Whole-system approaches to improving the health and wellbeing of healthcare workers: A systematic review

Background Healthcare professionals throughout the developed world report higher levels of sickness absence, dissatisfaction, distress, and “burnout” at work than staff in other sectors. There is a growing call for the ‘triple aim’ of healthcare delivery (improving patient experience and outcomes and reducing costs; to include a fourth aim: improving healthcare staff experience of healthcare delivery. A systematic review commissioned by the United Kingdom’s (UK) Department of Health reviewed a large number of international healthy workplace interventions and recommended five whole-system changes to improve healthcare staff health and wellbeing: identification and response to local need, engagement of staff at all levels, and the involvement, visible leadership from, and up-skilling of, management and board-level staff. Objectives This systematic review aims to identify whole-system healthy workplace interventions in healthcare settings that incorporate (combinations of) these recommendations and determine whether they improve staff health and wellbeing. Methods A comprehensive and systematic search of medical, education, exercise science, and social science databases was undertaken. Studies were included if they reported the results of interventions that included all healthcare staff within a healthcare setting (e.g. whole hospital; whole unit, e.g. ward) in collective activities to improve physical or mental health or promote healthy behaviours. Results Eleven studies were identified which incorporated at least one of the whole-system recommendations. Interventions that incorporated recommendations to address local need and engage the whole workforce fell in to four broad types: 1) pre-determined (one-size-fits-all) and no choice of activities (two studies); or 2) pre-determined and some choice of activities (one study); 3) A wide choice of a range of activities and some adaptation to local needs (five studies); or, 3) a participatory approach to creating programmes responsive and adaptive to local staff needs that have extensive choice of activities to participate in (three studies). Only five of the interventions included substantial involvement and engagement of leadership and efforts aimed at up-skilling the leadership of staff to support staff health and wellbeing. Incorporation of more of the recommendations did not appear to be related to effectiveness. The heterogeneity of study designs, populations and outcomes excluded a meta-analysis. All studies were deemed by their authors to be at least partly effective. Two studies reported statistically significant improvement in objectively measured physical health (BMI) and eight in subjective mental health. Six studies reported statistically significant positive changes in subjectively assessed health behaviours. Conclusions This systematic review identified 11 studies which incorporate at least one of the Boorman recommendations and provides evidence that whole-system healthy workplace interventions can improve health and wellbeing and promote healthier behaviours in healthcare staff.


Introduction
Healthcare professionals throughout the developed world have markedly high rates of sickness absence, burnout, and distress compared to other sectors [1][2][3][4][5][6][7]. With the added pressure on healthcare systems, and thus on healthcare staff, of rapidly aging populations and burgeoning chronic disease burdens [8], there is increasing interest in improving both the mental and physical health and wellbeing of healthcare professionals [9,10]. There is a growing call for the 'triple aim' (improving patient experience, patient outcomes, and efficiency) to become the 'quadruple aim', with the inclusion of improving healthcare staff experience of care delivery [11,12]. In the United Kingdom (UK) the National Health Service (NHS) in England's Five Year Forward View [9] identifies NHS staff health and wellbeing as a priority for the NHS.
Sub-optimal health behaviours of healthcare practitioners in the workplace are linked to stress, illness, increased healthcare costs, obesity, high staff turnover, errors, and poor quality healthcare delivery [4,13]. However, despite concerted policy and research efforts in the last decade designed to support and improve their health and wellbeing (for example [9,10]), the acute and long term sickness absence of UK healthcare deliverers remains high [14].
Interventions to improve healthcare staff health and wellbeing have primarily focused on supporting or improving individual coping skills rather than affecting the workplace environment such that it promotes healthier behaviours. Whilst personal coping skills mediate the effects of stressors at work on health and wellbeing, i.e. the ability to deal with environmental stressors at a personal level [5,7,15], research points to the potential preventative

Inclusion criteria
Studies were included if they reported interventions which were targeted at all staff within a healthcare setting (for example a whole hospital, health centre, or unit), were predominantly delivered as group rather than individual activities, and measured the impact on health behaviours or psychological wellbeing in healthcare professionals (outcomes chosen a-priori). Studies in which the intervention was solely aimed at a subgroup of the population (e.g. those with high cholesterol or smokers) were excluded.
Randomised controlled trials (RCT), before and after studies (with or without control), case control, cohort studies and survey designs were included.

