting stress, anxiety, and depression.
What is currently known about wearable devices which measure these mental states is the
primary focus of this review.
Specifically, this systematic review aims to answer the following questions:
1. What types of smart devices and wearable technologies are being used to detect or
monitor depression, anxiety, and stress?
2. What physiological or other process(es) do smart devices and wearable technologies
utilise to detect depression, anxiety, and stress?
3. Which of these devices have been manufactured and are available on the market as
technology?

2. Materials and Methods

2.1. Selection Criteria for the Present Review
Review Period. This systematic review was limited to the articles identified using
the search strategy outlined below to review wearable smart technologies’ most current
developments to detect and/or monitor depression, anxiety, and stress, and how this is
detected physiologically using such technology. The search was conducted initially on 27
December 2020 and updated on 4 January 2021, with a further updated search on 11 April
2021.
Types of studies and study design. Various study designs were included due to the
low yield of studies. In addition, studies that only used machine learning without their use
in a current wearable device or application were excluded.
Categories of effect. All studies were included in this review based on the primary
search measures stated in the Primary Effects below. Studies were not excluded based on
the reported method of investigation or research field.
Primary effects. The identification of wearable technologies and smart applications
that detect or monitor fluctuations in mental health (specifically depression, anxiety, and
stress) and the physiological measures used to detect these changes. The selection criteria
for inclusion of articles were agreed upon by the collaborative expertise of the authors
(B.H., T.C., P.N. and S.L.).
Sensors 2021, 21, 3461 3 of 17

2.2. Literature Search Methods

Peer-reviewed journal publications with complete full text, regardless of study design
and the country where the study was conducted, were the only articles included in this
review. Year of publication was not included in the exclusion criteria due to the focus on
modern technology, with the earliest included paper about this topic published in 2010.
Studies published in a language other than English, or those structured as editorials, news
releases, research highlights, or letters, review articles, commentaries, and technical papers,
were excluded.

2.3. Electronic Database Search

The CINAHL, MEDLINE, Cochrane Central, and PsycINFO databases were searched.
The primary search terms were limited to the title and abstract fields and included wearable
device, wearable technology, smart device, and wireless device. Secondary search terms
were limited to the title and abstract fields and included depression, anxiety, stress, and
fatigue. The search structure for the Medline (Table 1), CINAHL Plus (Table 2), Cochrane
Central (Table 3), and PsycINFO (Table 4) databases are provided below.

Table 1. The search structure for the MEDLINE database.

Search Term No. of Results

Step 1. Primary search term 1: Wearable device.ab,ti. 1496
Step 2. Primary search term 2: Wearable technology.ab,ti. 892
Step 3. Primary search term 3: Smart device.ab,ti. 1265
Step 4. Primary search term 4: Wireless device.ab,ti. 1429
Step 5. Combined primary search terms: Search term 1 or 2 or 3 or 4 2987
Step 6. Using only the secondary search terms below:
- Depression.ab,ti 73,999
- Anxiety.ab,ti 50,356
- Stress.ab,ti 162,184
Step 7. Combined Step 5 and Step 6 (where Step 6 was):
- Depression.ab,ti 34
- Anxiety.ab,ti 24
- Stress.ab,ti 113

Table 2. The search structure for the CINAHL Plus database.

Search Term No. of Results

Step 1. Primary search term 1: Wearable device.ab,ti. 454
Step 2. Primary search term 2: Wearable technology.ab,ti. 463
Step 3. Primary search term 3: Smart device.ab,ti. 345
Step 4. Primary search term 4: Wireless device.ab,ti. 346
Step 5. Combined primary search terms: Search term 1 or 2 or 3 or 4 1358
Step 6. Using only the secondary search terms below:
- Depression.ab,ti 35,468
- Anxiety.ab,ti 25,877
- Stress.ab,ti 42,329
Step 7. Combined Step 5 and Step 6 (where Step 6 was):
- Depression.ab,ti 18
- Anxiety.ab,ti 17
- Stress.ab,ti 45
Sensors 2021, 21, 3461 4 of 17

Table 3. The search structure for the Cochrane Central database.

Search Term No. of Results

Step 1. Primary search term 1: Wearable device.ab,ti. 535
Step 2. Primary search term 2: Wearable technology.ab,ti. 300
Step 3. Primary search term 3: Smart device.ab,ti. 434
Step 4. Primary search term 4: Wireless device.ab,ti. 313
Step 5. Combined primary search terms: Search term 1 or 2 or 3 or 4 1394
Step 6. Using only the secondary search terms below:
- Depression.ab,ti 78,922
- Anxiety.ab,ti 53,821
- Stress.ab,ti 60,314
Step 7. Combined Step 5 and Step 6 (where Step 6 was):
- Depression.ab,ti 93
- Anxiety.ab,ti 93
- Stress.ab,ti 85

Table 4. The search structure for the PsycINFO database.

Search Term No. of Results

Step 1. Primary search term 1: Wearable device.ab,ti. 164
Step 2. Primary search term 2: Wearable technology.ab,ti. 121
Step 3. Primary search term 3: Smart device.ab,ti. 123
Step 4. Primary search term 4: Wireless device.ab,ti. 79
Step 5. Combined primary search terms: Search term 1 or 2 or 3 435
Step 6. Using only the secondary search terms below:
- Depression.ab,ti 76,926
- Anxiety.ab,ti 62,904
- Stress.ab,ti 72,329
Step 7. Combined Step 5 and Step 6 (where Step 6 was):
- Depression.ab,ti 14
- Anxiety.ab,ti 13
- Stress.ab,ti 18

Selection of studies. Upon completion of the systematic database search, duplicate

articles were discarded. Two authors (B.H. and T.C.) screened the remaining articles
separately with use of Covidence systematic review software [10] which allows blind
review of articles by each reviewer to prevent bias, with unanimity on the articles removed.
No additional articles were added from external sources. Following inspection of titles
and abstracts, articles were extracted using the criteria above. The remaining articles
were subsequently sourced in full-text and critically appraised using the Joanna Briggs
Institute Critical Appraisal Checklist [11], a 10-question checklist appraising the research
methodology and level of bias of research articles.

3. Results
The overall systematic search strategy was conducted using Covidence systematic
review software, which allows for collation of citations, records articles not suitable for
review and the reasoning for rejection, and allows blind screening of articles between
reviewers. The results of the systematic search process performed on Covidence are
presented in Figure 1. A total of 567 studies were yielded by the systematic search across
the four databases mentioned above, with no additional studies added outside of this
search strategy. Of the identified studies, 223 duplicates were removed, leaving 342 studies.
 review and the reasoning for rejection, and allows blind screening of articles between re-
viewers. The results of the systematic search process performed on Covidence are pre-
Sensors 2021, 21, 3461sented in Figure 1. A total of 567 studies were yielded by the systematic search across the 5 of 17
four databases mentioned above, with no additional studies added outside of this search
strategy. Of the identified studies, 223 duplicates were removed, leaving 342 studies. A
further 265 studies were265
A further excluded
studiesfollowing
were excludeda review of their title
following and abstract,
a review of their with stud-abstract, with
title and
ies excluded studies
if the abstract did not mention the use of wearable technologies or
excluded if the abstract did not mention the use of wearable technologies or smart smart
devices for detection
devices forof stress,
detectionanxiety, or depression.
of stress, When the search
anxiety, or depression. Whenparameters anxi-
the search parameters anxiety,
ety, depression, and stress were used in addition to the primary search
depression, and stress were used in addition to the primary search terms, terms, the com-the combined
bined numbernumber
of articles yieldedyielded
of articles was small,was with
small,considerable overlap
with considerable of studies
overlap between
of studies between search
search terms terms
and databases. The search term “anxiety” yielded 91 articles for potential
and databases. The search term “anxiety” yielded 91 articles for potential review; of
review; of these, most
these, mostof the articles
of the were
articles wereeither notnot
either experimental
experimental studies, diddid
studies, notnot
useuse
the the wearable
wearable device to measure any physiological parameters, or anxiety was never assessed
device to measure any physiological parameters, or anxiety was never assessed in the study.
in the study. The search term
The search term “stress”
“stress”yielded
yielded121 121journal
journalarticles,
articles, with
with the
the potential
potential to to
be be
included in this
included in this review;
review; of these,
of these, the majority
the majority werewere removed
removed forusing
for not not using the wearable
the wearable device to measure
device to measure physiological
physiological data relating
data relating to stress,
to stress, with with remaining
remaining paperspapers
removedremoved
for being review
for being review papers,
papers, articles
articles withwith results
results not currently
not currently available,
available, or articles
or articles outlining
outlining the methodology
the methodology of studies
of studies yet toyetbetocompleted.
be completed. TheThe search
search termterm “depression”
“depression” yielded
yielded 97
97 journal articles,
journal articles, with the potential to be included in this review; of these,
with the potential to be included in this review; of these, the majority the majority were removed for
removed for notnotassessing
assessingdepression
depression in in the
theexperimental
experimental studystudy or
orforfornot
notcollecting
collectingphys-
physiological data
iological datawhich
whichwas wasused
usedtotoassess
assessfor fordepressive
depressivesymptoms.
symptoms.

