Greenspace interventions for mental health in clinical and non-clinical populations: What works, for whom, and in what circumstances?

https://doi.org/10.1016/j.healthplace.2020.102338 Received 3 October 2019; Received in revised form 18 February 2020; Accepted 8 April 2020 Health and Place 64 (2020) 102338 2 (van den Berg & van den Berg, 2014). Previous systematic reviews of greenspace interventions for mental health improvements have provided some evidence of their effectiveness (Bowen et al., 2016; Cipriani et al., 2017; Genter et al., 2015; Gorman and Cacciatore, 2017). Gorman and Cacciatore (2017) undertook the first systematic review of care farming, and highlighted that, while understudied, care farming appears to benefit people experiencing psychological distress and could be a feasible non-medicalised approach to improving mental health. Cipriani et al. (2017) found that 11 out of 14 horticultural therapy studies showed significant mood and performance improvements for people with mental health conditions while Genter et al. (2015) found that allotment gardening provided therapeutic benefits and improved health and wellbeing. While the effect size was small, Bowen et al. (2016) also found that wilderness adventure therapy produced statistically significant mental health improvements in young people over ten weeks. However, in these systematic reviews it is unclear why an intervention works and what mechanisms of change lead to the desired outcomes (Norton et al., 2014). For example, a greenspace intervention designed to decrease stress might be deemed effective if it led to quantitative differences in outcome measures such as blood pressure or cortisol levels (Roe et al., 2013). From a qualitative viewpoint, an intervention that has led to participants reporting positive changes, such as lower perceived stress levels, might also be deemed effective (Ellingsen-Dalskau et al., 2016). However, without knowing the necessary components, processes and influences needed for an intervention to work, it is impossible to understand why the programmes work and how best to replicate them. More in-depth reviews such as Lovell et al. (2015) and Husk et al. (2016) have produced more detailed conceptual models of the mechanisms by which engagement with nature impacts physical and mental health. We have built on Fig. 1. Key steps in a realist review as detailed in Pawson et al. (2005). W. Masterton et al. Health and Place 64 (2020) 102338 3 evidence in these reviews by focusing on context, and on what works ‘for whom’ and ‘in what circumstance’. Different contexts are likely to facilitate different mechanisms and outcomes, and what ‘works’ in one setting might not ‘work’ in a different one. To address this, realist methodology will be used to synthesise the evidence more broadly for greenspace interventions for mental health. A realist review is a defined as a ‘method for studying complex interventions in response to the perceived limitations of conventional systematic review methodology. It involves identification of contexts, mechanisms and outcomes for individual programmes in order to explain differences, intended or unintended, between them’ (Booth et al., 2016, p. 267). By using realist methodology, the underlying mechanisms and processes through which greenspace can improve mental health will be identified. This will allow a far greater theoretical understanding of the intervention process, rather than simply deducing whether an intervention is effective or not. Realist review methodology is becoming an increasingly popular way to synthesise public health interventions, given that they are complex by nature (Pawson et al., 2005). Wight et al. (2016) describe public health interventions as complicated and multicomponent, with many feedback loops, rather than simple, easily replicated entities. Greenspace interventions are an example of complex, public health interventions: the setting is in an uncontrolled environment, they are ideally run by multidisciplinary teams, and there are often many intervention components. The interventions may change in regard to context, and all programme components interact leading to outcomes that differ depending on such contextual factors (Wong et al., 2010). For these reasons, a realist review is the most appropriate methodology to synthesise existing greenspace interventions. Pawson et al. (2005) propose five steps which help guide the realist review process. These steps are iterative rather than sequential and each stage can influence another. For example, review questions might be refined after initial programme theory formulation, or the programme theory might be refined at any point when new evidence emerges. Steps 1–5 as reported in Pawson et al. (2005) are shown in Fig. 1. 1.1. Aims and objectives of review The aim of this realist review is to explore what greenspace interventions work to improve mental health, how they work, why they work, for whom do they work, how does context influence mechanisms of change, and how do mechanisms of change lead to outcomes. The objective of the review, therefore, is to develop initial programme theories and then test and refine these theories using both quantitative and qualitative evidence. 1.2. Review questions 1. What interventions, theories or strategies have been used in greenspace interventions that aim to improve mental health (as defined above) in both clinical and nonclinical samples? 2. What outcome measures (O) are associated with current greenspace interventions (e.g. quality of life, increased confidence, increased mood)? 3. What are the potential mechanisms (M) that influence outcomes? 4. What is the role of context (C) in enabling/constraining the above mechanisms? 5. What is the optimal C-M-O configuration that will lead to optimal outcomes in greenspace interventions to improve mental health? 2. Methods 2.1. Formation of initial programme theory Realist reviews aim to develop theories about how an intervention works. Central to a realist review is identifying the causal mechanisms that lead to an outcome, and in what contexts these mechanisms occur (Wong et al., 2013). This relationship is referred to as the C-M-O configuration (CMOc). By using this methodology realist reviews provide a theory-driven approach to analysing literature and identifying causal relationships. Unlike systematic reviews, meta-analyses or qualitative evidence syntheses, realist reviews analyse quantitative, qualitative, and mixed-method data, as well as grey literature (Abrams et al., 2018). Information about ‘what works’ is analysed using the findings of each paper, as well as through data extraction and synthesis from other sections of the paper which may inform theoretical understanding of causal pathways. The first step of this review was initial exploration of literature and theory formulation about how greenspace interventions for mental health might be effective. This involved comparing and synthesizing relevant theories and hypothesising how a greenspace intervention is thought to work to achieve desired outcomes. This initial theory mapping provided the proposed framework for the review about what works, for whom and in what circumstances. This framework (initial programme theories, IPTs) was then tested and refined throughout the realist review process as evidence emerged. The main IPTs were developed initially by the first author (WM) through reading existing literature on greenspace interventions for mental health, conversations with existing greenspace programme staff, and by reading relevant policy documents and reports which discuss conceptual frameworks in relation to practice. These IPTs were checked by the second author (HC) and then by the wider team (TP, KP). This ensured that all authors were involved, and in agreement with, the development of the IPTs. By using this approach, relevant contexts, mechanisms, and outcomes were identified for several different programmes and potential CMOcs developed. The guiding questions for initial theory formulation are ‘what outcome measures are associated with current greenspace interventions?‘, ‘what are the potential mechanisms that influence outcomes?‘, ‘what is the role of context in enabling/constraining potential mechanisms?‘, and ‘what is the optimal C-M-O configuration that will lead to optimal outcomes in greenspace interventions for mental health?’ Table 1 shows the eight IPTs proposed under three identified programme theory themes of Nature, Individual Self, and Social Self. To further clarify how contexts, mechanisms, and outcomes fit together in a causal relationship, ‘if-then-because’ statements are included under each IPT. 2.2. Testing the explanatory framework To test and refine programme theories a selection of relevant electronic databases were searched between May and July 2019 in order to achieve saturation of results. These were: MEDLINE, PsycINFO, GreenFile, SocINDEX, CINAHL, Health Source, SPORTDiscus, Scopus, Web of Science, Natural Science Collection, and Wiley Online Library. Searches were limited to studies published after 2000 to ensure that included evidence was current. Qualitative, quantitative, and mixed-methods papers were included. Several terms are used interchangeably for ‘greenspace’ and ‘mental health’ so a number of terms were included in the search string (see Table 2). Grey literature was searched in June 2019 through search engines (Google, Google Scholar), grey literature databases (OpenGrey, Social Care Online), relevant organisational websites and reports (see Table 3), social media platforms such as Twitter, and through word of mouth. 2.3. Inclusion/exclusion criteria Inclusion criteria aligned to both the research questions and IPT development, as suggested by Wong et al. (2013), refined in response to emerging data, and discussed as a team to reach agreement. All programmes had to be greenspace-based however this could include gardens, woodland, plots, parks, and other types of greenspace. All age groups were included. In terms of mental health, both non-clinical and clinical studies were included in the search strategy. Participants could have a mental health diagnosis or be self-diagnosed; as many greenspace W. Masterton et al. Health and Place 64 (2020) 102338 4 interventions are applied in a similar manner to specific and general populations. Programmes were included if improved mental wellbeing was an explicit intended outcome. The exclusion criteria were developed and refined in response to emerging data and again were discussed as a team to ensure consensus. A decision was made to exclude studies focused on dementia because, upon initial analysis, the CMOcs appeared very different to those for mental health. It was unclear whether those with dementia had the capacity to reflect meaningfully on their experiences and if these studies could effectively answer the review questions. Furthermore, many of these studies were implemented inside and could not be described as ‘greenspace’ programmes. 3. Results 3.1. Search results and study characteristics In the first stage of searching, after removing duplicates, 2119 titles and abstracts were screened against the inclusion and exclusion criteria: 2095 studies identified through database searching, 19 grey literature sources, and 5 studies through citation searching. In a realist review, the search process is iterative, and during a final search for evidence, another 8 empirical studies and 1 grey literature evaluation were identified. In total, 113 potentially eligible studies were identified in this Table 1 Initial programme theories identified to be tested and refined. Theme Initial Programme Theory (IPT) number Context (C) Mechanism (M) Outcome (O) Nature 1 Nature-based location Ease of access Feeling calm Feelings of escape Feeling removed from everyday life Decrease anxiety Decrease stress IPT 1: If there is easy access to a nature-based location, then participants may experience decreased anxiety and/or stress. This is because they can feel removed from everyday life, experience feelings of escape in nature, and feel calm. If the nature-based location is not easy to access, it is much less likely that people will go there. Nature 2 Nature-based location Ease of access Indirect attention used Attention restoration Decrease mental fatigue IPT 2: If there is easy access to a nature-based location, then participants may experience a decrease in their mental fatigue, as well as feel that their attention has been restored. This is because indirect, or effortless, attention, as described in Kaplan and Kaplan’s Attention Restoration Theory (Kaplan and Kaplan, 1989), is being used when immersed in nature rather than direct attention. If the nature-based location is not easy to access, it is much less likely that people will go there. Nature 3 Nature-based location Ease of access Time alone to reflect Increase in readiness to change lifestyle and/or coping strategies Increase in desire to change Increase in awareness of the need for change IPT 3: If there is easy access to a nature-based location, then participants may experience an increase in readiness to change, an increase in desire to change, and/or an increase in awareness of the need for change. This is because the nature-based location gives participants time along to reflect on their lives and what they want to change. If the nature-based location is not easy to access, it is much less likely that people will go there. Individual 4 Availability and resources for trained facilitators Access to resources Planned structured activities Enjoyment of activities Increased physical activity Increased physical health Improvement in mood IPT 4: If there is the availability and the resources to provide trained facilitators, and these trainers have access to a variety of resources (such as equipment), then there will be an increase in physical activity, and a subsequent increase in physical health and improvement in mood. This is because there will be the availability of a number of different planned, structured activities from the trained facilitators, and participants can pick what they would like to do best, and therefore enjoy the activity. Individual 5 Availability and resources for trained facilitators Learning new skills Feelings of self-efficacy Confidence Confidence in ability to change and cope with challenges in life IPT 5: If there is the availability and the resources to provide trained facilitators, then this will enable an increase in participant confidence, as well as in their confidence to change and cope with challenges in life. This is because participants are able to learn new skills from the facilitators, which lead to feelings of self-efficacy. Individual 6 Time on programme Availability and resources for trained facilitators Learning new skills Feeling responsible for something Increased self -esteem Increased vigour for life IPT 6: If there is the availability and the resources to provide trained facilitators, and if there is adequate time spent on the programme, then participants will show an increase in self-esteem and an increase in vigour for life. This is because participants are able to learn new skills from the facilitators, as well as feelings of responsibility. The longer that the participant is able to feel responsible for something, the bigger the increase in self-esteem and vigour for life. Social 7 Previous experience of patient-therapist relationship Existing facilitator attitudes and/or perceived attitudes of facilitator Feelings of rapport and trust Good relationship with facilitator Continued engagement with, and after, the programme. IPT 7: If facilitators have positive attitudes, then participants are more likely to engage with, and after, the programme. This is because, when participants perceive a positive attitude towards them, feelings of rapport and trust are more likely to develop, and a good relationship with the facilitator can be established. Previous experience of a patient-therapist relationship can also influence continued engagement with, and after, the programme. If there is a positive previous experience, then this can lead to engagement. This is because feelings of rapport and trust can be built quicker, and participants can more easily develop a good relationship with the facilitator. Social 8 Perception of how others are engaging on the programme Time on programme Team building/teamwork exercises Feeling safe and unjudged by others with similar backgrounds Feelings of rapport Opportunities to share Opportunities to learn from others Increased social abilities Improvements in interpersonal relationships IPT 8: If participants perceive others to be engaging well on the programme, then this can lead to increased social abilities and improvements in interpersonal relationships. This is because, when participants perceive others to be engaging, this increases feelings of rapport between participants. This can lead participants to feel safe and unjudged by others during team building/teamwork exercises where there are opportunities to share and learn from others. Even if others are perceived to be engaging well, time spent on the programme is also important in order to achieve outcomes. This is because social improvements do not occur quickly, and interpersonal relationships take time to build. W. Masterton et al. Health and Place 64 (2020) 102338 5 process so full texts were obtained. As a result of further close reading of full texts, 49 articles were identified and included. Literature searching and screening results are reported in Fig. 2 using PRISMA (Moher et al., 2009). Information provided in each study about the programmes and articipants varied, and key characteristics of all included studies were recorded (Table 4 and Table 5). 3.2. Relevance and rigour Following the guidance set in the quality standards for realist reviews (Wong et al., 2013), each study was appraised for relevance and rigour. Relevance was assessed in relation to three criteria: population, intervention, or study design; explanation of context, mechanism and outcome as individual aspects as well as in combinations; and explanation of theory. In realist reviews, studies can be included even if only a small part is relevant. This can mean that a certain amount of subjective judgement is necessary to ensure the number of included studies is not unmanageable. Similarly, in realist reviews, studies are assessed for rigour in a different way from systematic reviews: standard quality assessment tools are not used due to the risk of ‘nuggets of wisdom’ (Pawson, 2006a) being missed due to discarding papers deemed methodologically weak. As is advised in the quality standards (Wong et al., 2013), we identified whether the methods in each study were rigorous enough to be able to rely on the small percentage of findings that we needed to draw on and use in our review. However, as discussed in Pawson (2006a), even studies typically deemed methodologically weak can be included, with careful analysis and appraisal, since they may explicitly, or implicitly, allow insight into why an intervention did not work. To ensure that the risk of bias was reduced, a second reviewer (HB) checked a selection of included/excluded papers to ensure validity and consistency. Where there was inconsistency, a thorough discussion was held to decide whether to include or exclude the study. 