Objectives The shortage of healthcare staff is a global problem. UK mental health services have, on average, a higher turnover of staff than the NHS. Factors affecting retention of this staff group need to be explored in more depth to understand what is working for whom, for what reasons and in what circumstances. This review aims to conduct a realist synthesis to explore evidence from published studies, together with stakeholder involvement to develop programme theories that hypothesise how and why retention occurs in the mental health workforce and identify additional evidence to explore and test these theories thereby highlighting any persistent gaps in understanding. This paper develops programme theories that hypothesise why retention occurs and in what context and tests these theories thereby highlighting any persistent gaps in understanding.
Methods Realist synthesis was used to develop programme theories for factors affecting retention of UK mental health staff. This involved: (1) stakeholder consultation and literature scoping to develop initial programme theories; (2) structured searches across six databases to identify 85 included relevant literature relating to the programme theories; and (3) analysis and synthesis to build and refine a final programme theory and logic model.
Results Phase I combined findings from 32 stakeholders and 24 publications to develop six initial programme theories. Phases II and III identified and synthesised evidence from 88 publications into three overarching programme theories stemming from organisational culture: interconnectedness of workload and quality of care, investment in staff support and development and involvement of staff and service users in policies and practice.
Conclusions Organisational culture was found to have a key underpinning effect on retention of mental health staff. This can be modified but staff need to be well supported and feel involved to derive satisfaction from their roles. Manageable workloads and being able to deliver good quality care were also key.
- MENTAL HEALTH
- Human resource management
- Health policy
Data availability statement
Data sharing not applicable as no datasets generated and/or analysed for this study.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Strengths and limitations of this study
Realist synthesis focusses on interpretation rather than comprehensiveness.
The use of stakeholders early in the process was a strength.
The use of concepts rather than specific interventions in the synthesis improves the generalisability of the findings.
In 2022, the WHO called for all nations to invest in mental health workforce, with many facing shortages of trained mental health staff, with poor retention rates and low up take of new staff being trained.1 Shortages are being seen across mental health professions, for example across much of the world’s population there is just one psychiatrist to serve around 200 000 people, with the most severe shortages seen in rural areas where other mental healthcare providers who are trained to use psychosocial interventions are even scarcer.2
The problem of workforce retention in the UK National Health Service (NHS) is widely acknowledged and has worsened every year since 2011.3 UK mental health services have, on average, a higher turnover of staff than the NHS as a whole (13.4% of all mental health staff left in 2018/2019 compared with 11.9% across all NHS employers) and more vacant positions.4 5 Job satisfaction for mental health nurses is reportedly the lowest it has ever been, with many nurses emotionally and physically exhausted, leading them to consider leaving their profession.6 In December 2021, there were 1 110 000 vacancies in the NHS, with 22% of all vacancies in mental health trusts despite only 14% of NHS spending being on mental health.7 A recent report by the Royal Collage of Psychiatrist states that in England 9.3% vacancy rate in consultant posts and an increasing in use of locum consultant psychiatrist posts.8 These staff shortages are known to impact patient quality of care with approximately 1.5 million people in England waiting for mental health treatment and that two-fifths of those on mental health waiting lists, ended up receiving emergency or crisis services before receiving treatment.9 More immediate emphasis should be placed on retaining the staff already in post as retaining experienced staff is seen to not only have the potential to improve patient care and waiting times but also has the potential to inspire a new mental health workforce to join aiding retention and boost recruitment.6
Previous reviews have focused on the determinants of the healthcare workforce turnover and intention to leave rather than retention, within the physical healthcare workforce10–12 or only focused on specific mental health professions such as mental health nurses13 and psychiatrists14 rather than the entire workforce. Turnover and intention to leave has been associated with organisational factors such as resource allocation, leadership and pay as well as individual factors including stress and job dissatisfaction.10 11 Stress experienced by healthcare staff appears linked to the ability to give time to patients in need, missed care and the perception of the quality of care delivered, all of which potentially impact job satisfaction an intent to leave.15 Job dissatisfaction appears to be also strongly associated with poor retention of mental health nurses13 and one of the key predictors of turnover of all healthcare staff. Leadership appears to have an indirect role, in that it affects the organisational culture (the set of beliefs, values and behaviours that determine an organisation’s identity) which in turn affects job satisfaction.16 For example a transformational leadership style is related to supportive work environments and staff intention to remain,17 while a supportive and communicative leadership style leading to organisational commitment is a strong predictor of low turnover.12
Many of the stressors highlighted by previous reviews will be also relevant to the mental health workforce such as limited resources, job dissatisfaction, pay or poor leadership, but some stressors effecting mental health workforce retention maybe specific to this setting13 such as the use of restraint and seclusion,18 dealing with verbally aggressive and physical violent patient acts,19 continuous monitoring of patients at risk of self-harm and suicide.20 In 2018, the mental health workforce reported higher emotional exhaustion than emergency nurses and equal burnout to cancer professionals.21 Factors affecting all mental health staff retention need to be explored in more depth to understand what is working for whom, for what reasons and in what circumstances.
