Process evaluation of a multilevel intervention targeting emotional demands in a hospital

Emotional demands at work can be defined as “the extent to which the job requires sustained emotional effort because of interactional contact with clients” (Vegchel et al., 2004, p. 22). Some professions, particularly within healthcare, often experience emotional demands at work due to the intensive, direct care provided to patients (Madsen et al., 2010; Framke et al., 2019). Emotional demands have been associated with an increased risk of burnout (Borritz et al., 2010) depression (Madsen et al., 2022; Mikkelsen et al., 2021) and sickness absence (Framke et al., 2021, 2023), and there is a need to identify interventions that may prevent the potential negative consequences of emotional demands for employee health and wellbeing. In health care work, it is difficult to separate emotional demands from quantitative demands. Quantitative demands in this setting typically means a higher patient load that inevitably equally will intensify the emotional demands (Sjöberg, 2025). High workload stemming from a global healthcare system under pressure from staff shortage, an aging population, and increased patient complexity will likely be difficult to manage through internal organizational measures (Darzi and Evans, 2016; King et al., 2021; Naik et al., 2024, Sjöberg). As a high workload is a well-documented barrier to implementing organizational interventions (Roodbari et al., 2022; Daniels et al., 2021; Paterson et al., 2024), it is crucial to investigate further how to implement preventive interventions in this context. Focusing on effective implementation mechanisms in this context could provide valuable knowledge for the health care sector.

Few studies focus specifically on emotional demands in the research literature on workplace interventions. However, a vast literature on interventions to prevent burnout suggests that multilevel interventions may be effective (Aust et al., 2023). While multilevel interventions are promising, there is a need to couple this knowledge with implementation studies of organizations within this challenging context.

Following recommendations from the British Medical Council (Skivington et al., 2021) for evaluating complex interventions, this article forms part of a study using a mixed-methods study design to evaluate a multilevel intervention addressing emotional demands. We apply an observational design with a comparison group for the effect evaluation. The present paper reports the results of the process evaluation. Drawing on qualitative data from interviews and observations, we investigate (1) Key factors for the implementation of a multilevel intervention targeting emotional demands at an entire hospital and (2) How intervention activities affected social support, knowledge, action possibilities and prevention and management of emotional demands and traumatizing events.

We conducted a process evaluation drawing on concepts from realist evaluation (Bonell et al., 2024). In line with recommendations from the British Medical Council (Skivington et al., 2021) for evaluating complex interventions, the core element of the process evaluation was a program theory, and a continuous focus on context. To explore the context of the intervention, we applied the conceptualization from Greenhalgh and colleagues (Greenhalgh and Manzano, 2022, p. 590): “Contexts are not just things or people (material and social) but psychological, organizational, economic, technical and other relationships (forces) that interact and influence each other.”. Further, we integrated an assessment of implementation degree (fidelity, reach and dose) into the program theory, to be able to apply a comparative analytical approach to key factors for variation in implementation across cases. This comparative approach was inspired by recent developments in implementation science (Whitaker et al., 2020) and evaluation within organizational psychology (von Thiele Schwartz et al., 2024). We distinguished between implementation mechanisms as “the processes through which [implementation-] strategies exert their effects on outcomes” (Lewis et al., 2024, P. 2) and intervention mechanisms, as those that lead to the outcome of the intervention (Roodbari et al., 2022; Jaspers and Aust, 2025). Although specific implementation strategies were not designed for this intervention, we still judged it relevant to separate key factors for successful implementation and the impact of such implementation.

The study was conducted in Denmark, where all citizens have universal health coverage, and most hospitals are public. The intervention hospital was located in a rural area of Denmark with a relatively high percentage of senior citizens and citizens from a lower socio-economic status. It was a small emergency hospital receiving patients with acute injuries. The remote location posed challenges to the recruitment and retention of medical professionals. The demographic profile of the inhabitants tended to require long-term, resource-intensive care with high levels of emotional demands.