3.3. Testing and refinement of programme theory Detail on contexts, mechanisms, and outcomes of each included study were recorded in an Excel spreadsheet. Data extraction and synthesis were undertaken by the first author (WM), with results regularly discussed with the rest of the study team (HC, TP, KP) to ensure consistency, and reduce bias when refining programme theories. Ongoing conversations with greenspace organisation staff were held throughout the search and appraisal process to further ensure that programme theories accurately described the underlying mechanisms and causal pathways of the interventions. The development of IPTs into seven refined programme theories is described below. It became clear during data synthesis that IPTs did not adequately integrate the ‘for whom’ and ‘in what circumstance’ aspects of the realist method. Therefore, while the programme theory themes stayed similar, there was refinement and greater emphasis placed on these contextual factors given how essential they are for implementation and targeting. Fig. 3 shows a brief outline of how the identified programme theories fit in to three overarching themes. The seven programme theories are represented by headings which we believe best describe their core concept. 1. Escape/Getting Away Perhaps unsurprisingly, given the integral part nature plays in the programmes, most of the included studies mentioned the importance of immersion in greenspace for mental health benefits. Fernee et al. (2018) discuss how the role of the wilderness created a calming effect on participants, in contrast to their usually chaotic lives, and how the calming environment facilitated cognitive processes such as reflection. Participants in Kogstad, Agdal, and Hopfenbeck’s study (2014) described immersion in nature as feeling like a cloud had been lifted, while participants in the study by McIver et al. (2018) reported that immersion in nature helped reduce rumination and stressful, negative thoughts. One participant in O’Brien, Townsend, and Ebden’s study (2010) reported that he felt sitting on the hillside for 10 minutes was as effective as his antidepressant medication. A number of the studies made reference to well-established theories such as Attention Restoration Theory (Kaplan and Kaplan, 1989) which holds that when a person is immersed in nature this leads to feelings of calm and a reduction in mental fatigue. Previous reviews such as Berto (2014) and Hartig et al. (2014) have also supported the role of attention restoration and a reduction in mental fatigue as mechanisms in health improvement. One of the components of Attention Restoration Theory is that immersion in nature allows a person to feel removed from their everyday life and, therefore, from their everyday stressors. This feeling of escape, or ‘getting away’, is a key mechanism in the success of greenspace programmes. In the Nacadia Therapy Garden, service users described the garden as ‘a magical world of its own’ (Sidenius et al., 2017, p. 5), whereas other participants described being out in nature as ‘like another world’ and ‘sort of like part of the world but a pocket. A haven pocket’ (Stevens, 2018, p.7 & p.9 respectively). Refinement of programme theory IPT 1 and IPT 2, as shown in Table 1, were condensed into the above encapsulating programme theory of Escape/Getting Away. The greenspace setting was a key contextual factor as it provided the right supportive environment but also acted as the resource (mechanism resource), otherwise understood as the programme strategy or programme component introduced in a context. Programmes that utilise greenspace, and allow participants to feel as if they are escaping from their day-to-day lives, are shown to be particularly effective for participants with experience of trauma, anxiety, depression, suppressed anger, and other emotions, conflicts in relationships, as well as for people who explicitly state that they need help (Bettmann et al., 2011; Russell and Phillips-Miller, 2002). As well as existing diagnoses, the greenspace setting was particularly effective for participants who had previous experience of more typical treatments such as counselling (context), as Table 2 Search terms in published literature. Databases Searched Search Terms MEDLINE PsycINFO GreenFile SocINDEX CINAHL Health Source SPORTDiscus Scopus Social Care Online Web of Science Natural Science Collection Wiley Online Library greenspace OR “green space” OR “green care” OR greencare OR “nature therap*” OR “wilderness therap*” OR “outdoors behavi*ral healthcare” OR “outdoors behavi*ral therap*” OR “forest bathing” OR “shinrin yoku” OR “shinrin-yoku” OR “horticultur* therap*” OR “therapeutic horticulture” OR “green exercise” OR ecotherap* OR “conservation therap*" OR “care farm*" AND “mental health” OR “mental ill health” OR “mental illness” OR “mental disorder” OR “mental fatigue” OR psychiatric OR “psychiatric illness” OR stress OR depression OR anxiety OR recovery OR “low mood” OR wellbeing Table 3 Organisations included in search for grey literature. UK Europe International Venture Trust Asociacion Experientia (Spain) Enviros (USA) Phoenix Futures Shepherd’s Hill Academy (USA) The Wilderness Foundation Rites of Passage (USA) Forest Therapy Scotland Redcliff Ascent (USA) Cyrenians Venture Scotland The Green Team Youth Vision Venture Mor W. Masterton et al. Health and Place 64 (2020) 102338 6 they no longer felt as if they were confined within four walls (Fernee et al., 2018; Granerud and Eriksson, 2014; Sidenius et al., 2017; Woodford et al., 2017). It is possible, therefore, that nature-based programmes are most appropriate for participants who have previously attended traditional therapies which they believe were unsuccessful. The feeling of being away, relaxed, and removed from daily life, was shown to be further facilitated by sensory stimuli (context and mechanism resource) present in the environment (Adevi and Lieberg, 2012; Grahn et al., 2017; Harris, 2017; Rappe et al., 2008; Sidenius et al., 2017). There was some evidence that ease of access to the programme sites was a contextual factor, with one study highlighting that not owning a car to get to sites could be a barrier (O’Brien, 2018), and Husk et al. (2020) ) state that support to get to the location of the programme was necessary for success. Additionally, during a discussion with greenspace programme staff, one manager emphasised that access to minibuses could influence the ease by which the programme was attended so could be a potential contextual factor. Changes in participant reasoning (mechanism reasoning) occur as a result of introduced resources and together these constitute the programme mechanism. In this programme theory, stress levels and mental fatigue were reduced (outcomes) through indirect attention being used (mechanism reasoning), and through the participant feeling removed, relaxed, and ‘getting away’ from their stressors (mechanism reasoning). The WHO report Urban Green Space: A Review of the Evidence (WHO, 2016) discusses the importance of taking account of gender differences in response to exposure to greenspace, however, with a previous longitudinal study by van den Bosch et al. (2015) reporting positive associations between exposure to greenspace and mental health in women, but not men. Furthermore, Combs et al. (2016) reports that female participants showed a faster decrease in stress than male participants, suggesting that a shorter stay on a programme may work for female groups. Such findings suggest that men and women may respond differently to the greenspace environment on programmes so should be considered during programme development. It is also worth noting that cultural differences can influence how well a participant engages with a greenspace programme in the first instance. For example, during conversations with greenspace programme staff, we identified that uptake of greenspace programmes such as forest therapy is much higher in Japanese and Korean culture where time in forests is an integral part of that lifestyle. The normalisation of forest therapy in these cultures will likely have an influence on uptake and engagement of programmes compared to countries where there is stigma attached to such ideas. 2. Space to Reflect The contextual role of greenspace setting is discussed in the above Fig. 2. PRISMA diagram. W. Masterton et al. Health and Place 64 (2020) 102338 7 Escape/Getting Away programme theory and is also integral to this programme theory. In this programme theory, the greenspace environment acts as a catalyst for change, with McIver et al. (2018), and participants on the Living Wild programme (Venture Trust, 2019), describing nature as a mediator in preparing a person for a therapeutic experience. Sidenius et al.‘s study (2017) supports this describing nature as providing a ‘backdrop’ where therapeutic conversations and activities were more accessible. An integral part of this programme theory is that time alone in greenspace can allow participants to reflect on their lives. This is particularly important for those with coping strategies which may be harmful to them, such as using drugs, alcohol, or self-harm (Bettmann et al., 2011). Participants on a wilderness therapy programme (Fernee et al., 2018) spoke about the physical space allowing them to reflect in a prolonged and undisturbed way, both when sitting and walking. This, in turn, can increase their awareness of the need for change in their lives (Hassink et al., 2010; McIver et al., 2018; Russell and Phillips-Miller, 2002), and how to ‘live a better life’ (Fieldhouse, 2003, p.90). Refinement of programme theory As in the Escape/Getting Away programme theory, the greenspace setting provided the supportive environment for the programme (context and mechanism resource). The context of adequate time spent on the programme was a refinement to this programme theory since change and reflection did not happen quickly (Kogstad et al., 2014; P� alsd� ottir et al., 2014; Schreuder et al., 2014; Sidenius et al., 2017). Participants in Gabrielson et al. (2018) believed change happened due to the number of unique experiences participants have during programmes, but stated that change could take months to become apparent. Within these contexts, awareness of the need to change (outcome) was achieved by participants spending time alone and reflecting on their lives (mechanism reasoning). Additionally, the desire to change (outcome) could be facilitated by metaphors encountered within the programme (mechanism resource) and participants applying these to their own lives (mechanism reasoning). An example of this was a description of how trying to control a canoe and fight against its natural course proved more difficult than letting nature take its course around obstacles: a metaphor for trying to control life and avoid obstacles (McIver et al., 2018). Adevi and Lieberg (2012) also discussed how participants may seek out specific places in a therapy garden, depending on their emotional state. This is an important contextual factor as it highlights that the most therapeutic place for a participant to reflect is highly individual. It is, therefore, important that participants self-choose places that they have a connection with, or feel comfortable in (mechanism response). According to the staff in this garden programme, reliance on the self-chosen places appears to create greater confidence over time (outcome). Pre-existing diagnoses were also identified as a crucial contextual refinement for this programme theory, particularly important when designing programmes for specific populations. For example, extensive time alone for reflection is not appropriate for participants with existing diagnoses such as severe depression or psychosis (Fernee et al., 2017). 3. Physical Activity Enjoyment of physical activity appears to be the mechanism that best allows increased physical health and fitness. Two participants in Fernee et al.‘s study (2018) described how, even though they felt tired during physically challenging hikes, they still felt happier when taking part and therefore found it easier to push themselves. However, caution must be taken before generalising this finding: Caulkins et al. (2006) highlighted how young women in their study appeared to benefit less from wilderness hikes compared to other participants, due to higher levels of aversion to the outdoors. Evans (2013) suggests that greenspace programmes provide participants with unique, exciting experiences which encourage people to participate. Furthermore, with increases in physical activity, improvements in mood are also seen (Bryson et al., 2013; Eriksson et al., 2011; Fernee et al., 2018; Fieldhouse, 2003; Leck et al., 2015; O’Brien, 2018; Wilson et al., 2010). This supports existing systematic reviews and meta-analyses supporting the role of physical activity on mental health (Bize et al., 2007; Penedo and Dahn, 2005; Rosenbaum et al., 2014). Conversations with service managers identified that availability of resources for equipment is an important contextual factor for successful engagement with physical activities; as programmes must be fully equipped and functional. However, Surridge et al. (2004) discuss how resources from stakeholders can also be in the form of support and advice in areas such as risk assessment and group safety. Six studies stressed the importance of having confident, adequately trained facilitators to enable and lead activities (Bloomfield, 2017; Evans, 2013; Granerud and Eriksson, 2014; Kogstad et al., 2014; O’Brien et al., 2010; Surridge et al., 2004). With a diversity of activities available, participants are more likely find an activity that they enjoy and will engage with. Refinement of programme theory The availability of trained facilitators (context), and availability of resources to adequately support and fund programmes and their materials (context), is imperative to provide a variety of activities to service users (mechanism resource). This allows participants to engage with activities they enjoy (mechanism reasoning), and provides personcentred approaches (Barley et al., 2012; Bloomfield, 2017; Cole and Christie, 2016; Harris, 2017; O’Brien, 2018; Schreuder et al., 2014). This enjoyment of physical activity (mechanism reasoning) facilitates engagement (outcome) and, in turn, leads to increased physical activity, improvements in physical health, and improvements in mood (outcomes). However, particularly in winter weather (context), participants who did not like spending time outside found it difficult to enjoy any aspect of the programme due to discomfort (mechanism reasoning) (Harper et al., 2019), limiting positive outcomes. However, realistic expectations of anticipated challenges did appear to be an influencing context in the lead up to programme uptake in some circumstances. Gabrielsen et al., (2018) suggest that clearly informing and preparing participants for any challenges prior to the programme commencing is advisable, in particular, ensuring participants have the right equipment (context), such as waterproof clothing and shoes. 4. Self-Efficacy Twenty-eight studies reported that service users who learned and mastered new skills had increased self-esteem, pride, and confidence. Indeed, existing evidence supports continued learning as a mechanism for mental health improvement (Feinstein and Hammond, 2004; Fig. 3. Three programme themes and subsequent representative headings for the seven programme theories identified through data synthesis. W. Masterton et al. Health and Place 64 (2020) 102338 8 Ta bl e 4 In cl ud ed e m pi ri ca l s tu di es in th e re vi ew . A ut ho rs D at e of Pu bl ic at io n Co un tr y Ti tle In te rv en tio n A va ila bl e pa rt ic ip an t in fo M en ta l h ea lth o ut co m es M et ho ds IP T Su pp or te d A de vi & L ie be rg 20 12 Sw ed en St re ss r eh ab ili ta tio n th ro ug h ga rd en th er ap y: A c ar eg iv er pe rs pe ct iv e on fa ct or s co ns id er ed m os t e ss en tia l t o th e re co ve ry p ro ce ss H or tic ul tu ra l Th er ap y 5 pa rt ic ip an ts Im pr ov em en ts in s tr es sre la te d ill he al th In -d ep th in te rv ie w s an d fo cu sgr ou p in te rv ie w s w ith th e re ha bi lit at io n te am m em be rs o f A ln ar p Re ha bi lit at io n G ar de n in S w ed en lo ok in g at th ei r un de rs ta nd in gs o f s ig ni fic an t f ac to rs to th e st re ss r ec ov er y pr oc es s 1, 2 , 3 , 5 , 8 A de vi & M år te ns so n 20 13 Sw ed en St re ss r eh ab ili ta tio n th ro ug h ga rd en th er ap y: T he g ar de n as a pl ac e in th e re co ve ry fr om st re ss H or tic ul tu ra l Th er ap y 5 pa rt ic ip an ts 4 fe m al e, 1 m al e Im pr ov em en ts in s tr es sre la te d ill he al th In te rv ie w s w ith p ar tic ip an ts o f A ln ar p Re ha bi lit at io n G ar de n in S w ed en w ho de sc ri be th ei r ex pe ri en ce s of ho rt ic ul tu ra l t he ra py a nd w ha t t he y pe rc ei ve a s es se nt ia l f or th ei r re co ve ry 1, 2, 3, 4, 5, 6, 8 Ba rl ey e t a l. 20 12 U K Pr im ar y ca re -b as ed pa rt ic ip at or y re ha bi lit at io n: us er s’ v ie w s of a h or tic ul tu ra l an d ar ts p ro je ct H or tic ul tu ra l Th er ap y 16 p ar tic ip an ts 7 fe m al e, 9 m al e A ge s 38 -9 1 Im pr ov ed m en ta l w el lb ei ng in th os e w ith s ev er e m en ta l h ea lth d ia gn os es Se m i-s tr uc tu re d in te rv ie w s w ith pa rt ic ip an ts o f S yd en ha m G ar de n in So ut h Lo nd on lo ok in g at th ei r v ie w s a nd ex pe ri en ce o f t he h or tic ul tu ra l pr og ra m m e 1, 2, 3, 5, 6, 8 Be tt m an e t a l. 20 11 U SA A do le sc en ts in w ild er ne ss th er ap y: A q ua lit at iv e st ud y of at ta ch m en t r el at io ns hi ps W ild er ne ss Th er ap y 13 p ar tic ip an ts Im pr ov em en ts in in te rp er so na l re la tio ns hi ps a nd w el lb ei ng A n ev al ua tio n of th e na rr at iv es o f 1 3 ad ol es ce nt s in a w ild er ne ss th er ap y pr og ra m in a n at te m pt to u nd er st an d th ei r at tit ud es to w ar d pa re nt s, th er ap is ts , a nd o th er a du lt fig ur es 1, 2, 6 Bl oo m fie ld 20 17 U K W ha t m ak es n at ur eba se d in te rv en tio ns fo r m en ta l h ea lth su cc es sf ul ? Ec ot he ra py 48 p ar tic ip an ts Im pr ov ed g en er al m en ta l w el lb ei ng Su m m ar y of h ea lth a nd w el lb ei ng be ne fit s de ri ve d fr om A D os e of N at ur e pr oj ec t i n En gl an d, U K. K ey fa ct or s in vo lv ed in d ev el op m en t o f in te rv en tio ns w er e di sc us se d by a ut ho rs 5, 6, 7, 8 Br ys on e t a l. 20 13 Ca na da A n ex am in at io n of th e fe as ib ili ty o f a dv en tu re -b as ed th er ap y in o ut pa tie nt c ar e fo r in di vi du al s w ith p sy ch os is A dv en tu re Th er ap y 21 p ar tic ip an ts 4 fe m al e, 1 1 m al e A ge s 24 -4 8 En ga ge m en t i n th e re co ve ry p ro ce ss , an d im pr ov em en ts in e m ot io na l w el lb ei ng , e ne rg y, s el f-e st ee m , a nd gl ob al h ea lth in p eo pl e w ith Sc hi zo ph re ni a an d ot he r ps yc ho se s M ix ed -m et ho ds p re -p os t d es ig n to ex am in e th e fe as ib ili ty o f a 6 -w ee k Ri se U p A dv en tu re T he ra py in te rv en tio n in an o ut pa tie nt c ar e se tt in g. Q ue st io nn ai re s w er e gi ve n pr epo st th e in te rv en tio n an d in te rv ie w s w er e co m pl et ed p os t i nt er ve nt io n 4, 5, 7, 8 Ca ul ki ns e t a l. 20 06 U SA Th e ro le o f p hy si ca l e xe rc is e in w ild er ne ss th er ap y fo r tr ou bl ed ad ol es ce nt w om en W ild er ne ss Th er ap y 9 fe m al e pa rt ic ip an ts ( 6 on th e pr og ra m m e ag es 15 –1 6, a nd 3 in st ru ct or s) Em ot io na l, co gn iti ve , a nd p hy si ca l ch an ge s A q ua lit at iv e ca se s tu dy a pp ro ac h w as us ed to in ve st ig at e th e ex pe ri en ce o f s ix ad ol es ce nt w om en a nd th re e fe m al e w ild er ne ss in st ru ct or s at a n es ta bl is he d w ild er ne ss th er ap y pr og ra m . D at a w er e co lle ct ed th ro ug h pa rt ic ip an t ob se rv at io n, c lie nt p sy ch ol og ic al pr ofi le s, a nd se m i-s tr uc tu re d in te rv ie w s. Re se ar ch e xp lo re d th e em ot io na l, co gn iti ve , a nd p hy si ca l i m pa ct s at tr ib ut ed to th e ba ck pa ck in g co m po ne nt o f t he th er ap eu tic p ro ce ss an d th e re la tio ns hi ps b et w ee n th es e