Study identification
Inclusion and exclusion criteria were applied to all titles and abstracts by one reviewer (JP, SLB, AB or JTC) and double screened by a second (JP, SLB, AB, KW, LC, or JTC). Duplicates were identified, checked, and excluded. Discrepancies were resolved by discussion with a third reviewer (LF or KW) where necessary. The full text of potentially relevant articles was retrieved and screened independently by four reviewers (four of: JP, SLB, KW, LC, and JTC); discrepancies were resolved by discussion with a third reviewer (as appropriate, one of: AB, LF or KW).

Data extraction and quality assessment
A data extraction and quality assessment tool was developed and piloted for suitability on four papers by SLB and KW. Data extraction and quality assessment were undertaken by SLB and checked by KW; any disagreement was resolved through discussion.
Assessment of study quality was carried out using the Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses [21] and the EPOC guidance for randomized controlled trials, controlled before and after studies and interrupted time series (Cochrane Effective Practice and Organisation of Care Review Group [22]).
The following data were extracted from each eligible article (S2 Fig): study design; geographic location of study (country); numbers eligible to participate, numbers participating, loss to follow up; summary characteristics of the study population; details of the intervention; whether the intervention was designed to address a local need; whether and which stakeholders were involved in the development and implementation; whether senior management were involved and in what way, including whether there was visible leadership from or upskilling of management staff; treatment of any control group; duration of follow-up; primary and secondary outcomes, outcome measures and intervention effects. Details on whether the intervention was going to be continued at the site after the initial evaluation were also looked for.

Data analysis
As the studies as well as the workplace health and wellbeing interventions reviewed were heterogeneous in their design, implementation and outcomes, an overall meta-analysis was not appropriate, rather we aimed to describe the nature of the interventions, whether they engaged staff, and the outcomes. Study and intervention details were put in to tables, with columns for the whole-system recommendations and rows for description of the study/intervention. This supported us to identify patterns in relation to whether and how the studied intervention aimed to 1) engage staff at all levels of organisation in activities and be responsive to local need and context (relating to whole-system recommendations 1 & 2 [17]) and 2) engage, involve and up-skill leadership staff (whole-system recommendations 3, 4, & 5 [17]).
We provide a narrative review of overall patterns of whether and how we believe the interventions in the included studies take a whole-system approach as described in the Boorman recommendations [17], commenting on whether and how the groups of interventions improved the health and/or wellbeing and/or increased health behaviours of healthcare staff.

Identified studies
While the original searches retrieved a total of 14,526 records, the review process identified eleven studies to be included (Fig 1). After removing duplicates, 11,908 unique records were downloaded into the reference manager software Endnote to form the master library. The full texts of 379 papers were retrieved for closer examination. Three hundred and seventy five papers were excluded (Fig 1). Update searches (Oct 2013 and Sept 2016) identified a further 7 studies (6 from update searches and 1 from hand searching). A total of eleven studies [6,[23][24][25][26][27][28][29][30][31][32] were included, and are summarised in Tables 1-3.

