A systematic review of the health and well-being impacts of school gardening: synthesis of quantitative and qualitative evidence

Background School gardening programmes are increasingly popular, with suggested benefits including healthier eating and increased physical activity. Our objectives were to understand the health and well-being impacts of school gardens and the factors that help or hinder their success. Methods We conducted a systematic review of quantitative and qualitative evidence (PROSPERO CRD42014007181). We searched multiple databases and used a range of supplementary approaches. Studies about school gardens were included if they reported on physical or mental health or well-being. Quantitative studies had to include a comparison group. Studies were quality appraised using appropriate tools. Findings were narratively synthesised and the qualitative evidence used to produce a conceptual framework to illustrate how benefits might be accrued. Results Evidence from 40 articles (21 quantitative studies; 16 qualitative studies; 3 mixed methods studies) was included. Generally the quantitative research was poor. Evidence for changes in fruit and vegetable intake was limited and based on self-report. The qualitative research was better quality and ascribed a range of health and well-being impacts to school gardens, with some idealistic expectations for their impact in the long term. Groups of pupils who do not excel in classroom activities were thought to particularly benefit. Lack of funding and over reliance on volunteers were thought to threaten success, while involvement with local communities and integration of gardening activities into the school curriculum were thought to support success. Conclusion More robust quantitative research is needed to convincingly support the qualitative evidence suggesting wide ranging benefits from school gardens.


Background
School gardening and food growing have become popular activities in thousands of schools around the world. National school gardening programmes exist in some countries, such as the Royal Horticultural Society Campaign for School Gardening in the UK [1] and the Stephanie Alexander Kitchen Garden Program in Australia [2]. The individuals and organisations behind these programmes believe that school gardening has the potential to improve children's health, social development and academic attainment.
Since the 1990s, an increasing number of research studies have attempted to evaluate the effectiveness of school gardening programmes. More recently, several reviews of the literature on school gardening have been published [3][4][5][6][7][8]. Five of these reviews were limited to US studies and, whilst they found some empirical evidence for the health and well-being impacts of school gardening, some of their conclusions were based on theoretical rationale [3][4][5][6][7]. They recommended further research, using more rigorous study designs, on the effectiveness of school gardening programmes, mediation pathways and implementation factors.
The most comprehensive review, by the National Foundation for Education Research, included international studies and found evidence for positive impacts of growing activities on pupils' nutrition and attitudes towards healthy eating [8]. It also concluded there was modest evidence for social well-being benefits, especially for lower ability pupils or those who have become disengaged from learning. This review was described as being "underpinned by a systematic process for searching, selection, screening, coding, appraisal and synthesis" (page 3). However, it did not clearly and consistently report methods (inclusion criteria; quality appraisal criteria) and results (CONSORT flow diagram; number of studies with health and well-being outcomes; description of included studiesdesign, methods, quality etc.) as would be expected in a systematic review. In addition, since 2011, more studies have been published on the impact and meaning of school gardening programmes, including a large cluster randomised controlled trial of the Royal Horticultural Society Campaign for School Gardening in the UK [9,10] and several qualitative studies from the UK and the US [11][12][13]. Therefore, it was justified and timely to conduct a robust, mixed methods systematic review of the health and well-being impacts of school gardening, to support and inform the further development of this popular school-based intervention. This review aims to answer the following questions: What are the health and well-being impacts of school gardens? Are there different impacts for different age groups? What are the effects on other family and community members? What do school gardens mean to those who use them? Are there any factors that help or hinder the successful development, use or sustainability of school gardens?

Methods
We conducted a systematic review of quantitative and qualitative literature according to a pre-specified protocol that was registered with the International Prospective Register of Systematic Reviews (PROSPERO: CRD42014007181). We used the methods of thematic synthesis described by J. Thomas and A. Harden [14]. As this was an evidence synthesis of existing research, ethical approval was not required.

Search strategy
A search strategy was devised by the research team, led by our Information Specialist (AB), through examination of key studies and discussion. It captured the concepts of school gardening and horticulture activities. The following MeSH terms were used: school exp; gardening exp; child nutrition sciences. No methods filters were used. The master search strategy (Table 1) was adapted and run in the following electronic databases in February 2014 and updated in May 2015: MEDLINE, EMBASE, PsycINFO, HMIC and SPP (using the OVID interface); AEI, BEI, ASSIA, BNI 1994-current and ERIC (using the ProQuest interface); AMED and CINAHL (using the EBSCOHost interface). Additional grey literature databases were also searched: OpenGrey, EThOS and British Library Catalogue. The review by the National Foundation for Education Research was a useful source of includable references [8]. Reference lists of included studies were scrutinised for other relevant studies. Forward citation searches were undertaken on included studies. Citation searches were also performed in Web of Science using three key references [15][16][17].

