Hostname: page-component-78c5997874-4rdpn Total loading time: 0 Render date: 2024-10-31T22:17:51.184Z Has data issue: false hasContentIssue false

Wearables measuring electrodermal activity to assess perceived stress in care: a scoping review

Published online by Cambridge University Press:  24 March 2023

Agata Klimek
Affiliation:
School for Allied Health Professions, Fontys University of Applied Sciences, Eindhoven, The Netherlands
Ittay Mannheim*
Affiliation:
School for Allied Health Professions, Fontys University of Applied Sciences, Eindhoven, The Netherlands Tranzo, School of Social and Behavioural Sciences, Tilburg University, Tilburg, The Netherlands
Gerard Schouten
Affiliation:
School for Information & Communication Technology, Fontys University of Applied Sciences, Eindhoven, The Netherlands
Eveline J. M. Wouters
Affiliation:
School for Allied Health Professions, Fontys University of Applied Sciences, Eindhoven, The Netherlands Tranzo, School of Social and Behavioural Sciences, Tilburg University, Tilburg, The Netherlands
Manon W. H. Peeters
Affiliation:
School for Allied Health Professions, Fontys University of Applied Sciences, Eindhoven, The Netherlands
*
Author for correspondence: Ittay Mannheim, Email: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

Background:

Chronic stress responses can lead to physical and behavioural health problems, often experienced and observed in the care of people with intellectual disabilities or people with dementia. Electrodermal activity (EDA) is a bio-signal for stress, which can be measured by wearables and thereby support stress management. However, the how, when and to what extent patients and healthcare providers can benefit is unclear. This study aims to create an overview of available wearables enabling the detection of perceived stress by using EDA.

Methods:

Following the PRISMA-SCR protocol for scoping reviews, four databases were included in the search of peer-reviewed studies published between 2012 and 2022, reporting detection of EDA in relation to self-reported stress or stress-related behaviours. Type of wearable, bodily location, research population, context, stressor type and the reported relationship between EDA and perceived stress were extracted.

Results:

Of the 74 included studies, the majority included healthy subjects in laboratory situations. Field studies and studies using machine learning (ML) to predict stress have increased in the last years. EDA is most often measured on the wrist, with offline data processing. Studies predicting perceived stress or stress-related behaviour using EDA features, reported accuracies between 42% and 100% with an average of 82.6%. Of these studies, the majority used ML.

Conclusion:

Wearable EDA sensors are promising in detecting perceived stress. Field studies with relevant populations in a health or care context are lacking. Future studies should focus on the application of EDA-measuring wearables in real-life situations to support stress management.

Type
Review Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2023. Published by Cambridge University Press on behalf of Scandinavian College of Neuropsychopharmacology

Summation

  • The number of studies regarding wearable sensors measuring EDA and perceived stress are increasing, with more field studies being conducted. It is shown that this technology can be used to assess perceived stress.

  • Studies that show a relation between EDA measured by wearables and perceived stress sometimes use sophisticated techniques such as machine learning. However, the fact that data processing and analyses are mainly being done offline hampers the practical usability in the field.

  • Conclusions regarding their application in a healthcare context are limited, as this context is understudied.

Considerations

  • The methodology to process the EDA signal and relate it to perceived stress varies greatly across studies, hampering comparability between studies.

  • Multiple studies include other bio-signals besides EDA features in relation to stress with the same statistical model. It was not always possible to draw conclusions about the independent relationship between EDA (without other bio-signals) and stress.

  • No solid conclusion can be drawn regarding usability and adoption in healthcare populations, as studies in naturalistic healthcare contexts are scarce.

Introduction

Stress plays a significant role in our society and is described by the World Health Organization as the ‘Health Epidemic of the 21st Century’ (Fink, Reference Fink2017). Selye described stress as ‘the disturbance of the body’s homeostasis triggered by an in- or external factor’ (Selye, Reference Selye1956). Stress is seen as a broad concept, including both a physical and mental load (Van Houdenhove, Reference Van Houdenhove2007; Dhabhar, Reference Dhabhar2018). A stressful event involves a stimulus (stressor), which triggers a response in the brain (stress perception), which activates physiological fight or flight systems in the body (stress response). The stressor can be external or internal, physical (e.g. physical exertion or a painful stimulus) or psychological (e.g. work pressure). When stress is too intense or prolonged for a person, it can also have negative consequences and lead to psychological and/or physical health conditions, such as (but not limited to) cardiovascular problems (Steptoe & Kivimäki, Reference Steptoe and Kivimäki2012), anxiety disorders (Van Houdenhove, Reference Van Houdenhove2007) and persistent physical symptoms (Deary et al., Reference Deary, Chalder and Sharpe2007). This leads to major healthcare concerns in all types of populations in which stress management plays an important role, such as people with stress-related physical problems (Deary et al., Reference Deary, Chalder and Sharpe2007). Additionally, stress management is of utmost important in people living with dementia or intellectual disability (Melander et al., Reference Melander, Martinsson and Gustafsson2017), as stress-related behavioural problems are frequently experienced.

Up to 90% of the people with dementia present challenging behaviours at some point (Devshi et al., Reference Devshi, Shaw, Elliott-King, Hogervorst, Hiremath, Velayudhan, Kumar, Baillon and Bandelow2015), whereby ‘apathy, depression, irritability, agitation and anxiety, […] are the most frequent symptoms’(Cerejeira et al., Reference Cerejeira, Lagarto and Mukaetova-Ladinska2012). Stress is considered to be an antecedent leading to challenging behaviour, such as agitation (Ballard et al., Reference Ballard, O’Brien, Gray, Cormack, Ayre, Rowan, Thompson, Bucks, McKeith, Walker and Tovee2001; Melander et al., Reference Melander, Kikhia, Olsson, Wälivaara and Sävenstedt2018). Agitation is defined as the attempt of expressing unmet needs and unexpressed emotions as a response to in- or external stimuli (Kong, Reference Kong2005) and can negatively affect the health of both patients and caregivers (Ornstein & Gaugler, Reference Ornstein and Gaugler2012). Besides dementia, other psychopathological and/or intellectual disabilities are often associated with difficulty to understand stress-induced behaviour. Children with autism spectrum disorders (ASDs) have shown to have a prevalence of 56 to 94% to present challenging behaviours (Matson et al., Reference Matson, Matson and Rivet2007). Such challenging behaviours seem not to decrease with time, which is why ‘an optimized intervention’ could help to increase their quality of life (Rattaz et al., Reference Rattaz, Michelon, Roeyers and Baghdadli2017). Early detection of stress onset using wearable sensors may be an effective way to (self-)manage stress in order to treat or prevent health issues (Zangróniz et al., Reference Zangróniz, Martínez-Rodrigo, Pastor, López and Fernández-Caballero2017). Additionally, it facilitates the management of stress before it effects behaviour and may assist caregivers to mitigate and cope with stress-related challenging behaviour (Melander et al., Reference Melander, Martinsson and Gustafsson2017). However, it is still not commonly known how accurate wearables are in measuring and predicting stress, and how they can be utilised in caregiving.

Different bio-signals, which can be measured by wearables, are essential for the detection of acute and/or chronic stress. Physiological parameters that are influenced by the sympathetic nervous system (SNS) are as follows: 1) heart activity, 2) electrodermal activity (EDA), 3) muscle activity, and 4) pupil diameter (Wijsman et al., Reference Wijsman, Grundlehner, Liu, Hermens and Penders2011). Sierra et al. (Reference Sierra, Snchez-Avila, Guerra-Casanova and Bailador-Del2011) claim that EDA and heart rate (HR) are the two most important parameters enabling a high potential detection of acute stress (Sierra et al., Reference Sierra, Snchez-Avila, Guerra-Casanova and Bailador-Del2011). EDA, also known as galvanic skin response (GSR) or Skin Conductance, is the activity resulting from constant changes of the electrical properties of the skin. Sweating leads to a continuous change of the electrical properties of the skin. These electrical changes can be measured by placing two electrodes on the skin and give an insight into a person’s emotional state, psychological or physiological arousal, which indicate stress (Kurniawan et al., Reference Kurniawan, Maslov and Pechenizkiy2013).

Changes in EDA are considered to be one of the best indicators for real-time stress (Healey & Picard, Reference Healey and Picard2005). However, the measurement of EDA still includes technological and acceptance challenges. Electrodes are commonly placed on the fingers and wrist, although it is still debated which technology and bodily location is appropriate, with respect to (technical) validity (van Dooren et al., Reference van Dooren, de Vries and Janssen2012; Payne et al., Reference Payne, Dawson, Schell, Singh and Courtney2013). Importantly, validity can be assessed on different levels (i.e. on the ‘raw’, ‘parameter’ or ‘event’ level, see (van Lier et al., Reference van Lier, Pieterse, Garde, Postel, de Haan, Vollenbroek-Hutten, Schraagen and Noordzij2020). However, bodily location may also influence the acceptance of the wearer (Peeters et al., Reference Peeters, Schouten and Wouters2021).