Figure 1. Flow diagram of the systematic search strategy and results conducted using Covidence systematic review
Figure 1. Flow diagram of the systematic search strategy and results conducted using Covidence
software [10].
systematic review software [10].
Following a review of the 75 full-text studies remaining, 54 studies were excluded,
primarily due to not having stress, anxiety, or depression as a measured outcome, lack
of experimental results, the article focused primarily on the machine-learning aspect of a
device, or the study used a paediatric population. The remaining 21 studies were critically
Sensors 2021, 21, 3461 6 of 17

appraised using the Joanna Briggs Institute Critical Appraisal Checklist and included in
the review.
The key findings of these papers, including the type of wearable device, the physio-
logical process used to detect anxiety, depression and/or stress, and whether the device
is commercially available, are discussed below. A summary of the 21 included articles is
presented in Table 5.

Table 5. Summary table of included research articles which utilise wearable technologies to detect stress, anxiety, and
depression.

Study Wearable Technology Population Objective Primary Results

Combined EEG and HRV
14 male subjects, young To develop a wearable device were most accurate (87.5%) in
Novel, wearable EEG and
(mean age: 29.4 ± 3.3), consisting of four sensors, with the detecting stress in healthy
Ahn et al., 2019 [12] ECG head band that hangs
healthy, with no reported capability of detecting stress using subjects, compared with EEG
behind both ears.
comorbidities. dual EEG and ECG. alone (77.9%) and HRV
(75.0%).
Spire Stone (Spire Health, San The device was acceptable to
To assess the feasibility, acceptability,
Francisco, CA, USA) and 65 subjects (30 male, 90.3% of subjects. The use of
and efficacy of wearable respiratory
Spire Health Tag respiratory 35 female), middle-aged wearable respiratory
Alberts et al., 2020 [13] monitoring for chronic pain and
monitor (Spire Health, San (mean age: 44.1 ± 8.7), in monitoring may be useful in
associated stress, depression, and
Francisco, CA, USA) which adults with chronic pain. monitoring anxiety and
anxiety.
monitor respiratory rate. stress.
The use of wearable
Wearable brain-sensing 55 subjects (38 male, technology and mindfulness
To use wearable devices in
device, either the Muse™ 17 female), young (mean age: resulted in reduced perceived
conjunction with mindfulness
Balconi et al., 2019 [14] headband (InteraXon Inc.) or 23.2 ± 1.8), with no known stress and anxiety. Further,
techniques to reduce overall stress
the Lowdown Focus glasses diagnosed psychiatric illness HRV measures were reduced
and anxiety in a healthy population.
(SmithOptics Inc.). or cognitive impairment. by the technology-mediated
mindfulness exercise.
Post-ketamine treatment EEG
55 subjects (10 male,
Wearable forehead EEG To determine the response to signatures in the prefrontal
45 female), middle-aged
device the “Mindo-4S ketamine in patients with TRD using cortex showed improvements
Cao et al., 2018 [15] (mean age: ~mean age: 48),
Jellyfish” (Eee Holter a wearable forehead EEG and in EEG depressive signatures
with treatment-resistant
Technology Co.). Hamilton depression rating score. when compared with baseline
depression (TRD).
EEG (p < 0.05).
Aversive emotional responses
Sensor smart wristband Using a wearable device to detect if (i.e., stress, anger, fear)
30 subjects (10 male,
(Bodymonitor) which invasion of individual personal detected by EDA-based
Engelniederhammer 20 female), young (mean age:
monitors electrodermal space in crowded environments emotion response type, were
et al., 2019 [16] 24.8 ± 0.7) with no known
activity (EDA) and skin elicits an emotional response, increased when personal
diagnosed psychiatric illness.
temperature. including stress and aggression. space was encroached upon
on busy streets.
HRV indices from the R–R
To validate the Apple Watch in terms interval series reflected
Apple Watch compared 20 subjects, young in age with of HRV derived from the R–R changes brought on by mild
Hernando et al., against the Polar H7 chest no reported comorbidities interval series against the R–R stress compared with relaxed
2018 [9] device (Polar Electro Oy) both (little demographic interval provided by the Polar H7 states. HRV from the Apple
of which detect HRV. information provided). chest device and validated in both a Watch was not significantly
stressed and relaxed state. different to the Polar H7 chest
device.
HRV parameters had high
29 (25 male, 4 female) healthy, To verify the reliability of a wearable concordance correlations with
Dual electrode ECG device young (mean age = 29.62 ± ECG device to detect stress, by stress. Catecholamines had
Hong et al., 2010 [17] the Biopatch, worn on the 5.28) participants, no known correlating HRV to qualitative stress the highest correlation with
precordium. cardiovascular or indices, cortisol, and catecholamine qualitative stress level. HRV
neurological disease. levels. had weak correlation with
cortisol.
40 (17 male, 23 female) HRV parameters on the
Heart rate variability To investigate the validity of the
elderly (aged 65–79 years old) wearable device were highly
measured using Polar V800 Polar V800 heart rate monitor for
participants, with no correlated with ECG HRV,
Huang et al., 2021 [18] Heart Rate Monitor (Polar assessing R–R intervals and to
established cognitive and were capable of
Electro OY, Kempele, evaluate differences in HRV when
impairment or neurological adequately measuring HRV
Finland), under physical or cognitive stressors.
disease. under different stressors.
Galvanic skin response (GSR) To develop a statistical model that Classification accuracy of the
9 subjects (no further
Kim et al., 2020 [19] was recorded using Empatica accurately classifies driving stress by device to detect stress during
demographics provided).
wristband (E4, Empatica Inc.). monitoring GSR. driving was 85.3%.
99.1% detection of mild
36 subjects (24 male, To create a device which accurately
Single electrode EEG (no depression using k-nearest
Li et al., 2015 [5] 12 female), 9 of which had and objectively detects depression
specifics provided). neighbour kNN classifiers on
mild depression. and related risk factors.
EEG.
Sensors 2021, 21, 3461 7 of 17

Table 5. Cont.