The research team was solely evaluators. The intervention was developed collaboratively between the internal OSH advisor of the hospital and trained occupational psychologists from an external consultancy before the accompanying research was funded. Its content aligned with literature recommendations (described above) and followed the IGLO framework, targeting individual, group, leadership, and organizational levels (Nielsen and Christensen, 2021). Thus, we considered the entire hospital to be the intervention unit. The external consultants facilitated the main part of the intervention activities. The individual-level activities included psychological first-aid training for all staff. The group-level activities consisted of a digital course on emotional demands and traumatic events, with department-level group discussions, as well as de-escalation training and training in debriefing for frontline staff and first-line managers. The leadership-level activities included two workshop days for line and middle managers focused on preventing sickness absence due to emotional strain. At the organizational level, each department established a “TRIO” group—a well-established cooperation structure in the Danish labor market consisting of a line manager, a union representative, and a work environment representative. These TRIO groups were trained to facilitate group discussions for the digital course in the group-level intervention.

Action plans were developed in a participatory manner at each IGLO level. Additionally, 15 change champions (employees and managers) were selected, trained, and supervised across positions and professions. Their role was to support colleagues on site if, e.g. they needed help to facilitate the department-level group discussions. For an overview of planned intervention activities, see Figure 1: Initial program theory of the intervention.

Public funding supporting sickness absence prevention in public sector workplaces covered the costs of the intervention consultants. However, this funding did not cover the working hours of staff involved in the intervention activities. The planned duration of the intervention was approximately one and a half years.

In collaboration with the internal OSH advisor of the hospital and external consultants, we developed an initial program theory to form the basis of the evaluation (Figure 1). The context of the intervention was that the hospital faced high levels of sickness absence, turnover, and recruitment problems, which had been a driving factor in initiating the intervention. Both the hospital management and the consultants considered that these problems could be ascribed to the remote location, high workload, and emotional strain caused by high emotional demands at work. While location and patient uptake were factors that were out of the decision authority of the hospital, emotional demands and emotional strain were potentially alterable. The hospital already conducted regular risk assessments in line with OSH regulations. The aim was, therefore, to improve the prevention and management of emotional demands in the workplace to ultimately reduce absenteeism, turnover, and health expenses for employees and patients, thereby conserving their limited resources more effectively. Theoretically, introducing resources on many levels could turn the resource loss cycle as proposed in the conservation of resources (COR) theory (Hobfoll et al., 2018). These assumptions are illustrated in Figure 1.

A key assumption of the intervention was that increased shared knowledge and reflection on emotional strain would reduce the stigmatization of expressing emotions at work, thereby increasing the collective capacity to prevent and manage emotional demands. Consequently, the primary goal was to create collective awareness, language, learning, and increased capacity for action at the individual, group, management, and organizational levels (IGLO) regarding the emotionally demanding aspects of the work. Based on the theory of psychological safety (Edmondson and Lei, 2014), that is the shared belief that it’s safe to take interpersonal risks, de-stigmatizing of showing one's reactions to emotional demands and increased knowledge and possibilities to cope with emotional demands were some of the key mechanisms proposed. Gaining more knowledge and reflecting on the group members own typical reactions to emotional strain, should de-stigmatization showing one's reactions to emotional demands because it would increase psychological safety (Edmondson and Lei, 2014). In a group where it is possible to show your reaction to emotional stain and you have a common language for these reactions, it should be easier to manage the emotional demands.

We specifically developed assumptions on implementation mechanisms, as well-known barriers to implementation characterized the context. We hypothesized that top-management support would be essential for implementation and that relating strain to the core task/patient safety would facilitate implementation.

The research project was initiated at the suggestion of the intervention consultants, who approached the research team to propose conducting a research-based evaluation of their intervention. The hospital board agreed to participate in an externally funded evaluation. For data collection, we recruited participants for interviews conducted before and after the intervention period across five of the hospital's 31 departments. These departments were selected through stratified purposeful sampling based on the hospital's internal OHS advisors' evaluation of their readiness for change, the perceived relevance of the intervention, and the estimated level of emotional strain at work (Palinkas et al., 2015). The internal OSH advisor helped to facilitate contact with the targeted roles from the selected departments. From each department, we interviewed managers and two to three employees (e.g. physicians, nurses, social care workers, service assistants, physiotherapists, and occupational therapists).