Introduction and background
The beneficial effect of nature on human health and wellbeing is a concept that has been widely accepted since the 1800s (Hickman, 2013). Since then international agreements and organisations such as the World Health Organisation (WHO) have supported the establishment and maintenance of urban greenspaces to promote health and wellbeing, and have reviewed their effectiveness in contributing to healthy, sustainable cities (WHO, 2017). Within public health the positive effects of greenspace are becoming increasingly publicised (van den Berg & van den Berg, 2014). The WHO defines public health as 'the science and art of preventing disease, prolonging life and promoting health through the organized efforts of society, organisations, public and private, communities and individuals' (Acheson, 1988). Public health therefore encompasses all public and private organisations and all resources that aim to positively impact the health of the whole population. From a public health perspective greenspace can be defined and characterised by its ability to provide healing and 'green care' (Haubenhofer et al., 2010). Greenspace can be used to achieve health outcomes, such as a reduction in stress or an increase in positive mood, in a variety of settings from public parks and woodlands to gardens in hospitals and care homes (Frumkin, 2013). Understanding and recognition of how greenspace can contribute to public health is potentially significant for addressing numerous physical health-related issues, such as obesity, but is equally important to facilitating good mental health and addressing negative mental health.
The term 'mental health' is most commonly used to describe the state of a person's psychological wellbeing, running on a continuum from positive mental health to poorer mental health (Pilgrim, 2017). For this review, we are interested in how greenspace interventions might be effective in improving mental health in those who have a poor mental health diagnosis, or in those who have expressed concern about their own mental health. While 'mental health' is neither positive nor negative by definition (Pilgrim, 2017), the population inclusion criteria for our study means that the term, in this review, is more likely to represent a continuum of states from mild to moderate low mood to severe mental ill health. It is estimated that one in four people in the UK will experience a mental health problem at some point in their life, the most common being anxiety and depression (Kendrick et al., 2015). One of the benefits of using nature to aid mental health recovery is that it can be used alongside a more typical medicalised treatment plan such as talking therapy and interventions could potentially be implemented anywhere. Indeed, greenspace interventions could be a promising addition to both current health and social care provisions as they have the potential to be low-cost and widely accessible for people within their own communities (van den Berg & van den Berg, 2014).
Previous systematic reviews of greenspace interventions for mental health improvements have provided some evidence of their effectiveness (Bowen et al., 2016;Cipriani et al., 2017;Genter et al., 2015;Gorman and Cacciatore, 2017). Gorman and Cacciatore (2017) undertook the first systematic review of care farming, and highlighted that, while understudied, care farming appears to benefit people experiencing psychological distress and could be a feasible non-medicalised approach to improving mental health. Cipriani et al. (2017) found that 11 out of 14 horticultural therapy studies showed significant mood and performance improvements for people with mental health conditions while Genter et al. (2015) found that allotment gardening provided therapeutic benefits and improved health and wellbeing. While the effect size was small, Bowen et al. (2016) also found that wilderness adventure therapy produced statistically significant mental health improvements in young people over ten weeks. However, in these systematic reviews it is unclear why an intervention works and what mechanisms of change lead to the desired outcomes (Norton et al., 2014). For example, a greenspace intervention designed to decrease stress might be deemed effective if it led to quantitative differences in outcome measures such as blood pressure or cortisol levels (Roe et al., 2013). From a qualitative viewpoint, an intervention that has led to participants reporting positive changes, such as lower perceived stress levels, might also be deemed effective (Ellingsen-Dalskau et al., 2016). However, without knowing the necessary components, processes and influences needed for an intervention to work, it is impossible to understand why the programmes work and how best to replicate them. More in-depth reviews such as Lovell et al. (2015) and Husk et al. (2016) have produced more detailed conceptual models of the mechanisms by which engagement with nature impacts physical and mental health. We have built on Fig. 1. Key steps in a realist review as detailed in . evidence in these reviews by focusing on context, and on what works 'for whom ' and 'in what circumstance'. Different contexts are likely to facilitate different mechanisms and outcomes, and what 'works' in one setting might not 'work' in a different one. To address this, realist methodology will be used to synthesise the evidence more broadly for greenspace interventions for mental health.
A realist review is a defined as a 'method for studying complex interventions in response to the perceived limitations of conventional systematic review methodology. It involves identification of contexts, mechanisms and outcomes for individual programmes in order to explain differences, intended or unintended, between them' (Booth et al., 2016, p. 267). By using realist methodology, the underlying mechanisms and processes through which greenspace can improve mental health will be identified. This will allow a far greater theoretical understanding of the intervention process, rather than simply deducing whether an intervention is effective or not. Realist review methodology is becoming an increasingly popular way to synthesise public health interventions, given that they are complex by nature . Wight et al. (2016) describe public health interventions as complicated and multicomponent, with many feedback loops, rather than simple, easily replicated entities. Greenspace interventions are an example of complex, public health interventions: the setting is in an uncontrolled environment, they are ideally run by multidisciplinary teams, and there are often many intervention components. The interventions may change in regard to context, and all programme components interact leading to outcomes that differ depending on such contextual factors (Wong et al., 2010). For these reasons, a realist review is the most appropriate methodology to synthesise existing greenspace interventions.  propose five steps which help guide the realist review process. These steps are iterative rather than sequential and each stage can influence another. For example, review questions might be refined after initial programme theory formulation, or the programme theory might be refined at any point when new evidence emerges. Steps 1-5 as reported in  are shown in Fig. 1.