Different methods, protocols and systems which enable real-time stress detection by measuring biological signals are currently studied (Minguillon et al., Reference Minguillon, Perez, Lopez-Gordo, Pelayo and Sanchez-Carrion2018). Wearable sensors are of great value due to their ability to measure non-invasively, easily and wireless (Hanson et al., Reference Hanson, Powell, Barth, Ringgenberg, Calhoun, Aylor and Lach2009) and have the potential to support psychosocial approaches to manage challenging behaviours and thereby reduce pharmacological treatment (Foebel et al., Reference Foebel, Ballokova, Wellens, Fialova, Milisen, Liperoti and Hirdes2015; Berg-Weger & Stewart, Reference Berg-Weger and Stewart2017). The application of wearables is steadily increasing, and technology is constantly evolving. The integration of devices such as wearables in the healthcare department is crucial. Furthermore, these tools are offering great opportunities for patients and their caregivers in terms of monitoring vital signs on a daily life basis and therapy provision (Rienzo et al., Reference Rienzo, Di Rizzo, Parati, Brambilla, Ferratini and Castiglioni2005). However, today’s wearable devices still require improvements regarding data collection, as well as increasing usability in ambulatory settings (Posada-Quintero & Chon, Reference Posada-Quintero and Chon2020). As the majority of the studies are studying wearables in the lab, there is a lack of use in real-life situations.

A review summarising scientifically studied wearable technologies measuring EDA to detect and predict stress is currently missing. In order to develop a reliable transparent real-time stress detection system, it is important to collect fundamental knowledge about the current application and accuracy of using EDA measurements in daily life related to perceived stress. The goal of the current scoping review is therefore to create an overview of existing wearable devices used for measuring and monitoring EDA for stress detection. Such an overview will help researchers and developers to optimise stress detection using EDA, aiming for early stress recognition to support healthcare. To this end, the following three research questions are formulated:

  1. 1. Which wearable devices are scientifically studied in relation to perceived stress?

  2. 2. How are those wearables studied, regarding the research populations, measured bio-signals, which stressors (induced or observed) where studied, and in what kind of circumstances (lab vs. field studies)?

  3. 3. What is reported about the relationship between the EDA measurement of the wearable with perceived stress or stress-related (e.g. challenging) behaviours?

Methods

As the goal of this study is broad and aims to identify what is the most up-to-date knowledge regarding wearables measuring EDA, a scoping review was deemed as the most suitable method (Arksey & O’Malley, Reference Arksey and O’Malley2005; Levac et al., Reference Levac, Colquhoun and O’Brien2010; Munn et al., Reference Munn, Peters, Stern, Tufanaru, McArthur and Aromataris2018; Tricco et al., Reference Tricco, Lillie, Zarin, O’Brien, Colquhoun, Levac, Moher, Peters, Horsley, Weeks, Hempel, Akl, Chang, McGowan, Stewart, Hartling, Aldcroft, Wilson and Garritty2018). To assure that the most relevant studies were included, systematically reviewed and summarised, we followed the checklist of the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) (Tricco et al., Reference Tricco, Lillie, Zarin, O’Brien, Colquhoun, Levac, Moher, Peters, Horsley, Weeks, Hempel, Akl, Chang, McGowan, Stewart, Hartling, Aldcroft, Wilson and Garritty2018).

Search strategy

As the topic of this study strongly relates to healthcare as well as technological innovation, the databases used for this scoping review were PubMed, Medline, ACM digital and Web of Science to assure the relevance of the search.

The search string was composed of the main keywords ‘Electrodermal Activity’, ‘Technological tools’, ‘Measurement’ and ’Stress’. The search strategy was not case-sensitive, and keywords were paired together with the Boolean connector ‘AND’ and their synonyms with ‘OR’ (Table 1). An asterisk (‘*’) was used for words starting with the same letters but having a different ending. Separate words, belonging together, have been searched for by inserting quotation marks (‘…’) (Table 1).

Table 1. Set of keywords and related synonyms

Inclusion and exclusion criteria

This scoping review included peer-reviewed full-text articles written in English and studies that have been conducted on humans. To be included, articles were required to describe the measurement of EDA measured by a wearable (i.e. wireless) sensor, related to perceived stress. In this study, we considered perceived stress as an affective (i.e. emotional) state with a negative valence (Posner et al., Reference Posner, Russell and Peterson2005; Du et al., Reference Du, Huang, An and Xu2018). Studies were included when perceived stress was measured using a subject’s self-report or by observations of stress-related behaviour. Only original research (including original data collection) was included. Opinion papers were excluded from this research, because their level of scientific evidence was considered inadequate for the purpose of this study. Literature reviews were not included because the study performed was not a meta-analysis of other literature reviews. Finally, as the evolution of wearable technology has increased remarkably since 2013 (CCS Insight, 2016; Statista, 2017), all articles published before 2012 were excluded (Table 2).

Table 2. In- and exclusion criteria

Study selection

The final search was conducted on November 26 2022. Duplicates were removed using EndNote X9, and Covidence, a systematic review software (https://www.covidence.org/), was used to perform the study selection. Articles were first screened by title and abstract and by full text in the second phase by at least two independent researchers. The first author (AK) screened all papers, and the second assessment was divided between all other co-authors (IM, GS, EW, and MP). In case of a conflict, a third reviewer was involved. Calibration sessions were held with all authors in order to iteratively refine the selection process and solve ambiguities.

Data extraction

Relevant data were extracted by at least two independent researchers and compared. AK extracted all the articles, and the other co-authors (IM, GS, EW, and MP) extracted one-fourth of the total number of the included studies. Whenever a conflict was present, the reviewers met to discuss until agreement was reached. The data extraction table included author(s), publication year, study’s purpose, the wearable used, measured bio-signals, whether the study was done in the lab or in practice, type of stressor (induced or observed), population included, number of subjects, EDA features processed, whether the data were processed real time or post-processed, perceived stress assessment, and finally the main results regarding the relationship between EDA and perceived stress or stress-related behaviours.

Results

In total, 2299 records were identified, of which 941 were duplicates (see Fig. 1 for an overview). After the screening of the title and abstract, 993 articles were excluded. The remaining 365 articles were assessed for eligibility by reading the full text, after which 291 were excluded. Reasons for exclusion in the full-text phase were as follows: no wearable device was used and/or EDA was not related to perceived stress (n = 227); EDA was not measured (n = 11); not original research (n = 31); not written in English (n = 2); and other reasons (n = 20; such as no full-text availability). Finally, 74 articles fulfilled the required criteria to be included in this scoping review. The full data extraction table of all 74 studies can be found in the appendix 1.

Study population

Table 3 presents a summary of main study characteristics. The majority of studies (67.6%) measured stress exclusively on healthy subjects. 12.2% gave insufficient or unclear information about their study population. 9.5% reported stress measurements within children, aged from 0 to 17 years. Two of them included children diagnosed with ASD (Baker et al., Reference Baker, Fenning, Erath, Baucom, Moffitt and Howland2018; Goodwin et al., Reference Goodwin, Mazefsky, Ioannidis, Erdogmus and Siegel2019) or other neurodevelopmental problems (Betancourt et al., Reference Betancourt, Dethorne, Karahalios and Kim2017). Redd et al. (Reference Redd, Silvera-Tawil, Hopp, berg, Martiniuk, Dietrich and Karunanithi2020); Shao et al. (Reference Shao, Elsadek and Liu2020) and Weyn et al. (Reference Weyn, Van Leeuwen, Pluess, Goossens, Claes, Bosmans, Van Den Noortgate, Lutin, Bröhl, Chubar, Geukens and Bijttebier2022) included elementary school students in their research. One study included neonates on mechanical ventilation (Walas et al., Reference Walas, Halaba, Kubiaczyk, Piotrowski, Latka-Grot, Szczapa, Romul, Maroszynska, Malinowska, Rutkowska, Skrzypek and Smigiel2020). 5.4% of the studies, realised by Melander et al. (Reference Melander, Martinsson and Gustafsson2017), Lai Kwan et al. (Reference Lai Kwan, Mahdid, Ochoa, Lee, Park and Blain-Moraes2019), Deutsch et al. (Reference Deutsch, Patnaik and Greco2021) and Iaboni et al. (Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022) studied patients diagnosed with dementia. 32.4% of the studies reported the inclusion of students, one of them including 62 participants with 8 having a mental illness (Schlier et al., Reference Schlier, Krkovic, Clamor and Lincoln2019). 2.7% of the studies included patients with psychiatric problems (de Looff et al., Reference de Looff, Noordzij, Moerbeek, Nijman, Didden and Embregts2019; Goodwin et al., Reference Goodwin, Mazefsky, Ioannidis, Erdogmus and Siegel2019) and another 2.7% included subjects with a substance use disorder (Carreiro et al., Reference Carreiro, Chintha, Shrestha, Chapman, Smelson and Indic2020; Alinia et al., Reference Alinia, Sah, McDonell, Pendry, Parent, Ghasemzadeh and Clevel2021). 8.1% of the studies included subjects under stressful circumstances, such as post-operative pain (Aqajari et al., Reference Aqajari, Cao, Kasaeyan Naeini, Calderon, Zheng, Dutt, Liljeberg, Salanterä, Nelson and Rahmani2021), work-related stress (Phitayakorn et al., Reference Phitayakorn, Minehart, Hemingway, Pian-Smith and Petrusa2015; Georgiou et al., Reference Georgiou, Dimov, Boyanov, Zografos, Larentzakis and Marinov2019; Elsadek et al., Reference Elsadek, Liu and Xie2020; Hosseini et al., Reference Hosseini, Gottumukkala, Katragadda, Bhupatiraju, Ashkar, Borst and Cochran2022) and fear of spiders (Ihmig et al., Reference Ihmig, Neurohr-Parakenings, Schäfer, Lass-Hennemann and Michael2020). Cella et al. (Reference Cella, He, Killikelly, Okruszek, Lewis and Wykes2019) included patients with psychotic episodes for their study, and Kuijpers et al. (Reference Kuijpers, Nijman, Bongers, Lubberding and Ouwerkerk2012) worked with one female schizophrenic patient in her early 20s.

