Reducing work-related stress among health professionals by using a training-based intervention programme for leaders in a cluster randomised controlled trial

To randomly assign participating health organisations into the three arms (IG, CG, NOIG), a ranking of work-related stress was conducted using the results from the baseline measure (T0). Therefore, the mean values of the COPSOQ dimensions demands at work, work organisation and content, social relations and leadership and home–work interface were used for the ranking (see Supplement A for dimensions). Organisations with the highest extent of stressors at work were ranked top down. This was conducted separately for (1) acute care hospitals/rehabilitation hospitals, (2) psychiatric hospitals, (3) nursing homes and (4) home care organisations. Considering the participation rate of employees for each organisation, a minimum of 1500 participants was targeted to achieve the needed sample size separately for acute care / rehabilitation hospitals, psychiatric hospitals, nursing homes and home care organisations. In this process, the organisations were randomly assigned (top down – starting with organisations with the highest extent of stressors) into the intervention and control groups until the 1500 participants were reached using computer-based randomisation (randomizer.org). The person who performed the randomised allocation was blinded and not a member of the research team of this study. All remaining organisations (with a lower extent of work stressors) were assigned to the non-intervention group. This has the advantage that the intervention is mainly applied to those organisations with a higher extent of stressors and, therefore, are in a higher need for strategies to reduce work-related stress.
Figure 2 shows the organisations in- and excluded during the steps of enrolment, allocation, follow-up and analysis.

CONSORT flow diagram of the STRAIN study.
Despite randomisation, a group comparison between IG and CG of the T0 baseline measurement was carried out (using Mann–Whitney U-test, Bonferroni corrected) for the primary and secondary outcomes and participants’ sex. Significant differences between the IG and CG were found in acute care / rehabilitation hospitals for quantitative demands (p < 0.00), opportunities for development (p < 0.001), bond with the organisation (p < 0.01) and role clarity (p < 0.01). Significant differences between the IG and CG in psychiatric hospitals were found for the scale on bond with the organisation (p < 0.05) and for nursing homes for the scale on difficulties with demarcation (p < 0.05). Regarding home care organisations, significant differences between the IG and CG were found for opportunities for development (p < 0.001), quality of leadership (p < 0.01), rewards (p < 0.001), difficulties with demarcation (p < 0.01), the effort–reward ratio (p < 0.001) and participants’ sex (p < 0.05).
Development of the STRAIN intervention
The intervention programme was developed specifically for healthcare leaders at all hierarchical levels, as they have an essential role in implementing effective measures for a healthy work environment21,22,48. The programme was developed as evidence-based in accordance with the intervention mapping (IM) approach49. The approach is suitable for designing and developing complex workplace interventions and programmes for different settings and stakeholders as it describes the path from problem identification to problem solving or reduction. The intervention mapping approach consists of six steps and leads towards evidence-based programme development, implementation and evaluation49. Step 1 focuses on the problem and consists of a needs assessment; steps 2, 3 and 4 involve the initial development of the intervention programme; Step 5 consists of the implementation planning; and Step 6 involves evaluation and refinement of the programme (Fig. 3). Following those steps helps to addresses questions about how and when to use theory, empirical findings from the literature and data collected from a population. Its purpose is to create an intervention that leads to an effective behaviour or system change49.

Six steps of the intervention mapping approach (Bartholomew Eldredge et al., 2016)37.
A planning group of 10 people was installed and several workshops were held to develop the intervention programme. The planning group consisted of various researchers and health professionals (nurses, physiotherapists, physicians, occupational psychologists) from the German-, French- and Italian-speaking parts of Switzerland, most with a professional background in different work areas in the Swiss healthcare system and/or in a leadership position.
To develop an evidence-based intervention programme, various data sources were used in order to identify salient topics and contents (see Fig. 4). From the STRAIN baseline data set with 8,112 participating health professionals, several regression models were calculated for individual health disciplines, settings and outcomes11. The aim was to identify topics in which leaders have the greatest potential for reducing work stressors. In addition to these quantitative data, a total of 25 focus group interviews with various health professionals working in different settings of the Swiss healthcare sector were conducted to collect their ideas and recommendations on how to improve their working environment and reduce stressors at work50. Additionally, an extensive literature research was conducted in order to identify published international literature on effective prevention and intervention strategies to reduce stress at work. Therefore, 1400 studies (reviews, meta-analyses, intervention studies) and available guidelines were summarised and analysed as to their recommendations. Those three data sources (quantitative, qualitative, literature) served in developing the study intervention that focuses on the health professional leaders working in the lower-, middle- and upper-management levels.