Study Wearable Technology Population Objective Primary Results

Baseline HR significantly
23 subjects (16 male, mean increased during a shift when
To monitor changes in HR while
age 30.1, 7 female, compared to baseline, with
Apple Watch (series 1) to exposed to occupational stress and
Lucas et al., 2019 [20] mean age 30.6), majority not females’ HR significantly
track HR changes. to determine if gender or physical
physically active, with no higher than males (p < 0.001).
activity influences this change in HR.
chronic comorbidities. Physical activity level had no
significant impact.
Using a machine learning
To develop a Short-Term Depression model, subjects’ passive
Detector (STDD) framework that sensing data (mood, physical
Android smartphone and utilises a smartphone and activity, and sleep) was used
20 subjects (no demographic
smartwatch (Gear S3 Frontier) smartwatch to continuously monitor to classify depression
Narziev et al., data) with either no, mild,
accelerometer, significant self-reported symptoms and passive category and had high
2020 [21] moderate, or severe
motion, and step count data, sensing data, respectively. Mood correlation to the
depression (PHQ-9).
to assess physical activity. was ascertained by a combination of self-reported depression score.
physical activity measures and mean The STDD had an accuracy of
HR. 96 ± 2.8% in depressive
group classification.
Baseline HR was significantly
18 subjects (4 male, lower than the HR recorded
To evaluate if wearable technology
14 female), young (mean age: during a stressor (as
Fitbit® (equipped with can detect physiological responses to
Pakhomov et al., 20.6 ± 2.0) that were identified by EMA surveys),
photoplethysmography stress when an individual is exposed
2020 [22] non-smokers, not pregnant up to ~9 beats higher.
sensor). to both stress-inducing scenarios and
and with no comorbidities Concluded that Fitbit may be
stress in everyday life.
effecting HR. useful in identifying stressors
in everyday life.
The TSST stress task resulted
in significant changes in
® To assess the impact that stress has
Vital Jacket (1-Lead, 11 subjects (8 male, 3 female), several HRV parameters and
on cognitive performance in air
Rodrigues et al., Biodevices S.A), a mean age 46.7 ± 5.9, with no subjective stress level
traffic controllers and if this
2018 [23] medical-grade wearable ECG known cardiovascular (STAI-six item) compared to
decreased performance was related
device and actigraph. diagnoses. baseline readings, as well as a
to an autonomic stress response.
decline in cognitive
performance.
Two sensors on each wrist: To determine how accurately
Wearable sensor features (SC,
Q-sensor (Affectiva) to physiological and behavioural
ST) were more accurate in
measure skin conductance 201 college students measures recorded off wearable
classifying poor mental health
(SC), skin temperature (ST), (129 male, 72 female), young devices could detect stress and poor
Sano et al., 2018 [24] (87%) and stress (78.3%) than
three-axis acceleration (ACC), (aged 18–25) collected across mental health (anxiety/depressive
mobile phone and modifiable
and Motion Logger to 3 years (2013–2015). symptoms), and to evaluate which of
behavioural modalities when
measure acceleration and these physiological or behavioural
using machine learning.
ambient light data. measures were most accurate.
Multiparametric sensorised The Assessment in Real Time of the
HR and respiratory rate,
garment, containing one: Stress in Combatants (ATREC)
collected via ECG and
GSR, Temperature unit, ECG, project aimed to create a
electrical bioimpedance
Thoracic Impedance comprehensive wearable system to
42 subjects (no demographic measurements from the
Seoane et al., 2014 [25] recording unit, Sensorised detect the real-time emotional,
information provided). thorax, respectively, were
glove, upper-arm strap, chest physical, and mental stress load of
more useful in the assessment
strap system, and six soldiers during military combat by
of stress than GSR, skin
textrodes and smart phone monitoring various physiological
temperature, or speech.
for voice recording. parameters.
Microsoft Smartband 2™,
which measures skin
To assess the level of stress Multiple HRV parameters
conductance, body 82 Portuguese (63 male,
experienced by medical students were significantly different
Silva et al., 2020 [26] temperature, heart rate 19 female) young (17–38 years
during examination by monitoring during stressful condition
variability, calorie intake and old) medical students.
of HRV parameters. than baseline.
expenditure, sleep patterns,
and quality.
Of those that partook in the
stress management
Spire Stone (Spire Health) To determine if mental health
intervention, 28.2% felt a
which tracks respiratory rate 169 subjects (76 male, outcomes can be improved via a
reduction in stress and
Smith et al., 2020 [27] and variability. It identifies 93 female), young wearable stress management
anxiety. The respiratory data
stress, which is relayed to a (mean age: 33.2 ± 7.8). intervention and breathing
was unavailable due to poor
smartphone. biofeedback.
subject compliance in
wearing the device.
There is a significant
This study is part of the Stress in relationship between HR and
Chillband (wristband) to
Work Environment (SWEET), with the three-way interaction of
measure skin conductance 328 subjects (186 male,
vanKraaij et al., the aim to determine if chronic stress chronic stress, gender, and the
and temperature, and 142 female),
2020 [28] influences HR over time and to circadian harmonic. Female
ePATCH™ Extended Holter mean age 38.9 ±10.2.
assess if gender or age modulates gender was found to be
monitor.
this effect. associated with higher heart
rates.
Sensors 2021, 21, 3461 8 of 17

Table 5. Cont.

Study Wearable Technology Population Objective Primary Results

Subjects with depression
5895 subjects from the
Actiwatch (Cambridge To use feature extraction of motor were accurately detected in
Zanella-Calzada et al., “Depresjon Dataset” were
Neurotechnology Ltd.) activity level data to detect 86.7% of cases and those
2019 [29] used. No demographic data
detects activity levels. depressed subjects accurately. without depression were
was given.
identified in 91.9% of cases.
Subjects with depression
were accurately classified in
Brain oxy-hemodynamic 92.6% of subjects. This study
31 subjects, 14 with clinically
(HbO) responses were also identified mean HbO,
diagnosed major depressive This study aimed to assess the
recorded using a 16-channel full width half maximum and
disorders (6 male, 8 female), feasibility of fNIRS to assess and
Zhu et al., 2020 [30] wearable continuous-wave kurtosis, as specific
and 17 healthy controls classify depression using a motor
functional near-infrared neuromarkers, for predicting
(6 male and 11 female), rehabilitation task.
spectroscopy (fNIRS) device major depressive disorders
aged 20 to 80 years.
model 1000, (United States). across dorsolateral and
ventrolateral prefrontal
cortex.
ECG = electrocardiogram; EEG = electroencephalogram; EMA = ecological momentary assessment; HR = heart rate; HRV = heart rate
variability; SC = skin conductance; ST = skin temperature; STAI = state-trait anxiety inventory; TSST = trier social stress test.

Of the 21 studies reviewed, five wearables measured HRV to detect stress, two wear-
ables measured HRV and EEG concurrently to detect stress and anxiety, and two wearables
measured fluctuations in HR to detect stress. Utilising the physiology of the integumentary
system, three wearables measured either galvanic skin response (GSR) or electrodermal
activity (EDA) to detect stress, and one device measured skin conductance and HR to
detect stress. Depression was detected with two wearables which measured EEG signals,
one wearable which measured HR and podometry, one wearable which measured activity
using a podometer, and one wearable which recorded brain oxy-hemodynamic responses
using functional near-infrared spectroscopy (fNIRS). A multiparametric garment was one
wearable capable of measuring GSR, skin temperature, ECG, and respiratory rate to detect
stress in soldiers. Finally, two very recent wearable devices measured respiratory rate to
detect stress, but with one study reporting too few results.

4. Discussion
4.1. Anxiety
Anxiety is a common mental health issue, particularly in Australia where the preva-
lence is increasing [31]. It is defined as an unpleasant, emotional response out of proportion
to a particular stressor (or even in the absence of), the response of which may or may not
be prolonged, resulting in tension and physiological manifestations [32,33]. Episodes of
anxiety are triggered from unnecessary stimulation of the hypothalamic–pituitary–adrenal
axis, which stimulates the sympathetic limb of the autonomic nervous system (while simul-
taneously dampening the parasympathetic limb), which results in both psychological and
physiological manifestations [32]. Of the latter, alterations in heart rate, respiratory rate and
electrodermal activity reflect the function of the sympathetic nervous system [34]. Further,
heart rate variability, which is calculated from the R–R interval, has been previously shown
to represent the autonomic nervous system activity and is a good marker for stress and
anxiety, with anxiety resulting in decreased R–R interval time and increased heart rate due
to bolstered sympathetic response and reduced vagal inputs [35,36]. In a recent review by
Elgendi and Menon (2019), the validity of using ECG parameters using wearable devices
to detect different clinical diagnoses of anxiety was assessed [4]. The overall findings of
experimental papers were conflicting and controversial, and the authors concluded that it
was challenging to determine the impact ECG features had on determining anxiety with
a need for more robust studies moving forward [4]. These cardiovascular measures, as
well as respiratory and skin-related measurements, have been incorporated into wearable
technologies that were assessed in the studies below.
Sensors 2021, 21, 3461 9 of 17

