Workplace stress was reported nearly 200 years ago and has been described by Jackson [1] as “psychological pressures generated by the unfettered growth of industrial and technological capitalism.” When these pressures overwhelm, they cause distress (hereafter just referred to as stress) and stress continues as a pervasive phenomenon today. More than twice as many people report stress, depression, or anxiety causing them to miss work compared with any other reason, with teachers demonstrating nearly double the incidence rates of the general population [2]. Stress is a determinant of well-being and can be defined as “the adverse reaction people have to excessive pressures or other types of demand placed on them” [3]. Persistent, excessive stress is a risk factor for illness, including affective disorders such as anxiety and depression [4] and cardiovascular disease [5]; the latter is associated with burnout in even young female teachers [6]. Therefore, prevention or early intervention to reduce the effects of stress remains of huge importance.

Stress management interventions have proliferated, motivated by challenges including therapy costs, accessibility, and unmet needs of employees [7-11] with over 1000 stress apps found in web-based stores [12]. Increasingly, research shows digital tools can also be harnessed to detect, reduce, and manage stress. These technologies have used a variety of methods, including approaches broadly classified as personal informatics [13] that involve gathering self-report or automatic sensor data from everyday life to support reflection [14-16]. Other approaches include condition-based education [17], self-management through guided therapy [18,19], peer or social support [20,21], and gaming or entertainment [22,23].

Digital platforms mean that personal stress management interventions are now accessible throughout the working day via web- and computer-based programs [24], apps [25-28], and wearables [29,30]. In addition, popular websites have been repurposed to provide stress management interventions [10].

People’s ability to access stress management interventions is also highly reliant on their contexts. Methods such as contextual inquiry explore how technology fits into everyday life and tend to consider the context of the organization or sector [31,32]. Also important are the contextual influences on the ability to engage with digital interventions such as social norms, culture, time, and location [33]. Human-computer interaction analysis only occasionally extends to a distal understanding of the value chain in how interventions work in context and why [34]. Yet, this is important for people using stress management interventions in the workplace, both with novel designs [35-40] and consumer devices [41-44], as such an understanding of contextual influences on interaction with features would inform evaluation, iteration, and use. Such real-world studies are few, especially outside office environments.

There have been some multidisciplinary reviews of digital workplace interventions, yielding little insight into influences and features, but stress is usually grouped with other conditions in the continuum of mental health and well-being, and few studies are situated in the public sector [45-48]. Many studies in these reviews focus on novel system design or the evaluation of a single digital solution. Less is known about facilitators for and barriers to employees using consumer technologies to manage occupational stress, despite the proliferation of commercial tracking tools and the consequent amenability of studying their real-world use [16,49].

We know that many teachers already have their own stress management strategies [50,51] and some are already using consumer technologies to support them [52]. This study took a phenomenological approach to understand how digital technology could facilitate teachers’ stress management in their real-world, complex layered contexts of the school setting [53]. Our research questions were as follows:

In this paper, we first describe digital detection and management of workplace stress and then review what is known about facilitators and barriers to interaction with such digital tools. We then present our study, describing which technologies were used by the teachers and presenting the research findings. We discuss what features teachers found to help them with stress management and what were the impediments to self-care. Finally, we consider what this means for school leadership in the context of education.

The goal of this study was to investigate how consumer technologies could facilitate stress management in the school environment.

A longitudinal study was planned that used semistructured interviews, as well as open and closed-question surveys. Before the first interview, workshops were conducted to help teachers make an informed choice of consumer technology based on understanding their personal stress management strategy and technology preferences. Consumer technologies were framed as digital companions when presented to teachers in workshops. This framing drew on the framing of apps as “life companions” by Klasnja and Pratt [103], acknowledging users’ autonomy [104], emphasizing the mediating and collaborative role of the technology [105], and avoiding any commercial connotations of simply being a consumer Referred to by name or simply as technologies in this paper, those chosen by teachers for this study were the Withings Steel HR smartwatch (that uses PPG to detect heart rate changes); Wysa (Wysa Ltd), an artificial intelligence chatbot based on CBT techniques; Daylio, a mood diary and activity tracking app; and Teacher Tapp, a daily teachers survey with an educational blog. Support was given to the teachers in acquiring their choice of technology from those populating a rigorously compiled taxonomy based on suitability, availability, evaluation, security, validity, and cost; the design of which is described elsewhere [106]. Reasons for teachers’ choices are described in detail in a companion paper on their experiences but include apparent ease of use, to inform stress correlations; enable passive collation of data; and give insight into sleeping patterns [107].

