Feasibility of a theory-based intervention to reduce sedentary behaviour among contact centre staff: the SUH stepped-wedge cluster RCT

Sedentary behaviour as a public health problem

Sedentary behaviour is a serious occupational health hazard, which is linked with an increased risk of type 2 diabetes, cardiovascular disease, musculoskeletal issues and poor mental well-being. 1–6 Earlier research studies proposed that these risks are independent of physical activity7,8 given that, conceptually, sedentary behaviour and physical activity are different,9 with each thought to pose health risks independent of each other. 5,10–12 However, more recent research shows that, although physical activity can modify the associations between health risks and sedentary behaviour,13,14 health effects of sedentary behaviour are still evident even after controlling for physical activity. 4,5,12,15 A review16 of behaviour change strategies that are used for the reduction of sedentary behaviour among adults reported that interventions that showed the most promise in reducing sitting time were those that aimed to changed sedentary behaviour rather than increase physical activity. The reduction of sedentary behaviour is, therefore, not a consequence of effectively promoting physical activity and should be recognised independently when developing interventions, guidelines and legislation.

Workplace sedentary behaviour is placing a large burden on employers and the health-care system. In office-based roles, many employees have prolonged periods of sitting, which are enforced by workplace culture and ergonomic set-up. 17 This sitting behaviour can impact on the health of office-based workers. For example, musculoskeletal issues are a leading cause of disability worldwide and one of the most common health problems for desk-based workers. 18–20 Estimates of the prevalence of musculoskeletal symptoms in computer users are as high as 50%. 21 Lower back pain in particular is associated with prolonged sitting. 22

Factors associated with sedentary behaviour in workplaces

A cross-sectional study conducted among Australian adults23 found that the following were associated with higher levels of occupational sitting: being male; being younger; having higher education and income; both part-time and full-time employment; sedentary job tasks; white-collar or professional occupations; and higher body mass index (BMI). Some of these findings were corroborated by a systematic review of occupational correlates for sedentary behaviour. 24 The review identified full-time employment, working in a call centre, high level of leisure sitting time, high body weight, being older, high education and high income levels as positive correlates, and blue-collar occupations and smoking as negative correlates. 24 Political correlates (described as incentivisation, coercion and job description) that were identified included repetitive work, handling heavy loads and forceful exertion, which were found to be negatively associated with occupational sedentary behaviour. Low control or autonomy was found to have a positive association.

Qualitative studies conducted with employees and managers in a UK-based software engineering company25 and non-academic office staff at a university in Singapore26 have identified several barriers to reducing sedentary behaviour in workplaces, including workplace social and cultural norms (sitting is perceived to be a reflection of productivity by themselves and others, whereas standing is seen as disturbing others), work pressures (pressure to deliver work), the nature of the job requiring more sitting (e.g. working with computers) and physical environment and office infrastructure (small office spaces, no adjustable workstations and lack of facilities).

Contact centres as a setting for public health interventions

There are over 6200 contact centres in the UK, employing 734,000 agents,27 which is roughly 1 in 25 of the UK workforce. Scotland and the North East of England are home to some of the largest contact centres in Europe. Workplaces are often considered as homogeneous, but there are wide variations in worker demographics, the amount of worker autonomy, salaries, the environment and culture.

The call handlers (the highest proportion of contact centre staff) earn an average salary of £16,319 per year, compared with the national average of £26,500, which puts them in the bottom third of earners. 28 An industry report that collected data from 208 contact centre managers and directors using a structured questionnaire reported that turnover rates are high in UK contact centres, with an average attrition rate of 20% reported in 2020. 29 Short-term absence rates are also high, with 5.7% of agent workdays being lost to short-term sickness and unauthorised absences. 29 Career progression is limited, and a lack of promotion and development opportunities has been reported as the second greatest cause of employee attrition. 29

Contact centres and health

Contact centres are currently one of the most sedentary working environments, with higher levels of sedentary behaviour than other office-based work. 24,30 Studies have reported that contact centre staff spend 75–83% of their workday sitting. 30,31 The technology in contact centres prevents staff from regularly leaving their desk and many call handlers often report that work environments are stressful owing to low workplace autonomy, strict supervision of individual performance and commission-based salary systems. 32 Several international studies33,34 show that as many as 60–65% of contact centre employees experience musculoskeletal issues, and a study in the UK35 noted that one in four employees at a contact centre experienced upper body musculoskeletal issues.

A recent study32 found that a common factor for sedentary behaviour shared by contact centre agents, team leaders and senior staff was a considerable lack of knowledge and awareness of sedentary behaviour as a risk factor for poor health. In addition, there was a low level of knowledge among staff of guidelines and recommendations relating to sedentary behaviour and physical activity in the workplace, and therefore there is often no reflection of these guidelines and recommendations in organisational policies. 32,36

Contact centres and organisational drivers

Many organisations representing the contact centre industry are highly constrained by profit-based drivers, cost minimisation and economic outcomes based on productivity and high-quantity customer enquiry resolution. 37 Several contact centres adopt ‘mass production models’ to minimise cost and maximise output, resulting in mechanisation and standardisation of work methods, and reduced employee training and autonomy. 37 The work is such that employees do not engage with other team members and can be isolating, and employees may view themselves as replaceable. 37 These factors can also often influence organisational investment into workplace health initiatives; fears of cost-ineffective programmes and reduced productivity rates are commonly presented by senior team leaders within private contact centres. 32

Contact centre agents have voiced concerns over job security and performance monitoring, and a desire for increased autonomy over their working practices, as influential factors for their motivation to participate in strategies to reduce sedentary behaviour in the workplace. 32 However, organisational pressures to maintain high levels of productivity and meet targets frequently work against organisational investment into health and physical activity programmes within some contact centres. This is often due to perceptions that these activities will reduce the agents’ call-making time and lead to productivity losses. 38 One study reported that leaders and senior staff had ‘identified a conflict between promoting productivity and targets to call agents, while encouraging them to move more and sit less’. 32

Workplace intervention research to reduce sedentary behaviour

Over the last 5–10 years, there have been several systematic reviews of workplace interventions to increase physical activity and reduce sedentary behaviour. 24,39–41 Most recently, a systematic review of environmental interventions in workplaces (e.g. sit-to-stand desks)40 found evidence of significant reductions in sedentary behaviour in 14 out of 15 studies. This review found that multicomponent interventions that targeted more than one level of the socioecological framework, were most effective. In 2016, another systematic review41 assessed the effectiveness of white-collar workplace interventions to reduce sedentary time; it similarly found that multicomponent interventions had the greatest effect. Both reviews40,41 identified a need to assess whether policy-based measures or organisational change could further increase effectiveness. One study42 assessed the effect of sit–stand desks and ergonomic awareness on reducing sedentary behaviour in 15 Swedish contact centres and found that working at a sit–stand desk was associated with a slightly greater reduction in sitting time than sitting at a non-sit–stand desk. Regular interruptions to sitting time during the workday have previously been found to significantly reduce discomfort in the lower back and fatigue levels in overweight and obese office workers, without affecting productivity. 43 Another study44 found that the use of standing desks led to significant reductions in upper back and neck pain in office workers. A recent study45 exploring barriers to participation suggested that ‘barriers occurred at multiple levels of influence, and support the use of ecological or multilevel models to help guide future programme design/delivery’. 45

Multicomponent workplace interventions

Given the specified need to address cultural and organisational factors affecting workplace behaviour, we developed and piloted SUH (the intervention in this report). There are a limited number of other similar interventions and the evidence is still sparse, with only one UK-based study, which is ongoing. This UK-based study is funded by the National Institute for Health and Care Research (NIHR) Public Health Research (PHR) programme46 and is currently examining the effectiveness of the Stand More AT Work (SMArT Work) and life intervention aimed at reducing sitting time of office workers. Although this intervention is similar to our intervention in some respects, SUH specifically considers aspects of organisational change, targets contact centres and takes a systems-based approach. 47 This takes into account the complexity of the context, resources and assets of specific contact centres. The Stand Up Victoria study in Australia48,49 also had a multicomponent intervention. However, height-adjustable workstations were a main part of the intervention. Although we envisaged that some contact centres would take up this option, we recognised that not all contact centres have the resources, or desire, to implement them. A third study in Perth, Australia,50 used a similar participatory approach to SUH but had no theoretical basis and was assessed at only a 12-week time point.

Sedentary behaviour research in contact centres in the UK

Research on sedentary behaviour in contact centres is sparse and no reviews were identified on this topic. Two relevant primary studies are discussed here. 32,51 The first study32 is a qualitative study exploring factors that influence contact centre agents’ physical activity and sedentary behaviour in the workplace. The study conducted interviews with contact centre agents, team leaders and senior staff, and reported that agents said that several factors influenced their motivation to increase physical activity and reduce sitting time, including continuous performance monitoring, concerns over job security and a desire for increased autonomy in the workplace. As mentioned in Contact centres and organisational drivers, the study noted the conflict reported by team leaders and senior staff relating to productivity and promoting more movement and sitting less.

Another study51 used a mixed-methods approach to explore the acceptability and feasibility of a multicomponent intervention that targeted physical activity and prolonged sitting in contact centres. The study reported that contact centre agents perceived that the study assessments were acceptable and education sessions, height-adjustable workstations and e-mails were seen to be the most effective components. This intervention is also based on the socioecological model and is similar to the SUH programme. Although the intervention51 has several merits, SUH has a greater emphasis on organisational factors and takes an adaptive and flexible approach, taking into account the contexts and systems of each contact centre.

Relevant policy and practice

Policy regarding sedentary behaviour is insufficient considering the evidence of its associated risks. A recent study reviewed current national and international occupational safety and health policy documents (e.g. guidelines, legislation and codes of practice) for their relevance to occupational sedentary behaviour. 36 The review found that many workplaces and jurisdictions had legal frameworks that established a duty of care for occupational health, but discovered that no occupational health and safety authority had a policy specifically targeting occupational sedentary behaviour. Although some existing policies have aspects relevant to sedentary behaviour in the workplace, the authors identified a need to address the emergent hazard of excessive occupational sedentary behaviour by developing specific policies for this issue. They also highlight a need to support workplace-based initiatives that aim to minimise sedentary behaviour and the associated risks.