This review aims to generate an in-depth understanding of not only why mental health services find it challenging to retain its clinical staff, but what retention strategies are working and in what circumstance. The review’s objectives are to conduct a realist synthesis to explore evidence from published studies, together with stakeholder involvement to develop programme theories (PTs) that hypothesise how and why retention occurs in the mental health workforce and identify additional evidence to explore and test these theories thereby highlighting any persistent gaps in understanding.
We undertook a realist synthesis of the literature to create an overarching programme theory of the factors at an organisational, management and frontline level that influence retention and turnover of mental health staff in the UK.
A realist review is a theory-led method which seeks to understand a particular outcome through exploring the relationships between the context and the underlying causal processes (mechanisms) that lead to that outcome.22 The realist review process is used to generate ‘If–Then–Leading to’ statements that provide possible explanations for the outcome of interest. For example, ‘IF management ensure that staff have the time and support required to deliver needed care (Context) THEN patients engage with their care (Mechanism) and staff feel their role is meaningful and important (Mechanism) LEADING TO increased job satisfaction (Outcome) and enhanced patient outcomes (Outcome(s)). These statements, also known as Context–Mechanism–Outcome Configurations (CMOCs), are subsequently refined to produce PTs, which are then synthesised into a single overarching theory to be tested against existing studies (synthesis) or primary research (evaluation).
The realist approach was appropriate for this review as it recognises that the context within which mental health workforce retention occurs is ‘complex, multifaceted and dynamic’.23 Factors affecting this phenomenon/outcome are complex and changeable, involving various people, structures, policies and practice at different organisational levels. The effectiveness of any intervention to improve retention is contingent on diverse factors, and multiple explanations may be advanced for how these operate. This review seeks to identify the mechanisms that are activated across a variety of contexts operating at different levels (individual, team and organisational) within the healthcare system, which lead to outcomes relating to retention. Realist review methods enable exploration of this complexity and interconnectivity and enable development of a theoretical understanding of what factors influence retention within the UK mental health workforce.
The review was conducted in three phases1: exploratory scoping of key literature and stakeholder consultation to identify initial PTs2; structured searches, screening and data extraction of included studies related to each identified PT; and3 analysis and synthesis leading to refinement of the PTs and development of a final overall PT and logic model.
No changes to the protocol were introduced during the review process. While the phases were sequential, each phase itself was iterative and went through rounds of refinement which ultimately resulted in new interpretations and the development of theory. The RAMESES (Realist and Meta-narrative Evidence Syntheses: Evolving Standards) publication standards informed the reporting of this review.22
Patient and public involvement
Phase I of the study (described below) involved extensive patient and staff involvement in the design of the research questions, particularly around priorities and experiences and in the design of the study, particularly around the structured searches.
Phase I: exploratory scoping and stakeholder engagement
Exploratory scoping was conducted to locate and explore available evidence on what key factors might influence mental health workforce retention. Lines of inquiry included: exploring assumptions that underpin workforce retention interventions; the influence of job satisfaction on retention; and reported enablers or barriers to successful implementation of retention interventions at an organisation, management and frontline level. The purpose of this stage was to inform and shape subsequent search strategies and involved purposeful searching24 of specific bodies of evidence based on the teams’ clinical knowledge and review expertise.