Aims and objectives of review
The aim of this realist review is to explore what greenspace interventions work to improve mental health, how they work, why they work, for whom do they work, how does context influence mechanisms of change, and how do mechanisms of change lead to outcomes. The objective of the review, therefore, is to develop initial programme theories and then test and refine these theories using both quantitative and qualitative evidence.

Review questions
1. What interventions, theories or strategies have been used in greenspace interventions that aim to improve mental health (as defined above) in both clinical and nonclinical samples? 2. What outcome measures (O) are associated with current greenspace interventions (e.g. quality of life, increased confidence, increased mood)? 3. What are the potential mechanisms (M) that influence outcomes? 4. What is the role of context (C) in enabling/constraining the above mechanisms? 5. What is the optimal C-M-O configuration that will lead to optimal outcomes in greenspace interventions to improve mental health?

Formation of initial programme theory
Realist reviews aim to develop theories about how an intervention works. Central to a realist review is identifying the causal mechanisms that lead to an outcome, and in what contexts these mechanisms occur (Wong et al., 2013). This relationship is referred to as the C-M-O configuration (CMOc). By using this methodology realist reviews provide a theory-driven approach to analysing literature and identifying causal relationships. Unlike systematic reviews, meta-analyses or qualitative evidence syntheses, realist reviews analyse quantitative, qualitative, and mixed-method data, as well as grey literature (Abrams et al., 2018). Information about 'what works' is analysed using the findings of each paper, as well as through data extraction and synthesis from other sections of the paper which may inform theoretical understanding of causal pathways. The first step of this review was initial exploration of literature and theory formulation about how greenspace interventions for mental health might be effective. This involved comparing and synthesizing relevant theories and hypothesising how a greenspace intervention is thought to work to achieve desired outcomes. This initial theory mapping provided the proposed framework for the review about what works, for whom and in what circumstances. This framework (initial programme theories, IPTs) was then tested and refined throughout the realist review process as evidence emerged.
The main IPTs were developed initially by the first author (WM) through reading existing literature on greenspace interventions for mental health, conversations with existing greenspace programme staff, and by reading relevant policy documents and reports which discuss conceptual frameworks in relation to practice. These IPTs were checked by the second author (HC) and then by the wider team (TP, KP). This ensured that all authors were involved, and in agreement with, the development of the IPTs. By using this approach, relevant contexts, mechanisms, and outcomes were identified for several different programmes and potential CMOcs developed. The guiding questions for initial theory formulation are 'what outcome measures are associated with current greenspace interventions?', 'what are the potential mechanisms that influence outcomes?', 'what is the role of context in enabling/constraining potential mechanisms?', and 'what is the optimal C-M-O configuration that will lead to optimal outcomes in greenspace interventions for mental health?' Table 1 shows the eight IPTs proposed under three identified programme theory themes of Nature, Individual Self, and Social Self. To further clarify how contexts, mechanisms, and outcomes fit together in a causal relationship, 'if-then-because' statements are included under each IPT.