4.2. Findings of This Review

Four studies [13,14,24,27] were identified for inclusion in this review based on the
secondary search term “anxiety”. The study by Balconi et al. (2019) used wearable EEG and
ECG (for subsequent HRV computation) devices, including either the Muse™ headband
(InteraXon Inc., Toronto, ON, Canada) or the Lowdown Focus glasses (SmithOptics Inc.,
Clearfield, UT, USA) to determine the effects of mindfulness exercises on both an individ-
ual’s objective and subjective levels of stress and anxiety [14]. The study provided minimal
detail on which EEG signatures were used on subjects, though the authors suggested that
the wearable brain-sensing device has potential for promoting objective stress response
by increasing awareness of EEG signatures of dysfunctional hyperactivation [14]. With
respect to cardiovascular changes, HRV measures were reduced both at rest and during
the stressor task, in conjunction with subjective decrease in stress and anxiety. Reduction
of high-frequency components of HRV were found to be useful autonomic measures of
the impact of stressors or stress-inducing situations and therefore have implications for
the assessment of anxiety and stress [14,37]. Further, reduction in the high-frequency
component of HRV (which is a marker of parasympathetic influence on cardiac activity) is
consistent with the neurovisceral integration model of stress response [38], which outlines
the physiological association between parasympathetic vagal activity and improved execu-
tive function (alluding to sympathetic function induced by stress and anxiety, dampening
executive function).
Cardiac activity has been the predominant objectively measurable physiological pa-
rameter of anxiety in the literature; however, respiratory patterns have been reported to
robustly indicate cognitive emotional stress [39,40]. The second article included in the
present review, Smith et al. (2020), attempted to measure respiratory rate and variability
to compare physiological parameters with subjective scores of anxiety and stress, using
the only wearable device (Spire Stone (Spire Health, Stanford, California)) available at the
time that could measure these parameters unobtrusively [27]. Despite the capability of
the device in measuring respiratory rate and variability, there was a lack of compliance
by subjects in the experimental group (they only wore the device 52% of the study days),
despite the majority of subjects having reported high tolerability for the device. It was
noted that breathing exercises are often implemented to regulate anxiety in people and
were not assessed in the biofeedback model of the study [27], which may be worth investi-
gating in the future, as slow, deep breathing is useful in reducing anxiety. The third article
included in this review for anxiety, Alberts et al. (2020), also used the Spire Stone (Spire
Health, Stanford, CA, USA)) and an adapted version, the Spire Health Tag respiratory
monitor (Spire Health, San Francisco, CA, USA). Unlike the study by Smith et al. (2020),
the Spire Stone was found to be tolerable in 90.3% of participants, with respiratory rate
patterns found to be useful in detection of anxiety and stress [13]. Further studies testing
respiratory rate and variability using wearable technology alongside subjective stress and
anxiety results are required.
The relationship between the sympathetic nervous system and the integumentary is
well-known, with this physiological relationship being used to detect anxiety, stress, and
even depression. The fourth study included in the review for “anxiety”, Sano et al. (2018)
using two sensors, compared the accuracy of skin conductance (SC), skin temperature (ST),
and the three-axis acceleration using the wearable Q-sensor (Affectiva, Boston, MA, USA)
and the Motion Logger (AMI, Ardsley, NY, USA) which records activity levels, in detecting
mental health conditions and stress, with ST and SC being more useful in detection of
stress and mental health conditions [24]. This is unsurprising, as SC has been considered a
biomarker for stress [41] and also reflects the level of autonomic arousal, which can provide
a stress index during wakefulness. With an accuracy of 87% and 78.3% for detecting poor
mental health and depression, respectively, the dual sensor device Q-sensor (Affectiva) was
a success in the study by Sano et al. (2018), as one of the first wearable devices to detect
stress in a 24/7 daily life setting. Unfortunately, the findings of the three-axis acceleration,
which can be used to estimate activity levels and sleep or wake patterns, was not the overall
Sensors 2021, 21, 3461 10 of 17

focus of this study [24], though many smart phones currently use this technology, which is
useful in detecting depressive symptoms.

4.3. Stress
The bidirectional relationship between emotion and stress is well-known, with many
papers reporting the influence emotion has over the autonomic nervous system (Kreibig
et al., 2010). The physiological response from acute stress is often protective; however,
chronic stress is known to facilitate numerous physical and mental health illnesses, which
has a significant economic impact [17,28]. The understanding of chronic stress impact on
the body has driven researchers to continue to develop new ways to detect and monitor
stress, typically relying on the sympathetic nervous system physiological responses in-
duced by stress, including changes in heart rate, heart rate variability, skin temperature,
and conductance (van Kraaij et al., 2020). Algorithms developed based on these well-
researched parameters have high accuracy for detecting stress more than 90% of the time
in experimental conditions [42,43]. The use of various wearable devices and sensorised gar-
ments have been trialled to assess if they can accurately record the physiological responses
created by sympathetic nervous system activity, using non-invasive cardiac, respiratory
rate, skin conductance, and temperature [2,44].
According to a recent review, heart rate variability is the most studied [6] of all physio-
logical parameters. This study provides a very succinct review of over 60 different wearable
technologies, which assess a multitude of different physiological parameters including
those mentioned in this review, with the addition of sleep and cognitive function [6]. This
review article also contemporaneously reviews which wearable devices have been formally
validated for use in research for stress (10%), with only 5% of the wearable technologies
listed in the review having been formally validated as capable of accurately detecting
health parameters [6].
Electrodermal activity has also gained favour as a marker of sympathetic nervous
activity due to its emerging relationship neurophysiologically [45]. Skin conductance
responses are associated with the ventromedial prefrontal cortex involved in anticipatory
EDA responses, and the amygdala involved in EDA responds to the learned association
between stimuli and reinforcement [45], with EDA now seen as an index of attention and
not merely a measure of sympathetic activity.

4.4. Findings of This Review

A total of 15 studies were identified for inclusion in this review based on the search
term “stress”. Similar to anxiety, the use of cardiac metrics, namely heart rate and heart
rate variability, were the predominant physiological markers of stress detection in 10 of
the 15 studies which detected stress [9,12,14,17,18,20,22,23,26,28]. It has been reported
that altered HRV measurements are related to ANS dysregulation associated with many
cardiovascular diseases including cardiac ischemia, myocardial infarction and heart failure,
diabetes, and obesity, as well as mental health conditions including anxiety and depres-
sion [9,46]. Hernando et al. (2019), with use of the Apple Watch, reported that HRV
measurements (in particular R–R interval series) are superior in detecting stress compared
to HR alone, with most commercially available devices using average HR, which is heavily
controlled by the autonomic nervous system and can also be drastically altered in certain
physiological and pathological circumstances [9]. Further, in these situations where there
is altered autonomic function (depression), this will be reflected in HRV metrics but not
HR alone [47].
In the study by Rodrigues et al. (2020), the Vital Jacket® (1-Lead, Biodevices S.A,
Matosinhos, Portugal) was used to assess specific HRV metrics, namely the average of
normal-to-normal intervals (AVNN), standard deviation of all normal-to-normal intervals
(SDNN), root mean square of differences between successive rhythm-to-rhythm inter-
vals (RMSSD), and low frequency/high frequency (LF/HF) ratio [23]. During stress,
AVNN, RMSSD, and the percentage of successive R–R intervals that differ by more than
Sensors 2021, 21, 3461 11 of 17