A number of awards exist across the UK that are designed to recognise and encourage efforts made by organisations to improve health and well-being in the workplace. These include the North East Better Health at Work Award52 in England and the Healthy Working Lives award53 in Scotland. Although these awards have several categories for health promotion that can mitigate some of the negative outcomes of excessive sitting time, reducing sedentary behaviour as a specific outcome is not acknowledged.

Importance of this research

Current UK workplace legislation means that many members of staff in contact centres receive remedial ergonomic support as a mitigation measure to reduce existing musculoskeletal issues only after a chronic or musculoskeletal condition has been diagnosed. 54 However, current practices, compounded by workplace culture, inhibit initiatives that encourage contact centre staff to reduce sitting time. 55 Taking into account the lack of policies from authoritative bodies that are specific to sedentary behaviour, mean that it is important for workplaces to proactively develop their own organisational policies that include, and promote, opportunities for reducing occupational sitting time. 36 Given that contact centres are among the most sedentary workplaces,30 and that employees report higher levels of stress and depression than other desk-based work,56 it is key that preventative approaches are implemented in contact centres.

This work is currently needed to ensure that healthier working policies are distributed equitably across all workplaces, not just those that have more worker autonomy and better working conditions. Creating healthier contact centre environments may be more difficult than in other workplace settings, which is why such an intervention is necessary. In addition, building the capacity to develop and measure workplace-based interventions for health in contact centres is vital for developing a stronger business case to encourage and enhance organisational uptake and buy-in.

6SQuID step 1: defining and understanding the problem and its causes

The first step of developing an intervention is to understand the problems and their causes. Step 1 creates the foundation for a successful intervention and involves fully understanding the problem that an intervention is trying to address and developing an understanding of the contributing factors to the problem. From the literature review and the stakeholder consultations (with the pilot centre, Ipsos MORI) we collectively defined the problem as the high level of sedentary behaviour (sitting time) during working hours in contact centres.

Through the literature review and focus groups with 34 participants, four themes emerged and were grouped into four levels by the researchers: individual, social/community, structural environment and organisational (Table 1). 62

6SQuID step 2: identifying which factors can be modified

Modifiable factors with the greatest scope for change were considered to identify which specific factors should be targeted through intervention activities. These are described in the fishbone diagram (Figure 1).

FIGURE 1.

Fishbone diagram identifying the modifiable and non-modifiable factors. Black text, potentially modifiable factors; navy text, non-modifiable/hard-to-modify factors.

6SQuID step 3: deciding on the mechanisms of change (theory of change)

The intervention was based on two main theories: social cognitive theory63 and the social ecological model. 64 The intervention also aimed to create a sense of ownership to increase the likelihood of longer-term sustainability. 58 Although social cognitive theory addresses many personal determinants and socioenvironmental factors, the social ecological model takes the proposed multifaceted approach one step further to consider the intervention at not only the individual and interpersonal levels, but also the organisational, environmental and group levels, and takes into account the interactions between each of these. 65 By targeting multiple levels of the workplace, SUH aimed to foster an atmosphere that creates a social norm within the office community of sitting and standing within the workplace. The social ecological model justifies and predicts that SUH’s multifaceted approach will be effective, acceptable, feasible and sustainable. We also took a systems-based approach, by recognising that the implementation and sustainability of the intervention is dependent on how adaptive the contact centre system is to change.

Developing a theory of change involved developing intervention activities to target each modifiable factor leading to sedentary behaviour. This step details how each modifiable factor identified in step 2 could be addressed at each level of the contact centre by designing specific activities informed by the focus groups, literature review and workshop data. The literature review identified a number of systematic reviews of workplace interventions to increase physical activity and reduce sedentary behaviour. 17,24,39,40 Most recently, a systematic review of environmental interventions in workplaces (e.g. sit-to-stand desks)40 found evidence of significant reductions in sedentary behaviour in 14 out of 15 studies. This review found that multicomponent interventions that targeted more than one level of the socioecological framework were most effective. In 2016, another systematic review assessed the effectiveness of white-collar workplace interventions to reduce sedentary time. 41 It also found that multicomponent interventions had the greatest effect. Both reviews identified a need to assess whether policy-based measures or organisational changes could further increase effectiveness. SUH is designed to have organisational change as a key component of the intervention. A recent UK study funded by the NIHR PHR programme found that the SMArT Work intervention was effective in reducing sitting time of office workers within an NHS workforce using height-adjustable workstations, self-monitoring tools and behaviour change techniques. 46,66

We found no reviews of interventions to reduce sedentary behaviour specifically in contact centres. The use of sit–stand desks and ergonomic awareness, as well as multicomponent workplace interventions, are effective in increasing physical activity and decreasing sedentary behaviour in contact centres. 42,51 The programme theory that we developed is outlined in Figure 2.

FIGURE 2.

The SUH programme: theory of change.

6SQuID step 4: clarifying how the mechanisms of change will be delivered (theory of action)

After defining the theories of change, the researchers identified the necessary resources available for implementation of the activities (Figure 3) and illustrated the programme theory in a logic model to demonstrate how intervention activities would lead to intended outcomes (Figure 4). Inputs and resources were identified through focus groups and informal discussions with management. During this phase, a workshop was held at the test contact centre to introduce staff to examples of intervention activities and equipment for the workplace. The workshop activities were developed based on feedback from staff in the focus groups. Staff were asked to contribute their own ideas and prioritise the ideas for intervention activities/workplace equipment that they wanted to try out. The research team fed back the results to the SUH implementation group; the SUH implementation group then decided on the final intervention activities to be implemented. There was at least one activity from each theory of change and the activity took account of the resources, assets (e.g. local spaces, existing equipment and spare spaces) and budget available. Once the specific intervention activities were chosen, the team worked with the contact centres to decide on an action plan for delivery and implementation of the activities (who, what, when and where).

FIGURE 3.

The SUH programme: resource assessment.

FIGURE 4.

The SUH programme: theory of action.

6SQuID step 5: testing and adapting the intervention

Contact centre staff in the pilot centre were assisted and guided by the researchers during two wellness committee meetings to establish and set goals for the programme. Continuous feedback was gathered from staff throughout the duration of the intervention so that the delivery of activities could be adapted as necessary. The intervention that was initially developed in Ipsos MORI continues to be implemented (although with a reduced emphasis owing to COVID-19) nearly 3 years after it started. The contact centre has reported that 25% of staff now use standing desks (an increase from almost none). There are a number of different activities going on at any one time and the call centre staff are constantly thinking of new activities to implement. The contact centre also asked to ‘bolt on’ a mental health component: the research team worked with them to implement mindful activities that either were standing up interventions (e.g. jigsaw puzzles on a stand-up desk) or encouraged them to leave their desk and be mindful [e.g. Lego^® (Billund, Denmark) and knitting].

The final step of 6SQuID is to collect sufficient evidence of effectiveness to justify rigorous evaluation and implementation. This is the evaluation that is described in this report. However, as mentioned previously in Development of the Stand Up for Health programme theory, although the theory of change remains constant, an adaptive intervention needs to respond to contextual issues. Therefore, each centre was aided in developing a section of activities to meet its needs. Table 2 outlines which of the steps were undertaken prior to, and as part of, this study.

Developing theories of action for the contact centres in this study

6SQUiD steps 4 and 5 formed the basis of the tailored intervention for each contact centre. The hypothesised theories of change (step 3) remain constant, but the theories of action are specific to each contact centre. The first step of the intervention included a ‘workshop’ that took place at a contact centre in which staff could test, suggest and vote for various activities that they wanted to try. The staff also described the environment, location, existing equipment and other assets that were then used to plan the activities. The SUH research team provided some small pieces of equipment for centres to try out before any investment was made. Although the number of activities was not limited, we worked with the contact centres to implement at least one activity from each theory of change (see Figure 2). These activities were used to create an action plan to be implemented over several weeks. A SUH committee, made up of all levels of contact centre staff, was created to support activity implementation. At the end of the intervention period, the researchers ran a second workshop to understand which activities had worked (step 4 of 6SQuID). Further prioritisation and choosing of future activities was then undertaken. All equipment and activities were risk assessed and details of how to use the equipment/undertake the activities were provided. A website was developed with useful resources, and opportunities for the contact centres to blog/share their experiences and create a community of SUH contact centres.

Developing the programme activities: pre lockdown

To enable the activities to fit within the systemic, cultural and environmental constraints of individual contact centres, we used the following methods to develop and implement activities for the SUH programmes during the pre-lockdown period.

Duration of the intervention

The duration of the intervention was defined as the period that the centres took to develop their preferred activities for each theory of change, prepare an action plan for sustained engagement and test out some of the activities. This process took around 3 months but was dependent on the contact centres and COVID-19.

Scalability and translation

This intervention is designed to be scalable and transferable into other contact centre settings in the UK and internationally. It is also transferable to other workplace settings. The reason why we have not specified activities (or fidelity to particular activities) is to allow for flexibility, scalability and transferability to different contexts. We recognise that some contact centres will have more resources and assets than others, which is why we have suggested a range of activities that can all activate the theory of change. Given that the theories of change are the most adaptive aspects of the intervention, the intervention is adaptive to all contact centres, and it is designed so that contact centres can implement it at little or no cost.

Study aims and research questions

All of the original aims remained the same (see Overall research aims and objectives), but we added an additional aim: aim 4 – consider previous aims under the context of COVID-19.

All of the research questions also remained the same, but we included some additional ones to take account of the impact of COVID-19.

The research questions for aim 1 were:

• Can SUH be adapted for those working remotely?

• What is the acceptability and feasibility of this mode of delivery?

The research questions for aim 2 were:

• Is it feasible to adopt an online format and does this affect the use of a stepped-wedge study design?

• Does the mode of delivery impact on the feasibility of the study design (online vs. face-to-face delivery)?