Testing the explanatory framework
To test and refine programme theories a selection of relevant electronic databases were searched between May and July 2019 in order to achieve saturation of results. These were: MEDLINE, PsycINFO, Green-File, SocINDEX, CINAHL, Health Source, SPORTDiscus, Scopus, Web of Science, Natural Science Collection, and Wiley Online Library. Searches were limited to studies published after 2000 to ensure that included evidence was current. Qualitative, quantitative, and mixed-methods papers were included. Several terms are used interchangeably for 'greenspace' and 'mental health' so a number of terms were included in the search string (see Table 2).
Grey literature was searched in June 2019 through search engines (Google, Google Scholar), grey literature databases (OpenGrey, Social Care Online), relevant organisational websites and reports (see Table 3), social media platforms such as Twitter, and through word of mouth.

Inclusion/exclusion criteria
Inclusion criteria aligned to both the research questions and IPT development, as suggested by Wong et al. (2013), refined in response to emerging data, and discussed as a team to reach agreement. All programmes had to be greenspace-based however this could include gardens, woodland, plots, parks, and other types of greenspace. All age groups were included. In terms of mental health, both non-clinical and clinical studies were included in the search strategy. Participants could have a mental health diagnosis or be self-diagnosed; as many greenspace interventions are applied in a similar manner to specific and general populations. Programmes were included if improved mental wellbeing was an explicit intended outcome. The exclusion criteria were developed and refined in response to emerging data and again were discussed as a team to ensure consensus. A decision was made to exclude studies focused on dementia because, upon initial analysis, the CMOcs appeared very different to those for mental health. It was unclear whether those with dementia had the capacity to reflect meaningfully on their experiences and if these studies could effectively answer the review questions. Furthermore, many of these studies were implemented inside and could not be described as 'greenspace' programmes.

Search results and study characteristics
In the first stage of searching, after removing duplicates, 2119 titles and abstracts were screened against the inclusion and exclusion criteria: 2095 studies identified through database searching, 19 grey literature sources, and 5 studies through citation searching. In a realist review, the search process is iterative, and during a final search for evidence, another 8 empirical studies and 1 grey literature evaluation were identified. In total, 113 potentially eligible studies were identified in this IPT 1: If there is easy access to a nature-based location, then participants may experience decreased anxiety and/or stress. This is because they can feel removed from everyday life, experience feelings of escape in nature, and feel calm. If the nature-based location is not easy to access, it is much less likely that people will go there. Nature 2 Nature-based location Ease of access Indirect attention used Attention restoration Decrease mental fatigue IPT 2: If there is easy access to a nature-based location, then participants may experience a decrease in their mental fatigue, as well as feel that their attention has been restored. This is because indirect, or effortless, attention, as described in Kaplan and Kaplan's Attention Restoration Theory (Kaplan and Kaplan, 1989), is being used when immersed in nature rather than direct attention. If the nature-based location is not easy to access, it is much less likely that people will go there. Nature 3 Nature-based location Ease of access Time alone to reflect Increase in readiness to change lifestyle and/or coping strategies Increase in desire to change Increase in awareness of the need for change IPT 3: If there is easy access to a nature-based location, then participants may experience an increase in readiness to change, an increase in desire to change, and/or an increase in awareness of the need for change. This is because the nature-based location gives participants time along to reflect on their lives and what they want to change. If the nature-based location is not easy to access, it is much less likely that people will go there. Individual 4 Availability and resources for trained facilitators Access to resources Planned structured activities Enjoyment of activities Increased physical activity Increased physical health Improvement in mood IPT 4: If there is the availability and the resources to provide trained facilitators, and these trainers have access to a variety of resources (such as equipment), then there will be an increase in physical activity, and a subsequent increase in physical health and improvement in mood. This is because there will be the availability of a number of different planned, structured activities from the trained facilitators, and participants can pick what they would like to do best, and therefore enjoy the activity. Individual 5 Availability and resources for trained facilitators Learning new skills Feelings of self-efficacy

Confidence
Confidence in ability to change and cope with challenges in life IPT 5: If there is the availability and the resources to provide trained facilitators, then this will enable an increase in participant confidence, as well as in their confidence to change and cope with challenges in life. This is because participants are able to learn new skills from the facilitators, which lead to feelings of self-efficacy. Continued engagement with, and after, the programme.
IPT 7: If facilitators have positive attitudes, then participants are more likely to engage with, and after, the programme. This is because, when participants perceive a positive attitude towards them, feelings of rapport and trust are more likely to develop, and a good relationship with the facilitator can be established. Previous experience of a patient-therapist relationship can also influence continued engagement with, and after, the programme. If there is a positive previous experience, then this can lead to engagement. This is because feelings of rapport and trust can be built quicker, and participants can more easily develop a good relationship with the facilitator. Social 8 Perception of how others are engaging on the programme Time on programme Team building/teamwork exercises Feeling safe and unjudged by others with similar backgrounds Feelings of rapport Opportunities to share Opportunities to learn from others Increased social abilities Improvements in interpersonal relationships IPT 8: If participants perceive others to be engaging well on the programme, then this can lead to increased social abilities and improvements in interpersonal relationships. This is because, when participants perceive others to be engaging, this increases feelings of rapport between participants. This can lead participants to feel safe and unjudged by others during team building/teamwork exercises where there are opportunities to share and learn from others. Even if others are perceived to be engaging well, time spent on the programme is also important in order to achieve outcomes. This is because social improvements do not occur quickly, and interpersonal relationships take time to build. process so full texts were obtained. As a result of further close reading of full texts, 49 articles were identified and included. Literature searching and screening results are reported in Fig. 2 using PRISMA (Moher et al., 2009). Information provided in each study about the programmes and articipants varied, and key characteristics of all included studies were recorded (Table 4 and Table 5).

Relevance and rigour
Following the guidance set in the quality standards for realist reviews (Wong et al., 2013), each study was appraised for relevance and rigour. Relevance was assessed in relation to three criteria: population, intervention, or study design; explanation of context, mechanism and outcome as individual aspects as well as in combinations; and explanation of theory. In realist reviews, studies can be included even if only a small part is relevant. This can mean that a certain amount of subjective judgement is necessary to ensure the number of included studies is not unmanageable. Similarly, in realist reviews, studies are assessed for rigour in a different way from systematic reviews: standard quality assessment tools are not used due to the risk of 'nuggets of wisdom' (Pawson, 2006a) being missed due to discarding papers deemed methodologically weak. As is advised in the quality standards (Wong et al., 2013), we identified whether the methods in each study were rigorous enough to be able to rely on the small percentage of findings that we needed to draw on and use in our review. However, as discussed in Pawson (2006a), even studies typically deemed methodologically weak can be included, with careful analysis and appraisal, since they may explicitly, or implicitly, allow insight into why an intervention did not work. To ensure that the risk of bias was reduced, a second reviewer (HB) checked a selection of included/excluded papers to ensure validity and consistency. Where there was inconsistency, a thorough discussion was held to decide whether to include or exclude the study.

Testing and refinement of programme theory
Detail on contexts, mechanisms, and outcomes of each included study were recorded in an Excel spreadsheet. Data extraction and synthesis were undertaken by the first author (WM), with results regularly discussed with the rest of the study team (HC, TP, KP) to ensure consistency, and reduce bias when refining programme theories. Ongoing conversations with greenspace organisation staff were held throughout the search and appraisal process to further ensure that programme theories accurately described the underlying mechanisms and causal pathways of the interventions. The development of IPTs into seven refined programme theories is described below. It became clear during data synthesis that IPTs did not adequately integrate the 'for whom ' and 'in what circumstance' aspects of the realist method. Therefore, while the programme theory themes stayed similar, there was refinement and greater emphasis placed on these contextual factors given how essential they are for implementation and targeting. Fig. 3 shows a brief outline of how the identified programme theories fit in to three overarching themes. The seven programme theories are represented by headings which we believe best describe their core concept.