20 ms (pNN20) decreased, reflecting a depressed HRV, which is the expected response

to stress [48]. Additionally, during stress a significant increase in the LF/HF ratio was
reported, highlighting the impact of stress on the sympathovagal system [23]. These metrics
were useful in identifying stressful situations, and promote the need for the production
of quantified occupational health (qOHealth) devices to detect stress, as this study also
reported that during stressful episodes, cognitive performance declines.
In the study by Huang et al., the Polar V800 Heart Rate Monitor (Polar Electro OY,
Kempele, Finland), which monitors heart rate variability, was validated against ECG HRV
under differing stressors, with high correlations. This study found that this wearable
device is capable of monitoring stress to the same extent as an ECG, and therefore capable
of detecting acute stress [18].
During acute stress, the limbic system and thalamus are activated by the cerebral
cortex through the reticular activating system, which subsequently activates the hypotha-
lamus, triggering an autonomic nervous system and endocrine response, resulting in
catecholamine and cortisol secretion [49]. In the study by Hong et al. (2010), epinephrine,
the stress response hormone, was unsurprisingly reported as having the highest correlation
with qualitative stress levels [17]. Moreover, HRV index and LF/HF ratio were surprisingly
more accurate in stress detection than cortisol [17], alluding that HRV metrics detected
using wearable devices may be superior and more convenient than hormone and neu-
rotransmitter analysis in detecting stress. HRV parameters are reported to be the most
reliable in detecting stress, though many devices still use average HR alone, as reported
below.
In studies that only examined HR [20,22] it was found that there was a significant
difference between genders, with females having significantly higher average HR than
males when exposed to occupational stress, when measured using an Apple Watch (Series
1, Apple Inc., Cupertino, CA, USA) [20]. Interestingly, Lucas et al. (2019) also commented
that baseline cardiovascular fitness, determined by survey, had no significant impact on
HR, which is the inverse of what is physiologically expected [20]. Further, in the study
by Pakhomov et al. (2020), which used the Fitbit® (no model specifics provided) to detect
HR at baseline and during exposure to stressors, it was found that the Fitbit® is capable of
detecting stressors, with HR increasing an average of nine beats per minute [22]. Whilst
the findings by Lucas et al. (2019) and Pakhomov et al. (2020) suggest HR may be useful
in detecting stress, both studies were limited by the young age of their subjects; thus, the
impact of comorbidities on HR, and therefore stress detection, may not be reflective of
the general population [20,22]. The study by van Kraajj et al. (2020) supports this, using
two separate wearables: an unspecified chest patch for HR measurement and a wristband
(Chillband) for detecting activity, with the study reporting that there was a significant
relationship between HR and the three-way interaction between chronic stress, gender,
and circadian rhythm [28]. Further, it is known that maximum HR decreases linearly with
age [50], with sleep and stress levels fluctuating majorly throughout life, which further
supports this relationship. The influence of female gender over HR may require wearable
devices to have HR scaled to accommodate for this physiological difference, though it
appears that HRV metrics may make average HR detection obsolete.
EEG, as an adjunct to HRV in stress detection, was also assessed in two studies [12,14].
Asymmetric analysis of the frequency-band powers in the EEG, measured at the prefrontal
cortex, has been previously used to detect stress [51]. The creation of a novel EEG and ECG
system capable of simultaneously recording HRV features showed that EEG was more
accurate (87.5%) in the detection of stress compared with EEG (77.9%) and HRV (75%)
alone, thereby confirming that the simultaneous measurement of the EEG and HRV is more
effective for stress detection when combined [12]. Whilst EEG is reported as more sensitive
for stress detection in this study, its ability to be incorporated into a compact and visually
appealing wearable device is still limited; however, the Muse™ headband is capable of
doing this, though its tolerability as a wearable device is not known.
Sensors 2021, 21, 3461 12 of 17

The physiology of the integumentary system in responding to stress is a well-understood

phenomenon, with two studies [16,19] incorporating this physiology into a wearable device.
Engelniederhammer et al. (2019), who used a sensor smart wristband (Bodymonitor™,
Gesis Leibniz-Institute for the Social Sciences, Mannheim, Germany), reported that the
EDA is the most simplistic and accurate indicator of emotional arousal, notably stress or
aggression [16]. EDA is useful in the detection of stress but may pose challenges with
respect to reliability of results in populations who have comorbidities such as diabetes
mellitus or hyperthyroidism (though this can be overcome using models as outlined by
Kim et al., 2020 [19]). The study by Kim et al. (2020) used a wearable Empatica wristband
(E4, Empatica Inc., Boston, MA, USA) which recorded galvanic skin response (GSR) to
detect stress in drivers, with an accuracy of 85.3% [19]. Further, the study reported that
GSR sensors are currently the preferred method for stress detection, due to ease of setup,
its compact nature, and overall simplicity when compared with EGG and ECG [19]. One
study performed by Silva et al. (2020), using the Microsoft Smartband 2™, measured HRV
in conjunction with skin conductance, which, when incorporated into a machine learning
algorithm, could detect stress [26]. Multiple HRV parameters were significantly different
during stressful conditions than baseline, notably mean R–R and PSS13 scores.
Another study by Seoane et al. (2014) suggested that multiparametric testing (in-
cluding GSR, temperature, respiratory rate, and ECG) via a prototype wearable garment
had superior accuracy in detection of stress than EDA. This formed part of the “Assess-
ment in Real Time of the Stress in Combatants” project [25], which created a wearable
garment capable of detecting physical and mental stress within military combat soldiers
by monitoring HR, respiratory rate, and EDA [25]. Whilst the wearable device is capable
of detecting stress, there was a high rate of error across the metrics, with almost twice as
many GSR and skin temperature errors compared with ECG and respiratory rate [25]. This
wearable device has prompted other researchers [2] to develop future wearable devices
capable of detecting multiple stress-related metrics. In the study by Cho et al. (2017), the
research team wanted to create wearable technology that measured photoplethysmograms,
electrodermal activity, and skin temperature, with the aim of combining these parameters
to accurately detect stress throughout the day [2]. Further, incorporating wearables such as
this with feedback solutions to lower stress may aid in reducing the burden stress has on
people in everyday life.
In one systematic review, it was found that electrodermal activity is useful in measur-
ing neurocognitive stress, as skin conductance increases when individuals are stressed [52],
which reported a wearable not identified with the above search terms. The “shimmer
sensor” is a monitoring wearable sensor which uses EDA for stress monitoring, using two
finger sensors, capable in one reported study of detecting stress in 86% of subjects; however,
HRV and EEG data were also used in detection [52]. An additional study was also found
outside of the search criteria which measured EDA to determine the level of pre-surgery
stress, using the wrist wearable ADI-VSM (Analog Devices), with an accuracy of 85% [53].
A notable issue with EDA as a means of detecting stress was reported by Anusha
et al. (2019), who reported that devices reliant on EDA data are prone to motion artifacts;
further, varying pressure exerted on EDA electrodes related to the variable tightness of the
wearable and movement of the hand and wrists may also distort the data in a major way,
leading to potentially false readings [53].

4.5. Depression
The search term “depression” identified five studies [5,15,21,29,30] which assessed
only depression, and one study [24] which assessed “mental health” broadly. EEG is a
non-obtrusive, electrophysiological measure of the spontaneous electrical activity in the
brain and is widely used to study antidepressant treatment responses due to its availability
and low cost [54]. Two studies [5,15] reported the use of EEG in detection or monitoring
of depressive symptoms. Cao et al. (2018) tested the response of depressive symptoms to
ketamine by analysing EEG changes measured using a wearable forehead EEG device [15],
Sensors 2021, 21, 3461 13 of 17