The interview data were supplemented with feedback questionnaires, memos and note taking, and email responses (including from nonparticipants, with their permission). The details, rationale, and elements of these methods are now described.

The orientation of this research was experiential qualitative research [108]. The approach was not guided by an existing theory but combined inductive (bottom-up from the data), cognitive, and social constructivist (the participants’ viewpoints) reflexive thematic analysis [109,110]. The constructivist paradigm is familiar in both human-computer interaction and education research, thus enabling epistemic consistency.

The initial selection of participants was purposive. The inclusion criteria were to be a head of year or assistant head with pastoral responsibilities. Willingness to both talk about stress and to trial their chosen technology to support their stress management strategy; and not currently receiving treatment for a clinical stress disorder such as posttraumatic stress disorder or acute stress disorders were specified in the information sheet. The latter was stipulated to avoid interference between this study and potential prescribed interventions. No prior training in stress interventions was required.

Recruitment was from schools in one multi-academy trust (MAT) based in South London. The Trust had been approached for several reasons. First, MATs have a single governance structure, providing some reduction of internal organizational variation, but each has its own leadership team. Second, the MAT had 3 secondary schools in the South London region whose principals were willing to participate, providing a geographic locus and city context for the research. While there are always some differences in the local housing and labor markets, the good transport links, a predominance of business and service sector jobs along with low unemployment, gave some cohesion to area characteristics [111]. Third, the MAT had a track record of interest in staff well-being. By working in the same city with one MAT and teachers with the same role (head of year), the aim was to ensure some contextual consistency.

Of the 16 teachers who originally volunteered to take part before the pandemic (having been invited by the MAT directors and principals), 8 (50%) were still able to participate after the COVID-19 pandemic precipitated a national lockdown, with participation now facilitated on the web. None were previously known to the researcher. Four (N=16, 25%) of the teachers who dropped out gave reasons and consented to have them noted. These included extra childcare responsibilities, their partner working for the National Health Service and being overwhelmed, and an increase in teaching load. The other 4 did not communicate their reasons.

This research was approved by the University College London Research Ethics Committee (UCLIC_1920_004_Staff_ BlandfordManning). Information and consent forms were sent electronically to teachers before the workshops, followed up with reminder emails if they had not been received by the day before the workshop. The consent and information forms specified the purpose of the workshop and subsequent interviews and survey should they choose to take part, as well as the interviewer’s (JBM) department and university, and that the data would contribute to a final study being undertaken as part of her PhD. The data collected were anonymized, with personal names, identifying remarks, or identifiable topics mentioned in the interview audio recordings removed from the transcripts. Participation was voluntary, but a flat participation fee of £150 (US $183.10) was paid to cover the costs of the participant obtaining their chosen technology. Participants had the right to withdraw from the study at any time.

Due to the pandemic, only 2 participants spent more time in school than working from home during the summer leg of the study. Six (75%) of the 8 teachers extended their involvement from the “intended” study into the autumn term back in school. The 2 (25%) who did not continue gave no reason, but there was still a huge pressure on schools at that time. This autumn interview is described as the “extended” period in the results.

The choice of consumer technology was facilitated by the presentation of a taxonomy of stress management interventions [106]. The workshop was piloted and refined with 3 teachers not involved in the study. Workshops used Zoom (Zoom Video Communications), PowerPoint (Microsoft), and Mentimeter and were arranged at a convenient time for the teachers. Three teachers attended the first workshop, 2 the second, and 3 the third. Two people ran the workshops, a teacher and the researcher (JBM), who is an experienced clinician and policy researcher interested in professionals’ stress.