Data sources used to develop the STRAIN intervention programme.
The intervention was standardised and structured in two days of training and a half a day of coaching. For the content of the programme, the following key topics were identified:
(1) Reducing stressors in the upper-, middle- and lower-management levels,
(2) Enhance the compatibility of work and private life,
(3) Match requirements at work with skills and resources,
(4) Optimise leadership understanding and structures,
(5) Strengthen organisational commitment (reward, feedback, opportunities for development and salary),
(6) Create clear role profiles and competency-based deployment; and.
(7) Promote intra- and interprofessional communication and collaboration.
The key topics (1–3) were addressed on day 1 and the key topics (4–7) on day 2. The sequences for the individual key topics were developed individually based on the defined determinants, the corresponding objectives and methods (according to the intervention mapping approach49). At the beginning of the sequence, we often used methods to increase the ‘awareness and risk perception’ (e.g. consciousness raising, framing or personalised risk – using organisation-specific results from the baseline measure e.g. regarding the results of non-compliance with break and rest times within the organisation) to raise the awareness among leaders as to why the topic is important and what benefits a positive change could have. In addition, we used various methods to change attitudes, beliefs and outcome expectations (e.g. environmental re-evaluation, shifting perspective, cultural similarity), methods to change social influence (e.g. mobilising social support), methods to change skills, capability, self-efficacy and to overcome barriers (e.g. self-monitoring of behaviour, reattribution training, goal setting, planning copings responses) and to change environmental conditions, social norms and organisation (e.g. addressing team building and human relations training, enhancing network linkages, resistance to social pressure).
The intervention programme was developed in German and translated into French and Italian. The whole programme was primarily tested in one acute care hospital of the non-intervention group of the study (pilot).
STRAIN intervention programme
The study intervention included a two-day training programme and a 2–3-hour group-based coaching for health professional leaders (nurses, midwives, physicians, medical-technical and medical-therapeutic professionals) across organisations (e.g. leaders from different hospitals) and in multi-professional groups. The training programme included short presentations, group works and individual in-depth work (e.g. based on their own organisation-specific results from the baseline measure). In addition to presentations on current research results and measures to reduce workload, the face-to-face training sessions included many interactive parts to promote interprofessional and cross-organisational exchange between the leaders. This was followed by an additional 2–3 h of group coaching. During the intervention programme, health professional leaders were separated regarding their management level (upper- and middle-management levels in one group and lower-management level in another) to avoid the influence of organisational hierarchies in the intervention groups. The programme was also conducted for acute care / rehabilitation hospitals, psychiatric hospitals, nursing homes and home care organisations separately and took place between T1 and T2 (June 2019 until December 2019). The intervention programme was conducted in all language regions in Switzerland (German, French, Italian). After the programme, all leaders of the intervention group received an additional guideline including all the results / strategies to reduce stress at work presented in the intervention programme as a checklist (see Supplement B (German Version), Supplement C (French Version), Supplement D (Italian Version).
Data analysis
Data was analysed using SPSS 25® and R. All Items from the COPSOQ, EWCS, NEXT and Von Korff were transformed to having a value range from 0 (minimum value) to 100 points (maximum value), according to the original authors46,51. If fewer than half of the questions in a scale had been answered, no average score was calculated51. Further, the index for WAI (index scores from 7 to 49) and the effort–reward imbalance ratio (imbalance of high effort and low reward if effort–reward ratio > 1) were calculated according the original authors’ method42,44.
To test for significant differences between the IG and CG, an analysis of covariance (ANCOVA) using as outcomes the scales on effort-reward imbalance, quantitative demands, opportunities for development, bond with the organisation, quality of leadership, social community at work, role clarity, perceived rewards, difficulties with demarcation, work–private life conflict and intention to leave the profession at T2, and as predictors the same scales at T1 and a dummy variable for the treatment group assignment. This ANCOVA essentially models the (inflation-adjusted) difference of the outcomes between T1 and T2 and notably allows the extraction of the amount of the outcome increase from T1 to T2, which is more pronounced for the treatment group (time-by-group interaction). For analyses on the individual level of participants, we included random effects for the organisations and included only participants who participated at T1 and T2.
The analysis was carried out in two steps. In step one, the treatment effect was estimated, including all participating healthcare organisations on both the aggregated level of organisations and the individual level of participants. For this overall comparison of the IG and CG, an Intention-to-treat (ITT) as well as an As-treated (AT) analysis was conducted since in 23 (out of 60) organisations of the IG, no health professional leader took part in the intervention programme. In step two, treatment effects were analysed separately for acute care/rehabilitation hospitals, psychiatric hospitals, nursing homes and home care organisations. Since only a small amount of the respondents participated in both T1 and T2 (< 10%), models on the individual level could not be reliably estimated. Therefore, for simplicity, we performed the data analysis in step two on the aggregated level of organisation and using as-treated analysis only.
In a further step, more in-depth pre-/post-test analyses for each organisation were conducted separately, also regarding various stressors, stress symptoms and long-term consequences and considering the response rate of leaders in the intervention programme of each management level, the number of participants per organisation as well as how they were affected by COVID during the second measurement. Therefore, a Bonferroni-corrected Mann–Whitney U-test was used.
In addition, the additional questionnaire (positive and negative impressions of the intervention programme) was descriptively analysed and the written statements were analysed to form categories and subcategories, which were then condensed thematically in a further step.
link