the Mindo-4S Jellyfish (Eee Holter Technology Co. Zhubei District, Hsinchu, Taiwan).
The theta and low alpha activity signatures were used as the EEG metrics in this study,
which were significantly improved from baseline after ketamine treatment. In terms
of neurophysiology, it has been reported that depressive disorders are correlated with
a reduction in dorsolateral prefrontal cortex grey matter volumes, as well as unique
directional changes in the prefrontal cortex [55]. Li et al. (2015) also used a single-electrode
EEG (no specifics provided) to detect depressive symptoms, using specific classifiers,
including k-nearest-neighbour (kNN), naïve Bayes (NB), logistic regression (LR), support
vector machine (SVM), and random forest (RF) [5]. The kNN performed best out of
the outlined classifiers, detecting mild depressive symptoms in 99.1% with the study
concluding that a combination of linear and nonlinear EEG features proved to be effective
in improving the accuracy of detecting depression; however, the sample size of this study
was rather small [5]. One notable advancement in the wearable Mindo-4S Jellyfish by Cao
et al. (2018) is the use of dry electrodes and the reduction in preparation time; due to the
ease of wearing the device, this may eventuate into a means of monitoring depressive
symptoms daily [15].
One study by Zhu et al. (2020) used a 16-channel wearable continuous-wave func-
tional near-infrared spectroscopy (fNIRS) device model 1000 (United States) to measure
brain oxy-hemodynamic (HbO) response. The accuracy of the fNIRS in accurate classifi-
cation of depression was found to be 92.6% [30]. Further, this study also identified mean
HbO, full width half maximum, and kurtosis as specific neuro-markers for predicting
major depressive disorder across particular brain regions, notably the dorsolateral and
ventrolateral prefrontal cortex [30]. The information provided by fNIRS and EEG devices
is constantly improving the understanding of depression from a physiological stance, with
further investigations of fNIRS in a larger sample size required.
The remaining two studies [17,29] looked at activity levels for the detection of depres-
sion using a wearable actigraph watch or smart watch with or without a smartphone. In
the study by Zanella-Calzada et al. (2019), real-time measurements of behaviour, feelings,
and activity were recorded using an Ecological Momentary Assessment [56], through use
of smart phones and an actigraph watch; specifically, the Actiwatch (Cambridge Neu-
rotechnology Ltd., Cambridge, UK, model AW4) [29]. This assessment is necessary for
depression monitoring, as most depressive symptom monitoring methods rely on patient
reports, which are commonly biased. When blindly selecting depressed subjects from
non-depressed subjects, this method accurately detected depressed patients 86.7% of the
time [29]. Inversely, it also detected non-depressed subjects in 91.9% of cases. Detecting
depression based on the level of physical activity throughout a day through a smartphone
may expedite new diagnoses or recurrences in people with depression.
In another study assessing physical activity, Narziev et al. (2020) selected five depres-
sion symptom factors which were extracted from the DSM-5 questionnaire, with mood,
physical activity, sleep, social activity, and food intake (to ascertain appetite information)
and monitored to detect depression using the developed “Short-Term Depression De-
tector” (STDD) framework, which used smart watch (Galaxy S3) sensors and Android
smartphone [21]. Mood was determined by a combination of the above factors using
machine learning. For the focus of this review, it was noted that the smart watch used a
heart rate monitor and accelerometer to assess physical activity level in subjects, which is
typically lower in depression. The study reported that the STDD framework and passive
data collection had a strong correlation with the self-reported depression score, with the
STDD having an accuracy of 96% in depressive group classification (Narziev et al., 2020).
This study highlights the difficulties of objectively recognising depressive symptoms using
wearable technologies and promotes the idea of using smartphone apps to gather metrics
and qualitative data to assist in detecting depression.
Sensors 2021, 21, 3461 14 of 17

5. Conclusions
The incidence of anxiety, depression, and chronic stress are increasing globally, as is the
production of smart devices to help individuals monitor components of their health. Based
off the wearable devices yielded by the aforementioned search terms, HRV was identified
as the most useful physiological metric for detection of stress and anxiety [9,17,23]. Of
devices on the market currently, most of them utilise average HR, which can monitor
stress, albeit not as accurately as HRV parameters. Adjunctive EEG increases the accuracy
of stress detection [12]; however, future studies need to assess the applicability of dual
devices for long-term monitoring of chronic stress. EDA was considered a useful metric
for detecting stress, reported by one author to be the preferred wearable due to simplicity
and setup [19]; however, another author expressed some unreliability in results from
EDA measurement through wearable devices due to motion artefact [52]. Detection of
depression using wearable devices is an ongoing challenge, with wearable EEG [5,15] and
accelerometers [21,29] currently used for detection, with the prior capable of detecting
depression alone, whilst physical activity in a machine learning model can accurately
detect depression. Further research into combining an accelerometer and EEG into one
wearable device may yield promising results for depression diagnosis, though currently
smartphone incorporation is heavily present within the literature for detection using
subjective questionnaires and self-reporting [57].
Further research is needed into remote measurement technology, which can objectively
measure disrupted sleep, reduced sociability, physical activity, changes in mood, prosody,
and cognitive function—useful indicators of depression [58]. Passive remote measurement
technology uses sensors in activity monitors and smartphones to gather data automatically,
to assess if the wearer is demonstrating communication and activity patterns congruent
with a depressive episode [59]. At the time of this review, this technology had no observable
experimental data but, in the future, may be one of the leading technologies to detect
depression. Wearable technologies will continue to evolve in their ability to provide the
user with fast, health-related information. Whilst many of the aforementioned technologies
are not advertised as “medical devices”, emerging devices such as smart watches do
provide the wearer with a perception of their own health status, potentially influencing
an individual’s attitude and response to their perceived health status; therefore, further
investigation into validating wearable technologies and recording parameters on smart
devices for interpretation would be highly beneficial.
A potential limitation of this study is the omitted devices not yielded using the search
terms provided above, which subsequently may impede on the impact of conclusions
made by this study. Technologies that were not included in this review may have been
missed as a result of other authors’ choices of terminology, not consistent with our selected
search terms. Further research into the incorporation of other devices, which are able to
detect mental health changes and stress, not currently termed “wearables” may add to the
conclusions drawn in this review.
With mental illness incidence continually climbing, the desire for technology capable
of earlier detection or symptom management is highly desirable. While there has been
a large amount of work conducted on wearable devices used in stress detection, further
investigation into anxiety and depression detection is crucial in helping reduce the global
impact of mental health issues, with multiparametric devices likely to provide the greatest
likelihood for detection.

Author Contributions: Conceptualization, B.A.H., S.L., P.N., D.S. and T.C.; methodology, B.A.H. and
P.N.; investigation, B.A.H. and T.C.; writing—original draft preparation, B.A.H.; writing—review
and editing, B.A.H., S.L., P.N., D.S., T.C., C.-T.L., C.S.M., R.C.-B. and J.M.; supervision, S.L., P.N. and
T.C.; project administration, S.L. and P.N. All authors have read and agreed to the published version
of the manuscript.
Funding: This research was funded by the NSW Defence Innovation Network and NSW State
Government, grant number DINPP2019 S1-06, StressWatch Project.
Sensors 2021, 21, 3461 15 of 17