Search terms related to ‘healthcare personnel’ were variously combined with such terms as ‘retention, turnover, leavers, resignations and job satisfaction’ in the Cochrane and PubMed databases and Google Scholar for grey literature. The wider review team identified and prioritised information sources that could contribute to theory-building about factors influencing workforce retention. Twenty-seven sources (see table 1) were included, including policy and guidance documents, key research literature and grey literature. All reviewers initially independently extracted data from five sample papers to ensure consistency in approach and develop extraction tables. Findings were compared and discussed until consensus was reached. Data extraction of the remaining key papers was then divided between the team. Data were extracted as ‘If–Then–Leading to’ statements, that is, CMOCs. Extraction only occurred when at least two components of the If–Then–Leading to’ statements were present in a paper. Where complex ‘If–Then–Leading to’ statements were found with multiple components, these were extracted as individual statements on to an excel spreadsheet. When incomplete ‘If–Then–Leading to’ were extracted the team hypothesises on what the missing element might be during research meetings.
Key contexts, mechanisms and outcomes across the data were identified and agreed by the research team25 and mapped to explore their dynamic and interconnected nature.26 From this mapping process, initial PTs (IPT) were drawn out and refined. These were then discussed and prioritised by the reviewers, the research management group and a small panel of clinical academics (n=4), to identify a shortlist of eight initial PTs.
Two stakeholder consultation workshops were then held to obtain feedback on and further prioritisation and development of the initial PTs. Stakeholders were defined as people with experience of working in or using mental health settings and were approached via local NHS organisations research groups, professional bodies and service user groups. Thirty-one stakeholders were involved in this process: 23 mental health professionals (including doctors, nurses, occupational therapists and clinical psychologists) and 8 members of a local service user group. Both groups also suggested refinements to the wording of PTs. Following this process, six refined PTs were identified for further exploration in phase II.
Phase II: structured searches, screening and data extraction
Informed by phase I, search strategies to focus on each of the prioritised PTs were developed by an information specialist, who undertook the subsequent searches between March and July 2020. Search terms relating to the mental health workforce, the UK and qualitative research were combined with specific terms relating to each PT for each of the six searches. Full details of the search strategies are in online supplemental appendix tables 1 and 2. Medicine, nursing and health databases were supplemented by social science sources to cover human resource and workload aspects of the review question. The following databases were searched for literature published from 2004 onwards: MEDLINE (including Medline-in-Process and Epub ahead of print); EMBASE; CINAHL; Cochrane Library; PsycINFO; and ProQuest social science databases (including ASSIA). We included only qualitative studies of mental health staff to ensure we focused on participants’ experiences and perceptions and explored rich contextual data. Duplicate references were removed. The search and screening for each PT was documented separately to enable tracking of individual records through the stages of the process.
The review team worked with the wider project team to develop overall inclusion and exclusion criteria, as shown in table 2. These were supplemented by additional PT-specific criteria.
Four members of the team undertook title, abstract and then full text screening. Working in pairs they reviewed a common subset of approximately 20% of identified studies, comparing inclusion/exclusion and overall agreement. The high level of agreement from this process (95%) was considered justification for assigning the remaining screening to individual members of the team. Uncertainties regarding inclusion were brought to team meetings for discussion and resolution. The team met regularly throughout the screening process. Three key elements of richness, relevance and rigour were considered.27 Screening was undertaken separately for each PT and items considered relevant to other PTs were added to the other PTs for full review. References of all full text papers were checked for additional eligible studies, which were then retrieved and screened.
Study characteristics were extracted into bespoke data extraction forms (one for each PT) by three members of the team. Data extracted included: authors, year, nation (England, Wales etc), patient group, setting, type and number of staff involved in the research, data collection method, findings, limitations, and illustrative quotes.
Where papers were relevant to multiple PTs data, relevant data were added simultaneously to the different PT data extraction tables. The quantity and richness of relevant data varied considerably, with only very limited data being relevant in some cases for example, mixed-methods studies or surveys with open questions. Judgements on study quality were deferred until later stages of analysis.
Team members then used the data extraction table to identify key factors and issues from each PT. These key themes, informed by team member knowledge, were then used to inform the next stage of the review.