Impact of COVID-19: changes to data collection

All scheduled workshops and data collection until January 2020 were completed as planned. The research team did not make any in-person visits after this time. The SUH team had planned 13 additional visits from March 2020 to June 2020, all of which were cancelled. The team shifted to remote collection of data, including interviews, focus groups and questionnaires.

It is important to note that the normal routine in centres was disrupted and, hence, any follow-up data collected may have been biased. Outcomes such as mental health, work engagement and sitting behaviour are likely to have been influenced by the COVID-19 pandemic and may not accurately capture intervention effects. Given this, we still administered the questionnaire when it was easy to do so and endeavoured not to increase burden for centres.

Posting activPAL™ devices (PAL Technologies Ltd, Glasgow, UK) to the centres was considered as an option. However, this was not feasible because researchers needed to be present to explain attachment and filling of logbooks. There would also be high risk of loss of the devices.

Impact of COVID-19: changes to intervention activities and delivery

Study population

Given that this was a cluster RCT, contact centres rather than the staff employed in the contact centres (e.g. managers, supervisors and call handlers) were recruited. All staff in those contact centres had the opportunity to participate in the intervention. For the evaluation components, staff had the option of taking part on an opt-in basis.

Recruitment

Our original aim was to recruit 10 contact centres with more than 100 employees in Scotland and/or the North East of England. The target was to recruit 10 contact centres by month 3 of the project. A total of 11 contact centres were recruited by this time, two of which had under 100 employees (72 and 33 employees, respectively). We decided to include these two smaller centres in the place of one larger centre because both were very enthusiastic and we wanted to start randomisation as soon as possible.

All staff who had been working in the contact centres were invited (via e-mail through their contact centre, posters within the centre and in-person visits to the centre by the SUH research team) to be involved in the evaluation of the intervention. Staff who were interested in participating were sent an information sheet and consent form. It was made clear that participation or non-participation would not affect terms of their employment.

Randomisation

Randomisation occurred in month 3 of the study. The unit of randomisation was contact centres. A computer-generated block randomisation algorithm was used to randomly allocate each contact centre to start the intervention at one of five time points, 3 months apart. Randomisation of contact centres to sequences was conducted in May 2019, using stratification by centre size (≤ 500 vs. > 500 employees). Stratification by centre size ensured that the combined centre size was approximately balanced across time points. Randomisation in this way allowed us to introduce the intervention to each site in an unbiased way, unrelated to time or the particular circumstances of each site. It also aimed to ensure that there was approximate balance, on average, across all of the intervention start times in terms of participant or contact centre characteristics. Allocation concealment and centres were not informed of exactly when they were to start the intervention until close to the intervention visit, was used as far as possible.

A complete list of the contact centres that agreed to take part in the study was compiled by Divya Sivaramakrishnan and Jillian Manner. The contact centres were numbered 1 to 11, in fixed order, and this list was signed and dated by Divya Sivaramakrishnan and Jillian Manner. Randomisation was then conducted in May 2019 by Richard Parker, who was fully blinded to the names of the contact centres and who generated a list of centre numbers showing the sequences that each centre should be allocated to (see Figure 5).

Allocation

Allocation to trial groups was carried out by the statistician after recruitment, who was blind to the contact centre identity.

Sampling and sample size

We anticipated that all employees at a site were likely to take part in some or all of the intervention activities. However, employees also had the option of taking part in the research evaluation component. The sample size refers to the number of people taking part in the research evaluation (not those taking part in intervention activities). The sample size and target difference are the same as another similar study73 that proposed a sample size of 160 participants per group to detect a reduction in workplace sedentary behaviour of 45 minutes per day. Given that we aimed to have six control and post-intervention cross-sectional comparisons (see Figure 5), the target sample size was 160/6, which equals ≈ 27 participants per contact centre per data collection period. We anticipated that 10 contact centres would be recruited, so we aimed to recruit at least 270 employees to take part in the research evaluation. An aim of this feasibility study was to test sample size assumptions and produce a more accurate sample size calculation for a future study. We also aimed to recruit 6–8 individuals per contact centre for the qualitative data collection (60–80 individuals in total).

Ethics/regulatory approvals

The study was conducted in line with Medical Research Council and the Economic and Social Research Council ethics framework. Ethics approval was obtained from the School of Health in Social Sciences ethics committee, University of Edinburgh (ref STAFF142).

Acceptability and feasibility of the Stand Up for Health intervention

Feasibility of using study design and procedures

The feasibility of using a cluster-randomised controlled design was evaluated using qualitative methods along with quantitative indicators. This component of the process evaluation is not considered under the RE-AIM framework. The range of study procedures evaluated, and methods of analyses, were as follows:

• Recruitment and randomisation of contact centres and dropout– researchers’ reflections on completion of recruitment and randomisation of contact centres, including quantitative indicators on actual, compared with planned, recruitment and randomisation figures. Reasons for centre dropouts were explored using data from exit questionnaires completed by two centres.

• Recruitment of participants and dropout – (1) quantitative indicators on actual, compared with planned, recruitment numbers and data on dropouts; (2) qualitative data on recruitment procedures through focus groups and interviews with staff and stakeholders; and (3) statistical analysis of reasons for participant dropout.

• Data collection procedures and tools – assessed using qualitative data through focus groups and interviews with staff and stakeholders. Data from activPAL logbooks pertaining to detachment of the activPAL device and reasons for detachment are presented in this report, along with data on the number of activPAL devices issued and returned, and questionnaires completed.

A summary of the framework and measurement for the process evaluation is presented in Table 4.

Qualitative analysis

All focus group discussions had a moderator and co-moderator. The face-to-face focus groups and telephone interviews were recorded using an audio-recorder. The online sessions were recorded using the record function on Microsoft Teams. Data were transcribed verbatim by a transcription agency.

An initial coding framework was developed by Divya Sivaramakrishnan, Jillian Manner and Graham Baker based on the RE-AIM process evaluation framework (see Table 4). Five transcripts were coded by both Divya Sivaramakrishnan and Jillian Manner (two pre-lockdown staff focus groups, two pre-lockdown stakeholder interviews and one post-lockdown staff interview). The other transcripts were coded by one researcher (DS/JM). Graham Baker and Richard Parker acted as critical friends, who discussed the themes and subthemes with Divya Sivaramakrishnan and Jillian Manner; clarified and offered interpretation; and provided insights and suggestions to refine the themes. Transcripts were coded deductively based on the coding framework, with a deductive approach to capture themes within the broad framework. Thematic analysis was used to generate themes and subthemes following a six-step process. 80 Differences between centres within each theme were examined during analysis. A computer software package (NVivo version 11; QSR International, Warrington, UK) was used to code the transcripts and manage the thematic structure.

Other sources of data

Participant dropout

The SUH team considered individuals who participated in pre-lockdown data collection, but not post-lockdown data collection, as well as those who provided baseline measurements but not follow-up measurements (3 months later) in the post-lockdown phase. We also asked stakeholders the reasons for participant dropout (e.g. left the company, moved to a different part of the company, were unavailable, did not want to participate, technological barriers, did not receive communication, I don’t know, other – elaborate). Centres 3, 6, 10 and 11 were included in the pre-lockdown analysis and the following binary variables were considered: left company, moved job and lost to follow-up. A multiple logistic regression model was fitted to each of the three binary outcomes. The multiple logistic regression model had the following explanatory variables (all entering the model as fixed effects): contact centre, sex, part-time working, age and how long the participant had worked for the contact centre. Centres 3, 6, 7, 10 and 11 were included in the post-lockdown analysis. A multiple logistic regression model was fitted to the ‘lost to follow-up’ variable, adjusting for contact centre, sex, part-time working, age and how long the participant had worked for the contact centre.

Reporting

The process evaluation framework (see Table 4) was used for reporting and the results are organised under two main sections: (1) acceptability and feasibility of the SUH intervention and (2) feasibility of the study design and procedures. Within these two sections, pre- and post-lockdown results are examined separately.

Centre characteristics

Contact centre locations were London, Durham, Tyneside (Newcastle-upon-Tyne, Gateshead, South Shields and Jarrow), Sunderland and Edinburgh. The number of staff members in the centres ranged from 33 to 2000, with an average of 559 staff members.

Participant characteristics

In total, 33 staff and stakeholders from six centres participated in the process evaluation focus groups and interviews. Specifically, four focus groups (centre 2: eight participants; centre 6: two participants; centre 10: six participants; centre 11: six participants) and three interviews (centre 7: two interviews; centre 6: one interview) were conducted with staff members. Interviews were conducted with stakeholders from all six centres (eight participants).

Acceptability and feasibility of the Stand Up for Health intervention: pre-lockdown results

Acceptability and feasibility of the Stand Up for Health intervention: post-lockdown results

Feasibility of study design and procedures: pre-lockdown results

Feasibility of study design and procedures: post-lockdown results

Strengths

The process evaluation used a mixed-methods approach, utilising both qualitative data, obtained through focus groups and interviews, and quantitative data, which were also analysed as part of the feasibility outcome evaluation. The breadth of objective and subjective data collected, utilising the RE-AIM framework, allowed us to address the aims of the process evaluation robustly and rigorously. A range of views and experiences of SUH delivery, implementation and feasibility of study procedures were collected. This included detailed feedback and suggestions for improving participant recruitment, as well as intervention delivery and engagement, which will be useful for a future full-scale trial.

Using a multicomponent approach in the context of the socioecological model provided a strong foundation for the programme and allowed for ease of programme activity idea generation and implementation. Introducing intervention activities at different levels meant that there was a suitable activity for everyone and led to the recognition that changes at multiple levels supported sustained behaviour and cultural change within the workplace. The adaptive, non-prescriptive nature of the intervention meant that activities and initiatives could be tailored and implemented within the unique context of each contact centre.

The mixed-methods approach to data collection and analysis, as well as the multicomponent approach to the intervention, allowed the research team to address the research questions and continue rolling out the intervention despite the substantial limitations on in-person contact owing to the COVID-19 pandemic, which forced the UK into lockdown in March 2020. Interviews, focus groups and questionnaire data were still collected post lockdown and, as described above, the intervention was adapted to allow for online delivery.