The workshops were designed to first introduce the study concept. Second, they were designed to allow teachers the time and information to understand the different ways in which they consciously or subconsciously seek to manage stress. This was intended to facilitate the identification of personal stress management strategies. Third, the workshops were designed to introduce teachers to the different ways in which technology could conceptually support their stress management strategies. Finally, they were designed to describe each of the technology options open to them for the forthcoming study. Mentimeter, an interactive presentation software, was adopted to enable anonymous crowdsourcing of ideas from teachers. The software aggregates responses to questions and displays them in real time as a word cloud.

Data were gathered both through feedback forms and from the workshop recordings. Transcripts were made of the latter, but the data were mainly reflections on understanding stress and contextual information relating to COVID-19 pandemic changes. The feedback forms allowed teachers to record their choice of technology, along with why they chose it, and when and where they anticipated using it. Participants were reminded only to submit their feedback form if they wished to proceed. All participants returned their forms. The feedback form also captured some demographic data and contextual feedback, largely related to changes brought about by the pandemic. These data along with the comments from the workshop allowed for some tailoring of the semistructured interviews according to technology choice and the altered teaching arrangements due to the COVID-19 pandemic.

After receiving the feedback form, the researcher contacted the participant to arrange (1) payment of the flat participation fee and (2) schedule the time of their first interview. The staggered timings of the workshops meant that some teachers had their technology ahead of others. One teacher had considered trying an online CBT program but on examination reported it as too medical. Another had hoped to use the Fit2Teach Pro app but at the time, technology purchases needed to be made and the app had not been updated within the previous 6 months, so it was excluded as an option. Another teacher wanted to try both the Withings watch and Wysa app, but as the cost exceeded the flat participation fee, she opted for the Withings watch only (Table 1).

Eight teachers took part in 2 one-to-one interviews planned for the intended study; 6 (75%) teachers went on to participate in a further interview for the extended study in the autumn. At each interview, an overview of the study was first recapped and the teacher’s current circumstances were discussed. Open-ended questions were then framed around the teacher’s personal choice of technology followed by probing of contexts (eg, physical, social, and cultural). Questions tailored to the technology allowed in-depth explanations of the facilitation of and barriers to the use of features and their ideas for improvement. After the interview, each participant’s responses were copied into a file and added to a pseudonymized case study folder for that participant in NVivo (Lumivero). The length of each interview ranged from 32 to 76 minutes.

In the intended study, spanning 6 to 7 weeks of the summer term, surveys were sent after about 4 weeks to prompt thinking about activity, data gathered, frequency of use, and stress symptoms. Three surveys were set up using Microsoft Forms, and each participant was sent a link by email and staggered through the week. The questions were arranged around themes and included branching from responses to ensure subsequent questions were logical and related to their technology choice. Closed yes or no, multiple-choice, Likert scale, and open responses were included. The response rate varied from 100% (8/8, first and last day of the week) to 75% (6/8, midweek). The answers informed the second and third interview questions.

In the extended study, 14 closed questions were asked at the end of the interview as a way of capturing final participant thoughts on technology features and the influence of their occupational contexts on use. These were simple agree-disagree responses that allowed for greater confidence in the analysis. All survey responses also provided more data for triangulation.

All interview recordings were transcribed verbatim and identifiers were removed. Transcripts were not returned to participants, as (1) the researcher did not want to add a time burden to the teachers taking part, (2) the evidence suggests that doing so adds little to the accuracy of the transcript [112], and (3) this retains the responses precisely as they had been given at the time of the interview when the teacher had the time and space to respond naturally. Transcribed interviews were checked against the audio recordings for accuracy.