Table 3. Summary of study characteristics of the included studies (n = 74)

Sample sizes of included studies varied between 1 and 1002. The majority of the studies (63.5%) included 30 or less participants. 10.8% of the studies included over 100 participants, 4.1% had one subject (Kuijpers et al., Reference Kuijpers, Nijman, Bongers, Lubberding and Ouwerkerk2012; Vila et al., Reference Vila, Godin, Sakri, Labyt, Vidal, Charbonnier, Ollander and Campagne2019; Hewgill et al., Reference Hewgill, McGinnis and Frolik2020) and 1.4% did not indicate a number of included subjects (Seoane et al., Reference Seoane, Mohino-Herranz, Ferreira, Alvarez, Buendia, Ayllón, Llerena and Gil-Pita2014). When looking at the gender distribution, 24.3% of all included studies included more than 60% males, 18.9% included less than 40% males and 29.7% were more or less gender balanced with 40–60% males. 27.0% of the studies did not report the gender of the participants.

Overall, the age of participants of the included studies varied between 0 and 93 years. 9.5% of all studies included children (0–17 years). Most studies (60.8%) included participants between 18 and 40 years of age on average, with most of them (50% overall) being mainly participants in their 20s. 5.4% of reported mean age ranges including participants over 60 years, and 8.1% of all studies in total included people above 60 in their age range.

Wearables, bio-signals and their reported outcomes

Wearables

Sixteen different wearable devices were identified throughout the included studies, of which the most commonly used was the Empatica wristband in 40.1% of the articles, followed by the Shimmer3 sensor used in 6.8% of the studies (Xu et al., Reference Xu, Nwe and Guan2015; Williams et al., Reference Williams, Hodge and Wu2020; Wulvik et al., Reference Wulvik, Dybvik and Steinert2020; Giorgi et al., Reference Giorgi, Ronca, Vozzi, Sciaraffa, di Florio, Tamborra, Simonetti, Aricò, Di Flumeri, Rossi and Borghini2021) and Affectiva Q senor, which was used in 5.4% of the studies (Volonte et al., Reference Volonte, Babu, Chaturvedi, Newsome, Ebrahimi, Roy, Daily and Fasolino2016; Betancourt et al., Reference Betancourt, Dethorne, Karahalios and Kim2017; Baker et al., Reference Baker, Fenning, Erath, Baucom, Moffitt and Howland2018; Sano et al., Reference Sano, Taylor, McHill, Phillips, Barger, Klerman and Picard2018). Altogether, 67.6% of studies used a wristband sensor, 2.7% (Seoane et al., Reference Seoane, Mohino-Herranz, Ferreira, Alvarez, Buendia, Ayllón, Llerena and Gil-Pita2014; Lee et al., Reference Lee, Lee, Yang, Yoon and Kim2020b) used a glove with integrated sensors, 24.3% articles described devices with electrodes placed on the fingers, often with Velcro, for example, the Shimmer3 sensor, Biopac systems and the E243 inVivo metric systems corp. Marko (Reference Marko2016), Hollis et al. (Reference Hollis, Pekurovsky, Wu and Whittaker2018) and Bruun et al. (Reference Bruun, Law, Nielsen and Heintz2021) choose the Neurobit Optima 4 (finger electrodes), EmVibe (finger electrodes) and Mindplace Thoughstream (palmar electrodes), respectively, all of which were devices that needed to be carried (i.e. not attached to the body). 8.1% of the studies measured EDA on the palmar side of the hand, 2.7% measured on the chest (using the RespiBAN Professional; Jambhale et al., Reference Jambhale, Mahajan, Riel, Banerjee, Dutt, Kadiyala and Vinjamuri2022) and torso (custom build garment; Hewgill et al., Reference Hewgill, McGinnis and Frolik2020). Finally, Betancourt et al. (Reference Betancourt, Dethorne, Karahalios and Kim2017) measured EDA on the ankle using the affective Q sensor.

About half of the field studies used the Empatica E4 wristband (56.3%). The wristbands DTI2 (Melander et al., Reference Melander, Martinsson and Gustafsson2017) and Empatica E4 (Deutsch et al., Reference Deutsch, Patnaik and Greco2021; Iaboni et al., Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022) were used in clinical settings with patients with dementia. Additionally, Lai Kwan et al. (Reference Lai Kwan, Mahdid, Ochoa, Lee, Park and Blain-Moraes2019) attempted to detect ’significant moments’ in patients with dementia in a clinical field setting. De Looff et al. (Reference de Looff, Noordzij, Moerbeek, Nijman, Didden and Embregts2019) studied the aggressive behaviour of inpatient forensic psychiatric patients, and Goodwin et al. (Reference Goodwin, Mazefsky, Ioannidis, Erdogmus and Siegel2019) also analysed aggression in psychiatric ASD inpatients. Walas et al. (Reference Walas, Halaba, Kubiaczyk, Piotrowski, Latka-Grot, Szczapa, Romul, Maroszynska, Malinowska, Rutkowska, Skrzypek and Smigiel2020) and Aqajari et al. (Reference Aqajari, Cao, Kasaeyan Naeini, Calderon, Zheng, Dutt, Liljeberg, Salanterä, Nelson and Rahmani2021) conducted an observational study to pain. In children, bodily measurement locations used were the wrist (Betancourt et al., Reference Betancourt, Dethorne, Karahalios and Kim2017; Baker et al., Reference Baker, Fenning, Erath, Baucom, Moffitt and Howland2018; Goodwin et al., Reference Goodwin, Mazefsky, Ioannidis, Erdogmus and Siegel2019; Redd et al., Reference Redd, Silvera-Tawil, Hopp, berg, Martiniuk, Dietrich and Karunanithi2020; Weyn et al., Reference Weyn, Van Leeuwen, Pluess, Goossens, Claes, Bosmans, Van Den Noortgate, Lutin, Bröhl, Chubar, Geukens and Bijttebier2022), fingers (Shao et al., Reference Shao, Elsadek and Liu2020), palm (Walas et al., Reference Walas, Halaba, Kubiaczyk, Piotrowski, Latka-Grot, Szczapa, Romul, Maroszynska, Malinowska, Rutkowska, Skrzypek and Smigiel2020) and ankle (Betancourt et al., Reference Betancourt, Dethorne, Karahalios and Kim2017).

Bio-signals and processing of EDA

Besides the detection of EDA, the majority of studies (82.4%) simultaneously analysed other bio-signals such as HR, skin temperature (ST), respiration rate (RR), accelerometry, cortisol level, blood volume (BV), inter-beat interval (II) or other.

When analysing EDA, most included studies processed both the slow-changing component, the so-called skin conductance level (SCL), and the fast-changing component, the so-called skin conductance response (SCR). The SCL is related to a general arousal level, whereas SCR is a peak in EDA representing phasic activity, linked to for example novel or unexpected stimuli (Boucsein, Reference Boucsein2012). However, no gold standard exists for the decomposition of the EDA signal. Included studies showed variation in EDA components used for analysis in terms of features extracted from SCR, SCL or EDA. Additionally, the method of decomposition of the EDA signal varies. Finally, algorithmic specifications or software licences used to process EDA and extract features were scarcely mentioned. Most studies (94.6%) processed the EDA signal offline, and 5.4% studies processed their data in real time (Vila et al., Reference Vila, Godin, Sakri, Labyt, Vidal, Charbonnier, Ollander and Campagne2019; Anusha et al., Reference Anusha, Sukumaran, Sarveswaran, Surees Kumar, Shyam, Tony, Preejith and Mohanasankar2020; Dissanayake et al., Reference Dissanayake, Tang, Elvitigala, Wen, Wu and Nanayakkara2022; Hosseini et al., Reference Hosseini, Gottumukkala, Katragadda, Bhupatiraju, Ashkar, Borst and Cochran2022).