Limitations

Time was a persistent challenge for data collection, workshops, meetings and SUH activity participation. Each of these elements involved staff and/or stakeholder time, which was often scarce and centred around productivity and business needs. Staff found it difficult to get time away from the telephones (or sometimes other tasks) to attend workshops, meetings with the researchers and data collection sessions, as well as to participate in activities during the work day. Future trials should focus on scheduling communication, data collection and other time-intensive elements at the most convenient times for the centre. Encouraging centres to think long term and understand that change takes time, especially at the organisational level, should be another focus. This shift in thinking might offset the view of some centres that they were too busy to participate in certain elements of the intervention and allow them to consider SUH activities in the context of supporting employee well-being and, therefore, improving business in the longer term (e.g. by reducing sickness absence or increasing productivity).

Owing to the intervention’s flexible, multicomponent approach, and in the light of the COVID-19 pandemic, the SUH team were able to adapt the intervention for fully online delivery through offering consults and providing support and ideas for activities that were safe to do during lockdown. The post-lockdown results outline the success of these efforts. Many participants felt that the team did all they could given the difficult circumstances, and participation in activities helped to mitigate some of the negative mental and physical impacts of the pandemic. However, the inability to conduct workshops, host meetings and collect data in person and within the contact centres themselves limited the team’s ability to introduce the intervention and build rapport with stakeholders and staff. It also limited what data collection and recruitment measures could be used, and what programme activities could be carried out owing to home working and social distancing. Although feedback was received from a few centres that participated in the pre-lockdown programme, results related to the use of equipment and other office-based activities are limited. Some centres dropped out during the pandemic, which lessened the pool of staff who were able to participate. These circumstances reduced activity and data collection participation rates, which suggests that the overall impact of the programme, and the evidence to support this, could have been more significant had the team continued with the pre-lockdown project plan.

Key recommendations for programme adaptation

• Set more realistic expectations for the speed of changes to behaviour and implementation of activities and changes at the various levels.

• Ensure appropriate placement and labelling of, and communication about, equipment for staff.

• Incorporate or suggest more competition-type activities for staff within and between teams and organisations, and using the website to showcase this.

• Share results and benefits of the programme, especially at the individual level, to motivate staff.

• Have a higher amount and variety of communication and advertising measures to create awareness of the programme and check in with staff. This advertising and communication should include information relating to the risks of high rates of sedentary behaviour and low levels of physical activity to emphasise the programme’s importance and increase motivation to participate.

• Visit the centres more often and at different time points to maximise in-person contact with staff (if possible).

• Encourage active participation in programme activities from stakeholders and others at the various management levels.

Key recommendations for study evaluation procedures

• The use of a cluster-randomised stepped-wedge design meant that some centres were randomised to a much later date, which had an impact on their ability to take part. If this study design is used, ensure communication and engagement with all centres during the time gap between expressing interest and participation.

• Have more and a greater variety of recruitment measures to create awareness of the evaluation.

• Encourage active participation in recruitment from stakeholders and others at the various management levels using tailored, individualised communication.

• Work more closely with stakeholders to effectively and flexibly work around staff schedules to maximise opportunities for evaluation participation.

• Emphasise to participants the importance of completing follow-up measures.

• If in-person visits are safe (as with the pre-lockdown evaluation), provide more detailed information on activPAL device risks and how to fill out logbooks.

Study design

The original stepped-wedge study design is shown in Figure 5. Clusters were sites (call centres). The original study design included five sequences (A, B, C, D and E) of at least two clusters each. One of the sequences (A) contained three clusters because there were 11 sites in total. Each ‘step’ in the stepped-wedge design consisted of 3 months, with a total of five periods (i.e. data collection time points). The participants assessed in different periods were expected to be mostly the same people (although we did not prevent new individuals being recruited and assessed at the different time points). For this reason, this stepped-wedge study has an ‘open cohort’ design.

It was planned for data to be collected at a maximum of three visits per site: at the end of the control period and at 3 and 6 months after the end of the intervention. This limited number of data collection visits was in place to minimise cost and participant burden. Participants were also given £5 vouchers each time they participated in data collection to increase response rate.

However, not all data collection visits were completed because of some centres dropping out, and we were unable to continue with the intervention and collect in-person data beyond 20 March 2020 because of the COVID-19 pandemic. Nevertheless, even before the pandemic began in the UK, we experienced difficulties with the stepped-wedge design. In particular, we experienced a difficulty in maintaining contact centre interest and it proved difficult to ensure that data collection took place on schedule. There were also issues with collecting device-measured data from all participants in all centres using accelerometers (e.g. activPAL devices).

The actual stepped-wedge design for the pre-lockdown phase was as shown in Figure 5.

For the pre-lockdown questionnaire analysis, there were six centres involved that belonged to three different randomised sequences (A, B and C). These sequences are described as follows:

• sequence A – started the intervention in July 2019 and had data collected at the 6-month follow-up (no within-site control group comparison) (site numbers 2 and 11 in Figure 5)

• sequence B – started the intervention in October 2019 with baseline data collected only (sites 3 and 10 in Figure 5)

• sequence C – started the intervention in January 2020 with baseline data collected only (sites 1 and 6 in Figure 5).

The concurrent cross-sectional comparison involves comparison between sequence A and sequence C only, for which data were collected at a similar time (approximately January 2020). Sites in sequence B had baseline (control) data collection 3 months previously in October 2019.

Owing to the COVID-19 pandemic and associated lockdown restrictions, we changed the design of our trial after March 2020 to include a simple before-and-after comparison. In the post-lockdown phase, questionnaire data were collected in participants at baseline and 3 months later (see Figure 5).

There were five centres involved in the post-lockdown data analysis. The aim was to collect questionnaire data in participants at baseline and at the 3-month follow-up. However, although a total of 51 participants completed at least some information at baseline or follow-up, only 25 participants provided questionnaire outcome responses at both baseline and the 3-month follow-up.

Intervention and control conditions

The intervention is an adaptive theory-based complex intervention called SUH (full details are provided in Chapter 1). The intervention was delivered once in each phase (pre lockdown and post lockdown) at the level of the contact centre, with data collected at both the contact centre and individual levels. Participants recruited were employees of the contact centres, including managers, supervisors and call operators in Scotland and England (primarily in the north of England).

Under the control condition, no component of the SUH intervention was delivered.

Outcomes

The primary outcome in a definitive trial was the activPAL device-measured sedentary time in the workplace, measured over a continuous 7-day period. This will be referred to as the primary outcome for the remainder of this report. 67

Secondary outcomes included:

• subjectively measured sedentary time in the workplace, as measured by the Occupational Sitting and Physical Activity Questionnaire (OSPAQ)81

• activPAL device-measured () prolonged sitting time in the workplace (bouts of ≥ 30 minutes)

• activPAL device-measured total sedentary time (i.e. including time outside the workplace, such as at home and leisure time)

• activPAL device-measured workplace and total standing time

• activPAL device-measured workplace and total physical activity (based on stepping)

• activPAL device-measured workplace and total sit-to-stand transitions

• Utrecht Work Engagement Scale (UWES)82

• mental well-being, as measured by the Warwick–Edinburgh Mental Well-being Scale (WEMWBS)59

• musculoskeletal health, as measured by the Musculoskeletal Health Questionnaire (MSK-HQ)83

• Scottish Physical Activity Screening Question (PASQ)84

• activities questionnaires to measure use of activities and preference

• staff turnover – number of people leaving and number of new joiners over the study period.

Participant sampling and recruitment

Data collection and outcome measures

Given that this is a feasibility study, we tested out the methods of collecting data on outcomes, as well as preliminary estimates of effectiveness. Table 10 outlines the methods/instruments used.

We also collected data on demographics, such as age, sex and any medical conditions, that may impact on sedentary behaviour.

Data collection instrumentation and procedures

Progression criteria

In the original study protocol,67 we reported that we could proceed to a future larger study if the following criteria were satisfied:

• The 95% CIs for the primary outcome include a clinically relevant reduction in sedentary time of 45 minutes per day or greater in favour of the intervention. This would reflect substantial progression towards accumulating the recommended quantity of 2 hours per day standing/light activity during working hours55 for employees in predominantly desk-based occupations.

• Intervention successfully delivered in at least five of the sites within the study period, if at least one person in each site was able to use/experience at least one activity.

• At least 10% of employees at a site being able to use/experience at least one of the intervention activities in the sites at which the intervention was successfully delivered.

• Primary and secondary outcome data being collected for at least 75% of participants overall.

• Contamination between sites being low or else it being envisaged that contamination can be addressed in the study design of a future study.

• It being feasible that any practical difficulties in delivering the intervention across multiple sites or in measuring effectiveness can be overcome when conducting a future large-scale study.

Introduction

The statistician was blind to the allocation of the contact centres. The analysis and presentation of results adhered to Consolidated Standards of Reporting Trials (CONSORT) guidance,90 and a statistical analysis plan. 67 A stepped-wedge version of the CONSORT checklist was added as a supplementary file (see Report Supplementary Material 9). The analyses were all exploratory to investigate if it is feasible to fit such a model in a future trial.

The quantitative analyses were stratified according to whether data were collected before or after the UK lockdown measures were introduced because of the COVID-19 pandemic. In what follows, we distinguish between pre-lockdown analyses/results and post-lockdown analyses/results. The ‘post-lockdown’ phase refers to any data collected after the lockdown measures began in March 2020, whereas ‘pre lockdown’ refers to data collected prior to March 2020. Our study design and pre-specified statistical analysis plan has already been described in detail in an open access publication. 67 However, we had to make changes to the planned statistical analysis plan owing to the impact of the COVID-19 pandemic.

A total of 155 individuals were recruited pre lockdown and 70 individuals were recruited post lockdown for the feasibility outcome evaluation.