Study characteristics and quality
Outcomes. All reviewed studies included self-reported measures of individual health behaviours and health outcomes, and four studies [6,23,28,29] included self-reported measures of the psychosocial workplace environment (see Table 3). Two studies [28,32] reported Body Mass Index (BMI).
Study design and quality. The overall quality of included studies was considered to be poor ( Table 1). The main reasons relate to the outcome measures, which were generally low in reliability, variable in validity, and heterogeneous; lack of controls; variable follow-up length; and high attrition rates.
Reliability and validity of outcome measures: Six [6, 23-25, 27, 28, 31] out of the eleven studies had partial or low reliability of outcome measures. Three studies [6,23,24] did not use validated outcome measures. Petterson and colleagues [6] used self-report scales based on the findings of a factor analysis. They report that in general these scales have high internal consistency, though two (job demands and work pressure) had lower internal consistency, leading them to question the ability of these scales to measure unitary dimensions. Some studies used self-report measures. Sorenson and colleagues [23] used self-report survey and process tracking measures (including self-reported number of activities taken part in). Jasperson and colleagues [24] constructed and used a health questionnaire that measured self-reported physical activity and diet and a survey measuring self-reported walking event attendance.
Heterogeneity of outcome measures: Three studies [25,27,28] used an objective outcome measure of health (BMI). The other nine studies used subjective self-report measures, and no two of them used the same subjective self-report measures.
Study design: In addition (see Table 1) five of the eleven studies lacked a control group [6,24,25,27,30], two had no follow-up [24,31], and one did not report follow-up information because baseline and follow-up questionnaires were not linked by person [23]. Follow up periods varied considerably; in six studies [23][24][25][26][29][30][31] follow-up (or single time-point) data were collected immediately post intervention. In the remaining studies follow-up data were collected at between 3 months [26,32] and 5 years [27] after the start of the programme. Follow-up rates also varied as the workforce itself changed over the follow-up period.

Randomised-Controlled
Trial • Change in BMI. Body mass index (BMI) was calculated from measured weight and height. Weight measurement was taken on digital scales and rounded to the nearest 2/10 of a pound. Heights were measured to the nearest l/8inch using portable stadiometers. The average BMI across baseline 1 and baseline 2 assessments was used in this analysis.
• Fruit and vegetable and fat consumption. Fruit and vegetable and saturated fat consumption were measured by the Block rapid food screener, a brief food frequency type measure that assessed commonly eaten foods: 10 items summarised as servings of fruits and vegetables per day. The fat screener consists of 17 items summarised as percentage of total calories from saturated fat.
• Physical activity. Self-administered long-form of the International Physical Activity Questionnaire (IPAQ), developed by the World Health Organization, with demonstrated reasonable psychometric properties for assessing population levels of self-reported physical activity. Vigorous, moderate, and walking activity in 4 domains, work, household, free time, and transportation, were assessed.
• Perceived organizational commitment to employee health. 4-item subscale of the worksite health climate survey (WHC), which demonstrated strong internal reliability (a = .88). Respondents rated each item on a 5-point scale.
• Perceived co-worker normative behaviours. Modified versions of the WHC subscales for health norms measured employee perceptions of eating and physical activity behaviours of co-workers. Individual items were selected and adapted to focus on at-work behaviours. Four items asked about coworkers' physical activity behaviours at work, and 5 asked about co-workers' eating habits at work. Seven response categories (almost none to almost all) estimated the proportion of co-workers who practice specific behaviours. Negative items were reverse coded, with higher scores corresponding to healthier behaviours. Psychometric testing of each scale indicated very good internal consistency (a = .78, healthy eating; a = .74, physical activity)

102
Multivariable linear regression models for survey data to assess associations of demographic and job characteristics with the 3 worksite perceptions scales and relationships of the 3 worksite perceptions scales with BMI, fruit and vegetable consumption, saturated fat consumption, and physical activity, controlling for demographic and job characteristics. • Perception of stronger organizational commitment to employee health was associated with lower BMI (B = 0.73, p = 0.03).
• Higher perception of co-worker normative healthy eating behaviours was associated with greater fruit and vegetable consumption and less fat consumption (B = .33 p < .001).
• Higher perception of co-worker normative physical-activity behaviours was associated with greater total physical activity (18.2%, p = 0.003).
• Participation dose-response effect: The more intervention activities people participated in the greater the reduction in BMI: When intervention exposure was used as the independent variable BMI decreased for each unit increase in intervention participation at 24 months (p = 0.006

Results Summary
Petterson,

[6], Sweden
Before-after study (no control) • Based on results from overall factor analysis, six indices of perceived work quality (competence and skills development, job demands, work pressure, optimal workload, organizational climate and goal clarity), three indices of supporting resources (social climate, job control and coping) and two health indices (psychosomatic symptoms and exhaustion) were measured at baseline and used as outcome measures to evaluate effects of the intervention program. In general, scales used have high internal consistency. Job demands and work pressure had lower internal consistency, questioning their ability to measure unitary dimensions.