Acknowledgments: The authors would like to acknowledge the Neuroscience Research Unit, School
of Life Sciences, the University of Technology Sydney, through which this review was conducted. We
also thank the NSW Defence Innovation Network and NSW State Government for financial support
of this project through grant DINPP2019 S1-06, StressWatch Project.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Can, Y.S.; Chalabianloo, N.; Ekiz, D.; Ersoy, C. Continuous Stress Detection Using Wearable Sensors in Real Life: Algorithmic
Programming Contest Case Study. Sensors 2019, 19, 1849. [CrossRef]
2. Cho, D.; Ham, J.; Oh, J.; Park, J.; Kim, S.; Lee, N.-K.; Lee, B. Detection of Stress Levels from Biosignals Measured in Virtual Reality
Environments Using a Kernel-Based Extreme Learning Machine. Sensors 2017, 17, 2435. [CrossRef] [PubMed]
3. Vogel, J.; Auinger, A.; Riedl, R.; Kindermann, H.; Helfert, M.; Ocenasek, H. Digitally enhanced recovery: Investigating the use of
digital self-tracking for monitoring leisure time physical activity of cardiovascular disease (CVD) patients undergoing cardiac
rehabilitation. PLoS ONE 2017, 12, e0186261. [CrossRef] [PubMed]
4. Elgendi, M.; Menon, C. Assessing Anxiety Disorders Using Wearable Devices: Challenges and Future Directions. Brain Sci. 2019,
9, 50. [CrossRef] [PubMed]
5. Li, X.; Hu, B.; Shen, J.; Xu, T.; Retcliffe, M. Mild Depression Detection of College Students: An EEG-Based Solution with Free
Viewing Tasks. J. Med Syst. 2015, 39, 187. [CrossRef]
6. Peake, J.M.; Kerr, G.; Sullivan, J.P. A Critical Review of Consumer Wearables, Mobile Applications, and Equipment for Providing
Biofeedback, Monitoring Stress, and Sleep in Physically Active Populations. Front. Physiol. 2018, 9, 743. [CrossRef]
7. Weenk, M.; Bredie, S.J.H.; Koeneman, M.; Hesselink, G.J.; Goor, H.v.; Belt, T.H.v.d. Continuous Monitoring of Vital Signs in the
General Ward Using Wearable Devices: Randomized Controlled Trial. J. Med. Internet Res. 2020, 22, e15471. [CrossRef]
8. Chan, M.; Estève, D.; Fourniols, J.-Y.; Escriba, C.; Campo, E. Smart wearable systems: Current status and future challenges. Artif.
Intell. Med. 2012, 56, 137–156. [CrossRef]
9. Hernando, D.; Roca, S.; Sancho, J.; Alesanco, Á.; Bailón, R. Validation of the Apple Watch for Heart Rate Variability Measurements
during Relax and Mental Stress in Healthy Subjects. Sensors 2018, 18, 2619. [CrossRef] [PubMed]
10. Covidence Systematic Review Software, Veritas Health Innovation; Covidence: Melbourne, Australia. Available online: http:
//www.covidence.org (accessed on 27 December 2020).
11. Lockwood, C.; Munn, Z.; Porritt, K. Qualitative research synthesis: Methodological guidance for systematic reviewers utilizing
meta-aggregation. Int. J. Evid. Based Healthc. 2015, 13, 179–187. [CrossRef]
12. Ahn, J.W.; Ku, Y.; Kim, H.C. A Novel Wearable EEG and ECG Recording System for Stress Assessment. Sensors 2019, 19, 1991.
[CrossRef]
13. Alberts, N.M.; Leisenring, W.M.; Flynn, J.S.; Whitton, J.; Gibson, T.M.; Jibb, L.; McDonald, A.; Ford, J.; Moraveji, N.; Dear, B.F.;
et al. Wearable Respiratory Monitoring and Feedback for Chronic Pain in Adult Survivors of Childhood Cancer: A Feasibility
Randomized Controlled Trial from the Childhood Cancer Survivor Study. JCO Clin. Cancer Inform. 2020, 4, 1014–1026. [CrossRef]
14. Balconi, M.; Fronda, G.; Crivelli, D. Effects of technology-mediated mindfulness practice on stress: Psychophysiological and
self-report measures. Stress 2019, 22, 200–209. [CrossRef] [PubMed]
15. Cao, Z.; Lin, C.; Ding, W.; Chen, M.; Li, C.; Su, T. Identifying Ketamine Responses in Treatment-Resistant Depression Using a
Wearable Forehead EEG. IEEE Trans. Biomed. Eng. 2018, 66, 1668–1679. [CrossRef]
16. Engelniederhammer, A.; Papastefanou, G.; Xiang, L. Crowding density in urban environment and its effects on emotional
responding of pedestrians: Using wearable device technology with sensors capturing proximity and psychophysiological
emotion responses while walking in the street. J. Hum. Behav. Soc. Environ. 2019, 29, 630–646. [CrossRef]
17. Hong, S.; Yang, Y.; Lee, J.; Yang, H.; Park, K.; Lee, S.; Lee, I.; Jang, Y. Ambulatory stress monitoring with a wearable bluetooth
electrocardiographic device. Stud. Health Technol. Inform. 2010, 161, 66–76.
18. Huang, C.J.; Chan, H.L.; Chang, Y.J.; Chen, S.M.; Hsu, M.J. Validity of the Polar V800 Monitor for Assessing Heart Rate Variability
in Elderly Adults under Mental Stress and Dual Task Conditions. Int. J. Environ. Res. Public Health 2021, 18, 869. [CrossRef]
19. Kim, J.; Park, J.; Park, J. Development of a statistical model to classify driving stress levels using galvanic skin responses. Hum.
Factors Ergon. Manuf. Serv. Ind. 2020, 30, 321–328. [CrossRef]
20. Lucas, B.; Grayson, S.; Hamidu, H.; Han, A.; No, S.; Varghese, A.; Campisi, J. Sex differences in heart rate responses to occupational
stress. Stress: Int. J. Biol. Stress 2019, 23, 13–18.
21. Narziev, N.; Goh, H.; Toshnazarov, K.; Lee, S.A.; Chung, K.-M.; Noh, Y. STDD: Short-Term Depression Detection with Passive
Sensing. Sensors 2020, 20, 1396. [CrossRef] [PubMed]
22. Pakhomov, S.V.S.; Thuras, P.D.; Finzel, R.; Eppel, J.; Kotlyar, M. Using consumer-wearable technology for remote assessment of
physiological response to stress in the naturalistic environment. PLoS ONE 2020, 15, e0229942. [CrossRef] [PubMed]
23. Rodrigues, S.; Paiva, J.S.; Dias, D.; Aleixo, M.; Filipe, R.M.; Cunha, J.P.S. Cognitive Impact and Psychophysiological Effects of
Stress Using a Biomonitoring Platform. Int. J. Environ. Res. Public Health 2018, 15, 1080. [CrossRef]
24. Sano, A.; Taylor, S.; McHill, A.W.; Phillips, A.J.; Barger, L.K.; Klerman, E.; Picard, R. Identifying Objective Physiological Markers
and Modifiable Behaviors for Self-Reported Stress and Mental Health Status Using Wearable Sensors and Mobile Phones:
Observational Study. J. Med. Internet Res. 2018, 20, e210. [CrossRef]
Sensors 2021, 21, 3461 16 of 17