The CONSORT flow diagram

A CONSORT flow diagram is shown for the pre-lockdown phase in Figures 6 and 7, and for the post-lockdown phase in Figure 8.

FIGURE 6.

The CONSORT flow diagram (pre-lockdown phase).

FIGURE 7.

The CONSORT flow diagram: non-participation and missing primary outcome data (pre-lockdown phase).

FIGURE 8.

The CONSORT flow diagram (post-lockdown phase).

Any missing values for the questionnaire data analysis at baseline were because of incomplete responses. Not everyone who completed the questionnaires used the activPAL devices. For the primary outcome analysis (based on the activPAL data), reasons for missing data included failure to complete logbooks (recording time spent at work and sleeping), incomplete logbooks or failure to extract valid data from the system (perhaps because of device faults or the way that the activPAL device was attached) (see Figure 7).

Baseline demographics

Participant demographic details (age, sex, job title, employment type, length of time working at the contact centre, any health problems reported) were summarised using descriptive statistics. Table 12 shows the baseline demographic information for participants recruited in the pre-lockdown data collection period, split by sequence (see Report Supplementary Material 10, Table S10.1, for the same table but split by contact centre). In total, 154 out of 155 participants recorded at least some demographic information pre lockdown, whereas this information was complete for 150 participants.

Sites in sequence A (intervention sites) had slightly older participants, on average, with a higher proportion of females and people working in the company for more than 3 years than sequence C (concurrent control sites). None of the participants had previously worked for a company that used the SUH intervention.

There were five centres involved in the post-lockdown data analysis in a before-and-after study design. Questionnaire data were collected from participants at baseline and at the 3-month follow-up.

Participant demographic details (age, sex, job title, employment type, length of time working at the contact centre, any health problems reported) were summarised using descriptive statistics split by contact centre. Although 51 participants completed at least some information at baseline or follow-up, only 25 provided questionnaire outcome responses at both baseline and the 3-month follow-up. Therefore, the baseline descriptive statistics were also computed for the subset of 25 participants providing outcome responses at both baseline and the 3-month follow-up (henceforth called the ‘adherent participants’). Table 13 shows baseline data for all adherent participants involved in post-lockdown data collection (n = 25). Baseline data for all participants recording baseline data in the post-lockdown phase (n = 51) can be found in Report Supplementary Material 10 (see Table S10.2).

Descriptive statistics of outcome data

The device-measured activPAL outcome data and continuous participant self-reported outcomes recorded at baseline or the 6-month follow-up were summarised using descriptive statistics [mean, median, standard deviation (SD), minimum, maximum, interquartile range] and box plots split by sequence.

The purpose of this analysis was to help answer research question 6:

• What are the preliminary estimates of the variability of primary and secondary outcomes within and between contact centres?

Table 14 shows descriptive statistics for the primary outcome (device-measured sedentary time per day in the workplace) and all secondary outcomes for the pre-lockdown analysis, stratified by sequence. The anonymised data sets for the secondary outcomes (questionnaire data) collected pre and post lockdown have been uploaded into a University of Edinburgh repository. 91

For the adherent participants in the post-lockdown phase, descriptive statistics (mean, median, SD, minimum, maximum and interquartile range) were computed for the subjective continuous questionnaire outcomes split by time point (baseline or 3-month follow-up) for comparison (Table 15).

Simple statistical and graphical analyses of outcome data

Pre-lockdown phase

Figures 9 and 10 show box plots of the primary outcome (sedentary time per day in the workplace measured by the activPAL device), stratified by centre and sequence, respectively, for the pre-lockdown phase.

FIGURE 9.

Box plot of primary outcome (sedentary time in the workplace, as measured by the activPAL device), stratified by centre, in the pre-lockdown phase.

FIGURE 10.

Box plot of primary outcome (sedentary time in the workplace as measured by the activPAL device), stratified by sequence, in the pre-lockdown phase.

In the pre-lockdown phase, the number of participants (%) meeting physical activity guidelines was 30 (60%), 41 (69%) and 26 (62%) in sequences A, B and C, respectively. Overall, 97 (64%) participants met physical activity guidelines. The proportion meeting physical activity guidelines in each centre is shown in Table 16. The percentage meeting physical activity guidelines per centre ranged from 57% to 79%.

TABLE 16 - Percentage meeting physical activity guidelines in each centre

Centre	n (%)
1	13 (61.9)
2	16 (57.1)
3	15 (57.7)
6	13 (61.9)
10	26 (78.8)
11	14 (63.6)
Total	97 (64.2)

Box plots for the subjective secondary outcomes are shown in Report Supplementary Material 10 (see Tables S10.3–12).

Post-lockdown phase

Regarding the post-lockdown phase analysis, crude paired t-tests of each outcome were computed to compare baseline with follow-up (Table 17). Mean differences and 95% CIs were reported. Only participants responding at both baseline and the 3-month follow-up (n = 25) were included in the descriptive analysis and paired t-test.

TABLE 17 - Crude paired t-test results (post lockdown)

Outcome	n	Mean difference (3-month follow-up – baseline)	95% CI	p-value
Hours sitting at work per week	25	0.47	–2.77 to 3.71	0.768
Minutes sitting at work per working day	23	13.62	–20.65 to 47.90	0.418
WEMWBS total score	25	–0.40	–3.18 to 2.39	0.771
UWES mean score	23	0	–0.37 to 0.37	1.00
MSK-HQ total score	23	–2.09	–4.83 to 0.65	0.129

For the binary outcome of ‘meeting physical activity guidelines’ according to the Scottish PASQ, a cross-tabulation of the number of participants meeting physical activity guidelines at baseline and follow-up was constructed (Table 18).

TABLE 18 - Cross-tabulation of baseline against the 3-month follow-up for the numbers meeting physical activity guidelines

		3-month follow-up (n)		Total (n)
		No	Yes
Baseline	No	3	3	6
	Yes	4	15	19
Total		7	18	25

A crude McNemar’s test for paired data was performed as an exploratory test to determine if there was any statistically significant difference in the proportion of participants meeting physical activity guidelines between baseline and follow-up. This McNemar’s test produced an exact p-value of 1.00, indicating no evidence of a difference between baseline and the 3-month follow-up. There was no change for 18 out of 25 (72%) participants.

A scatterplot of the number of hours sitting at work for baseline compared with the 3-month follow-up is shown in Figure 11. The solid line represents the line of equality. Similar plots are shown for the other outcomes in Report Supplementary Material 10, Figures S10.13–16.

FIGURE 11.

Scatterplot showing number of hours sitting at work at baseline and the 3-month follow-up.

Mixed-effects model analysis

Calculating estimates of variability

The assumptions underlying the sample size calculation that was carried out by a similar study73 were tested and compared with the estimates that we calculated using real data from the feasibility trial (e.g. SD, within-site correlation and intracluster correlation coefficient) to inform the power calculation for a future trial. The heterogeneity of quantitative outcomes across sites (e.g. estimated via intracluster correlation coefficients) and the variability in intervention delivery and processes across sites was of particular interest. Our aim was to determine the feasibility of conducting a future multisite study with a larger number of contact centres.

The purpose of this analysis was to help answer research questions 2 and 6:

• Research question 2 – how many clusters and participants per cluster are required for a confirmatory trial?

• Research question 6 – what are the preliminary estimates of the variability of primary (reduction of sedentary behaviour in the workplace) and secondary outcomes within and between contact centres?

Using the pre-lockdown data, we calculated indicative approximate estimates of empirical between-centre SD, within-centre SD and coefficient of variation of means between clusters (κ) to inform sample size calculations for future cluster randomised trials. 93 This analysis was carried out for the subgroup of centres recording information at baseline (pre intervention), as well as overall (across all centres). These estimates were computed to identify objective and subjective outcomes, with the lowest or highest levels of between-centre variation to help identify likely candidates for the primary and secondary outcomes of a future study, and also to inform sample size calculations for future cluster randomised trials. Empirical estimates of SD and coefficients of variation were computed for the pre-lockdown analysis (Table 21). High values of kappa indicated high levels of between-centre variation. The results suggest that device-measured total sedentary time exhibits potentially low levels of between-centre variation.

Attrition and staff turnover

Evaluation of relevant research questions based on study findings

1. Is the study design (cluster RCT) feasible for a confirmatory trial of an intervention to reduce sedentary behaviour in staff working in contact centres?

   We found that a stepped-wedge design would not be suitable for a future larger-scale trial. We experienced difficulties in maintaining site interest, which would be exacerbated in a trial of longer duration (e.g. if some sites were randomised to receive the intervention 12 months later), and it was challenging to ensure that each site adhered to the scheduled data collection time points. In addition, 22% of participants moved jobs after 10 to 12 months. Instead of a stepped-wedge trial design, we suggest that a parallel-group cluster-randomised trial, in which sites are recruited in pairs (or groups) over time, would be much more practical.

2. How many clusters and participants per cluster are required for a confirmatory trial?

   This research question can be answered using the data presented in Table 21. For example, for the primary outcome (total sedentary time in the workplace per day in minutes), the overall coefficient of variation was 0.12 minutes, and the within-site SD was 100.08 minutes overall and 96.04 minutes at baseline. In addition, the mean value of the primary outcome at baseline was 359.6 minutes and we recruited a mean of 15.6 participants providing valid primary outcome data per cluster (see Figures 6 and 8). Therefore, using equation 6 from Hayes and Bennett,93 we calculate that we would need 25 clusters in a future cluster-randomised trial to detect a clinically relevant difference in sedentary time of 45 minutes with 90% power and two-sided 5% significance level. We suggest that a future cluster trial, with clusters recruited in pairs, would have a low cluster-level dropout rate of 10% because in this future trial, design clusters would not have to wait long to begin the intervention. We, therefore, suggest that a future trial should aim to recruit 28 clusters, taking centre-level dropout rates into account. We also recommend that a future trial aims to recruit 40 participants per centre, on average, to use the activPAL devices and to have at least 15 or 16 participants per centre recording valid primary outcome data.

3. What is the recruitment rate of participants in each cluster and how many are lost to follow-up (e.g. owing to staff turnover)?