Stressors studied

The majority of studies (71.6%) were conducted in the context of general stress detection, such as job stress (e.g. Hosseini et al., Reference Hosseini, Gottumukkala, Katragadda, Bhupatiraju, Ashkar, Borst and Cochran2022), drivers stress (e.g. Oh et al., Reference Oh, Ryu, Jeong, Yang, Hwang, Lee and Lim2021) or the development of a general stress detection algorithm (e.g. Jambhale et al., Reference Jambhale, Mahajan, Riel, Banerjee, Dutt, Kadiyala and Vinjamuri2022). Furthermore, 14.9% of the studies were conducted in the context of healthcare, such as predicting pain intensity in post-operative patients (e.g. Aqajari et al., Reference Aqajari, Cao, Kasaeyan Naeini, Calderon, Zheng, Dutt, Liljeberg, Salanterä, Nelson and Rahmani2021). Finally, 13.5% of all studies were conducted in the context of challenging behaviour such as detecting ‘combat behaviour’ in people with dementia (Deutsch et al., Reference Deutsch, Patnaik and Greco2021) or psychotic behaviour (Schlier et al., Reference Schlier, Krkovic, Clamor and Lincoln2019). Studies were categorised on whether the stressor was induced (i.e. manipulated) or observed. 66.3% of the studies induced their stressor, for example, a surgical simulation (Georgiou et al., Reference Georgiou, Dimov, Boyanov, Zografos, Larentzakis and Marinov2019) or watching film clips inducing emotions (Wang et al., Reference Wang, Yu, Zhao, Guo, Chen and Yu2020). 35.1% of the studies examined EDA in the context of observed (i.e. non-manipulated) stressors, such as agitated behaviours on a psychiatric ward of a nursing home (Iaboni et al., Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022). 56.7% of included studies were performed in a laboratory setting, 39.2% were field studies and 4.1% combined the laboratory setting with a field study (de Arriba Pérez et al., Reference de Arriba Pérez, Santos-Gago, Caeiro-Rodríguez and Iglesias2018; Kyriakou et al., Reference Kyriakou, Resch, Sagl, Petutschnig, Werner, Niederseer, Liedlgruber, Wilhelm, Osborne and Pykett2019; Can et al., Reference Can, Chalabianloo, Ekiz, Fernez-Alvarez, Repetto, Riva, Iles-Smith and Ersoy2020). When considering the year of publication, a rapid increase of field studies is seen, especially over the last 5 years (see Fig. 2).

Fig. 2. Lab and field studies over time relating electrodermal activity measured with a wearable and perceived stress.

Relationship between EDA and perceived stress or stress-related behaviours

In 8.1% of the studies, no relationship between EDA and perceived stress was found. 5.4% of the studies described the relationship between EDA and perceived stress qualitatively. The remaining 86.5% of the included studies showed a (positive) association between EDA and perceived stress. Of those studies, 40.6% reported the accuracy (i.e. percentage of correct classifications regarding stress), which ranges between 42 and 100%, with an average of 82.6%. All studies conducted in the context of challenging behaviour reported a relationship between features of EDA and perceived stress, such as an increase in EDA features 20 min before the onset of aggression in people with dementia (de Looff et al., Reference de Looff, Noordzij, Moerbeek, Nijman, Didden and Embregts2019) and EDA features combined with other bio-signals (such as heart activity, movement and ST) were successful in prediction agitation in people with dementia (Iaboni et al., Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022). In most studies using machine learning (ML), EDA features were combined with other features, making it not possible to draw conclusions about the independent ability of EDA features to predict perceived stress. 32.4% of studies using ML showed that stress can be predicted by using EDA features alone. Reported accuracies of classification models using other features next to EDA ranged from 74.2 to 100% with an average of 87.7%, whereas reported accuracies of classification models using solely EDA features ranged from 42 to 94% with an average of 71.2%. 13.5% of all studies only studied the relationship between EDA and perceived stress indirectly, by showing that features of EDA and perceived stress were increasing simultaneously, but not relating them directly.

Stress assessment, in the form of self-report of perceived stress or the observation of stress-related behaviour (such as challenging behaviour), differed across studies. Additionally, the way of synchronising EDA features with perceived stress or stress-related behaviours differed between studies. 12.2% of the studies synchronised the perceived stress with the measured EDA signals continuously, others measured perceived stress before, during, after and/or in intervals during the performance of different tasks, experiments or the observation. In every study, either a self-reported stress questionnaire (e.g. State Trait Anxiety Inventory (STAI)), a behavioural analysis or a survey was performed. 46.0% of all included studies used ML to predict stress using EDA features next to other bio-signals, the majority (>50%) of them were conducted after 2020 (see Fig. 3). All of the algorithms used implemented EDA features in their classification model. 61.7% of all studies using ML were conducted in research populations aged 18–40 years, and 8.8% in 40–60 year old research populations, 8.8% in 60+ years of age research populations, 2.9% in children, and finally 17.6% of the studies using ML did not report the age of their research population.

Fig. 3. Machine Learning (ML) studies over time relating electrodermal activity measured with a wearable and perceived stress.

Melander et al. (Reference Melander, Martinsson and Gustafsson2017) showed a correlation between SCL and the prediction of agitation (73%), and also Iaboni et al. (Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022) showed that agitation could be predicted with a high accuracy when ML was used, combining EDA features with other physiological variables (measured by the Empatica E4). To assess perceived stress, most studies used a measure of self-report (87.8%), of which the STAI was frequently used (21.5%), as were single items using Likert Scale (16.9%) and Self-Assessment Manikin and Profile of Mood States (6.2%). Of all studies using observations to assess stress, 42.8% studies analysed their findings via video analysis, looking at stress-related behaviour such as agitation (Lee & Chung, Reference Lee and Chung2016; Betancourt et al., Reference Betancourt, Dethorne, Karahalios and Kim2017; Kyriakou et al., Reference Kyriakou, Resch, Sagl, Petutschnig, Werner, Niederseer, Liedlgruber, Wilhelm, Osborne and Pykett2019; Lai Kwan et al., Reference Lai Kwan, Mahdid, Ochoa, Lee, Park and Blain-Moraes2019; Lee et al., Reference Lee, Choi, Jebelli, Ahn and Lee2020a; Iaboni et al., Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022).

Georgiou et al. (Reference Georgiou, Dimov, Boyanov, Zografos, Larentzakis and Marinov2019) and Hosseini et al. (Reference Hosseini, Gottumukkala, Katragadda, Bhupatiraju, Ashkar, Borst and Cochran2022) stated that the wristband could be accepted to provide accurate information about stress. An increase of EDA could be observed during increased anxiety (e.g. Georgiou et al., Reference Georgiou, Dimov, Boyanov, Zografos, Larentzakis and Marinov2019; Tao et al., Reference Tao, Huang, Shen and Sun2022), and ML algorithms are capable of predicting stress in real-life settings (e.g. Hosseini et al., Reference Hosseini, Gottumukkala, Katragadda, Bhupatiraju, Ashkar, Borst and Cochran2022). Melander et al. (Reference Melander, Martinsson and Gustafsson2017) suggested that the Philips DTI-2 wristband sensor can correctly predict 73.5% of agitation, before the caregivers’ behavioural observation. It was also found that data collected by the sensor were good for the visualisation of the perceived signals, supporting the nursing staffs’ observations and helping for a better understanding of the results. As for the studies conducted with children, Baker et al. (Reference Baker, Fenning, Erath, Baucom, Moffitt and Howland2018) stated that low EDA is related to externalising behaviour problems. Kuijpers et al. (Reference Kuijpers, Nijman, Bongers, Lubberding and Ouwerkerk2012) found an increased SCL before the onset of aggressive behaviour in their study. In contrast, Betancourt et al. (Reference Betancourt, Dethorne, Karahalios and Kim2017) faced many challenges to be addressed when measuring EDA in the field. Also in the lab, data were lost due to sensor failure. For example Bruun et al. (Reference Bruun, Law, Nielsen and Heintz2021) had to discard data from 16 out of 61 participants due to sensor failure using the Mindplace ThoughtStream. Finally, it is shown that EDA features can be used not only to predict momentary stress but also daily or even monthly stress (Can et al., Reference Can, Chalabianloo, Ekiz, Fernez-Alvarez, Repetto, Riva, Iles-Smith and Ersoy2020).

Discussion

Stress management plays an important role in health and care of people with stress-related physical (Deary et al., Reference Deary, Chalder and Sharpe2007) and/or behavioural problems (Matson et al., Reference Matson, Matson and Rivet2007; Melander et al., Reference Melander, Martinsson and Gustafsson2017). As insight into stress levels may facilitate stress management, the aim of this scoping review was to collect and summarise knowledge about currently existing wearable technologies measuring EDA designed to detect perceived stress. Seventy-four articles, published between 2012 and 2022, were identified in this study. The majority of the studies were performed in a laboratory setting with an induced, that is, manipulated, stressor, in a research population that was mostly young and healthy. In most studies, sample sizes were small (<30). The technologies used were predominantly wristbands, which, next to EDA, measured other physiological parameters. Most studies processed EDA offline (i.e. post-processing) and included slow changing (SCL) and rapid changing (SCR) features. The exact (i.e. mathematical) way of extracting EDA features varied across studies. Finally, the exact method of assessing perceived stress (i.e. questionnaires and behavioural observation schemes used) varied considerably across studies.