25. Seoane, F.; Mohino-Herranz, I.; Ferreira, J.; Alvarez, L.; Buendia, R.; Ayllón, D.; Llerena, C.; Gil-Pita, R. Wearable biomedical
measurement systems for assessment of mental stress of combatants in real time. Sensors 2014, 14, 7120–7141. [CrossRef]
26. Silva, E.; Aguiar, J.; Reis, L.P.; E Sá, J.O.; Gonçalves, J.; Carvalho, V. Stress among Portuguese medical students: The eustress
solution. J. Med. Syst. 2020, 44, 1–6. [CrossRef]
27. Smith, E.N.; Santoro, E.; Moraveji, N.; Susi, M.; Crum, A.J. Integrating Wearables in Stress Management Interventions: Promising
Evidence from a Randomized Trial. Int. J. Stress Manag. 2020, 27, 172–182. [CrossRef]
28. van Kraaij, A.W.J.; Schiavone, G.; Lutin, E.; Claes, S.; Van Hoof, C. Relationship Between Chronic Stress and Heart Rate over Time
Modulated by Gender in a Cohort of Office Workers: Cross-Sectional Study Using Wearable Technologies. J. Med. Internet Res.
2020, 22, e18253. [CrossRef] [PubMed]
29. Zanella-Calzada, L.A.; Galván-Tejada, C.E.; Chávez-Lamas, N.M.; Gracia-Cortés, M.; Magallanes-Quintanar, R.; Celaya-Padilla,
J.M.; Galván-Tejada, J.I.; Gamboa-Rosales, H. Feature Extraction in Motor Activity Signal: Towards a Depression Episodes
Detection in Unipolar and Bipolar Patients. Diagnostics 2019, 9, 8. [CrossRef] [PubMed]
30. Zhu, Y.; Jayagopal, J.K.; Mehta, R.K.; Erraguntla, M.; Nuamah, J.; McDonald, A.D.; Taylor, H.; Chang, S.H. Classifying major
depressive disorder using fNIRS during motor rehabilitation. IEEE Trans. Neural Syst. Rehabil. Eng. 2020, 28, 961–969. [CrossRef]
31. AIHW (Australian Institute of Health and Welfare). Australian Burden of Disease Study: Impact and Causes of Illness and Death in
Australia 2015; Series no. 19. Cat. no. BOD 22; Australian Burden of Disease Study; Australian Institute of Health and Welfare:
Canberra, Australia, 2019.
32. Liu, H.; Wen, W.; Zhang, J.; Liu, G.; Yang, Z. Autonomic nervous pattern of motion interference in real-time anxiety detection.
IEEE Access 2018, 6, 69763–69768. [CrossRef]
33. Semple, D.; Smyth, R. Oxford Handbook of Psychiatry; Oxford University Press: Oxford, UK, 2013.
34. Kreibig, S.D. Autonomic nervous system activity in emotion: A review. Biol. Psychol. 2010, 84, 394–421. [CrossRef]
35. Adheena, M.; Sindhu, N.; Jerritta, S. Physiological Detection of Anxiety. In Proceedings of the 2018 International Conference on
Circuits and Systems in Digital Enterprise Technology (ICCSDET), Pathamuttom, India, 21–22 December 2018.
36. Wen, W.; Liu, G.; Mao, Z.H.; Huang, W.; Zhang, X.; Hu, H.; Yang, J.; Jia, W. Toward constructing a real-time social anxiety
evaluation system: Exploring effective heart rate features. IEEE Trans. Affect. Comput. 2018, 11, 100–110. [CrossRef]
37. Subhani, A.R.; Kamel, N.; Mohamad Saad, M.N.; Nandagopal, N.; Kang, K.; Malik, A.S. Mitigation of stress: New treatment
alternatives. Cogn. Neurodyn. 2018, 12, 1–20. [CrossRef] [PubMed]
38. Thayer, J.F.; Åhs, F.; Fredrikson, M.; Sollers, J.J.; Wager, T.D. A meta-analysis of heart rate variability and neuroimaging studies:
Implications for heart rate variability as a marker of stress and health. Neurosci. Biobehav. Rev. 2012, 36, 747–756. [CrossRef]
39. Hovsepian, K.; al’Absi, M.; Ertin, E.; Kamarck, T.; Nakajima, M.; Kumar, S. cStress: Towards a Gold Standard for Continuous
Stress Assessment in the Mobile Environment. In Proceedings of the 2015 ACM International Joint Conference on Pervasive and
Ubiquitous Computing, Osaka, Japan, 9–11 September 2015.
40. Vlemincx, E.; Taelman, J.; De Peuter, S.; Van Diest, I.; Van den Bergh, O. Sigh rate and respiratory variability during mental load
and sustained attention. Psychophysiology 2011, 48, 117–120. [CrossRef]
41. Boucsein, W. Electrodermal Activity; Springer Science & Business Media: Berlin, Germany, 2012.
42. Boonnithi, S.; Phongsuphap, S. Comparison of heart rate variability measures for mental stress detection. In Proceedings of the
Computing in Cardiology, Hangzhou, China, 18–21 September 2011.
43. Castaldo, R.; Montesinos, L.; Melillo, P.; Massaro, S.; Pecchia, L. To What Extent Can We Shorten HRV Analysis in Wearable
Sensing? A Case Study on Mental Stress Detection. In EMBEC & NBC 2017, Proceedings of the Conference of the European Medical
and Biological Engineering Conference (EMBEC), Tampere, Finland, 11–15 June 2017; Eskola, H., Väisänen, O., Viik, J., Hyttinen, J.,
Eskola, H., Väisänen, O., Viik, J., Hyttinen, J., Eds.; Springer: Singapore, 2018; pp. 643–646.
44. Seoane, F.; Ferreira, J.; Alvarez, L.; Buendia, R.; Ayllón, D.; Llerena, C.; Gil-Pita, R. Sensorized garments and textrode-enabled
measurement instrumentation for ambulatory assessment of the autonomic nervous system response in the ATREC project.
Sensors 2013, 13, 8997–9015. [CrossRef]
45. Posada-Quintero, H.F.; Chon, K.H. Phasic Component of Electrodermal Activity is more Correlated to Brain Activity than Tonic
Component. In Proceedings of the 2019 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), Chicago,
IL, USA, 19–22 May 2019; pp. 1–4.
46. Beauchaine, T.; Thayer, J. Heart rate variability as a transdiagnostic biomarker of psychopathology. Int. J. Psychophysiol. 2015, 98,
338–350. [CrossRef] [PubMed]
47. Valenza, G.; Garcia, R.; Citi, L.; Scilingo, E.; Tomaz, C.; Barbieri, R. Nonlinear digital signal processing in mental health:
Characterization of major depression using instantaneous entropy measures of heartbeat dynamics. Front. Physiol. 2015, 6, 74.
[CrossRef]
48. Castaldo, R.; Melillo, P.; Bracale, U.; Caserta, M.; Triassi, M.; Pecchia, L. Acute mental stress assessment via short term hrv analysis
in healthy adults: A systematic review with meta-analysis. Biomed. Signal Process. Control 2015, 18, 370–377. [CrossRef]
49. Mtui, E.; Gruener, G.; Dockery, P. Fitzgerald’s Clinical Neuroanatomy and Neuroscience; Elsevier: Amsterdam, The Netherlands, 2020.
50. Tanaka, H.; Monahan, K.D.; Seals, D.R. Age-predicted maximal heart rate revisited. J. Am. Coll. Cardiol. 2001, 37, 153–156.
[CrossRef]
51. Hou, X.; Liu, Y.; Sourina, O.; Tan, Y.R.E.; Wang, L.; Mueller-Wittig, W. EEG Based Stress Monitoring. In Proceedings of the 2015
IEEE International Conference on Systems, Man, and Cybernetics, Hong Kong, China, 9–12 October 2015; pp. 3110–3115.
Sensors 2021, 21, 3461 17 of 17

52. Chen, J.; Abbod, M.; Shieh, J.S. Pain and Stress Detection Using Wearable Sensors and Devices—A Review. Sensors 2021, 21, 1030.
[CrossRef]
53. Anusha, A.S.; Sukumaran, P.; Sarveswaran, V.; Shyam, A.; Akl, T.J.; Preejith, S.P.; Sivaprakasam, M. Electrodermal activity based
pre-surgery stress detection using a wrist wearable. IEEE J. Biomed. Health Inform. 2019, 24, 92–100.
54. Baskaran, A.; Milev, R.; McIntyre, R.S. The neurobiology of the EEG biomarker as a predictor of treatment response in depression.
Neuropharmacology 2012, 63, 507–513. [CrossRef] [PubMed]
55. Chang, C.C.; Yu, S.C.; McQuoid, D.R.; Messer, D.F.; Taylor, W.D.; Singh, K.; Boyd, B.D.; Krishnan, K.R.R.; MacFall, J.R.; Steffens,
D.C.; et al. Reduction of dorsolateral prefrontal cortex gray matter in late-life depression. Psychiatry Res. Neuroimaging 2011, 193,
1–6. [CrossRef]
56. Shiffman, S.; Stone, A.A.; Hufford, M.R. Ecological momentary assessment. Annu. Rev. Clin. Psychol. 2008, 4, 1–32. [CrossRef]
57. Taj-Eldin, M.; Ryan, C.; O’Flynn, B.; Galvin, P. A Review of Wearable Solutions for Physiological and Emotional Monitoring for
Use by People with Autism Spectrum Disorder and Their Caregivers. Sensors 2018, 18, 4271. [CrossRef] [PubMed]
58. Warmerdam, L.; Riper, H.; Klein, M.C.; van de Ven, P.; Rocha, A.; Henriques, M.R.; Tousset, E.; Silva, H.; Andersson, G.; Cuijpers,
P. Innovative ICT solutions to improve treatment outcomes for depression: The ICT4Depression project. Stud. Health Technol.
Inform. 2012, 181, 339–343. [PubMed]
59. Matcham, F.; di San Pietro, C.B.; Bulgari, V.; De Girolamo, G.; Dobson, R.; Eriksson, H.; Folarin, A.A.; Haro, J.M.; Kerz, M.; Lamers,
F.; et al. Remote assessment of disease and relapse in major depressive disorder (RADAR-MDD): A multi-centre prospective
cohort study protocol. BMC Psychiatry 2019, 19, 72. [CrossRef] [PubMed]