Additionally, we interviewed the board of management, the internal OSH advisor, the change champions, and external intervention consultants, totaling 39 interviews. Participants were all approached by e-mail. SJ contacted some directly, and some were recruited through their managers. As all interviews were conducted within participants' working hours, initial contact was always established with the manager. In after-interviews, the target was to interview the same persons from the pre-intervention interviews, however this was only possible for two informants due to the high rate of turnover. Instead, managers were asked to point to employees working at the hospital during the entire intervention period. Few could accommodate this request due to high turnover, a bad fit with roostering, and vacant positions. In one department, we could not contact a physician who had been interviewed before the intervention, and the manager could not find a replacement. To guide the recruitment process and sample size, we chose not to use saturation as a criterion but rather “information power” as defined by Malterud et al. (2016). In certain instances, specific roles were unavailable due to understaffing, and consequently, we supplemented with interviews from another department, as shown in Table 1.

We collected data from several sources. Primary data sources were individual semi-structured interviews (39 in total) with managers (14), employees (17), the board of management (2), the internal OHS adviser (2), change champions (1), and intervention consultants (3). Some job groups were difficult to reach and to align the interviews; we conducted two interviews in a sixth department. We conducted 68 h of observation of intervention activities, which SJ, AJ, and EA carried out. We observed one session of each type of planned intervention activity (see Table 2) along with selected additional activities. We collected attendance registrations for all planned intervention activities as an indicator of reach, dose and fidelity to estimate the degree of implementation (Saunders et al., 2005). We obtained data on participation rates for 28 out of 35 planned intervention activities. Supplementary data consisted of notes from phone calls with the hospital project manager (the internal OSH advisor), intervention materials, and e-mail exchanges about contextual factors.

Most interviews were conducted in person at the hospital and lasted 30–60 min. A few interviews were conducted by phone or virtually, but all in an uninterrupted setting with only the participant and the researcher(s) present. Interview guides were structured systematically to assess all factors in the initial program theory and were inspired by realist interviewing. Pre-intervention interview guides included questions on existing prevention practices and participants' expectations for the intervention. Post-intervention interview guides focused on participants' perceptions of their work environment during and after the intervention and changes in related preventive practices. SJ conducted most of the interviews (35) and observations (62 h) and received supervision from a licensed psychologist with experience in traumatology to minimize any risk of re-traumatization among participants. AJ, EA, and MM supported the data collection. None of the researchers had prior relations with the informants. All interviewees and observed individuals were informed about the study's purpose and terms of participation. For interviews interviewees were informed both through a written information distributed either by email or a physical print as well as verbally at the beginning of the interviews. The observer at the beginning of the activity informed observed individuals attending intervention activities verbally. All interviews were audio recorded and transcribed verbatim; a participant did not wish to be recorded in one instance, and field notes were taken instead.

Based on the initial program theory, we developed a coding tree covering all factors except distal outcomes. We used thematic analysis (Braun and Clarke, 2006) and principles from comparative case studies (George and Bennett, 2005) and coincidence analysis (Whitaker et al., 2020). Initial coding was conducted individually by authors SJ, AJ, EA, and MM. Subsequently, factors were condensed or their names adjusted by SJ. Finally, all the adjusted factors were rated through discussion among SJ, AJ, EA, and MM until agreement was reached on their low, medium, or high degree presence in the five cases. We developed scoring criteria for each factor, e.g. a Low degree of staff stability was assigned when informants consistently reported a low degree of staff stability (sickness absence, unfilled positions, turnover) and medium when informants reported mixed high and low degrees or consistently across informants a medium degree of staff stability. Physicians stood out through the rating discussions for the five departments, evaluating the intervention activities as less effective than the rest of the care personnel in the departments. Consequently, we conducted an additional sub-analysis rating physicians separately. We used Nvivo12 for data management and coding. The analysis was conducted blinded to the results of the quantitative effect evaluation.