In included studies, the exact software used to process and extract EDA features was barely mentioned, and only a minority of studies were conducted outside of the lab, although field studies are increasing over the years. Using wearables in daily life is relatively new and influences the way of measuring and processing data. In natural environments, data processing should account for different types of measurement artefacts compared to clean and standardised laboratory settings. Therefore, lab data and outcomes cannot be translated directly to daily life situations. To ensure ecological validity, it is of utmost importance to validate the stress measurements of wearables in natural environments (Smets et al., Reference Smets, De Raedt and Van Hoof2019).

Moreover, as the ultimate goal is to integrate wearable technologies in a healthcare context, it is important to include a diversity of target populations in research. This means that people with neurocognitive (e.g. dementia) and developmental disorders (e.g. autism) as well as people with stress-related physical complaints (e.g. medically unexplained physical symptoms; MUPS) should be included in research populations. In this scoping review, only 20% of the studies included (sub)clinical populations and only 28% was conducted in the context of health or challenging behaviour. Around half of the studies included ML to predict stress. However, only 8% included older persons and a similar number of studies explored younger persons with disabilities. Critically, older persons (Mannheim et al., Reference Mannheim, Schwartz, Xi, Buttigieg, McDonnell-Naughton, Wouters and van Zaalen2019) and people with disabilities (Shew, Reference Shew2020) are too often excluded from research and design of digital technology and are significantly underrepresented in datasets used to train ML algorithms (Park et al., Reference Park, Bernstein, Brewer, Kamar and Morris2021). A main consequence is that such newly developed technologies may not fit their needs (Mannheim et al., Reference Mannheim, Wouters, Köttl, Van Boekel, Brankaert and Van Zaalen2022), or even that ML algorithms may be biased in their assessment of populations that are not young and healthy (Mehrabi et al., Reference Mehrabi, Morstatter, Saxena, Lerman and Galstyan2019; Joyce et al., Reference Joyce, Smith-Doerr, Alegria, Bell, Cruz, Hoffman, Noble and Shestakofsky2021).

An adequate detection of EDA would lead to a better stress identification, offering a variety of possibilities for its implementation in the healthcare department. In the study of Dias & Cunha (Reference Dias and Cunha2018), wearable health devices were reviewed on their current state and likely progression in the following years. They concluded that these technologies can support healthcare but are currently still limited in many ways: existing wearables are not available everywhere, are too expensive and are not yet suitable for ambulatory usage. Nevertheless, the integration of these technologies has the potential to lower healthcare costs. It could, amongst other, facilitate a non-pharmaceutical approach in stress management, support caregivers in monitoring patients, facilitate an optimised therapy provision and overall improve the patient’s quality of life (Cahill et al., Reference Cahill, Macijauskiene, Nygard, Faulkner and Hagen2007). Furthermore, caregivers are acknowledged to ‘play an important role in monitoring patient acceptance of the device’ (Mahoney & Mahoney, Reference Mahoney and Mahoney2010). If wearables are to fulfil this potential, we need more research including the perspectives of patients and their caregivers.

A limitation found in this scoping review is that the assessment of perceived stress was performed in many different manners: the exact proceeding method, such as a questionnaire and/or the behavioural observation, the exact time point, when perceived stress was actually measured, and synchronisation with the continuous EDA measurements highly varied between studies. In addition, there is no gold standard for measuring stress, as various perceived stress scales were used. This hampers the comparability of the studies. As reported, stress was mostly assessed, before, after or in intervals during the performed study, an exact (i.e. real time) synchronisation and immediate comparison with the measured EDA signals was rare. Furthermore, not all studies directly compared perceived stress and EDA. Mostly, an offline analysis was done, while stress detection in daily life should be (pseudo)real-time to be relevant. The latter is influencing the possibilities of data processing, that is, offline data processing will be different from real-time data processing. This means that, in order to get an exact insight into the perceived stress of an individual in relation to their EDA levels during a so called ‘stressful’ task or situation, further investigation is needed considering real-time processing of measurements, also regarding the ML pipeline (Vos et al., Reference Vos, Trinh, Sarnyai and Azghadi2022).

Interpreting skin conductance and thus detecting stress in real time on a personal level would have a high impact on healthcare. The development within the field of artificial intelligence (AI) can contribute greatly to this fact (see Vos et al. (Reference Vos, Trinh, Sarnyai and Azghadi2022) for a review). The use of ML has increased in the last years to detect stress using bio-signals, including EDA features, appreciating interpersonal variation. Studies like Anusha et al. (Reference Anusha, Sukumaran, Sarveswaran, Surees Kumar, Shyam, Tony, Preejith and Mohanasankar2020) and Iaboni et al. (Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022) show that ML algorithms are very valuable when providing feedback regarding stress levels in a real-life setting, and stress could be detected before the caregivers observe it. Importantly, the ‘human in the loop’ remains crucial for annotating data (Melander et al., Reference Melander, Kikhia, Olsson, Wälivaara and Sävenstedt2018; Iaboni et al., Reference Iaboni, Spasojevic, Newman, Schindel Martin, Wang, Mihailidis and Khan2022) and interpretation of the measurement (Kikhia et al., Reference Kikhia, Stavropoulos, Andreadis, Karvonen, Kompatsiaris, Sävenstedt, Pijl and Melander2016). During the development of AI applications, it is crucial that common biases such as sampling bias, aggregation bias or longitudinal data fallacy are mitigated (see Mehrabi et al. (Reference Mehrabi, Morstatter, Saxena, Lerman and Galstyan2019) for an overview). Furthermore, the acceptance by healthcare professionals is of great importance for their use in practice. To ensure acceptance of AI-enabled solutions, ‘explainable’ AI (XAI) is paramount. Whereas traditional AI deals with the prediction or decision support by using ‘smart’ algorithms, XAI on top of that addresses the question ‘why is the AI predicting a certain outcome?’ XAI includes the application of unbiased, fair and transparent algorithms (Adadi & Berrada, Reference Adadi and Berrada2018). Although future research is shifting towards this aim, this scoping review identified no articles focusing on XAI in the context of predicting perceived stress.

Another important factor in the acceptance and ultimate usage of wearables is the placement on the body. Studies included in this review used wearables mainly on the wrist, the hand or different fingers. Several studies have been performed to determine on which part of the body EDA can be detected in the best possible and most effective manner. A study from 2013 compared measuring EDA on the foot and fingers (Payne et al., Reference Payne, Dawson, Schell, Singh and Courtney2013) and van Dooren et al. (Reference van Dooren, de Vries and Janssen2012) analysed 16 different bodily locations to measure EDA while provoking an emotional reaction. Both studies concluded that the fingers showed the best results when it comes to measuring EDA. However, finger sensors are used most often under laboratory conditions and not in daily life settings, limiting them in their application. Importantly, while jewellery-like rings have been introduced in the recent years (e.g. the Moodmetric smart ring (Vigofere Ltd, Finland), as described in Pakarinen et al., Reference Pakarinen, Pietilä and Nieminen2019), no studies using rings were found during the search of this scoping review, suggesting that despite their availability they are scarcely studied in relation to perceived stress. In addition, the level of evidence is only moderate for the fingers or the wrist to be the best placement option, suggesting focusing eventually more on other body parts such as the forehead, feet, shoulders, neck and chest. These bodily locations have shown good EDA responsiveness, with shoulders being the best solution (van Dooren et al., Reference van Dooren, de Vries and Janssen2012). Subsequently, wearable technologies can be developed for these bodily locations, making them more suitable for ambulatory usage.

In addition to the placement on the body, also considering the exact context and especially the target group that is measured is important when selecting a sensor. For example, a wristband sensor in the form of a simple black watch might be a good solution when measuring in patients outside the lab. However, when working with people experiencing mental disorders such as dementia, there are some obstacles: the look of the watch-like wristband might confuse the patients and they would be able to either switch or take it off on their own. Furthermore, in case of a wrong placement of the device, signal processing might be disturbed and distorted. In the current review, the Empatica E4 wristband is the most often used tool in the included studies. Peeters et al. (Reference Peeters, Schouten and Wouters2021) have shown that the use of this exact wearable is limited in real-life situations such as with people with dementia. This applies not only to the suitability of the wearable but also for the results regarding stress-related parameters, as well as the implementation of the wristband in daily life settings. Other technologies, such as ring sensors, could present similar issues in terms of causing confusion and interfering with daily life actions, making them not always suitable for use in practice. Less obtrusive tools, such as socks or patches, should therefore be further developed and analysed in future studies. However, the purpose of most of the studies included in this review was not to analyse the usability of the device, which made it difficult to assess their usability in real-life settings.