   As shown in Figure 6, the CONSORT flow diagram (pre-lockdown phase), there was a mean of 26 participants per cluster, of whom 16 per cluster provided valid primary outcome data and 25 per cluster provided valid secondary outcome data (i.e. data from the OSPAQ), on average. Considering specific outcomes relating to attrition, of the four centres invited to participate in the post-lockdown data collection (centres 3, 6, 10 and 11; n = 106), 13 out of 106 staff were recorded as having left the company after the pre-lockdown data collection (12%, 95% CI 7% to 20%). In total, 23 out of 106 staff were recorded as having moved jobs after the pre-lockdown data collection (22%, 95% CI 14% to 31%).

4. What are the preliminary estimates of the variability of primary (reduction of sedentary behaviour in the workplace) and secondary outcomes within and between contact centres?

   Estimates of variability for the primary and secondary outcomes are presented in Table 21. The OSPAQ showed very substantial variability in our study (SD at baseline of 7.2 hours of sitting), although the primary outcome SD at baseline was lower (1.6 hours).

Rationale and guiding principles

Typical cost-efficiency analyses used in health technology assessment rely on estimating impacts on ‘hard’ health outcomes and associated improvements in morbidity, mortality and health and social care service utilisation. 91,93,94 Although it may be plausible that increased fitness could in the long term lead to such improvements, the trial follow-up period was insufficient to reasonably result in measurable `differences. Direct surveying of individual participants was also not possible given that only centre-level data were recorded. Hence, health economic assessment was instead designed to scope the feasibility of future trials. Of additional interest was identifying outcomes (including costs) from an employer’s perspective that could be important at an implementation stage to leverage commercial support and scoping potential methods of measurement, should the programme demonstrate benefits. Economic perspectives for analyses, therefore, comprised a mix of an NHS and PSS perspective [as per the standard National Institute for Health and Care Excellence (NICE) Reference Case98] and employer’s perspective assessments.

In March 2020, national lockdown restrictions because of the COVID-19 pandemic changed the nature of the study and the nature of office working. It was not possible to know what future office environments would look like relative to before or during the lockdown period. Hence, working environments may be considered on a spectrum between pre-lockdown office environments and full home-working office environments. Many of the analyses were, therefore, split into results before and after lockdown, representing the two possible extremes. In principle, should the world move to some form of long-term hybrid working or different offices adopt different rates of hybrid working, each may reasonably be considered to be somewhere on a spectrum between these two points, although it is recognised that such assumptions are necessarily speculative and post hoc. A CHEERS (Consolidated Health Economic Evaluation Reporting Standards) checklist highlighting key principals (where relevant) and where they are located in this report has been included in Report Supplementary Material 12.

Qualitative survey of prospective outcome measures from an employers’ perspective

A questionnaire was developed with aid from, and administered by, qualitative researchers in the team (see Report Supplementary Material 11). This aimed to identify employers’ priorities for outcomes of the programme that might drive decisions to adopt or form criteria for judging success and, therefore, continuation of a programme in a hypothetical future rollout. The questionnaire included free-format questioning regarding the metrics they use:

• to measure productivity, absenteeism, presenteeism, staff turnover or other metrics of importance to them

• how and how frequently these were measured

• whether or not they attribute a financial cost to these routinely and, if so, how

• whether or not they benchmark these measures between teams within the site, against other sites, other organisations, industry standards and/or other groups

• what arrangements were in place to assess ergonomics and mental health

• what measures would be of use to measure in a future follow-up trial and which were of highest priority

• would they, hypothetically, be willing and able to share any of their internally collected metrics and, if so, what form of data might be possible.

Surveys were issued after the trial observation period. Notably, in all cases, survey responses decreased post lockdown and resulting changes in office working patterns. As such, respondents may reasonably be assumed to have attempted to take these into account in their replies. However, the specifics of these would not have yet been established and, therefore, any adjustments were necessarily speculative.

Completed forms were analysed using descriptive statistics and reported narratively. Qualitative (free-form text) data from the five returned forms were summarised and general themes were reported. We did not use a more rigorous analysis (such as thematic analysis) owing to the small pool of data and the short length of the individual answers. The themes and key points identified were then incorporated into the model-scoping exercise (component 3) to examine whether or not there may be any potential to build in any identified outcomes, such as absenteeism/productivity measures, alongside more traditional health-related cost-efficiency outcomes.

Activity-based costing

A taxonomy of activities involved in programme provision was generated by the research staff providing the service. This included equipment, SUH staff time, call centre workforce and management staff time, travel costs, and stationery and sundries. In accordance with standard ABC principles,95,97,98 staff were asked to record items that may reasonably be expected to occur in a hypothetical mass rollout of the programme only, that is excluding activities specific to research, such as interviewing or data collection. This was initially planned to be measured prospectively; however, before this could be undertaken, lockdowns were initiated, which prevented direct measurement. Instead, a breakdown of activity was generated from records and invoices, with durations of workshops based on professional judgement and experience of what would be needed in practice.

An opportunistic post hoc analysis repeated the exercise for workshops held during lockdown. This was undertaken to investigate how changes to remote working might influence the intervention and costable activities, potentially reflecting two different options for programme provision and testing the feasibility of ABC measurement in a future trial.

We initially planned to record details of devices purchased by centres on completion of the SUH programme. In practice, sites initially borrowed devices from a central pool of equipment held by the SUH programme, but adoption/purchase of their own devices became a moot point with the need to prioritise home-working arrangements as a result of lockdowns. The feasibility of attributing costs to equipment use was, therefore, based on SUH programme costs alone.

The primary focus of assessment was to estimate the rates of various forms of activity rather than their implicit costs in a hypothetical rollout, given that there is a great deal of heterogeneity in the components of price weights between sites. Examples include differences in rates of facilities available, differences in wage or other labour cost elements for staff at different organisations, and scope to have different types of staff providing different programme implementation activities, were the intervention implemented at scale. The time horizon considered cost/activities on a per-annum, per-site basis.

Finally, it should be noted that these results should be considered indicative and not definitive owing to small samples and low power.

Model-scoping exercise

Health economic decision analytic modelling methods frequently rely on grouping individuals into common pooled ‘health states’ or events of broadly similar impact on health-related quality of life and health-care resource implications. Models typically include a clinically based core that simulates patient transitions between health states and/or critical health events. A set of health and cost outcomes are then attached to each health state/event to convert these into economic outcomes of interest, usually accumulated over a prespecified time horizon. Both elements are often based on established probabilities or risk equations and synthesise a wide range of evidence. 93,97

To scope the potential for modelling of both of these aspects of the model, two principal steps were undertaken:

1. Potential health states or groups of health states were identified – experts within the team were asked to describe potential long-term outcomes that the intervention might reasonably aim to improve if successful.

2. A non-systematic search of the literature was undertaken to look for potential similar models; literature linking surrogate outcomes, such as secondary behaviour, to these target health states; and potential studies attributing cost, quality of life [as measured using quality-adjusted life-years (QALYs)] or outcomes identified by qualitative consultation with sites as desirable to employers to these health states.

Consideration was then given to how well these data might fit the available options for collection alongside a future trial.

Implications for future health economic evaluations

This section considers the implications of these results with respect to onsite data collection within a hypothetical future trial. Implications with respect to potential modelling are discussed separately in Model-scoping exercise. Respondents measured productivity with a variety of metrics. Although many of these revolved around call-handling time and quality, the lack of primary specific common metrics between sites may make this challenging to measure quantitatively, at least directly. It is, therefore, important to first consider its components individually.

Absenteeism was commonly measured in terms of some form of percentage full-time equivalent (FTE) lost, whether this was converted from days or hours, probably reflecting the relative simplicity of the concept. Those not stating this directly at least implied it to be in use within the organisation, for example via a central department. A common method of quantifying this (when stated at all) was to multiply percentage FTE lost by wage. This method is consistent with the methodological and theoretical framework available in a human capital approach from the economic literature. 92 Thus, a simple theoretical and methodological framework for future analysis is at least available. 94 There were no indications of anything akin to, or usable with, the alternative friction costing approach,95 probably because of its relative complexity in calculation. Such an approach may require synthesis with external data or bespoke surveying; however, a specific method to do so beyond standard measures was not identified.

Few sites were able to provide data on staff turnover directly. Several did indicate that staff turnover was recorded elsewhere within their organisation, and it was implicit that these data were recorded in a consistent manner, although recoded reasons for turnover differed. It is possible that a future trial may be able to obtain rates of total turnover by contacting the appropriate department or collecting them prospectively. With differences between organisations driving per-trainee cost implications and sites themselves not able to provide known estimates of cost per trainee, reporting simple rates of turnover or change in percentage turnover may be sufficient information for prospective organisations considering a SUH rollout. This would enable them to interpret this in their own context, rather than consolidating this to an abstracted financial figure.

Although sites recognised that presenteeism was an important factor in productivity, they also recognised that it was challenging to measure, and few collected any metrics or converted it to a cost. In economic analyses, this often carries an inherently subjective element of perceived self-reported percentage loss in productivity. This component both carries risk of being perceived as less credible than days/hours absent and requires individual-level surveying, which was not possible in this style of study. By contrast, ‘harder’ outcomes, such as days/hours absent, are recorded centrally. Several sites also considered staff engagement to be important, but similarly did not offer a means to convert this to cost. There was an implicit understanding that engagement has the potential to lead to improved productivity and staff turnover savings, even if the specific mechanism(s) involved may not be easy to disentangle. Sites also indicated desirability for engagement improvements beyond a purely financial/productivity level. It is unclear how much weight this would be given in a rollout decision, although it remains an option for measurement where possible in its own right, irrespective of purpose.

First, we look at the productivity angle of presenteeism and engagement collectively. Given their acceptance of the challenges of recording metrics to enable conversion to cost, organisations may be open to lower grades of evidence to inform their decisions, should direct measurement prove impractical. Options could include:

1. Allowing sites to record productivity using their metric(s) of preference, which could be reported in terms of percentage change.