In Denmark, the scientific ethics committee only requires approval to collect biological data, not qualitative data. Studies based only on registry data in Denmark do not require participant consent or ethical approval from The Danish National Committee on Health Research Ethics. The study has been notified to and approved by the Danish Data Protection Agency under the joint notification of the National Research Centre for the Working Environment (approval no. 2015-57-0074). For further details, refer to the Danish Research Ethics Committee site: https://researchethics.dk/information-for-researchers/overview-of-mandatory-reporting, specifically the subsection: “What Should Not Be Reported.”. The hospital's deputy director provided written consent to publish the evaluation results in a collaboration agreement before data collection.

The results are presented in three parts. First, we present the five purposely sampled departments (cases). Then, we report the extent to which the intervention was implemented, the changes in preventive practices experienced, and the mechanisms driving these changes. Finally, we synthesize these results to present a revised program theory of the intervention.

The five cases were from different departments across the hospital: anesthesiology, geriatric, emergency, service, and physiotherapy. The five departments had different work tasks. The anesthesiology department was centered around anesthesia before, during, and after surgeries, and patient contact mainly occurred when preparing for surgery. It was sampled as a department with low readiness for the intervention. The geriatric department focused on the elderly population, with typical diseases being delirium, dementia, and malnutrition, and was sampled as a department with higher emotional demands. The emergency department frequently treated acute injuries and urgent cases of illness and met patients in crisis. It was sampled as a department with high readiness for the intervention. Unlike the other three departments, the service, physiotherapy and occupational therapy departments were spread across the hospital. The service department was responsible for moving patients around the hospital, sitting watch if patients were at risk of self-harm or harming others, waste handling, food delivery, and cleaning, and was sampled as a department with higher emotional demands but a different organization of work. Lastly, the physiotherapy and occupational therapy department treated and rehabilitated ambulatory and hospitalized patients and was sampled as a department with lower emotional demands and a different work organization.

The intervention was implemented across the entire hospital and comprised interventions on all IGLO levels. Table 2 provides an overview of intervention activities. The shade of the bars indicates whether the activity was planned or added during the intervention. The intervention period was expanded substantially (from one and a half to three and a half years) due to overcrowding, unfilled positions throughout the period, and COVID-19, with shutdowns in winter 2021 and winter 2022.

The planned activities were all conducted, although some were delayed from the initial schedule. Activities had a high participation rate ranging from 57-100%, with an average of 85%. As indicated in the table, the additional activities were mainly on an organizational level, comprising an onboarding concept, developing locally tailored policies for handling traumatic events and violence and threats, and a coaching session for the management board focusing on emotional strain. Some of these activities were planned independently of the intervention, but the models and terms from the intervention were then integrated into the content by the hospitals' internal OHS advisor.

The observations of intervention activities also allowed for a qualitative assessment of fidelity to the intervention concept. In this extract, the external consultants ask what successes they have had so far:

Field note, management theme day, mid ways in the implementation period.

This illustrates how the participants themselves perceive the core idea of the intervention activities – to support de-stigmatization, and improve action possibilities in this case through psychological safety and participation.

To compare the development of mechanisms and outcomes across the five selected departments, we systematically rated all factors described in the program theory for each department. The rating of the factors is shown in Figure 2.

The overall pattern across departments showed that a challenging implementation environment characterized the context, as expected in the initial program theory. Specifically, informants highlighted instability in staffing, time pressure, and overcrowding as important factors shaping the implementation. These deviated slightly from the initial program theory, e.g. in that Covid-19 did not appear as significant a factor for the participants as expected. The implementation mechanism that was rated as present to the highest extent across cases was “relating emotional strain to the core task and patient safety” followed by “prioritization from the top management”. The rating of prioritization from the top management varied across the five departments. Although the hospital management showed strong prioritization because the intervention supported their strategic focus on employee recruitment and retention, some line managers felt that management prioritization lacked in assigning hours for watching the digital course and carry out the group reflection discussions. None of the interviewed line managers experienced an increased focus on managers' emotional strain during the intervention period and was rated as low for all cases.