Escape/Getting Away
Perhaps unsurprisingly, given the integral part nature plays in the programmes, most of the included studies mentioned the importance of immersion in greenspace for mental health benefits. Fernee et al. (2018) discuss how the role of the wilderness created a calming effect on participants, in contrast to their usually chaotic lives, and how the calming environment facilitated cognitive processes such as reflection. Participants in Kogstad, Agdal, and Hopfenbeck's study (2014) described immersion in nature as feeling like a cloud had been lifted, while participants in the study by McIver et al. (2018) reported that immersion in nature helped reduce rumination and stressful, negative thoughts. One participant in O'Brien, Townsend, and Ebden's study (2010) reported that he felt sitting on the hillside for 10 minutes was as effective as his antidepressant medication. A number of the studies made reference to well-established theories such as Attention Restoration Theory (Kaplan and Kaplan, 1989) which holds that when a person is immersed in nature this leads to feelings of calm and a reduction in mental fatigue. Previous reviews such as Berto (2014) and Hartig et al. (2014) have also supported the role of attention restoration and a reduction in mental fatigue as mechanisms in health improvement. One of the components of Attention Restoration Theory is that immersion in nature allows a person to feel removed from their everyday life and, therefore, from their everyday stressors. This feeling of escape, or 'getting away', is a key mechanism in the success of greenspace programmes. In the Nacadia Therapy Garden, service users described the garden as 'a magical world of its own' (Sidenius et al., 2017, p. 5), whereas other participants described being out in nature as 'like another world' and 'sort of like part of the world but a pocket. A haven pocket' (Stevens, 2018, p.7 & p.9 respectively).
Refinement of programme theory IPT 1 and IPT 2, as shown in Table 1, were condensed into the above encapsulating programme theory of Escape/Getting Away. The greenspace setting was a key contextual factor as it provided the right supportive environment but also acted as the resource (mechanism resource), otherwise understood as the programme strategy or programme component introduced in a context. Programmes that utilise greenspace, and allow participants to feel as if they are escaping from their day-to-day lives, are shown to be particularly effective for participants with experience of trauma, anxiety, depression, suppressed anger, and other emotions, conflicts in relationships, as well as for people who explicitly state that they need help (Bettmann et al., 2011;Russell and Phillips-Miller, 2002). As well as existing diagnoses, the greenspace setting was particularly effective for participants who had previous experience of more typical treatments such as counselling (context), as greenspace OR "green space" OR "green care" OR greencare OR "nature therap*" OR "wilderness therap*" OR "outdoors behavi*ral healthcare" OR "outdoors behavi*ral therap*" OR "forest bathing" OR "shinrin yoku" OR "shinrin-yoku" OR "horticultur* therap*" OR "therapeutic horticulture" OR "green exercise" OR ecotherap* OR "conservation therap*" OR "care farm*" AND "mental health" OR "mental ill health" OR "mental illness" OR "mental disorder" OR "mental fatigue" OR psychiatric OR "psychiatric illness" OR stress OR depression OR anxiety OR recovery OR "low mood" OR wellbeing they no longer felt as if they were confined within four walls Granerud and Eriksson, 2014;Sidenius et al., 2017;Woodford et al., 2017). It is possible, therefore, that nature-based programmes are most appropriate for participants who have previously attended traditional therapies which they believe were unsuccessful. The feeling of being away, relaxed, and removed from daily life, was shown to be further facilitated by sensory stimuli (context and mechanism resource) present in the environment (Adevi and Lieberg, 2012;Grahn et al., 2017;Harris, 2017;Rappe et al., 2008;Sidenius et al., 2017). There was some evidence that ease of access to the programme sites was a contextual factor, with one study highlighting that not owning a car to get to sites could be a barrier (O'Brien, 2018), and Husk et al. (2020) ) state that support to get to the location of the programme was necessary for success. Additionally, during a discussion with greenspace programme staff, one manager emphasised that access to minibuses could influence the ease by which the programme was attended so could be a potential contextual factor. Changes in participant reasoning (mechanism reasoning) occur as a result of introduced resources and together these constitute the programme mechanism. In this programme theory, stress levels and mental fatigue were reduced (outcomes) through indirect attention being used (mechanism reasoning), and through the participant feeling removed, relaxed, and 'getting away' from their stressors (mechanism reasoning). The WHO report Urban Green Space: A Review of the Evidence (WHO, 2016) discusses the importance of taking account of gender differences in response to exposure to greenspace, however, with a previous longitudinal study by van den Bosch et al. (2015) reporting positive associations between exposure to greenspace and mental health in women, but not men. Furthermore, Combs et al. (2016) reports that female participants showed a faster decrease in stress than male participants, suggesting that a shorter stay on a programme may work for female groups. Such findings suggest that men and women may respond differently to the greenspace environment on programmes so should be considered during programme development. It is also worth noting that cultural differences can influence how well a participant engages with a greenspace programme in the first instance. For example, during conversations with greenspace programme staff, we identified that uptake of greenspace programmes such as forest therapy is much higher in Japanese and Korean culture where time in forests is an integral part of that lifestyle. The normalisation of forest therapy in these cultures will likely have an influence on uptake and engagement of programmes compared to countries where there is stigma attached to such ideas.

Space to Reflect
The contextual role of greenspace setting is discussed in the above Escape/Getting Away programme theory and is also integral to this programme theory. In this programme theory, the greenspace environment acts as a catalyst for change, with McIver et al. (2018), and participants on the Living Wild programme (Venture Trust, 2019), describing nature as a mediator in preparing a person for a therapeutic experience. Sidenius et al.'s study (2017) supports this describing nature as providing a 'backdrop' where therapeutic conversations and activities were more accessible. An integral part of this programme theory is that time alone in greenspace can allow participants to reflect on their lives. This is particularly important for those with coping strategies which may be harmful to them, such as using drugs, alcohol, or self-harm (Bettmann et al., 2011). Participants on a wilderness therapy programme  spoke about the physical space allowing them to reflect in a prolonged and undisturbed way, both when sitting and walking. This, in turn, can increase their awareness of the need for change in their lives McIver et al., 2018;Russell and Phillips-Miller, 2002), and how to 'live a better life' (Fieldhouse, 2003, p.90).
Refinement of programme theory As in the Escape/Getting Away programme theory, the greenspace setting provided the supportive environment for the programme (context and mechanism resource). The context of adequate time spent on the programme was a refinement to this programme theory since change and reflection did not happen quickly (Kogstad et al., 2014;P� alsd� ottir et al., 2014;Schreuder et al., 2014;Sidenius et al., 2017). Participants in Gabrielson et al. (2018) believed change happened due to the number of unique experiences participants have during programmes, but stated that change could take months to become apparent. Within these contexts, awareness of the need to change (outcome) was achieved by participants spending time alone and reflecting on their lives (mechanism reasoning). Additionally, the desire to change (outcome) could be facilitated by metaphors encountered within the programme (mechanism resource) and participants applying these to their own lives (mechanism reasoning). An example of this was a description of how trying to control a canoe and fight against its natural course proved more difficult than letting nature take its course around obstacles: a metaphor for trying to control life and avoid obstacles (McIver et al., 2018). Adevi and Lieberg (2012) also discussed how participants may seek out specific places in a therapy garden, depending on their emotional state. This is an important contextual factor as it highlights that the most therapeutic place for a participant to reflect is highly individual. It is, therefore, important that participants self-choose places that they have a connection with, or feel comfortable in (mechanism response). According to the staff in this garden programme, reliance on the self-chosen places appears to create greater confidence over time (outcome). Pre-existing diagnoses were also identified as a crucial contextual refinement for this programme theory, particularly important when designing programmes for specific populations. For example, extensive time alone for reflection is not appropriate for participants with existing diagnoses such as severe depression or psychosis (Fernee et al., 2017).

Physical Activity
Enjoyment of physical activity appears to be the mechanism that best allows increased physical health and fitness. Two participants in Fernee et al.'s study (2018) described how, even though they felt tired during physically challenging hikes, they still felt happier when taking part and therefore found it easier to push themselves. However, caution must be taken before generalising this finding: Caulkins et al. (2006) highlighted how young women in their study appeared to benefit less from wilderness hikes compared to other participants, due to higher levels of aversion to the outdoors. Evans (2013) suggests that greenspace programmes provide participants with unique, exciting experiences which encourage people to participate. Furthermore, with increases in physical activity, improvements in mood are also seen (Bryson et al., 2013;Eriksson et al., 2011;Fernee et al., 2018;Fieldhouse, 2003;Leck et al., 2015;O'Brien, 2018;Wilson et al., 2010). This supports existing systematic reviews and meta-analyses supporting the role of physical activity on mental health (Bize et al., 2007;Penedo and Dahn, 2005;Rosenbaum et al., 2014). Conversations with service managers identified that availability of resources for equipment is an important contextual factor for successful engagement with physical activities; as programmes must be fully equipped and functional. However, Surridge et al. (2004) discuss how resources from stakeholders can also be in the form of support and advice in areas such as risk assessment and group safety. Six studies stressed the importance of having confident, adequately trained facilitators to enable and lead activities (Bloomfield, 2017;Evans, 2013;Granerud and Eriksson, 2014;Kogstad et al., 2014;O'Brien et al., 2010;Surridge et al., 2004). With a diversity of activities available, participants are more likely find an activity that they enjoy and will engage with.
Refinement of programme theory The availability of trained facilitators (context), and availability of resources to adequately support and fund programmes and their materials (context), is imperative to provide a variety of activities to service users (mechanism resource). This allows participants to engage with activities they enjoy (mechanism reasoning), and provides personcentred approaches (Barley et al., 2012;Bloomfield, 2017;Cole and Christie, 2016;Harris, 2017;O'Brien, 2018;Schreuder et al., 2014). This enjoyment of physical activity (mechanism reasoning) facilitates engagement (outcome) and, in turn, leads to increased physical activity, improvements in physical health, and improvements in mood (outcomes). However, particularly in winter weather (context), participants who did not like spending time outside found it difficult to enjoy any aspect of the programme due to discomfort (mechanism reasoning) (Harper et al., 2019), limiting positive outcomes. However, realistic expectations of anticipated challenges did appear to be an influencing context in the lead up to programme uptake in some circumstances. Gabrielsen et al., (2018) suggest that clearly informing and preparing participants for any challenges prior to the programme commencing is advisable, in particular, ensuring participants have the right equipment (context), such as waterproof clothing and shoes.

Self-Efficacy
Twenty-eight studies reported that service users who learned and mastered new skills had increased self-esteem, pride, and confidence. Indeed, existing evidence supports continued learning as a mechanism for mental health improvement (Feinstein and Hammond, 2004;Fig. 3. Three programme themes and subsequent representative headings for the seven programme theories identified through data synthesis.

Table 4
Included empirical studies in the review.      Hammond, 2004). Learned skills can be practical tasks, for example, learning how to look after plants was very effective for those with stress-related illness (Adevi and Lieberg, 2012;Eriksson et al., 2011), and for those without a clinical diagnosis wanting to improve wellbeing in general (O'Brien et al., 2010). Learning practical skills on wilderness therapy programmes was shown to be a particularly positive experience for young people Warber et al., 2015), and for those who were in the wilderness alone for the first time (Russell and Phillips-Miller, 2002). However, learned skills can also be skills such as self-regulation of emotion (Adevi and Mårtensson, 2013), and coping strategies (Barley et al., 2012). These psychological skills are particularly important in facilitating self-efficacy post-programme enabling service users to integrate new skills into their lives (Bryson et al., 2013;Howarth et al., 2018;Phoenix Futures, 2019;Sidenius et al., 2017). As with the programme theory for Physical Activity, the availability of adequately trained facilitators is necessary to enable participants to learn new skills (Bloomfield, 2017;Evans, 2013;Granerud and Eriksson, 2014;Kogstad et al., 2014;O'Brien et al., 2010;Surridge et al., 2004). Refinement of programme theory The availability of trained and confident facilitators (context) is necessary to enable service users to learn both practical and psychological skills (mechanism resource). As well as improved self-efficacy (mechanism reasoning), another identified change in reasoning was an increased feeling of empowerment when learning new skills (Cole and Christie, 2016;Combs et al., 2016;Howes et al., 2018;Fernee et al., 2018;Harris, 2017;Lehmann et al., 2018;McIver et al., 2018;O'Brien, 2018;P� alsd� ottir et al., 2014;Woodford et al., 2017). Through this mechanism, skills development can lead to increases in pride, self-esteem, and confidence (outcome), as well as in self-efficacy for individuals to implement new skills in their life outside the programme (outcome). To ensure that this programme theory is representative of all greenspace programmes it is necessary to highlight that, whilst teaching skills such as coping with challenges were present across programme type, the type of challenges varied. For example, in wilderness therapy, adventure therapy, and care farming programmes, coping strategies focused on overcoming physical challenges , while on horticultural therapy programmes, coping strategies might focus on dealing with how to manage plants or vegetables that were failing to grow or dying (P� alsd� ottir et al., 2014). As previously mentioned, realistic expectations of anticipated challenges also seemed to be an important context in the lead up to programme uptake .