Phase III: analysis and synthesis
Data synthesis was undertaken by three team members. Each team member independently summarised the key CMOCs from the data extraction and developed logic models to describe the PTs for which they had undertaken data extraction and were therefore most familiar with. This was done by starting with one CMOC and then looking for interconnections with others, until all had been incorporated into one large logic model. This was repeatedly refined to ensure overlapping elements were removed (eg, risk and safety are two sides of the same issue so were merged). Some of the CMOCs led on to each other meaning one outcome became the context for other CMOC. The logic model was refined to address this complexity. Findings were discussed within the wider team and logic model development was undertaken in consultation with the team methodologist.28 29 These discussions assisted the refinement of the PTs, ensuring they were plausible, grounded in the evidence and comprehensible. The team also reviewed whether the findings from this process suggested the need to modify or further develop the existing PTs, including identifying connections and overlaps between the PTs. As a result, the six individual PTs were refined and assimilated within three integrated key PTs which the team considered more accurately provided a whole systems perspective on the issues relating to staff retention. While other PTs could be seen to be operating, these three appeared to be the key drivers, and a single overarching logic model was then developed to illustrate the interconnectedness and complexity of these refined PTs. One team member initially drafted this model, which was progressively refined and simplified through multiple iterations within the review team and wider advisory team to clarify and accentuate the key findings. This process in turn led to the reformulating of the six ‘If–Then–Leading to’ statements created in phase I to create three new more clearly articulated core PTs. Throughout this process, the team ensured that the evolving PTs remained underpinned by the literature and informed by the findings of the previous stages of the review.
Phase I findings: identification of initial PTs
Phase I led to prioritising of six PTs for further exploration in phase II searches: perceived quality of patient care delivered; workload and staffing levels; team relationships and cohesion; leadership; development opportunities; and supervision. These PTs were formally articulated in the CMOC format of IF–THEN–LEADING TO, see table 3.
Phase II findings: characteristics of included studies from structured searches
In total 3277 titles and abstracts were screened leading to full text review of 383 articles and final identification of 88 relevant papers across all six PTs, with many included papers being relevant to more than one PT. The main exclusion reasons at this stage were lack of identifiable results relating to mental health staff, lack of qualitative findings and not relevant to the UK context. Only three full texts were excluded due to being unobtainable. Numbers of full text articles included for each PT were as follows: Workload: 26; Quality of Care: 37; Team Cohesion: 41; Leadership: 13; Development: 36; Supervision: 27 (categories are not mutually exclusive). The Preferred Reporting Items for Systematic Reviews and Meta-Analyses diagram in figure 1 shows numbers of papers identified for each search and for final inclusion.
The large majority (71, 81%) of papers had been published since 2010, with the remainder (17, 19%) between 2004 and 2009. Papers predominantly reported studies in England (65, 74%), with another 13 (15%) covering more than one nation or the UK as a whole. Very few related specifically to the other nations (Scotland 2, Wales 2, Northern Ireland 1). The predominant data collection methods were interviews (54, 61%), surveys or questionnaires (21, 24%) or focus groups (18, 20%), with some studies employing more than one approach. Study size varied widely depending on the method used, with some surveys including several hundred people while some interview studies included less than 10 participants. Participants were often drawn from a variety of professions (38, 43%); where they focused on a single group, nurses were most common (26, 30%). Eight studies included patient views. Where the setting was specified, studies were approximately evenly split between inpatient and community. Details in relation to all the above characteristics were not always reported, but was only missing for a few studies except for study setting (33, 38%).
A full list of included papers mapped to their respective PTs is presented in online supplemental appendix table 3.
Phase III findings: final PTs and logic model
The three core PTs finally identified were interconnectedness of workload and quality of care, investment in staff support and development and involvement of staff and service users in development of policies and practice. Figure 2 presents these PTs within the final overarching logic model developed to offer a holistic picture of what is required to encourage best practice for retention of the mental health workforce.
Revised programme theory
From the initial PTs in phase I the hypothesis was that if interventions were focused on specific contexts (such as supervision, quality of care, team leadership) it would result in improving retention within the mental health workforce. However, while refining the PTs it became apparent that many of these, we had labelled contexts, were actually occurring further downstream as mechanisms and outcomes, proceeded by organisational contexts fundamental to change.