Nevertheless, this medium to low degree of presence of the implementation mechanisms in the cases was enough to secure a high implementation degree across departments. Consequently, the interviewees reported improvements in intervention mechanisms. In the following sections, we unfold the intervention mechanisms of social support, increased knowledge, and the proximal outcome: “improved prevention”.

The first intervention mechanism increased knowledge, language, and possibilities of action was experienced by most of the interviewees across the five cases. The most widespread model that interviewees spontaneously recalled without priming was a concept for communicating about emotional strain. This model offered a common language for reactions to emotional strain. The model stipulated that when one was not exposed to emotional strain, one could have a balanced connection with the patients (called the green space). Whenever one was exposed to emotional strain, one could either distance oneself, becoming “cold” and factual, similar to the state of de-personalization (called the blue space), or become overinvolved and, e.g. get angry or cry in front of the patient (called the red space). In one department, the model had been embedded in their everyday practice, providing the possibility of early and sophisticated social support and redistribution of tasks:

-Employee, post-intervention

To note the subtle expressions of emotional strain was a pre-condition for adjusting the workload in a timely manner. While the model was designed to provide a language for relational work with the patients, informants widely used it to handle cross-disciplinary collaboration. This was evident in the case of the service center that employed domestic service staff that also held the task of hospital porters. They worked with all departments and often worked alone, coming into an established staff group in run-up situations. They used the common language to avoid conflicts with colleagues from other departments because it helped them to take their colleagues' perspectives. In this way, the model provided an attribution model where the negative attitudes or rude behaviors were always caused by strain. This explanation model had a de-escalating effect in assuming that colleagues were not rude on purpose, and it helped prevent conflicts in their interdisciplinary work.

The second mechanism, “social support - horizontal and vertical,” was experienced as improved by most interviewees across cases. In some departments, they worked with a concept of emotional support they called “putting a full stop to the shift”. The aim was to have a forum by the end of each shift to share emotionally demanding situations and team work-related issues, to address such issues during working hours so they did not take up mental space after work or “contaminate” the next shift with a stressful environment. One employee described how this practice strengthened their ability to ask for help and provide help to their colleagues during shifts:

-Employee, post-interview

The intervention consultants assisted this department in developing the “putting a full stop to the shift” as a tool to acknowledge vulnerability and strengthen social support collectively.

De-stigmatization was difficult to assess through the interviews. We asked informants if they knew about colleagues who hid their emotional reactions, and some answered that they knew that some colleagues did so, but it was hard to assess whether it improved through the intervention. The above-cited quotes about social support and common language indicate that it had become more legitimate to speak about colleagues' reactions to emotional strain and to ask for help. Data indicated that this might have been different for physicians. We therefore explored physicians separately, and as seen in Figure 2, physicians generally experienced fewer improvements.

Several physicians highlighted the hierarchical structures they worked under as a barrier to showing vulnerability and asking for help. This was, e.g. expressed during the digital course on emotional demands, where they were to discuss their reactions to emotional strain in groups afterward:

-Chief physician, post-intervention

The hierarchy was rooted in an educational structure and an approval system for physicians from other countries that depended on the approval of their Danish physician license from the hospital to stay in the country. This hierarchical context led to only few signs of de-stigmatizing amongst the physicians and may have been decisive in the lower extent of changes in preventive practices observed for physicians specifically (see Figure 2).

Due to the multi-stranded intervention approach, the changes reported across the five selected departments were diverse. Few departments succeeded in changing several preventive practices (cases 1, 4, and 5—see Figure 2), but all five departments experienced improvements in at least one preventive practice. In the following, we delineate some preventive measures that improved or were introduced due to the intervention.