2. Measuring perceived productivity change in categorical format, such as a Likert scale-based questionnaire issued to management.

3. Use of some form of discrete choice experimental design. 97 Non-productivity/financial changes in engagement might then be recorded separately using bespoke metrics, possibly within the same questionnaires to management if used.

For those sites willing, as well as for sites unwilling or unable to share internal data, particularly around absenteeism for which metrics are likely to be more comparable between sites, discussions should be held with sites at recruitment and disclosure agreements put in place where needed.

Cardiovascular

The methodology for economic modelling in cardiovascular health is well established compared with many disease areas. Therefore, many pre-existing models provide a ready list of candidate health states and events for modelling, with associated cost and QALY weightings derived from well-established sources and/or literature review (Table 27).

At the most complex end of the modelling spectrum, transitions between long-run rates are modelled based on sets of validated risk equations derived from large population data sets, such as Framingham Risk Score(s),96 the QRESEARCH Cardiovascular Risk Algorithm (QRISK) series (which has current NICE recommendation)98 or similar studies. 101 Although such models and equations are published in the public domain to enable their reuse or the reuse of their subcomponents in future modelling, a number of the input parameters for the risk equations underpinning the models require biomarker data, preferably at the individual level. For example, the QRISK3 series includes age, sex, ethnicity, deprivation, smoking status, diabetes status, family history of angina/heart attack, presence of stages 3–5 chronic kidney disease, atrial fibrillation, several variables on current blood pressure treatment and comorbidities, cholesterol/high-density lipoprotein ratio, blood pressure, and height and weight to calculate BMI.

It is possible in some cases to use these equations to estimate the effect of a change in one component of risk holding all others equal if reasonable assumed/mean population values can be found for the other parameters.

However, the challenge at conception of the SUH trial was that several key aspects of these parameters, such as blood pressure, cholesterol or weight, are challenging to record in a trial without direct measurement, which would add an impractical need for physical examination that may deter some participating centres and participants. Hence, such an approach would necessarily rely on more tertiary inputs with smaller effect sizes. Additional barriers to this method include staff turnover and the sharing of equipment, which are not easily measurable. Model scoping for this area, therefore, focused on identifying alternative options for modelling the identified health states that were compatible with the restricted data collection options.

Older modelling has been based around dividing participants into mutually exclusive binary groups of active versus sedentary, and then applying differential relative risks to the development of health conditions. 99 Although this approach is, in principle, simple to apply with a literature review, a more granular breakdown of degree of activity would be desirable where possible.

A recent systematic review by Lutz et al. 102 of economic evaluations of interventions to increase workplace physical activity and decrease sedentary behaviour found 18 studies, including one (also recent) fully modelled study by Gao et al. 100 which may provide a pragmatic approach. 100 Gao et al. 100 provide a Markov model of different workplace interventions to reduce standing time. Differences in energy expenditure were estimated for each intervention based on meta-analysis of changes in standing time, converted via metabolic equivalent units, a measure of estimating energy expenditure derived from body weight and increase in physical activity/standing time. Population-level rates of coronary heart disease and/or stroke, and associated outcomes, were then adjusted for relative risks of associated with estimated improvements in energy expenditure. A similar approach may be possible utilising metabolic equivalent unit-based on activity accelerometer data. Although the relative risks reported may not all be applicable to the specific SUH style of intervention or accelerometer data, the same method could be applied, potentially using a new or updated meta-analysis. Owing to the acknowledged impracticality of physical examination(s), some assumptions around the use of population-level weight/BMI may still be necessary, but this has the advantage of focusing on the most directly modified aspect of cardiovascular risk of a SUH-style programme. The model also includes Australian health-care perspective costs and QALY waiting sets for the health states, although UK equivalent remain available for substitution (see Table 27) or could similarly be derived by fresh/top-up review.

Although, in principle, significant cardiovascular events would reasonably have long-term implications for absenteeism, presenteeism, engagement and staff turnover (the primary areas of interest identified from an employer’s perspective from the consultation exercise detailed in Qualitative survey of prospective outcome measures from an employers’ perspective), modelling these may be challenging. The health states and events identified are low frequency, long-term outcomes. For this reason, meaningful differences may be observable at a population level only, and a given employer is unlikely to observe modifiable differences in their targeted metrics. It is not implausible that there could be behavioural changes in staff turnover and engagement when staff are aware of the impacts of sedentary behaviour on long-term health and make conscious efforts to avoid it. However, these may be better examined within trial using the methods discussed in Qualitative survey of prospective outcome measures from an employers’ perspective above.

Musculoskeletal

To identify candidate health states and other parameter data to populate a model, physical activity experts within the team were asked for details of any known literature relating to musculoskeletal disorders caused or worsened by office working that were plausibly modifiable by a SUH-style programme. Specific focus was given to matters that might reasonably alter productivity, absenteeism, health-care resource use and/or quality of life and, thus, justify categorising as a ‘health state’ in a simulation model. Further ad hoc literature searches were carried out for studies of economic burden, cost of illness, cost–efficiency/–utility and/or long-term modelling of musculoskeletal disorders in a work/office environment.

Potential modifiable health states identified two main categories of outcomes commonly applied by both fields of literature to group disorders:

• back pain – often divided into upper or lower103–105

• upper limb disorders, such as neck or shoulder issues. 103–105

There were also plausible links to their development being in part because of combinations of ergonomics, prolonged sitting and computer use. 103 Should a programme be able to demonstrate an ability to modify the appropriate risk factors and appropriate other parameter data available, this provides a potentially plausible mechanism to model long-run musculoskeletal disorders health state trajectories. Alternatively, one model of long-term musculoskeletal disorder absence grouped employees into health states denoting 6-monthly increments of absence rather than medical cause, and simulated cost-efficiency of different workplace interventions based on differences in rates of return to work reported in the literature. 106

In general, attempts to locate usable economic literature highlighted some challenges. The terms back or upper limb pain can be used to refer to a range of heterogeneous symptoms104 and searches can be confounded by a number of other conditions, including work-related pain as a result of non-office activities, such as manual labour and associated arthritis. 105 Crucially, patients’ treatment for a different form of pain may follow a different profile of health-care resource use or QALYs and, at least within the economic literature, there was no obvious common lexicon used to categorise degrees of severity of symptom. Hence, in addition to an increased degree of complexity needed in literature searching, synthesising data from multiple sources into common health states may be more challenging than in cardiovascular health and it may be advisable to aim to build health states around results from a singular study, in which we can be more confident that health-care resource use and health-utility weightings reported are attributable to the same patient group.

Cost-of-illness surveys, which have the potential to offer broader categories of care depending on study design, wielded few UK studies, with a recent systematic review107 of methodology of cost of illness work for back pain identifying only three UK studies. Two of these were over 20 years old and potentially not reflective of current care108 (including the primary health-care resource use source for the long-term absence model106,108) and the third was from 2013, but available in abstract only. 109

Several musculoskeletal disorder trials including cost–utility components were found, including recent UK-based trials in which health-care resource use are likely to be most relevant and/or trials using UK-based health-utility tariffs from the EuroQol-5 Dimensions (EQ-5D) or Short Form questionnaire-6 Dimensions tool groups. 110 Those with UK health-care resource use did provide detailed breakdowns of care, the most promising of which provides EQ-5D utilities and health-care resource use profiles for groups of patients defined using the Roland Morrison Disability Questionnaire111 as being at low, medium or high risk of developing long-term disability. 110

Therefore, although on face value there appears to be a reasonable evidence base to populate a potential model, the need to synthesise evidence from a heterogeneous literature base may provide challenges, and the specific model structure may be best selected in part to fit the available parameters rather than derived top down, as was attempted here. Such an approach may be better started with a systematic review, which was beyond the scope of this study.

Time horizon considerations

A key assumption for modelling is that any modelled benefits are sustained beyond the observed duration. Gao et al. 100 apply an assumed 5 years based on previous modelling in the wider area of public health100,112 and note an absence of evidence of decaying effects;100 however, to our knowledge, this assumption has not been validated and an absence of evidence cannot rule it out. 100,112 Similarly, any modelling of a SUH-style programme would need to take into account staff turnover, as those who change jobs would not have a continuation of SUH unless by chance a similar programme was in place in their next employment. A simple solution to this would be for a model to instead solve for the minimum duration of continued benefits needed for the intervention to be cost-effective.

Value-for-money considerations

Although potential options for modelling cardiovascular health and musculoskeletal disorders were identified, both would rely on some form of systematic, or at least in-depth, review to identify the most suitable parameters. Such reviews are often time-consuming and costly aspects of research, and need to demonstrate likelihood of value for money. Any intervention that increases costs needs to offset these with sufficient effect size if it is to be considered not only cost-effective but also effective.

Although the trial was never powered to be definitive, it was hoped that at least a broad indicator of the direction and degree of activity improvement might be achieved at this stage, around which at least subjective judgement of the potential value of modelling could have been speculated. Unfortunately, COVID-19 disruptions prevented this. With broader literature on programmes to improve office-based physical activities having mixed results, depending greatly on the nuance of the specific intervention method,102,104 an argument can be made that such modelling should be deferred until a programme has successfully demonstrated an element of effectiveness to minimise potentially sunk research expense.

Aim 1: to test the acceptability and feasibility of implementing the Stand Up for Health intervention in contact centres

Research questions for aim 1

What is the acceptability, feasibility, and utilisation of the various components of the intervention in a range of contact centres?

For the pre-lockdown programme, process evaluation findings showed that the staff members found the activities at the individual, social and environmental levels acceptable, and at least 50% of all contact centre staff took part in at least one activity. This high level of engagement is probably because of the adaptative nature of SUH, and its focus on ownership and co-production. The adaptive nature of SUH meant that preferences and contexts were considered when developing the activities, and further activities were developed as SUH became more embedded in the culture and systems as staff and management started to see positive changes to their working lives. These were not always related specifically to changes in sedentary behaviour but more of a healthier workplace culture and environment. The barriers to delivery in some centres stemmed from lack of communication between SUH team and centres, as well as between individuals and teams within centres, not setting up a SUH committee, and lack of strong buy-in at the organisational level.