Interconnectedness of workload and quality of care
In the early stages of the review, workload and the quality of care were identified as related, but ultimately different, concepts and assigned to separate initial PTs. As the review progressed and the wider literature was explored, we found that the two concepts were inseparable and were two aspects of the same critical driver of retention. When workload is high, due to acuity and staff shortages, staff feel they have to deliver lower quality care in order to be able to manage extended demands on their time. In relation to workload, the included studies considered not only the number of qualified staff on the ward or in community mental health services but also the experience level of those staff and if they were regular or temporary such as bank or agency staff. If the organisation did not prioritise strategies to manage these aspects of workload, then mechanisms such as perception of delivering high-quality care, ability to make a difference and therapeutic relationships were negatively affected. As most health professionals are motivated, at least in part, by a desire to help their patients, the realisation that working conditions are having a detrimental effect on their ability to make a difference leads to reduced job satisfaction and in the longer term if it continues, reduced staff morale and intention to leave.
Table 4 contains the final statements and supporting evidence.
Investment in staff support and development
Where organisations have developed effective structures for staff support such as supervision and offer training and development opportunities, staff felt more valued and confident in their clinical skills and subsequent quality of care provided. In phase I, we considered that different elements of investment in staff, such as high-quality supervision and training and development opportunities, constituted separate PTs. Subsequent review of the wider literature led us to conclude that these are all elements of the same organisational contextual factor. An organisation that invests in its staff, to ensure that they are supported, enriched and feel valued, is likely to also invest in clinical supervision and training and development opportunities. The level of investment and the priority given to it therefore provides a key context, which then triggers mechanisms relating to staff feeling valued and supported and more able to deliver high-quality care. Outcomes may be as diverse as impacts on patient care (better trained staff deliver better care) and staff morale, both of which lead to greater job satisfaction and improved retention.
Involvement of staff and service users in development of policies and practice
The processes of organisational policy development were not initially identified as a PT in phase I, however closer exploration of the literature revealed that these are key and underpin many other activities. If policies and practices of an organisation are developed using input from frontline staff and patients then they will be perceived to be directly relevant to what is important to those groups, they will feel included, valued and supported. In contrast, in situations such as where there is a perceived clash of values, for example when management are perceived as being very focused on financially driven targets rather than patient or staff well-being, these triggers mechanisms relating to not feeling valued, and a lack of consistency between their own values and those of the organisation. This leads to poorer job satisfaction and morale, both of which directly impact on retention.
The aim of our realist review was to identify, from the published literature and stakeholder priorities, which factors are affecting the retention of mental health staff in UK adult services, in what contexts these factors operate and for what reasons. This can potentially increase understanding to help organisations design and implement more effective, evidence-based retention strategies and interventions. The review findings enabled the development of a programme theory incorporating three key processes which influence the retention of mental health staff. First, if organisations enable safe workload/staffing levels, staff perceive that they can provide high-quality patient care. Second, if organisations invest in training, development and protected supervision, staff feel valued and more able to provide good patient care. Finally, if the organisation’s policies and practices are informed by and congruent with staff values and prioritise patient and staff well-being, staff feel valued and supported. These processes re all vital to job satisfaction and successful staff retention.
The programme theory also allows the development of hypotheses of why things might not work and in what circumstance. Much of the included literature in this review focused on interventions targeted at individuals rather than at a team or an organisational level. The lack of workforce retention interventions focusing on the interconnectedness of individual, team and organisation has previously been highlighted in reviews targeting single professions such as nursing10 and doctors.30
In line with the Kings Fund 2022 statement that workforce strategies cannot be looked at in isolation, the programme theory developed in this review suggests that retention interventions which are focused on frontline individual staff stress and resilience levels are likely to have minimal success and only short-term gain.31 This is supported by Foster et al32 who found that while individual retention interventions were positively received by frontline staff they led only to short-term gains and highlighted the need for organisational issues to be addressed.32 Therefore, while interventions focusing on individuals or teams may have some effect on job satisfaction, to be effective in the longer-term healthcare workforce retention strategies must be mapped back through the IF, THEN, LEADING to statements and refocused further ‘upstream’.
This review suggests organisational culture is the foundation for successful healthcare workforce retention and the key context for any mechanisms to be activated or not.
If this is the case, then significant change in retention levels requires organisational culture to be the primary focus of positive change. Previous literature has identified that interventions that address organisational and professional issues simultaneously are more likely to be successful30; our review would go one step further to suggest that organisational change is a necessary precondition for positive and effective change downstream. This potentially has significant implications for healthcare organisation decisions on where to prioritise their workforce retention interventions.