Reflexive thematic analysis was undertaken of the whole data set (transcripts, surveys, researchers notes, and emails) [113] and informed by interpretive phenomenological analysis. NVivo (version 12) and SimpleMind (versions 1.22 and 1.31) software were used for the organization and development of codes, categories, relationships, and themes. Codes were shared with the coauthors (AB and JE-C), and a codebook was generated. In line with the advice of Braun and Clarke [110], a self-reflexive approach was taken throughout the research, including creating a rolling memo of reflections from the time of the interviews until the end of the study. These notes were shared with the coauthors during the stages of thematic analysis. Table 1 summarizes the teachers’ technology choice and supporting concepts, duration of use, strategy and mechanisms, and extra features desired.

A total of 8 teachers completed the summer interviews, and 6 (75%) were involved in the extended study in the autumn of 2020. Most teachers were working from home during the summer term due to COVID-19 pandemic restrictions. They were all back in school in the autumn but in “bubbles” segregated by year group, moving between classes instead of students moving around the school and with extended school days due to staggered student arrival times. The sample was ethnically diverse, with just 1 male participant in both summer and autumn studies.

Teachers described how they began to use their technologies and which features facilitated their interaction with the technology and awareness of their stress. Quick and discreet interactions were complemented by passive data collection, making use simple, convenient, and private. Automated prompts and data collation facilitated easy engagement and awareness of stress patterns and associations. However, the teachers described having to be persistent with adoption, as inadequate instructions required them to experiment and use social media to discover more features. There was a particular issue reported with the Withings watch measuring sleep when they were awake and a desire that this wearable could automatically detect activity related to their job or their stress-relieving exercise. For fuller use and benefits to be realized, teachers described creating a culture in which it is socially acceptable to manage and talk about stress, where privacy was facilitated and self-care with technology was normal and personalized from the start of a teacher’s career. This would legitimize experiences of pressure and permit self-care.

The participant code is included at the end of quotations in the form of a teacher number (T1, T2, etc); the technology; and the term (summer or autumn) from which the quotation was taken.

This theme captured how the design and features supported rapid technology use. Teachers conveyed a sense of confidence that interaction could therefore take place without anyone knowing.

To the best of our knowledge, this is the first study to describe facilitators and barriers to personally chosen technology support for stress self-management in the educational context, along with insight into how schools could facilitate such interventions. The findings complement those of teachers’ experiences, which demonstrate how a more holistic understanding of stress, and permission to manage it is generated as teachers found meaning in their data and interactions [107].

As shown in our findings, the themes generated indicate specific features, data collation, and technology-supported concepts that facilitated teachers’ use during their working day. In contrast, the barriers to use centered around a lack of information and signposting and external, school-related mesocontexts that inhibited use or putting their stress-relieving action into practice as planned. These themes are now unpacked further.

Teachers expressed their appreciation of the wearable appearing to be a watch, not an obvious tracking device. They also liked the rapidity with which Withings watch data could be viewed and the Daylio app could be used, thus preventing potentially stigmatizing exposure. These themes reflect the persistent taboo of experiencing stress, and teachers’ fear of not being perceived to be able to cope. Discreet “design” has been noted previously as a facilitator for recording feelings [49,114] and for the use of technologies by employees for stress management [93,115]. It has also been observed that the esthetic of design of wearables is important [116]. Added to the knowledge that short interventions are important for stress management for providers and users [95], including in constrained workplaces [98], we suggest designers should ensure that where the perceived risk of stigma in the workplace remains, design camouflage would encourage engagement for teachers.

Immediacy is increasingly being facilitated by wearables and sensors [117]. The Withings watch enabled stress awareness with cognitive ease both through passively collated data and through haptic notifications. Haptic stimulation has been shown to have a positive influence on attitude to accompanying messages and such a physical cue is motivational [118]. Our findings suggest that for teachers in their polychronically demanding environments, such a physical cue facilitates engagement up to a point, dependent on the individual and competing demands. The exception to immediacy was Wysa, which required deliberate, sustained engagement. Temporal constraints precluded use in school, but Wysa (the chatbot) was described as a “best friend” for giving contextually applicable advice that was both educational and transferable. Such high-effort, problem-solving interventions have been found to be more effective for IT workers than low-effort, positive distractions [8]. Therefore, we suggest that cognitive ease can also enable digitally supported stress management by virtue of ease of application of learning, not just the immediacy of information.