One of the core intervention activities was to train and establish unit-level “TRIO” OHS groups to provide a forum for systematic collaboration on work environment issues. These groups were widely implemented, and their continuous work entailed different changes in preventive practices, some of which are described below. An example of its implementation is this OHS representative who was promoted to line manager during the intervention and explained:

-Line manager, post-intervention

In this hospital, the unit-level “TRIO” OHS groups were separate for physicians and other staff, and we have no data suggesting that they improved on the physician side.

The intervention aimed to stimulate changes in managers' primary prevention of employee sickness absence. The managerial practice, e.g. how well managers could detect early signs of emotional strain, only changed marginally according to managers and employees. This was likely due to the finding that managers did not experience an increased focus on managers own strain. However, the distinction between employee and management levels was blurry in collective initiatives. Where unit-level “TRIO” groups succeeded, the manager was inherently part of the new collective preventive practices. Also, in cases where the manager's ability to detect early signs of emotional strain in individual employees did improve, the line manager attributed the change to the improved collaboration with his employees. Employees became more aware of their colleagues and felt safe informing the manager about early signs of emotional strain. The line manager explains:

-Line manager, post-interview

Most departments introduced a tool for prioritizing work tasks to manage the strain induced by overcrowding and unfilled positions. Two tools were developed: one called a “traffic light”, estimating with colors (green, yellow, and red), the ward workload, and the individual employee's caseload during a shift. Some departments combined this approach with a second tool, “planned timeout” twice per shift, where they paired employees with a “green” caseload with employees with a “red” caseload to distribute tasks and minimize strain. Another version of this tool used unplanned timeouts to prioritize tasks when needed (especially in the emergency ward, where the caseload was unpredictable). One employee described the effects of these initiatives:

-Social care worker, post-intervention

There were differences in whether interviewees attributed the “traffic light” method to the intervention or as something that would have been developed in any case to handle the hospital's overcrowding. Data indicate however, that the prioritization initiatives would have been challenging to implement without improving de-stigmatization and social support, as a pre-requisite for distributing tasks was that it was legitimate and safe to ask each other for help.

Lastly, the intervention introduced training in verbal de-escalation through a train-the-trainer approach. These techniques were used to prevent patient conflicts, e.g. when surgery had to be postponed on the day. Some physicians also reported using de-escalation techniques. One physician, who had just learned Danish during the past year, explained that the training helped to structure and systematize the de-escalating communication:

-Chief physician, post-intervention

The de-escalation intervention was substantially delayed, and therefore, by the time of the follow-up interviews, only a few informants had had the chance to use it in practice.

The analysis resulted in a revised version of the program theory (see Figure 3).

The analysis showed that the intervention could be implemented in a resource-scarce environment. This implementation was likely due to a high degree of experienced top-management support and relating strain to the core task. The resources provided by the intervention (tailored concepts and process support) led to a change in participant reasoning and behavior around social support, knowledge, and perceived possibilities of action and probably de-stigmatization, although data is weaker for this factor. Interviewees across cases also described improvements in proximal outcomes. However, the specific changes differed given the multi-stranded intervention approach.

For managers, changes were not as pronounced, which may be related to missing activation of the implementation mechanism: “focus on manager strain”. For physicians, it seemed that a specific context of a more hierarchical structure blocked the implementation mechanisms, and consequently, fewer changes in practice were elicited for this group (see Figure 2). Few factors were omitted in the revised program theory, as data showed these factors were less important than those included, e.g. COVID-19 was less important than assumed because the hospital already experienced many acute patients and a high workload – irrespectively of COVID-19.

This study investigated key factors for implementation of a multilevel intervention to prevent and manage emotional demands. We found that in the context of staff instability, overcrowding, and staff groups with low power distance, implementation mechanisms of top-management support and a strong connection to patient safety supported a high degree of implementation. We found that the intervention activities led to participants experiencing improvements in social support, knowledge, action possibilities and prevention practices. Our findings align with previous reviews of implementation factors for organizational interventions to improve the psychosocial work environment in identifying top-management support through strategic prioritization of the intervention and alignment with the other organizational goal: patient safety as central to successful implementation (Roodbari et al., 2022; Daniels et al., 2021; Paterson et al., 2024; Yarker et al., 2022; Jaspers and Aust, 2025). The literature and this study suggest that these factors are pivotal to successfully implementing work environment interventions even in resource-scarce environments.