52
ANOVA pre-versus postand high-versus lowactivity uptake departments. Based on activity ratings, departments were separated into two groups, one highly active (n = 20) and one less active (n = 17) in the change process. The groups were compared regarding measures of work quality, supporting resources, and health. • Between baseline and follow-up a notice of staff cut-back was announced which was considered a reason for the general deterioration in most measures. All hospital staff were exposed to the same information.
• Participation dose-response effect: Staff in departments rated as highly active in improvement activities did not deteriorate during the follow-up period compared to a worsening in departments rated as less active in work pressure (active: pre = 14.

Results Summary
Wieneke, 2016 [31], USA Cohort study • A (web-based) survey (validated) was conducted to assess whether the objectives of the wellness champion programme were achieved. The objectives were to increase awareness and participation in healthy living programmes, promote positive health behaviours among employees, and provide a supportive work environment among employees: • Wellbeing: A validated single-item question was used to assess well-being.
• Awareness and participation: 2 Likert scale items were designed by a health and wellness expert panel to assess awareness and participation in the wellness champions programme.
• Self-rated health: Participants were asked to rate their overall health and wellness on a scale of 0 to 10 (0 being the "worst health and wellness" and 10 being the "best health and wellness").
• The effect of the wellness champions programme on health behaviours: A team of medical and wellness experts created 2 additional study specific items that asked participants to rate on a 5-point Likert scale, from strongly agree to strongly disagree, their answers to: (1) My co-workers and I support one another in our effort to practice a healthy lifestyle; and (2)   Creating healthy workplaces for healthcare workers  Creating healthy workplaces for healthcare workers rates varying from 22 to 33%; one study reported 50% attrition, one [23] did not report attrition rates; and two did not have baseline measurements [24,31]. Participation: Nine studies [6,23,24,[26][27][28][29][30][31][32] offered the intervention to all hospital/unit/ health centre staff; one study [25] offered the intervention to everyone, but operated on a first come first served basis as there were only 400 places available to the 2900 staff; one study [31] offered the intervention to all staff working in a work area that had a workplace wellness champion working in it. Given the nature of the included interventions (i.e. aiming to affect wholesystem change), it is hard to estimate overall participation rates other than in the specific activities which were delivered within the intervention programme.
None of the studies described the interventions in sufficient detail to allow replication. One study [30] offered the manual for their brief MBSR intervention upon request.
Due to the heterogeneity of types of study and measures used, it is difficult to make meaningful comparisons between the studies. We describe the interventions in relation to the degree to which they included the whole-system recommendations for healthy workplace interventions in healthcare settings [17].
Included interventions. Interventions varied in terms of whether and how they incorporated the five whole-system recommendations [17] (Table 3) and their overall effectiveness (as reported by the authors of each study) in improving healthcare staff health and wellbeing and/ or health behaviour change (Table 3:

Recommendations 1&2: Identifying and responding to local need and engaging staff at all levels
The eleven studies varied considerably in how they tailored their interventions to local need and engaged staff at all levels (Table 3; [17]). Interventions were: 1) pre-determined and fixed from the outset without choice of activity [25,30]; 2) pre-determined with choice of activity [26]; 3) had choice of a wide range of activities and some adaptivity of the programme, with further activities added in response to take-up [23,24,27,32]; and, 4) adaptive and responsive workplace programmes, taking a participatory approach from the beginning and creating programmes responsive and adaptive to staff needs, in which the implementation process was part of the intervention [6,28,29,31].