Does the programme theory and process of implementing the intervention work as intended?

The initial draft of the SUH programme theory was based on data from the pilot centre (see Figure 2). Although the outcome evaluation did not yield adequate evidence, the pre-lockdown process evaluation results indicated that several aspects of the programme theory were supported. Pre-lockdown process evaluation findings showed that, for most centres, the SUH programme triggered changes at each of the levels (organisational, environmental, social, individual, ownership and information). Figure 12 highlights the aspects of the programme theory model that were supported by process evaluation data. In some centres, the mechanisms of change were not activated and the reasons are discussed in detail in Chapter 3. Lack of communication (between the SUH team and the centres, as well as between individuals and teams within centres), not creating a SUH committee and organisational rigidity in some centres (centres 2 and 6) had a negative effect on ownership and awareness about the programme, sedentary behaviour and physical activity among staff. These factors also impacted individual, environmental and social levels and overall programme success. Two new medium-term outcomes were identified: increased physical activity outside work and the adoption of a healthy diet.

FIGURE 12.

Refining programme theory: pre lockdown. Note that the figure does not include outcome data as data collection was incomplete and did not provide evidence to accept or refute hypotheses.

During the post-lockdown period, several contact centre staff members were working from home. Although the initial programme theory model and activities were not designed for home working, the programme was adapted to suit the new working pattern. However, the post-lockdown process evaluation results showed that the mechanisms of change at all levels of the programme theory model were activated to a lesser degree than pre lockdown (Figure 13). In particular, post lockdown, the SUH programme did not produce a strong change with respect to organisational, social and environmental levels, or create a sense of ownership among staff. An improved understanding of activities that trigger changes at these levels is required to set the mechanisms of change in motion for staff working from home. Adopting a healthy diet was identified as an unintended consequence and a medium-term outcome for the post-lockdown programme also.

FIGURE 13.

Refining programme theory: post lockdown.

Does the programme theory/intervention need adapting and if so, in what ways?

Although the pre-COVID programme theory worked as intended, the SUH programme could not achieve change at some levels during the post-COVID programme. More work is required to understand the hybrid working models and refine the programme theory in accordance with these. Working with contact centres during the COVID-19 pandemic provided an opportunity, and we are now at the forefront of thinking about these new models in the context of workplace health interventions.

Are there differences in delivery of the intervention between different contact centres? If so, what are the reasons for these?

There were differences found in the delivery of the intervention between contact centres. Some identified reasons were communication, the existence (or not) of a SUH committee and organisational support. Effective communication and co-ordination between the SUH team and the centres, as well as between individuals within centres, aided implementation. The SUH committee created a sense of ownership among staff and supported implementation activities. Centres that had these elements saw more successful implementation. Organisational support was also a key element to successful implementation, and rigid organisational structures and bureaucracy were posed as barriers to implementation.

Can Stand Up for Health be adapted for those working remotely, and what is the acceptability and feasibility of this mode of delivery?

The intervention was successfully adapted for online delivery by replacing the in-person workshops with online consults and conducting the step count challenge to address the social level. Several resources for individual activities were also provided, including mindfulness and other online resources to reduce sedentary behaviour and increase physical activity. The process evaluation findings from the post-COVID programme showed that staff members found the activities at the individual and social levels acceptable and feasible.

The data indicate that SUH can be adapted successfully for hybrid working. However, more activities are required to trigger the environmental and organisational levels of change and maximise participant engagement. A greater understanding of working from home is required to adapt the programme theory for hybrid working.

Aim 2: to assess the feasibility of using a cluster randomised controlled trial study design

Is the study design (cluster RCT) feasible for a confirmatory trial of an intervention to reduce sedentary behaviour in staff working in contact centres?

Aim 3: to scope the feasibility of a future health economic evaluation of Stand Up for Health

Aim 4: consider previous aims under the context of COVID-19

The post-COVID-19 programme has been discussed in detail within aims 1–3 above. The post-COVID-19 period was challenging for the contact centres, many of whom were struggling to set up new systems of working while being busier than usual. Despite this, several activities were implemented and were found to be acceptable within certain levels of change (such as the step count challenge and consults). However, it was challenging for the post-COVID-19 programme to create significant changes at the environmental and social levels. For this reason, further work is required to explore how to adapt the programme to ensure that it is suitable for organisations with hybrid workers. It is possible that our contact centres have tried more than ever to support well-being (or at least staff engagement) but that taking part in a formal research trial was too much of a commitment or their focus was on something else (e.g. mental health or more basic staff engagement).

Strengths

The intervention was co-produced with significant input from contact centres and their staff from the outset. It was developed with them and for them and used the 6SQuID intervention development framework. It was piloted in one contact centre, and the intervention has been sustained in that centre for over 3 years now. To enable the intervention to be adaptive and respond to both the opportunities and the constraints within individual contact centres, we stressed fidelity to the theory of change rather than the theory of action. This is a significant departure from most interventions, which stress fidelity to a particular activity and can specify how often, who by and how much of an activity is needed. From initial conversations with the contact centre, it was obvious that this approach would not work to make it sustainable and acceptable over the longer term. Instead, we worked with them to develop activities and approaches that were adaptive and could be implemented to fit in with the system constraints, such as budget, shift patterns, the working environment and the contact centre culture. We believe that creating an adaptive intervention that contact centres could feel proud of and take ownership of was key to successful implementation in some of the contact centres.

In terms of the research design, we used a mixed-methods approach to assess the feasibility and acceptability as well as the potential impact of the intervention. Rigorous methods were used for data collection and analysis of this feasibility trial, including device-based measures to assess sedentary behaviour and self-report measures for other outcomes. Methods to inform a future cost-effectiveness analysis were also identified.

The process evaluation aimed to explore the feasibility and acceptability of the intervention and trial methods as well as to obtain feedback on and suggestions for improving the intervention and adapting the programme theory. The qualitative methods were robust and the large number of interviews and focus groups gave extensive, in-depth accounts of the experience of the intervention. Further strengths of the trial were that 3 out of 4 key progression criteria were achieved (at least partially), recruitment rates were adequate, retention was good and engagement with the intervention was acceptable.

Limitations

A key limitation was the study design, which we would not consider for a larger evaluation study. The stepped-wedge design, although pragmatic and cost-effective, presented challenges in maintaining engagement with the centres and their staff, and ensuring that the intervention start times followed the complex, pre-planned timeline. Owing to the long study duration, there was also an increased risk of contamination and dropout.

Limitations of data collection included the use of activPAL devices and organising focus groups for the process evaluation. The use of activPAL devices was not logistically feasible, particularly when in-person visits to the contact centres stopped during the pandemic. Although focus groups provided rich insight and generated good discussions around intervention and data collection experiences, it was difficult for the research team and the centres to arrange them as only so many staff could be off the telephones at once. This was also an issue for implementation of the intervention.

Limitations of the intervention included the time that it took for implementation and the communication and promotion of the intervention and specific intervention activities. Implementation of the intervention involved running two workshops with each centre (open to all staff), as well as various meetings with stakeholders and with the SUH committee. These implementation activities required substantial staff time, which was often scarce in productivity-driven contact centre environments, in which a certain number of staff needed to be on the telephones at all times. It was challenging for staff to partake in non-desk-based SUH activities because their shifts were quite rigid and they had little flexibility for taking time away from their desks. Communication about, and promotion of, the intervention took an unexpected amount of time and co-ordination between the researchers and the centres, and also within the centres themselves. For some centres, it was challenging to ensure that communication was received by all levels and to effectively promote the intervention, especially during the COVID-19 pandemic.

If SUH is to be evaluated in a larger study, a more traditional trial study design and more flexible approaches to intervention implementation and data collection would be used, taking into account the specific needs and context of each participating contact centre. The intervention implementation and data collection activities that require interaction between the researchers and the centre staff will be offered at various times of day, more frequently and for longer chunks of time so that they are more accessible for staff who may be busy. The researchers will work with each centre to determine what dates and timings work best for them. Interviews may be used instead of focus groups so that only one staff member is removed from the telephones at once. In addition, more attention will be paid to the SUH website in attempting to engage staff in intervention participation.

Centre 5 exit questionnaire summary

There were two main reasons for dropping out:

1. Pandemic.

   The pandemic was the main factor for dropping out. The centre was very comfortable with accommodating the key elements of the programme before the pandemic but felt that the safety and security of colleagues and customers’ needs were the priority during the pandemic. It felt that it would not get enough participants for the programme during the pandemic owing to its priorities and workload. The centre had moved to a hybrid model of home and office working, and felt that it would not be able to participate fully in the programme given that on-site activities were no longer possible.

2. In-house initiatives for staff well-being.

   The centre had already organised several initiatives to support the mental and physical well-being of staff (e.g. online bingo, step challenge, mental health first aid training, support for homeworkers, reduced meeting times, health and well-being course for managers, other tools and communication).

Centre 8 exit questionnaire summary

There were three main reasons for dropping out:

1. Found project requirements difficult and expensive and had concerns about reduction in sales revenue.

   The centre felt that, as an outsourced sales centre, it was not able to invest in projects that may reduce sales revenue. The centre found it difficult to commit to (1) allocating 27 employees for research evaluation (despite SUH offering to do this in batches), as it meant over 3 hours away from the telephone, (2) interviews with participants and (3) retention of participants, as it felt that this would be challenging.

2. Changes to circumstances since registering interest.

   The centre had registered interest in April 2019 but was allocated (randomly) to start only in March 2020. Although the centre had a consistent workforce and positive levels of recruitment when it registered, it faced recruitment challenges in March 2020 and needed to carefully consider the time allocated to external projects.

3. Inadequate communication between registering interest and actually starting.

   The centre felt that the SUH team did not provide sufficient communication between April 2019 (first registered) and February 2020. If the centre had known earlier, it could have planned for the project to start in March 2020. The lack of communication and allocation to a later start date also made the centre feel that it was not prioritised and not essential.