Having a Purpose
A number of participants on care farms spoke about the responsibility of looking after animals and how their success with this task allowed them to feel satisfied due to having a purpose (Elings and Hassink, 2008;Schreuder et al., 2014). This appears to be particularly applicable to participants on care farms who have psychiatric or addiction histories, where the work and community-like environment enables them to fill their day and have a routine (Elings and Hassink, 2008;Hassink et al., 2010). Participants in Cole and Christie's study (Cole and Christie, 2016) spoke about feeling valued and appreciated for their work, which was motivating. The availability of adequately trained facilitators is necessary for both leading and enabling participants to learn new skills, and provide guidance relating to their responsibilities (Bloomfield, 2017;Evans, 2013;Granerud and Eriksson, 2014;Kogstad et al., 2014;O'Brien et al., 2010;Surridge et al., 2004). Time spent on a programme has been shown to be correlated with achieving outcomes since change occurs slowly Harris, 2017;Kogstad et al., 2014;P� alsd� ottir et al., 2014;Schreuder et al., 2014;Sidenius et al., 2017), so it seems logical to suggest that time spent on the programme might be a contextual factor in this programme theory too; the longer service users were responsible for something, the higher their self-esteem. Both this programme theory, and the Self-Efficacy programme theory above, provide some explanation of why greenspace interventions may fail, in that when there is an absence of confident, trained facilitators, or an absence of programme components which allow participants to learn new skills, interventions will not be effective. While the need for programme variety has already been covered, routine and planning are also necessary. Although there is some evidence that people on greenspace programmes for leisure purposes can benefit from passive immersion in nature (Lovell et al., 2015), this might not be enough to achieve changes in mental wellbeing in those with high levels of stress/mental ill health. In a previous review by Hunter et al. (2015), greenspace interventions were shown to be most effective when there were structured programmes in place, rather than simply changing the physical environment.
Refinement of programme theory The availability of trained and confident facilitators (context) and adequate time spent on the programme (context) are both necessary to enable individuals to learn practical and psychological skills (mechanism resource), which facilitate feelings of responsibility and purpose (mechanism reasoning), and in turn lead to increases in self-esteem and vigour for life (outcomes). The mechanisms of feeling responsible and purposeful were seen across all programme types, however, these mechanisms were facilitated by different contexts. For example, in care farm programmes, participants felt responsible for animals and farm activities (Schreuder et al., 2014), and in horticultural therapy programmes, participants felt responsible for plants and other produce . Managers of wilderness therapy and adventure therapy programmes, as well as facilitators in Surridge et al.'s study (2004), also discussed how service users felt responsible for carrying resources, even when this was challenging. Feelings of purpose were also gained from the routine that programmes provided (mechanism resource). Hassink et al. (2010) reported that almost all participants who accessed a care farm to improve their mental wellbeing acknowledged the positive effect that routine had. Similarly, service users in Iancu et al.'s study (Iancu et al., 2014) reported feeling that structure was something they were lacking before the programme. As well as feelings of purpose, participants reported increases in feelings of empowerment (mechanism reasoning). Twelve studies mentioned how this increase in empowerment led to participants feeling more hopeful and excited about life in general (outcome) (Cole and Christie, 2016;Combs et al., Fig. 4. A novel conceptual model developed on the basis of our review findings to show the overarching programme theory for greenspace interventions for mental health. Evaluation of two projects Relevant information regarding enjoyment of activities, sense of belonging, impact of learning, and relaxation in nature 2016; Gabrielsen et al., 2018;Harris, 2017;Howarth et al., 2018;Lehmann et al., 2018;O'Brien, 2018;P� alsd� ottir et al., 2014;Schreuder et al., 2014;Sidenius et al., 2017;Wilson et al., 2010;Woodford et al., 2017).

Relationship with Facilitators
Five studies highlighted the influence of previous relationships with healthcare professionals as a contextual factor in how well participants initially engaged with programmes (Cole and Christie, 2016;Fernee et al., 2018;Granerud and Eriksson, 2014;Stevens, 2018;Woodford et al., 2017). Existing attitudes of programme facilitators were crucial in enabling mechanisms to achieve outcomes: facilitators who appeared non-judgemental, open, and genuine, enabled relationships to be built quickly with participants. Participants in one study discussed the importance of being treated without prejudice and as a person, rather than a diagnosis . Through this relationship, participants were able to build rapport, trust, and confidence in facilitators, particularly crucial given many programme users had experienced difficult interpersonal relationships and problems developing trust (Evans, 2013;Fieldhouse, 2003;Iancu et al., 2014;McIver et al., 2018). Overall, the stronger the relationship between programme user and facilitator, the more likely participants were to fully engage with programmes and available aftercare support (Cole and Christie, 2016;Combs et al., 2016;Gabrielsen et al., 2018;RedCliff Ascent, 2019;Schreuder et al., 2014;Sidenius et al., 2017;Stevens, 2018).
Refinement of programme theory Previous positive experience with healthcare professionals such as therapists (context), as well as existing positive attitudes and attributes of programme facilitators (context), influenced good initial engagement with programmes (outcome). These contextual factors enabled relationships between facilitator and participant (mechanism resource) to be built quickly, resulting in trust, confidence in the facilitator, and rapport (mechanism reasoning). However, previous negative experiences (context) can negatively impact engagement, particularly for adolescents who may show lower levels of trust towards all adult relationships after a negative experience (Bettmann et al., 2011). There is some evidence that this might be mitigated by ensuring that adequate information about the programme is provided prior to the start, and that each participant is met by a confident and friendly facilitator at the start to help engage participants (O'Brien et al., 2010). Another contextual factor was effective programmes having a culture of 'doing with' not 'doing for' people. Involvement of the facilitators in the same tasks as the service users (mechanism resource) led to decreased perceived power inequality and increased empowerment (mechanism reasonings). For example, some study participants described how facilitators would ask them what they wanted to do, eat, and which way to go allowing participants to feel empowered, decreasing power imbalances, and promoting inclusion (McIver et al., 2018). These mechanisms made continued engagement with the programme more likely (outcome), as well as acceptance of any available support after the programme (outcome).

Shared Experiences
As highlighted in almost all the included studies, greenspace programmes are typically undertaken in environments that promote social cohesion. It is through these shared experiences that participants experience increased social skills and improvements in interpersonal relationships. Hassink et al. (2010) reported that the community feel of the greenspace programme was the most valued aspect mentioned by the majority of participants, a finding supported by many other studies (Adevi and Mårtensson, 2013;Barley et al., 2012;Bryson et al., 2013;Cole and Christie, 2016;Combs et al., 2016;Cook, 2008;Dolgin, 2014;Fieldhouse, 2003;Gabrielsen et al., 2018;Harris, 2017;Howes et al., 2018;Iancu et al., 2014;Leck et al., 2015;O'Brien, 2018;Rappe et al., 2008;Surridge et al., 2004;Stevens, 2018;Wilson et al., 2010;Woodford et al., 2017), as well as by conversations with service managers. Mechanisms identified in this programme theory that led to improved social outcomes were: the group environment feeling safe (Kogstad et al., 2014;Sidenius et al., 2017), lack of stigma and judgement (Combs et al., 2016;McIver et al., 2008;Stevens, 2018), increased rapport (Evans, 2013;Fernee et al., 2018;Warber et al., 2015), and trust between people on the programme, with participants feeling comfortable to express themselves (Adevi and Mårtensson, 2013). Previous systematic reviews support the role of social capital in improving mental health, both in young people (McPherson et al., 2014), and in older adults (Nyqvist et al., 2013). However, the engagement of others on the programme can be a contextual factor to be aware of, with the success of social interactions being a two-way process : if a participant does not believe that other participants are engaging, they are less likely to do so. Time spent on the programme is another key contextual factor for social cohesion to occur . As previously mentioned, change does not happen instantly and social changes, in particular, can take longer to occur compared to psychological, physical, physiological, or cognitive changes .
Refinement of programme theory Perceived engagement of others on the programme (context), and time spent on it (context), can both facilitate a participant's engagement with team-building activities, and other shared experiences (mechanism resource). Furthermore, the group environment might be described as more like 'real life' than traditional therapy (context) (Fieldhouse, 2003). As a result of these shared experiences in a 'real life' environment, with peers who are perceived to be engaging with the programme, participants begin to feel safe and unjudged, build trust and rapport, and feel more comfortable in trying to interact with others (mechanism reasonings). This in turn leads to improvements in social skills and in interpersonal relationships (outcomes). Three studies (Barley et al., 2012;Cook, 2008;Woodford et al., 2017) also noted that this increase in social skills also led to an increase in self-esteem (outcome). Conversely, if participants do not see others engaging well on the programme (context), then this can hinder increases in social abilities or improvements in relationships (outcome): participants feel less safe/comfortable during team exercises (mechanism reasonings), and do not have the opportunity to share or learn from others (mechanism reasonings). Age (context) can also influence the applicability of this programme theory. For example, adolescents may be more susceptible to peer influence in general, and while perceived social support can have a buffering effect on stress in adolescents, low satisfaction with social support in adolescents can increase anxiety (Dolgin, 2014). Therefore, perceived engagement and social support in programmes may be much more important in adolescent programmes compared to those with older adults. Furthermore, O'Brien et al. (2010) discussed how having people of the same age (context) on the programme facilitated cohesion (outcome), as they were able to speak about similar interests, encouraging rapport (mechanism reasoning).