1) Pre-determined interventions with no choice of activities.
Two studies offered a fixed set of activities, including some element of group activities, to all staff in one workplace. These activities were not created in response to local need, and nor was there choice about which activities to participate in [25,30].
An eight-week Mentalisation-Based Stress Reduction (MBSR) intervention for staff in a 12 bed mental health inpatient unit (MBSR practice focuses on individual coping but the team delivery design of the intervention also enabled whole-system change; Table 3) resulted in a significant decrease in self-reported psychological distress, including reduced levels of selfreported anxiety [30]. There was no overall increase in the Kentucky Inventory of Mindfulness Skills, suggesting that these changes in distress and anxiety may have resulted from the increased communication and activity-sharing between the work unit [30]. Implementing a pre-determined 12 week intervention to improve physical activity and nutrition behaviours across a hospital site using a team-based approach and peer support (Table 3) resulted in those completing the intervention reporting significantly higher physical activity, fruit and vegetable consumption, water intake, and feeling less stressed than the noncompleters [25].
2) Pre-determined interventions with some choice of activities. One study [26] had a fixed set of activities and some choice about which activities to participate in. In response to an identified need, a self-care plan and holistic learning programme on one hundred and three nurses' health-promoting behaviours in intervention units over twelve months resulted in a significant difference pre-and post-intervention in nurses in intervention versus control units in overall Health Promoting Lifestyles Promotion (HPLP) II scores, and the sub-scales of stress management, nutrition, and spiritual growth [26].
3) Choice and some adaptivity of the programme (supplementary activities). Five interventions [23,24,27,28,32] offered an initial range of activities for the workforce to participate in, as well as providing supplementary activities during the implementation of the interventions.
Three of the five interventions involved "Workplace Champions" whose roles were delivery as well as gathering feedback and planning further activities [23,24,27]. In one intervention [23], the role of the workplace champion was to further refine and adapt the intervention activities delivered depending on identified need and context. Employee leadership and advisory boards were also created to develop site-specific strategies and approaches.
Three interventions [23,28,32] had an explicitly participatory approach both in the design and the delivery. Two [23,28] included External Advisory Boards at each intervention site to engage the workforce and tailor activities to their needs and two had a strong emphasis on engagement of leadership and staff in the development and tailoring of intervention activities [28,32]. The latter two contained activities designed to engage the whole worksite and develop relationships to support healthy behaviours, e.g. a directors' team-building course, and activities for all staff (including leadership) to improve team coordination, communication and stress management [32].
Uptake of the intervention was not determined for any of the studies, probably as all offered a variety of activities as well as making some environmental changes; one study [24] reported that a third of all the employees participated in a competition organised as part of the intervention.
Four of the five studies [23,24,26,27] reported an improvement in health and wellbeing behaviours (Table 3); increase in fruit and vegetable consumption (three-arm randomised controlled trial [23]); increased physical activity (before and after study with no control [27]; self-report, survey design, no control [24]); increase in healthy eating post-intervention [26]; and improved self-reported nutrition post-intervention in a controlled before and after study [26]. Two studies found no change in BMI [27,28].
Three studies found improvements in employee mental health: improved job satisfaction (but no effect on mood or work perceived work performance [27]); more self-reports of staff satisfaction in post-intervention survey [24]; improvements in stress management and spiritual wellbeing post-intervention compared to controls [26].
One study [28] observed improved employee perception of worksite commitment to their health and wellbeing and that changes in perceptions of co-worker norms changed outcomes for participants: higher perception of co-worker normative healthy eating behaviours was associated with greater fruit and vegetable consumption and less fat consumption; and higher perception of co-worker normative physical-activity behaviours was associated with greater total physical activity. Perceived co-worker support also increased in the intervention arm in another study [23].
One study [28] observed a significant participation dose-response effect: When intervention exposure was used as the independent variable BMI decreased for each unit increase in intervention participation at 24 months (Table 3).