Job satisfaction has been well documented as a strong predictor of turnover intent.33 34 This paper employed job satisfaction as a proxy to retention, its position in the final programme theory as a director precursor outcome to staff burnout, sickness and finally to staff retention. All three ‘If–Then–Leading to’ statements tracked through job satisfaction, suggesting it may be a useful measure for measuring effectiveness of any retention interventions.
Strengths and limitations
Using a realist approach is considered a strength as realist synthesis privileges relevant, rich and rigorous literature with the focus on depth and breadth of understanding. An added strength was the inclusion of job satisfaction in the search terms as a proxy for retention, which widened the search strategies to enable a more comprehensive search of relevant literature. However, the team acknowledged that the priority for synthesis was interpretation rather than ensuring the most comprehensive identification of information and is possible that other relevant evidence may have been missed or overlooked.
Another strength of the study is the embedding of stakeholders within phase I to aid development, prioritisation, refining and testing of the initial emerging PTs. The inclusion of diverse healthcare personnel, service users and academics ensured that the widest range of differing perspectives as possible was taken into account within the review process.
Although the mental health workforce is highly diverse in profession and setting, much of the included research was conducted in inpatient settings, and the resulting theory is therefore mostly based on this context. Despite the review only including mental health settings, and most of that being inpatient, the findings are potentially generalisable to other settings. For example, staffing is a retention issue in many healthcare settings. However, staffing issues in mental health contexts bring specific challenges that may not exist elsewhere, such as increased risk of violence, aggression, and self-harm.
The focus on UK NHS staff experience helped to reduce the volume of potentially relevant literature and provided a specific context for the findings but may also have limited international generalisability. However, the exploration of underlying mechanisms, rather than specific interventions, helps to increase the potential transferability of the findings. Healthcare staff retention is a complex phenomenon and other causal relationships may be present in the data and more influential in other contexts (including different countries and different healthcare settings).
Future research and recommendations
Research is needed to test the programme theory, looking at how upstream organisational changes may have a knock-on effect allowing a multiple of mechanisms to be fired in to action and looking at ways to measure the downstream effect such as job satisfaction as a proxy for staff retention. The reviewed highlighted significantly less published literature on leaderships impact on job satisfaction and retention than the other five IPTs despite the high weighting stakeholders gave this IPT.
Additionally, all the included studies were conducted in the UK, it is not known if this programme theory is applicable to both low and high resource settings. The PTs require further testing against data from low-income and middle-income countries.
Research papers looked at the opinions and satisfaction of staff who were still employed within their respective trust. No studies included in the review focussed on those staff who had already left and why.
This realist review facilitated the construction of robust, evidence-based and stakeholder informed PTs about the mechanisms underlying mental health workforce retention. These explanations hold the potential to support the future development and delivery of effective retention strategies and interventions. The logic model illustrates the interconnectivity of the three overarching theories. Findings highlight the importance of organisation context as preceding and activating mechanisms to produce particular retention outcomes. The findings further suggest that workforce retention interventions may need to first focus on organisational policy changes which can influence context to achieve an effect downstream.
Retention interventions which focus on ‘outcomes’ such as team cohesion may have some influence on outcomes. However, much of the evidence explored workload and staff development, so often the product of organisational decision-making. Frontline reconfigurations and individual-level interventions cannot help in the long term if organisations do not demonstrate to their staff that they are valued and supported to make a difference to patients. ‘Demonstrate' is key here, if staff do not feel valued and supported, then changes will be resisted.
Data availability statement
Data sharing not applicable as no datasets generated and/or analysed for this study.
Patient consent for publication
The authors would like to thank Mark Clowes for developing the search strategies and conducting the searches and Sue Ridgway for assisting with screening the search results. We also thank the patient and staff stakeholders for their invaluable contributions at the design and prioritisation stages of the study.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Contributors SO, EW, SW and AB designed the study, JL, EW and MS conducted screening, data extraction and analysis, SO, AB and EW conducted the initial searches and stakeholder involvement. JL, SO and EW wrote the paper. EW is the guarantor. All authors contributed to theory and logical model design and commented on the paper.
Funding The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article. This work was supported by The Health Foundation (grant number 1275277).
Competing interests None declared.
Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting or dissemination plans of this research. Refer to the Methods section for further details.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.