Surprisingly, we found that managers did not experience an increased focus on their own emotional strain, although several intervention activities were designed to support this. Earlier studies within occupational health and safety management have highlighted the focus on managers' strain as important for implementation success, but have not succeeded in improving it through intervention (Busch et al., 2017; Lundqvist et al., 2024, Justesen et al., 2027). In our study, the lack of experienced increased focus on managers strain may have influenced that managers reported no improvements in prevention practices specific to management and there is still a need to understand this issue better in future research.

The study's second aim was to investigate intervention mechanisms and potential changes in the prevention and management of emotional demands. We found that the intervention activities led to participants experiencing improvements in social support, knowledge, action possibilities and prevention practices.

Our results shed light on the process of breaking a vicious cycle of resource depletion in organizations in line with the conservation of resources (COR) theory (Hobfoll et al., 2018; Nielsen and Noblet, 2018). The theory states that organizations must invest resources to utilize new resources. Investing staff time in implementation of OSH initiatives can be viewed as such an investment. The study shows how investing in intervention activities improved resources such as social support, collaboration, and work organization. The results are in line with a recent study that identified profiles of organizational conditions for successful occupational health and safety management (Roczniewska et al., 2025). They found that the most prosperous foundations of successful OSH management consisted of a combination of supervisor provision of resources, presence of leadership support to invest in improving OSH, required budget allocations and knowledge and appropriate working methods. By assessing both implementation mechanisms and intervention mechanisms in this process evaluation, we were able to look at several of these conditions and how they were connected and interacted. Investigating configurations of key success factors (irrespectively of research tradition: implementation science, realist evaluation, process evaluation or other) has been promoted as a promising way for evaluating organizational or multilevel interventions where multiple mechanisms will interact and several implementation strategies will be at play simultaneously (von Thiele Schwartz et al., 2024). The results of this process evaluation suggest that our refined program theory can serve as a specification of how the COR theory applies in the specific area of implementation strategies for multilevel interventions to improve prevention and handling of emotional demands. The refined program theory suggest that by investing resources in terms of personnel time spent on intervention activities it is possible to improve organizational resources such as social support, collaboration, and work organization. This investment could be facilitated by top-management support and alignment with other organizational goals.

Regarding the transferability of the findings, some of the identified implementation and intervention mechanisms could apply to multilevel interventions to prevent emotional demands in other organizations where emotional demands are frequently reported, such as primary schools, elder care, or social workers.

A strength of this study was that we used several data sources to map implementation and improvements in preventive practices. The data from observations of intervention activities allowed tailoring the interview guides for post intervention interviews to efficiently track relevant changes in practice. This was important as the high level of turnover in the organization prevented us from recruiting the same participants for pre- and post-interviews. We used a stratified purposeful sampling to get variation in the five selected cases and allow for a comparative analysis. Results should therefore be relatively representative for departments with various characteristics. A limitation is the lack of solid measures for the hypothesized intermediary outcomes (e.g. quality of the core task, decreased work-related emotional strain, and decreased sickness absence caused by work-related emotional strain). This limitation might weaken the explanatory link between the process evaluation findings and the effect evaluation of distal outcomes.

We found that top-management support and a strong connection to patient safety could support a high degree of implementation. We also found that the intervention activities led to participants experiencing improvements in social support, knowledge, action possibilities and prevention practices. Our findings suggest that the intervention improves specific resources in the form of collaboration, efficient work organization, and social support. Further quantitative analyses are needed to establish the effects of the intervention on more distal outcomes.

The Danish Working Environment Research Fund funded the study, grant number 30-2021-04 20205100382. The fund only granted the funding and had no subsequent involvement in the project. The authors had no conflicts of interest. A special thank you to all the informants who shared their important experiences in a challenging work situation.