4) Adaptive and responsive workplace programmes.
Three large hospital studies [6,29,31] viewed the process of developing the intervention to be a part of creating a healthy workplace: a before and after study involving over three thousand employees from thirty seven regional hospital departments in Sweden [6]; a cluster-RCT in twenty four hospital departments in two hospitals in Japan [29]; and a cohort study comparing people exposed and not exposed to a workplace wellness champion intervention (self-reported) within a large academic medical centre [31]. The three interventions were adaptive and responsive to local needs and context from development and implementation, right through until the end of the study and beyond. This responsivity provides the opportunity for sustainability after the end of the study, and for processes involved in the intervention to become part of the workplace culture.
All three: aimed to improve the psychosocial work environment by utilising a participatory approach, asking each department or work area to identify the enablers and barriers to workplace wellbeing, and to set goals and identify areas for improvement; appointed key people or "champions" to support the interventions and to act as communicators within and across departments; and used feedback to develop activities responsive to local need (two [6,29] fed back survey results to local staff and one [31] had workplace wellness champions (Table 3) design activities based on local feedback) Two studies showed some evidence of a dose-response effect in which greater participation produced greater benefits. When a notice of staff cut-back was announced between baseline and follow-up staff in departments rated as highly active in improvement activities did not deteriorate during the follow-up period in work pressure, organizational climate and coping whereas staff in departments rated as less active did deteriorate [6]. Similarly, participants in a local work area who did versus did not participate in activities rated their overall health and wellness as significantly higher and significantly more of participating versus not participating in a local area agreed that their co-workers support one another in practicing a healthy lifestyle [31].
One [29] found varying levels of participation across the departments, with staff citing "realising that change was possible" and responding to identified needs as positive ways of improving the psychosocial work environment; concomitantly, the lack of time and common understanding, staff changes, and a feeling that activities were not responding to staff needs were given as reasons why the environment did not change. They observed no overall effect on mental health status, but significant increases in participatory management, co-worker support, and job control versus control.

Recommendations 3, 4, & 5: Engagement, involvement and upskilling of management and board-level staff
Three of the five recommendations involve the engagement and support of management and board-level staff in intervention activities, including strong visible leadership, support for the health and wellbeing of staff, and the targeting of resources on improving management capability and capacity to deliver this increased visible leadership and support. Despite this emphasis in the recommendations these activities were notably lacking in seven interventions reviewed. Promisingly, five focused considerable resource on engagement, involvement and upskilling of management staff [6,28,29,31,32]. Two significantly improving mental health and wellbeing of healthcare staff [29,31], one reducing deterioration in mental health measures for people with high versus low participation rates [6], one significantly improving physical health (BMI) and health behaviours [28]. The fifth found no significant effect on mental health and wellbeing [32].
Four interventions involved extensive management involvement as champions [28,29,31,32]. Sun and colleagues [32] Engaging higher management staff in local staff activities and creating opportunities for increased communication, group solidarity and group coordination, alongside intervention activities to improve leadership and management and communication skills of higher levels of management (e.g. directors of intervention centres were engaged and educated about the importance of activities in groups for staff, and then were involved in implementing and taking part in the group activities, having the responsibility to plan, coordinate, and monitor the group's activities and to convey a vision, inspiring team collaboration) resulted in no significant impact of their intervention on workplace social capital, the measures of which included items on vertical and horizontal trust and communication. In another intervention [31], following efforts to involve management, supervisors and HR in supporting the workplace wellness champions to deliver and implement their locally adapted intervention, twenty-three percent of people engaging with the intervention reported an improved work atmosphere, and significantly more of those participating strongly agreed or agreed that the organisation provides a supportive environment to live a healthy lifestyle compared to those not familiar with the wellness champions. Enthusiastic"buy in" at the upper level of administration and visible strong leadership support, when it improves cooperation by other staff to implement changes [28], produced a significant association between perception of stronger organisational commitment to employee health and a reduction in BMI. When local leadership staff were directly supported to develop their capability and capacity to improve staff health and wellbeing (Table 3) [29], the intervention group showed a statistically significant increase in the psychosocial work environment questionnaire sub-scale of 'participatory management', along with self-reports of improved work environment in the process evaluation; however, there was no significant difference in the intervention groups' scores on the depression scale pre-to post-intervention.
When there was a focus on management visibility and involvement in the feeding back of local results and the active implementation of changes related to locally raised needs, and improving vertical communication between managers and staff, staff in departments that actively participated showed significantly less deterioration in perceived organisational climate compared with staff in departments with low participation [6].