What does not work
With the increase in awareness of the benefits of being outside for mental health, more greenspace programmes are embedding mental health outcomes into their aims. This increases the risk that some programmes could be claiming all types of benefits, with little evidence to support claims. Without clarity of what approaches may or may not consist of, it is difficult to distinguish practice that is ethical and effective, from programmes that over-claim benefit and put users at risk of potential harm. This potentially makes it difficult to know which programmes to enrol on, or which programmes care providers should recommend. Richards, Hardie, and Anderson (2019) suggest an Outdoor Mental Health Intervention Model outlining the importance of the combination of competence, professional responsibility, and leadership in each intervention. The model maintains that, for best practice, a multidisciplinary team approach is adopted, and professionals work collaboratively in the delivery of an integrated approach. The authors state that programme providers should represent themselves, and their practice, using terms that can be justified and evidenced by professional training and qualifications, rather than using terms such as 'therapy' too loosely. Only then can programmes enhance opportunities for improved mental health and wellbeing and offer a best-fit intervention for individual clients.
There is a myriad of contextual factors which will likely influence the success of greenspace interventions, and it is not feasible to attempt to identify the many individual factors which might make a programme work, or not. However, there are certain factors which seem particularly influential in programme success. For example, as has been identified through this review, programmes based in the wilderness are often undertaken by adolescents or young adults. What seems less clear, is whether wilderness-based programmes are successful for older adults. During a conversation with a greenspace project staff member, the fear of injury or fear of falling was highlighted as the top barrier to engagement. Furthermore, although there are programmes designed specifically for older adults, such as horticultural therapy programmes, specific contextual barriers can limit their effectiveness. For example, the Greenspace and Health Strategic Framework for Edinburgh and Lothians (Greenspace Scotland, 2019) discusses how staffing numbers on hospital wards means that patients cannot leave the hospital to access greenspaces with the necessary support. Without staff available to support people who need assistance to and from greenspace programmes, programmes cannot be successful. While this report is specific to one geographical area, it is feasible to see how systemic understaffing will affect any greenspace programme reliant on support staff.
This review also identified that some circumstances, such as time alone in a wilderness environment, might not be appropriate for people with pre-existing diagnoses like psychosis. However, there are other circumstances where certain greenspace programmes might not be appropriate, for example, residential greenspace programmes for those on daily pick-up prescriptions. Livingston et al. (2011) discuss how people on methadone prescriptions can be excluded because early start times mean they cannot pick up their medication beforehand. Another example raised during a meeting with a member of staff was electronic tagging. Greenspace programmes have been successful in supporting people who have been involved in offending (Venture Trust, 2019), but they may be limited to where they can go if a programme does not account for this.
A person's belief about the programme is also a driving contextual factor in initial enrolment. While some people with previous treatment experience may welcome a new approach, particularly if they feel that current treatment has not worked, others may be cynical about its reliability. Husk et al. (2020)) reported barriers such as concerns about adequate facilities, and adequate staff experience/training. They also reported concern about the greenspace environment and whether it was an appropriate environment for people with complex needs. Davis--Berman and Berman (2012) state that participants on greenspace programmes need to want to be part of the programme and have some level of self-motivation. If a person does not want to enrol on a greenspace programme because they do not believe that it will be beneficial for them, it is unclear how this can be changed, and even if it should. For example, this review has identified that one of the key mechanisms by which greenspace programmes are effective is through an increase in feelings of empowerment. In contrast, coercion and involuntary treatment has shown to threaten effectiveness of treatment (de Valk et al., 2019). Harper et al. (2019) raise concerns about how this may impact the effectiveness of youth programmes, where parents have enrolled their children, or in hospitals where primary care staff may have enrolled patients on their behalf. Husk et al. (2020) highlight how the power dynamic between care provider and patients can be equally influential, with some patients viewing social prescriptions, such as greenspace programmes, as an order rather than a choice. If empowerment and agency are mechanisms that lead to successful outcomes, then by taking these away it seems unlikely programmes will be effective. However, as identified in Husk et al. (2020) this does not equate to leaving all responsibility for enrolment to the person potentially accessing the programme. Instead, it highlights the importance of dialogue between care provider and participant, as well as the necessity of the provider knowing what is available for recommendation. One of the concerns in this regard, however, is that short term funding makes it difficult for providers to recommend greenspace programmes, due to lack of continuity of services.
Aside from issues which impact uptake of greenspace programmes, it is important to recognise that not everyone will benefit or enjoy programmes when on them. This review has covered the necessity of a variety of activities to initially engage participants Wilson et al. (2010), but O'Brien et al. (2010 also discuss that activities that are repetitive can cause participants to lose interest and quit. Even participants who enjoy programmes, but see no change in their condition, can become demotivated and quit (Husk et al., 2020). In Husk et al.'s study (Husk et al., 2020), participants explicitly said that the main reason for drop-out was lack of change in health status leading to them questioning if the interventions were effective and worthwhile. Similarly, those with higher, or unrealistic, expectations of the intervention were more likely to drop out.
While we have discussed a number of challenges that may hinder the implementation of greenspace interventions, it is necessary to highlight that reporting bias was evident throughout data extraction, in many of the included studies. This finding is supported in a review of wilderness therapy programmes, where Fernee et al. (2017) reported that almost all included studies reported positive results, and some even explicitly reported a reluctance to analyse negative experiences in detail. Without in-depth understanding of negative, or even neutral, experiences, and with no advice or support about how to overcome challenges, then it is unclear how beneficial research can be in informing future practice. In future research, more evidence is therefore needed regarding alternative examples which counter successful case narratives.

Discussion
This review contributes to international empirical research as it is a novel approach to both understanding and evaluating how greenspace interventions can be used to improve mental health. Through an iterative process, data were collected and analysed which allowed continuous development of programme theories as new data emerged. The synthesis of empirical findings allows a greater theoretical understanding of the intervention process itself, rather than reporting whether an intervention is effective or not. The theoretical findings are therefore transferable across a range of interventions and are more useful for the logical, evidence-based development of other effective interventions. To identify the context-mechanism-outcome configuration for each programme theory, the IPTs were first tested against the literature and then refined to explain how, for whom, and in what circumstances, do greenspace interventions for mental health work, or do not work. Russell and Farnum (2004) have previously suggested a programme theory for wilderness therapy that incorporated three interrelated factors of Wilderness, Physical Self, and Social Self. This programme theory was noted, but did not prematurely influence our review since our review was of greenspace programmes in general, and not of a specific type. In-depth reviews by Lovell et al. (2015) and Husk et al. (2016) have also produced detailed conceptual models of how engagement with nature can impact physical and mental health. These models were helpful for building a deeper understanding of mechanisms and outcomes, as well as touching on some of the contextual factors which may influence programme development. Our review expands on some of the work in these models through further focus on context, additional mechanisms, and the focus on 'for whom ' and 'in what circumstance'. Overall, we found that Nature, Individual Self, and Social Self aptly described the three overarching themes under which our programme theories fell. The headings of the seven programme theories identified through a thorough engagement with 49 included studies and discussions with greenspace service providers are shown in Fig. 3 under the three identified themes of Nature, Individual Self, and Social Self. The in-depth synthesis of each of the programme theories, as covered in the results, allows an understanding of the causal relationships which make up each programme theory. While it is indeed possible for programme theories to exist independently from each other, it is feasible to deduce that greenspace programmes work best in the circumstances where CMOcs are activated under each programme theory simultaneously. As Pawson (2006b) states, transformation may be achieved by the fact that CMOcs happen together in a process over time. It is worth mentioning that participants may focus their development in one area of the programme which could lead to trade-offs in outcomes. For example, participants who spend time honing a particular independent skill, and therefore increasing in self-efficacy, may actually decrease in social skills due to time spent alone. Further exploration of how mechanisms in one programme theory may affect outcomes in a separate theory is therefore recommended. However, through data synthesis, we found that 27 of the included studies explicitly reported that the interaction of nature, individual changes, and social changes, was related to best outcomes. Therefore, programmes should include adequate opportunities for development in both individual and social skills, in order to mitigate any negative effects of trade-offs.
Based on the seven programme theories, Fig. 4 visually depicts a novel conceptual framework developed on the basis of our review findings. The key differences between this conceptual framework and previously mentioned models are: firstly, showing that this framework could be an overarching programme theory for all types of greenspace programmes, and not just one type of programme; and secondly, as well as identifying programme theories about how greenspace interventions may be successful in improving mental health, within the seven programme theories we have synthesised context, mechanism, outcome configurations which allow a better causal understanding of the pathways to mental health improvement.

Strengths and limitations of the review
To our knowledge, this review is the first to use realist methodology to examine greenspace interventions for mental health where studies were not excluded based on intervention type. This allowed different types of greenspace programmes to be analysed with a realist lens and similar CMOcs to be identified across programmes. The findings highlight that greenspace programmes appear to be successful as a result of three interacting themes; Nature, Individual Self, and Social Self, regardless of programme type. In future work, interventions such as care farming, wilderness therapy, or horticultural therapy could be analysed in separate reviews which could allow the overarching conceptual framework outlined in this review to be tested and refined further. Another strength is that studies covered nine countries allowing the findings of this review to be internationally relevant.
Limitations must also be recognised when using realist methodology, particularly relating to reviews being based on guiding principles rather than standardised rules . Although we have endeavoured to ensure transparency at all points of our review, for example by submitting our protocol to PROSPERO, adhering to robust quality standards (Wong et al., 2013), and through thorough documentation and in-depth discussion of key decisions, the realist review process is inherently interpretive and subjective, especially in regard to relevance and rigour assessment. Secondly, as previously mentioned, further exploration of how mechanisms in one programme theory may affect outcomes in a separate theory is recommended via future research. Thirdly, although not within the scope of the current review, future research should identify whether greenspace programmes are successful in the longer term, and whether the proposed programme theories can also explain long term success. Finally, realist approaches can synthesise data from quantitative and qualitative methods (Wong et al., 2016;Pawson et al., 1997), and analysis is guided by data that are best suited to answer research questions. We found that in the examples we examined, the qualitative studies were regarded as higher relevance for informing programme theories compared to quantitative data due to CMOc information in these studies being more accessible. Future research should examine how best to integrate more quantitative data into programme theories, for example with physiological mechanisms and outcomes such as salivary cortisol changes, body mass, and heart rate.

Conclusion
This realist review has examined the contexts and mechanisms in greenspace programmes which can lead to outcomes in mental health to show what works, for whom, and in what circumstances. These configurations have been developed into an original overriding theory involving seven programme theories under three themes of Nature, Individual Self, and Social Self. The interaction of these three factors represents a new conceptual framework for greenspace interventions for mental health. The findings of this review are not only theoretically novel, but also have practical relevance for those designing such interventions, providing recommendations on how to optimise, tailor, and implement, existing interventions. These will be particularly relevant for academic researchers, health professionals, and mental health multidisciplinary teams, and for those working in the third sector, developing and delivering such interventions.

Declaration of competing interestCOI
None.

Data access statement
This study was a review of existing data, which is openly available at locations cited in the reference section. Further documentation about data access and records are available from the authors.