This applicability of the chatbot advice also links to the finding of the importance of contextually mediated reflection on patterns for engagement. This aligns with research hinting that more contextualization facilitates engagement [119] and the noted lack of contextualization in mood-tracking apps used by the general population [49]. A combination of data patterns that could be correlated by the teacher to their circumstances, and the legitimacy of the intervention as facilitated by the school leadership, enabled teachers to reflect on their stress associations and plan mitigating actions.

Teachers identified automated prompts for taking more steps or meditating, or for messages of affirmation when an activity had been undertaken, as helpful. This complements teachers’ reports of appreciating congratulatory text messages of support via WhatsApp [120]. Elsewhere, police officers using wearables to track activity [121] and working-age women with technologies meeting self-set goals [122] have also reported the motivational aspect of digital messages of encouragement. Prompts or reminders for stress management can be preplanned or automated through algorithmic analysis of data [35,82]. Conversely, when prompts are not received, stress awareness among Chinese adults was reduced as was their ability to manage their stress [41]. Even for people in less-restricted roles than teaching, receptivity to automated supportive messaging is highly dependent on contextual variables [123]. Future research could consider how messaging could be further enhanced to enhance stress management among employees in temporally demanding roles.

This study validated the established need for good data visualization with improvements suggested for the Daylio app and readability of the Withings watch’s digital display screen. Stress data visualization has been tested at the organizational level [39], but this would be difficult to translate to schools. The paucity of setup information supplied with the Withings watch was commented on by 5 (83%) out of the 6 users. Teachers resorted to YouTube (not the Withings website) for guidance, although not knowing what constituted “normal” data remained a problem for 2 of them. Social sensemaking with personal informatics has demonstrated that value can be gained from contextualizing data through comparison with peers [124], but the COVID-19 pandemic would have been a barrier to this. The app users also cited apparent design assumptions of intuitive use, particularly with Daylio, which delayed full adoption. These barriers to adoption extend the finding of inadequate technical instruction in a digital mental health review [125], adding a specific example of a wearable consumer technology adopted for stress management (out of 208 papers, only n=2, 0.01% were for wearables in the review by Borghouts et al [125], and these were for tracking serious mental illness). Wearable connectivity was found to be limited by Bluetooth range, a challenge noted generally with digital stress support [126], which for teachers specifically related to the proximity of their smartphone, often not kept with them [52]. Designers therefore need to remember that users still need comprehensive navigational advice and that they cannot assume proximity between wearables and smartphones for all occupations.

Some teachers were disappointed that the data indicated sleep when they were awake, despite Withings’ reputation for accuracy [127]. This has been noted in studies of Fitbit [61,128]. Liang et al [128] suggested the companion app should allow the user to edit their inaccurate data, an important consideration because inaccurate data is generally a reason for abandonment [129]. This remains a consideration as users perceive stress algorithms to be at odds with their subjective experiences [64].

No other inaccuracies were noted by teachers in this study. Several teachers wished that their technology offered alternative strategies, for example, relaxation as well as education, partly because stress symptoms can change over time. This adds to similar reports for a greater plurality of features or strategies from other studies [93,130-132]. However, this must be balanced against the extent of choice [133], as too much choice in digital tools could be overwhelming [134].

Temporal constraints led to reduced frequency of interaction with technology for most of the staff when back in school, a barrier to interaction that can only be discerned in real-world evaluations [135]. Notably, this was only detected because of the extraordinary circumstances of teachers able to compare working from home with working back in school and because sustained engagement was still demonstrated months after initiation. The text-heavy Wysa (never accessed in school) and Teacher Tapp (significantly reduced interaction in school) apps were the most affected. Counson et al [98] found that vocational intensity was a barrier to engagement with a stress management app cocreated with doctors and professional demands constrained time and enthusiasm for digital health interventions generally [136]. Thus, organization-level solutions are likely to be required as well, connecting to the final finding that teachers’ desire is for a workplace culture that facilitates self-care and caring. Such a culture requires the facilitation of mesocontexts, such as time and physical environment. Although other studies have used machine learning to judge the best timings for stress interventions for students and in offices [10,38,82,131], the polychronic nature of teachers’ work would make finding patterns in behavior and just-in-time interventions more challenging. In addition, during the school day, stress management can require privacy, for example, in the form of a staff room, with lack of privacy being a known deterrent to interaction generally for digital mental health [125].