Discussion
This systematic review identified eleven studies of workplace health promotion interventions which sought to enhance the health and wellbeing of healthcare staff by using a whole-system approach to interventions. The low number of identified studies highlights that the impact of whole-system healthy workplace interventions for healthcare workers, as recommended by Boorman [17] is under-researched, and we feel this gap is important for future research to address.
Although the studies were of mixed (mid to low) quality and the intervention designs varied considerably, the reported results suggest that interventions taking a whole-system approach can improve physical and mental staff health and wellbeing and promote healthier behaviours.
Interventions that incorporate at least one of the five whole-system recommendations for improving healthcare worker health and wellbeing [12] resulted in improvements in physical and/or mental health and promoted healthier behaviours in healthcare staff. However, one study [32] incorporated all five of the recommendations in their workplace social capital intervention and did not find any significant change in measures of mental health (workplace social capital measure, see Table 2). It is not possible to draw conclusions regarding the specificity of the interventions as they varied widely in terms of their context, development, design, and implementation, but it is interesting to note that there seems to be no relation between the greater number of recommendations incorporated in interventions and the effectiveness of the study. However, heterogeneity in outcome measures makes this a tentative comparison.
Four studies of interventions offering choice of a range of activities to participate in [6,23,28,31] individually offer some evidence that the greater the level of participation, the greater the individual benefit: greater participation improved: resilience to organisational change [6]; self-rated overall health and wellness [31]; BMI [28]; and fruit and vegetable consumption [23]. The latter study also found enhanced effect of a worksite intervention when there was family participation suggesting that widening activities beyond the workplace to include family and friends may further enhance engagement and improvements to wellbeing.
The suggestion of potential individual "dose-response effect" in these four studies (i.e. more benefit derived from more participation [6,23,28,31] has several implications: firstly, it suggests that attention needs to be given to creating intervention activities that healthcare staff want to engage in and offering a selection of a range of activities, some team-and some individual-based, for participants to choose between. Interestingly, some workplace interventions [28] had the least take-up of team-based activities, whereas others [24] found this the most participative aspect of the intervention. The four studies that reported findings from their process evaluations [23,25,28,29] all suggested that time was one of the greatest barriers to employee participation in workplace health and wellbeing interventions. Understanding the barriers to participation (such as time, resources, and poor communication about activities) should be part of the process of evaluating any workplace intervention, and having an intervention able to adapt to allow different times and ways of participating should be beneficial.
Although it is hard to make any meaningful comparisons regarding effectiveness, the studies which assessed intervention activity participation [23,25] suggest that the interventions in which staff were involved from the beginning in determining the activities had greater participation. There was lower participation in interventions with more pre-determined activities, even when there was an opt-in process for staff and hence potentially had a more motivated workforce participating. One study [25] invited 400 of 2900 staff to participate in a 12 week intervention and had 61% participation, compared with the 81% participation in the workplace intervention implemented by another study [23] where employees were involved in the development and implementation of their workplace intervention.