Conclusion

Using wearables to measure features of EDA to predict perceived stress is promising as an assistive technique to support stress management. Nevertheless, research with people with health or care conditions and real-life settings is still limited. Additionally, studies conducting real-time stress prediction are scarce. Advances in ML are offering opportunities to account for interpersonal differences and apply real-time stress detection algorithms, using EDA features measured by wearables, under daily life conditions. However, to avoid biased applications, future research should include daily life situations, samples from diverse populations, as well as usability testing.

Supplementary material

To view supplementary material for this article, please visit https://doi.org/10.1017/neu.2023.19

References

Adadi, A and Berrada, M (2018) Peeking inside the black-box: a survey on explainable artificial intelligence (XAI). IEEE Access 6, 5213852160.CrossRefGoogle Scholar
Alinia, P, Sah, RK, McDonell, M, Pendry, P, Parent, S, Ghasemzadeh, H and Clevel, MJ (2021) Associations between physiological signals captured using wearable sensors and self-reported outcomes among adults in alcohol use disorder recovery: development and usability study. JMIR Formative Research 5(7), e27891.CrossRefGoogle ScholarPubMed
Anusha, AS, Sukumaran, P, Sarveswaran, V, Surees Kumar, S, Shyam, A, Tony, JA, Preejith, SP and Mohanasankar, S (2020) Electrodermal activity based pre-surgery stress detection using a wrist wearable. IEEE Journal of Biomedical and Health Informatics 24, 92100.Google Scholar
Aqajari, SAH, Cao, R, Kasaeyan Naeini, E, Calderon, MD, Zheng, K, Dutt, N, Liljeberg, P, Salanterä, S, Nelson, AM, Rahmani, AM (2021) Pain assessment tool with electrodermal activity for postoperative patients: method validation study. Journal of Medical Internet Research Mhealth Uhealth 9(5), e25258.CrossRefGoogle ScholarPubMed
Arksey, H and O’Malley, L (2005) Scoping studies: towards a methodological framework. International Journal of Social Research Methodology 8(1), 1932.CrossRefGoogle Scholar
Baker, JK, Fenning, RM, Erath, SA, Baucom, BR, Moffitt, J, Howland, MA (2018) Sympathetic under-arousal and externalizing behavior problems in children with autism spectrum disorder. Journal of Abnormal Child Psychology 46(4), 895906.CrossRefGoogle ScholarPubMed
Ballard, C, O’Brien, J, Gray, A, Cormack, F, Ayre, G, Rowan, E, Thompson, P, Bucks, R, McKeith, I, Walker, M, Tovee, M (2001) Attention and fluctuating attention in patients with dementia with lewy bodies and Alzheimer Disease. Archives of Neurology 58(6), 977.CrossRefGoogle ScholarPubMed
Berg-Weger, M and Stewart, DB (2017) Non-pharmacologic interventions for persons with dementia. Missouri Medicine 114(2), 116119.Google ScholarPubMed
Betancourt, MA, Dethorne, LS, Karahalios, K and Kim, JG (2017) Skin conductance as an in situ marker for emotional arousal in children with neurodevelopmental communication impairments: methodological considerations and clinical implications. ACM Transactions on Computer Systems 9, Article8.Google Scholar
Boucsein, W (2012) Electrodermal activity, 2nd edn. New York: Springer US.CrossRefGoogle Scholar
Bruun, A, Law, ELC, Nielsen, TD and Heintz, M (2021) Do you feel the same? On the robustness of cued-recall debriefing for user experience evaluation. ACM Transactions on Computer-Human Interaction 28(4), 145.CrossRefGoogle Scholar
Cahill, S, Macijauskiene, J, Nygard, AM, Faulkner, JP and Hagen, I (2007) Technology in dementia care. Technology & Disability 19(2-3), 5560.CrossRefGoogle Scholar
Can, YS, Chalabianloo, N, Ekiz, D, Fernez-Alvarez, J, Repetto, C, Riva, G, Iles-Smith, H and Ersoy, C (2020) Real-life stress level monitoring using smart bands in the light of contextual information. IEEE Sensors Journal 20, 87218730.CrossRefGoogle Scholar
Carreiro, S, Chintha, KK, Shrestha, S, Chapman, B, Smelson, D and Indic, P (2020) Wearable sensor-based detection of stress and craving in patients during treatment for substance use disorder: a mixed methods pilot study. Drug and Alcohol Dependence 209, 107929.CrossRefGoogle ScholarPubMed
Cella, M, He, ZM, Killikelly, C, Okruszek, L, Lewis, S and Wykes, T (2019) Blending active and passive digital technology methods to improve symptom monitoring in early psychosis. Early Intervention in Psychiatry 13(5), 12711275.CrossRefGoogle ScholarPubMed
Cerejeira, J, Lagarto, L and Mukaetova-Ladinska, EB (2012) Behavioral and psychological symptoms of dementia. Frontiers in Neurology 3, 73.CrossRefGoogle ScholarPubMed
de Arriba Pérez, F, Santos-Gago, JM, Caeiro-Rodríguez, M and Iglesias, MJF (2018) Evaluation of commercial-off-the-shelf wrist wearables to estimate stress on students. Journal of Visualized Experiments 57590.Google Scholar
de Looff, P, Noordzij, ML, Moerbeek, M, Nijman, H, Didden, R and Embregts, P (2019) Changes in heart rate and skin conductance in the 30 min preceding aggressive behavior. Psychophysiology 56(10), e13420.CrossRefGoogle ScholarPubMed
Deary, V, Chalder, T and Sharpe, M (2007) The cognitive behavioural model of medically unexplained symptoms: a theoretical and empirical review. Clinical Psychology Review 27(7), 781797.CrossRefGoogle ScholarPubMed
Deutsch, CK, Patnaik, PP and Greco, FA (2021) Is there a characteristic autonomic response during outbursts of combative behavior in dementia patients? Journal of Alzheimer’s Disease Reports 5(1), 389394.CrossRefGoogle Scholar
Devshi, R, Shaw, S, Elliott-King, J, Hogervorst, E, Hiremath, A, Velayudhan, L, Kumar, S, Baillon, S and Bandelow, S (2015) Prevalence of behavioural and psychological symptoms of dementia in individuals with learning disabilities. Diagnostics 5(4), 564576.CrossRefGoogle ScholarPubMed
Dhabhar, FS (2018) No titThe short-term stress response - mother nature’s mechanism for enhancing protection and performance under conditions of threat, challenge, and opportunity. Frontiers in Neuroendocrinology 49, 175192.CrossRefGoogle ScholarPubMed
Dias, D and Cunha, JPS (2018) Wearable health devices—vital sign monitoring, systems and technologies. Sensors 18(8), 2414.CrossRefGoogle ScholarPubMed
Dissanayake, V, Tang, V, Elvitigala, DS, Wen, E, Wu, M and Nanayakkara, S (2022) Troi: towards understanding users perspectives to mobile automatic emotion recognition system in their natural setting. Proceedings of the ACM on Human Computer Interaction, pages 6.CrossRefGoogle Scholar
Du, J, Huang, J, An, Y and Xu, W (2018) The relationship between stress and negative emotion: the mediating role of rumination. Clinical Research and Trials 4.CrossRefGoogle Scholar
Elsadek, M, Liu, BY and Xie, JF (2020) Window view and relaxation: viewing green space from a high-rise estate improves urban dwellers’ wellbeing. Urban Forestry & Urban Greening 55, 126846.CrossRefGoogle Scholar
Fink, G (2017) Stress: concepts, definition and history , The curated reference collection in neuroscience and biobehavioral psychology. Elsevier Science Ltd.Google Scholar
Foebel, A, Ballokova, A, Wellens, NIH, Fialova, D, Milisen, K, Liperoti, R and Hirdes, JP (2015) A retrospective, longitudinal study of factors associated with new antipsychotic medication use among recently admitted long-term care residents. BMC Geriatrics 15(1), 128.CrossRefGoogle ScholarPubMed
Georgiou, KE, Dimov, RK, Boyanov, NB, Zografos, KG, Larentzakis, AV and Marinov, BI (2019) Feasibility of a new wearable device to estimate acute stress in novices during high-fidelity surgical simulation. Folia Medica 61(1), 4960.CrossRefGoogle ScholarPubMed
Giorgi, A, Ronca, V, Vozzi, A, Sciaraffa, N, di Florio, A, Tamborra, L, Simonetti, I, Aricò, P, Di Flumeri, G, Rossi, D, Borghini, G (2021) Wearable technologies for mental workload, stress, and emotional state assessment during working-like tasks: a comparison with laboratory technologies. Sensors 21(7), 2332.CrossRefGoogle ScholarPubMed
Goodwin, MS, Mazefsky, CA, Ioannidis, S, Erdogmus, D and Siegel, M (2019) Predicting aggression to others in youth with autism using a wearable biosensor. Autism Research 12(8), 12861296.CrossRefGoogle ScholarPubMed