The findings on both facilitators and barriers to teachers’ engagement suggest that school leadership could leverage technology to support stress management. Some alerts can prompt teachers’ just-in-time stress relief strategies or simple self-care such as having a drink. Yet, teachers also need supportive mesocontexts such as a private space in which to relax, protected time when they are undisturbed, and a culture in which stress is seen as a warning and not a weakness. This need for a positive organizational culture for stress support has been noted before by office workers considering the use of personal digital tracking and e-coaching [137]. As several teachers commented, what is the point of drinking water to keep hydrated, if there is no time or provision (in their building) or even acceptance that they will then need to use the bathroom.

Rich, contextualized data on the adoption of and influences on technology use were collected over an extended period of up to 27 weeks, enabling deep and detailed insights into teachers’ engagement. All 8 heads of year completed the summer study and just 2 (25%) were lost to follow-up in the extended period. This longitudinal study provides evidence of the acceptability of consumer technology to teachers for stress management support and which features were of value for self-care in an occupation renowned for its stressfulness.

Taking part in the research inevitably meant there was some extrinsic motivation and would have raised awareness of stress and self-care; therefore, there could have been a digital placebo effect [138-140]. The pandemic led to the exceptional circumstances of both more attention being given to teachers’ well-being by the school leadership and the teachers beginning to use their technologies while working from home. This meant that all participants had time to adopt their technology in a quieter environment and could compare facilitators and barriers between the 2 settings. Further research could test digital stress support against an active control.

The choice of technology was framed by a specific taxonomy which enabled choice based on strategy and conceptualization of interaction. Given the pandemic, workshops were held on the web which prevented hands-on exploration of technologies before selection. However, the main author was available for queries throughout the study. Therefore, another limitation was that autonomy was limited to the technologies that populated the taxonomy, selected based on suitability, availability, usability, security, validity, and cost. In addition, there was no classification of technologies according to detection, management or prevention, just-in-time recovery, or resilience building. These concepts and characteristics could be tested to see if they assist in technology selection and self-care. Apart from Wysa, no technology was chosen that facilitated treatment or guided self-management nor was entertainment based. The reasons for this as well as encouraging a wider variety of interventions could be investigated further.

Another strength of this longitudinal study was that it was undertaken with teachers who taught many socioeconomically deprived students, typically a more stressful setting. Teachers in more affluent areas or working with more economically advantaged children are known to experience less stress [141-143]. Thus, the findings should benefit teachers in more challenging situations. It was poignant to hear teachers in the autumn term talking about children returning to school after the pandemic without even a pen, let alone lunch money.

The participant sample cannot be claimed to be representative of all secondary school heads of year, although this study makes an important contribution, given the dearth of research with middle managers. The sample was self-selecting and only one head of year was male, but the racial mix was diverse.

Consumer technologies are quick, easy, and discreet to use; offer encouragement; and enable personalization. They could have an explicit role in facilitating teachers’ understanding and managing their stress. Developers should not make design assumptions of users intuitively discovering all features, and the ability to manually amend some data should be facilitated. School leadership has an important part to play in facilitating stress management, and technology use could contribute to a culture in which self-care is normative and discussions of stress are socially acceptable. The convenience and ease of technology engagement could be leveraged at the organizational level to integrate stress support. This invites the proposition that a teacher’s ability to act on technology data is a proxy measure for a caring school culture. This would require both protected time and space on the school premises. We hope our findings are of interest to policy makers and encourage developers of digital interventions to work with educators to ensure contextual insights inform the design of health and well-being technology.