Implications for policy/practice
The Boorman Review [17] called for healthcare workplaces that: support local staff needs; have staff engagement at all levels; have strong visible leadership and support at senior management and board level on health and wellbeing; have a focus on management capability and capacity to improve staff health and wellbeing. Our systematic review shows that interventions incorporating these whole-system approaches can improve healthcare staff health and wellbeing and increase health behaviours.
Only five of the eleven studies focussed on management capability and capacity. There was some evidence from subjective author reports that this focus resulted in enthusiastic engagement from leadership [28,29,32]. Of these five interventions, four involved management-level staff as healthy workplace champions. It is interesting to note that the findings of these four studies all involved perceptions of improved workplace culture or atmosphere in participants: improved work atmosphere and environment, more supportive environment to live a healthy lifestyle, more 'participatory management', less deterioration in perceived organisational climate, and stronger organisational commitment to employee health (the latter of which was significantly associated with reductions in objectively measured BMI). However, the fifth study found no significant impact of their intervention on measures of vertical trust and communication. These findings provide some evidence that interventions including efforts to engage and involve management staff, such as in the feedback of local results of health and wellbeing surveys and involvement in discussions with local staff of how they would like to address the, in being workplace champions themselves, to make their leadership on health and wellbeing more visible, and to provide training on skills to support the health and wellbeing of their staff, can impact the perception by those staff that management are on their side and that they work in a place with a positive workplace environment.
The finding of potential "dose-response" effects in the four studies that report participation rates suggest that participation and engagement are important in designing and implementing healthy workplace interventions: A flexible intervention with continuous employee involvement and an ongoing evaluation to highlight facilitators and barriers to participation has greater potential to positively affect and sustain health and wellbeing for the healthcare workforce and thus to improve staff health and wellbeing.

Strengths and limitations
The review was conducted according to the principles published by the NHS Centre for Reviews and Dissemination (CRD) and is reported according to PRISMA guidelines (S3 Fig). The review was comprehensive, searching across electronic and grey literature sources to identify studies. There were no language or date restrictions in the searches.
Due to the nature of the topic under consideration, the inclusion criteria in this review were open to a degree of subjective interpretation. For this reason we took all reasonable steps to ensure that eligibility criteria were applied consistently across all identified articles by 1) piloting the criteria on a subset of papers, 2) having two reviewers independently assess the eligibility of all articles with discussion of all disagreements, and 3) involvement of a third reviewer to resolve disagreements where necessary.
Comparison across the approaches utilised to improve health and wellbeing of healthcare professionals was challenging due to the lack of detail provided regarding the specific nature of the components and the mechanisms making up the interventions.
Variable methodological quality, mostly related to the outcome measures used, which were in general low in reliability, variable in validity, and heterogeneous, along with the nature of the study designs prevented any conclusions related to the effect on health and wellbeing outcomes of incorporating more versus less of the five whole-system recommendations being made. Nor were we able to compare the effectiveness of different patterns of whole-system recommendation implementation.

Recommendations for future research
Despite extensive and systematic searching of the literature, we were only able to identify eleven studies that met our inclusion criteria. The low number of identified studies highlights that there is currently limited evidence regarding the effectiveness of whole-system approaches to enable staff health and wellbeing for healthcare professionals in healthcare settings, as recommended by Boorman [17]. Ten out of eleven included studies provide evidence that wholesystem approaches to healthcare workplace health interventions that include at least one of the five whole-system recommendations [17] improve physical and/or mental health and promote positive health behaviours in healthcare staff, suggesting this is an area of potentially fruitful inquiry.
The methodological quality of the studies was mostly low, with only five out of eleven studies included being rated as "medium" quality. This systematic review clearly identifies a need for good quality primary research using similar and validated outcome measures to evaluate whole-system approaches to health and wellbeing interventions in healthcare worker populations. Comparative studies of the effectiveness of individual-focused versus whole-systemfocused approaches would clarify their relative effectiveness and cost-effectiveness. Longterm follow-up is necessary to evaluate the sustainability of observed change. More systematic reporting would allow more definitive conclusions about how the conditions for sustainable healthy workplaces for healthcare workers can be created.
The low number of studies and heterogeneous intervention designs and outcome measures used in those studies, makes it challenging to pin down whether and in what way whole-system approaches improve healthcare worker health and wellbeing. Realist reviews of the literature, in which context-mechanism-outcome configurations rather than whole interventions are the unit of analysis, would be of use in establishing what it is about whole-system interventions that works to improve health and wellbeing for healthcare workers, who for, under what circumstances, and in what way.

Conclusion
This systematic review identified 11 studies which incorporate at least one of the Boorman recommendations and provides evidence that whole-system healthy workplace interventions can improve health and wellbeing and promote healthier behaviours in healthcare staff.