Hanson, MA, Powell, HC, Barth, AT, Ringgenberg, K, Calhoun, BH, Aylor, JH and Lach, J (2009) Body area sensor networks: challenges and opportunities. Computer 42(1), 5865.CrossRefGoogle Scholar
Healey, JA and Picard, RW (2005) Detecting stress during real-world driving tasks using physiological sensors. IEEE Transactions on Intelligent Transportation Systems 6(2), 156166.CrossRefGoogle Scholar
Hewgill, B, McGinnis, RS and Frolik, J (2020) A modular open source health monitoring garment. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference, pages 45104513, 2020CrossRefGoogle Scholar
Hollis, V, Pekurovsky, A, Wu, E and Whittaker, S (2018) On being told how we feel: how algorithmic sensor feedback influences emotion perception. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2(3), Article114.CrossRefGoogle Scholar
Hosseini, S, Gottumukkala, R, Katragadda, S, Bhupatiraju, RT, Ashkar, Z, Borst, CW and Cochran, K (2022) A multimodal sensor dataset for continuous stress detection of nurses in a hospital. Scientific Data 9(1), 255.CrossRefGoogle ScholarPubMed
Iaboni, A, Spasojevic, S, Newman, K, Schindel Martin, L, Wang, A, Mihailidis, Ye B, Khan, A and SS (2022) Wearable multimodal sensors for the detection of behavioral and psychological symptoms of dementia using personalized machine learning models. Alzheimer’s & Dementia 14(1), e12305.Google ScholarPubMed
Ihmig, FR, Neurohr-Parakenings, HAG, Schäfer, F, Lass-Hennemann, SK, Michael, J and T (2020) On-line anxiety level detection from biosignals: machine learning based on a randomized controlled trial with spider-fearful individuals. PloS One 15(6), e0231517.CrossRefGoogle ScholarPubMed
Jambhale, K, Mahajan, S, Riel, B, Banerjee, N, Dutt, A, Kadiyala, SP and Vinjamuri, R (2022) Identifying biomarkers for accurate detection of stress. Sensors 22(22), 8703.CrossRefGoogle ScholarPubMed
Joyce, K, Smith-Doerr, L, Alegria, S, Bell, S, Cruz, T, Hoffman, SG, Noble, SU and Shestakofsky, B (2021) Toward a sociology of artificial intelligence: a call for research on inequalities and structural change. Socius 7, 2378023121999581.CrossRefGoogle Scholar
Kikhia, B, Stavropoulos, TG, Andreadis, S, Karvonen, N, Kompatsiaris, I, Sävenstedt, S, Pijl, M and Melander, C (2016) Utilizing a wristband sensor to measure the stress level for people with dementia. Sensors 16(12), 1989.CrossRefGoogle ScholarPubMed
Kong, EH (2005) Agitation in dementia: concept clarification. Journal of Advanced Nursing 52(5), 526536.CrossRefGoogle ScholarPubMed
Kuijpers, E, Nijman, H, Bongers, IMB, Lubberding, M and Ouwerkerk, M (2012) Can mobile skin conductance assessments be helpful in signalling imminent inpatient aggression? Acta Neuropsychiatrica 24(1), 5659.CrossRefGoogle ScholarPubMed
Kurniawan, H, Maslov, AV and Pechenizkiy, M (2013) Stress detection from speech and Galvanic Skin Response signals. Proceedings of the 26th IEEE International Symposium on Computer-Based Medical Systems. IEEE, pages 209214.CrossRefGoogle Scholar
Kyriakou, K, Resch, B, Sagl, G, Petutschnig, A, Werner, C, Niederseer, D, Liedlgruber, M, Wilhelm, F, Osborne, T, Pykett, J (2019) Detecting moments of stress from measurements of wearable physiological sensors. Sensors 19(17), 3805.CrossRefGoogle ScholarPubMed
Lai Kwan, C, Mahdid, Y, Ochoa, RM, Lee, K, Park, M and Blain-Moraes, S (2019) Wearable technology for detecting significant moments in individuals with dementia. BioMed Research International.CrossRefGoogle ScholarPubMed
Lee, BG and Chung, WY (2016) Wearable glove-type driver stress detection using a motion sensor. IEEE Transactions on Intelligent Transportation Systems 18(7), 18351844.CrossRefGoogle Scholar
Lee, G, Choi, B, Jebelli, H, Ahn, CR and Lee, S (2020a) Noise reference signal-based denoising method for eda collected by multimodal biosensor wearable in the field. Journal of Computing in Civil Engineering 34(6).CrossRefGoogle Scholar
Lee, S, Lee, T, Yang, T, Yoon, C and Kim, SP (2020b) Detection of drivers’ anxiety invoked by driving situations using multimodal biosignals. Processes 8(2), 155.CrossRefGoogle Scholar
Levac, D, Colquhoun, H and O’Brien, KK (2010) Scoping studies: advancing the methodology. Implementation Science 5(1), 69.CrossRefGoogle ScholarPubMed
Mahoney, EL and Mahoney, DF (2010) Acceptance of wearable technology by people with Alzheimer’s disease: issues and accommodations. American Journal of Alzheimer’s Disease & Other Dementias 25(6), 527531.CrossRefGoogle ScholarPubMed
Mannheim, I, Schwartz, E, Xi, W, Buttigieg, SC, McDonnell-Naughton, M, Wouters, EJM and van Zaalen, Y (2019) Inclusion of older adults in the research and design of digital technology. International Journal of Environmental Research and Public Health 16(19), 3718.CrossRefGoogle ScholarPubMed
Mannheim, I, Wouters, EJ, Köttl, H, Van Boekel, LC, Brankaert, R and Van Zaalen, Y (2022) Ageism in the discourse and practice of designing digital technology for older persons: a scoping review. The Gerontologist gnac144.CrossRefGoogle Scholar
Marko, M (2016) Development of a protocol for simultaneous assessment of cognitive functioning under psychosocial stress. Journal of Psychology 150(7), 916929.CrossRefGoogle ScholarPubMed
Matson, JL, Matson, ML and Rivet, TT (2007) Social-skills treatments for children with autism spectrum disorders. Behavior Modification 31(5), 682707.CrossRefGoogle ScholarPubMed
Mehrabi, N, Morstatter, F, Saxena, N, Lerman, K and Galstyan, A (2019) A survey on bias and fairness in machine learning. ACM Computing Surveys 54.Google Scholar
Melander, C, Martinsson, J and Gustafsson, S (2017) Measuring electrodermal activity to improve the identification of agitation in individuals with dementia. Dementia and Geriatric Cognitive Disorders Extra 7(3), 430439.CrossRefGoogle ScholarPubMed
Melander, CA, Kikhia, B, Olsson, M, Wälivaara, BM and Sävenstedt, S (2018) The impact of using measurements of electrodermal activity in the assessment of problematic behaviour in dementia. Dementia and Geriatric Cognitive Disorders Extra 8(3), 333347.CrossRefGoogle ScholarPubMed
Minguillon, J, Perez, E, Lopez-Gordo, M, Pelayo, F and Sanchez-Carrion, M (2018) Portable system for real-time detection of stress level. Sensors 18(8), 2504.CrossRefGoogle ScholarPubMed
Munn, Z, Peters, MDJ, Stern, C, Tufanaru, C, McArthur, A and Aromataris, E (2018) Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Medical Research Methodology 18(1), 143.CrossRefGoogle ScholarPubMed
Oh, G, Ryu, J, Jeong, E, Yang, JH, Hwang, S, Lee, S and Lim, S (2021) DRER: deep learning-based driver’s real emotion recognizer. Sensors 21.CrossRefGoogle ScholarPubMed
Ornstein, K and Gaugler, JE (2012) The problem with “problem behaviors”: a systematic review of the association between individual patient behavioral and psychological symptoms and caregiver depression and burden within the dementia patient-caregiver dyad. International Psychogeriatrics 24(10), 15361552.CrossRefGoogle ScholarPubMed
Pakarinen, T, Pietilä, J and Nieminen, H (2019) Prediction of self-perceived stress and arousal based on electrodermal activity. 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 21912195.CrossRefGoogle Scholar
Park, JS, Bernstein, MS, Brewer, RN, Kamar, E and Morris, MR (2021) Understanding the representation and representativeness of age in AI data sets. AIES. 2021 - Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society, pages 834842.CrossRefGoogle Scholar
Payne, AFH, Dawson, ME, Schell, AM, Singh, K and Courtney, CG (2013) Can you give me a hand? A comparison of hands and feet as optimal anatomical sites for skin conductance recording. Psychophysiology 50(11), 10651069.CrossRefGoogle Scholar
Peeters, MWH, Schouten, G and Wouters, EJ (2021) Wearables for residents of nursing homes with dementia and challenging behaviour: values, attitudes, and needs. Gerontechnology 20, 113.CrossRefGoogle Scholar
Phitayakorn, R, Minehart, RD, Hemingway, MW, Pian-Smith, MCM and Petrusa, E (2015) Relationship between physiologic and psychological measures of autonomic activation in operating room teams during a simulated airway emergency. American Journal of Surgery 209(1), 8692.CrossRefGoogle ScholarPubMed
Posada-Quintero, HF and Chon, KH (2020) Innovations in electrodermal activity data collection and signal processing: a systematic review. Sensors 20(2), 479.CrossRefGoogle ScholarPubMed
Posner, J, Russell, JA and Peterson, BS (2005) The circumplex model of affect: an integrative approach to affective neuroscience, cognitive development, and psychopathology. Development and Psychopathology 17(03), 715734.CrossRefGoogle ScholarPubMed
Rattaz, C, Michelon, C, Roeyers, H and Baghdadli, A (2017) Quality of life in parents of young adults with ASD: EpiTED cohort. Journal of Autism and Developmental Disorders 47(9), 28262837.CrossRefGoogle ScholarPubMed
Redd, CB, Silvera-Tawil, D, Hopp, D, berg, D, Martiniuk, A, Dietrich, C and Karunanithi, MK (2020) Physiological signal monitoring for identification of emotional dysregulation in children. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference, pages 42734277 CrossRefGoogle Scholar
Rienzo, M, Di Rizzo, F, Parati, G, Brambilla, G, Ferratini, M and Castiglioni, P (2005) MagIC system: a new textile-based wearable device for biological signal monitoring. Applicability in daily life and clinical setting. 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference. IEEE, pages 71677169.Google Scholar
Sano, A, Taylor, S, McHill, AW, Phillips, AJK, Barger, LK, Klerman, E and Picard, R (2018) Identifying objective physiological markers and modifiable behaviors for self-reported stress and mental health status using wearable sensors and mobile phones: observational study. Journal of Medical INTERNET Research 20(6), e210.CrossRefGoogle ScholarPubMed
Schlier, B, Krkovic, K, Clamor, A and Lincoln, TM (2019) Autonomic arousal during psychosis spectrum experiences: results from a high resolution ambulatory assessment study over the course of symptom on- and offset. Schizophrenia Research 212, 163170.CrossRefGoogle ScholarPubMed
Selye, H (1956) The stress of life. McGraw-Hill, New York: McGraw-Hill Book Company. McGraw-Hill.Google Scholar
Seoane, F, Mohino-Herranz, I, Ferreira, J, Alvarez, L, Buendia, R, Ayllón, D, Llerena, C and Gil-Pita, R (2014) Wearable biomedical measurement systems for assessment of mental stress of combatants in real time. Sensors 14(4), 71207141.CrossRefGoogle ScholarPubMed
Shao, YH, Elsadek, M and Liu, BY (2020) Horticultural activity: its contribution to stress recovery and wellbeing for children. International Journal of Environmental Research and Public Health 17(4), 1229.CrossRefGoogle ScholarPubMed
Shew, A (2020) Ableism, technoableism, and future AI. IEEE Technology and Society Magazine 39, 4085.CrossRefGoogle Scholar
Sierra, A, Snchez-Avila, C, Guerra-Casanova, J and Bailador-Del, G (2011) Real-time stress detection by means of physiological signals. Recent Application in Biometrics 58, 48574865.Google Scholar
Smets, E, De Raedt, W and Van Hoof, C (2019) Into the wild: the challenges of physiological stress detection in laboratory and ambulatory settings. IEEE Journal of Biomedical and Health Informatics 23, 463473.CrossRefGoogle ScholarPubMed
Steptoe, A and Kivimäki, M (2012) Stress and cardiovascular disease. Nature Reviews Cardiology 9(6), 360370.CrossRefGoogle ScholarPubMed
Tao, K, Huang, Y, Shen, Y and Sun, L (2022) Automated stress recognition using supervised learning classifiers by interactive virtual reality scenes. IEEE Transactions on Neural Systems and Rehabilitation Engineering: a Publication of the IEEE Engineering in Medicine and Biology Society 30, 20602066.CrossRefGoogle ScholarPubMed
Tricco, AC, Lillie, E, Zarin, W, O’Brien, KK, Colquhoun, H, Levac, D, Moher, D, Peters, MDJ, Horsley, T, Weeks, L, Hempel, S, Akl, EA, Chang, C, McGowan, J, Stewart, L, Hartling, L, Aldcroft, A, Wilson, MG and Garritty, C, et al. (2018) PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Annals of Internal Medicine 169, 467473.CrossRefGoogle ScholarPubMed
van Dooren, M, de Vries, JJG and Janssen, JH (2012) Emotional sweating across the body: Comparing 16 different skin conductance measurement locations. Physiology & Behavior 106(2), 298304.CrossRefGoogle ScholarPubMed
Van Houdenhove, BJ (2007) Stress, het lijf, en het brein. Lannoo.Google Scholar
van Lier, HG, Pieterse, ME, Garde, A, Postel, MG, de Haan, HA, Vollenbroek-Hutten, MMR, Schraagen, JM and Noordzij, ML (2020) A standardized validity assessment protocol for physiological signals from wearable technology: methodological underpinnings and an application to the E4 biosensor. Behavior Research Methods 52(2), 607629.CrossRefGoogle Scholar
Vila, G, Godin, C, Sakri, O, Labyt, E, Vidal, A, Charbonnier, S, Ollander, S and Campagne, A (2019) Real-time monitoring of passenger’s psychological stress. Future Internet 11(5), 102.CrossRefGoogle Scholar
Volonte, M, Babu, SV, Chaturvedi, H, Newsome, N, Ebrahimi, E, Roy, T, Daily, SB and Fasolino, T (2016) Effects of virtual human appearance fidelity on emotion contagion in affective inter-personal simulations. IEEE Transactions on Visualization and Computer Graphics 22, 13261335.CrossRefGoogle Scholar
Vos, G, Trinh, K, Sarnyai, Z and Azghadi, MR (2022) Machine learning for stress monitoring from wearable devices: a systematic literature review, arXiv preprint arXiv: 2209.15137.Google Scholar
Walas, W, Halaba, Z, Kubiaczyk, A, Piotrowski, A, Latka-Grot, J, Szczapa, T, Romul, M, Maroszynska, I, Malinowska, E, Rutkowska, M, Skrzypek, M, Smigiel, R (2020) Skin conductance measurement for the assessment of analgosedation adequacy in infants treated with mechanical ventilation: a multicenter pilot study. Advances in Clinical and Experimental Medicine 29(9), 11171121.CrossRefGoogle ScholarPubMed
Wang, Z, Yu, Z, Zhao, B, Guo, B, Chen, C and Yu, Z (2020) EmotionSense: an adaptive emotion recognition system based on wearable smart devices. ACM Transactions on Computing for Healthcare 1(4), 117.Google Scholar
Weyn, S, Van Leeuwen, K, Pluess, M, Goossens, L, Claes, S, Bosmans, G, Van Den Noortgate, W, Lutin, E, Bröhl, AS, Chubar, V, Geukens, F, Bijttebier, P (2022) Individual differences in environmental sensitivity at physiological and phenotypic level: two sides of the same coin? International Journal of Psychophysiology: Official Journal of the International Organization of Psychophysiology 176, 3653.CrossRefGoogle ScholarPubMed
Wijsman, J, Grundlehner, B, Liu, H, Hermens, H and Penders, J (2011) Towards mental stress detection using wearable physiological sensors. 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE, pages 17981801.CrossRefGoogle Scholar
Williams, D, Hodge, VJ and Wu, CY (2020) On the use of AI for generation of functional music to improve mental health. Frontiers in Artificial Intelligence and Applications 3, 497864.CrossRefGoogle ScholarPubMed
Wulvik, AS, Dybvik, H and Steinert, M (2020) Investigating the relationship between mental state (workload and affect) and physiology in a control room setting (ship bridge simulator). Cognition Technology & Work 22(1), 95108.CrossRefGoogle Scholar
Xu, Q, Nwe, TL and Guan, C (2015) Cluster-based analysis for personalized stress evaluation using physiological signals. IEEE Journal of Biomedical and Health Informatics 19(1), 275281.CrossRefGoogle ScholarPubMed
Zangróniz, R, Martínez-Rodrigo, A, Pastor, J, López, M and Fernández-Caballero, A (2017) Electrodermal activity sensor for classification of calm/distress condition. Sensors 17(10), 2324.CrossRefGoogle ScholarPubMed
Figure 0

Table 1. Set of keywords and related synonyms

Figure 1

Table 2. In- and exclusion criteria

Figure 2

Fig. 1. Flow chart diagram based upon Tricco et al. (2018).

Figure 3

Table 3. Summary of study characteristics of the included studies (n = 74)

Figure 4

Fig. 2. Lab and field studies over time relating electrodermal activity measured with a wearable and perceived stress.

Figure 5

Fig. 3. Machine Learning (ML) studies over time relating electrodermal activity measured with a wearable and perceived stress.

Supplementary material: File

Klimek et al. supplementary material

Klimek et al. supplementary material

Download Klimek et al. supplementary material(File)
File 43.5 KB