Inclusion criteria
Studies were considered eligible for inclusion if they met the following criteria: Population: School children, school staff, family and community members (all ages) were included. Studies conducted in OECD countries and published in English were included. Table 1 Search strategy for the health and well-being impacts of school gardening (as used in Medline) 1 school*.tw.
Interventions: Studies were included if they reported the effects of participation in school gardening activities. The definition of 'school' included all educational settings up to 18 years, including special schools. The definition of 'gardening' included growing or cultivating any kind of plants (such as vegetables, fruits, trees, shrubs and flowers). Gardening activities included preparing the soil, planting, weeding, watering, harvesting and garden-related cooking and tasting activities. These gardening activities were either integrated into the curriculum, or conducted outside of lesson time (e.g. lunchtime clubs, after school clubs, school-organised trips to community allotments). Gardening activities for school-age children that did not involve schools were not included (e.g. summer holiday clubs or community youth interventions).
Comparators: Quantitative studies were only included if groups participating in school gardening activities were compared with control groups or groups participating in alternative activities (such as nutrition education without gardening activities). This criterion was not relevant for qualitative studies.
Outcomes: Studies were included if they reported quantitative or qualitative health and well-being outcomes including dietary intake; food-related knowledge, attitudes and preferences; physical, mental or emotional health; quality of life indicators. Qualitative findings also included themes, concepts and metaphors relating to the experience and meaning of school gardens, and any perceived factors that help or hinder their success. Additional outcomes, including adverse or unintended outcomes, were only considered where they were reported alongside health and well-being outcomes.
Study design: Suitable quantitative study designs included randomised controlled trials (RCT), nonrandomised controlled trials, and other 'controlled before and after' studies. Suitable qualitative study designs included any recognised methods of data collection and analysis from any discipline or theoretical tradition. The types of data collection methods included (but were not limited to): focus groups, individual interviews, participant or systematic observation, documentary analysis, audio/visual/note collection. Methods of analysis included (but were not limited to): grounded theory, narrative analysis, thematic analysis, phenomenological analysis, discourse analysis.

Selection process
References identified through the search strategy were uploaded into ENDNOTE (X7, Thomson Reuters). References (titles/abstracts) were independently double screened using the eligibility criteria (by reviewers HO/ RW/SG). Studies appearing to meet the criteria were obtained as full text articles. Full texts were independently double screened using the same criteria (by reviewers HO/RW/SG/RG). Any disagreements were resolved through discussion with the whole team.
During full text screening, additional inclusion criteria were developed as an iterative process. For example, studies in which school gardening was one of multiple components in a 'whole school approach' intervention (in combination with farm visits, school food policy development, school meals and catering reforms, nutrition education or cookery programmes) were not included if the reported outcomes were not specifically attributable to school gardening. Some qualitative studies did report health and well-being impacts attributable to school gardening (i.e. distinct from other components) and these studies were included. Studies that did not report sufficient information about data collection and/or analysis methods to enable critical appraisal were excluded (such as conference abstracts for which the full text was not available).

Data extraction
Standardised, piloted data extraction sheets were developed for the review to ensure consistency between studies and between reviewers. The data extracted for each study included, where possible: study design, sample characteristics, description of school gardening activities (intervention group), description of alternative activities (comparison group), duration of intervention, data collection methods, analysis methods and duration of follow up. For quantitative studies, health and well-being outcomes (and other secondary outcomes) were extracted. For qualitative studies, findings in the form of participants' quotes and author themes and concepts were extracted. Data were extracted by one reviewer (HO) and independently double checked (by reviewers RW/SG/RG). Any disagreements were resolved through discussion with the whole team.

Quality appraisal
The quality of included quantitative studies was appraised using the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies. Each study was rated strong, moderate or weak for the following components: selection bias, study design, confounders, blinding, data collection, withdrawals and drop outs. A 'global' or overall rating was then allocated using the standard system: strong (no weak ratings), moderate (one weak rating), or weak (two or more weak ratings).
The quality of included qualitative studies was appraised using criteria suggested by Wallace et al. [18]. In addition to the standard ratings of yes, no, can't tell, we used 'partial' in some cases, for example studies in which some ethical issues were addressed and others were not (Wallace criterion 12). We also generated overall quality ratings using our own system: strong (11-12 ratings of yes); moderate (6-10 ratings of yes); weak (1-5 ratings of yes).
Studies were appraised by one reviewer (HO) and independently double checked (by reviewers RW/SG/RG). Any disagreements were resolved through discussion with the whole team.

Data synthesis
It was not possible to meta-analyse any of the quantitative data we extracted due to study design and data limitations (further information provided in Tables). Authors were contacted to clarify some study details (such as precise methods of dietary assessment) before this decision was confirmed. Data from quantitative studies were therefore tabulated, grouped by similar outcomes and the effectiveness of the school gardens in influencing health and wellbeing outcomes was described narratively.
The qualitative data were synthesised thematically in order to understand the perceived well-being and wider impacts of gardens, as well as factors influencing success and sustainability, in order to make practical recommendations for interventions [14,19]. Narrative synthesis has three overlapping stages: 1) coding of the findings of primary qualitative studies; 2) organising codes into descriptive themes and concepts; 3) generating analytical themesthis final stage goes beyond the interpretations of the original studies and aims to generate new understandings or hypotheses in relation to the review questions [14]. Initially, all study findings were coded, including those that did not relate directly to health and well-being impacts. Where multiple publications reported findings from the same qualitative study, we were careful not to 'double count' findings. Therefore, where multiple publications by the same team of authors reported similar themes, only one publication has been referenced for each theme (the oldest one). In stages 1 and 2, we used an inductive approach to code the data (line by line) and identify common themes and categories of themes between studies. This was done manually rather than using software, using the principles of constant comparison and reciprocal translation [20]. Care was taken to recognise divergent data and to interpret the raw data (where presented) rather than uncritically accept the authors' interpretations. In stage 3, we developed a conceptual model to summarise and illustrate proposed mechanisms for the health and well-being impacts of school gardening. The qualitative synthesis was completed by two reviewers (HO/SG) in discussion with the rest of the team. Throughout the synthesis, we considered study quality, context and transferability as we developed our conceptual model.

Search results
Electronic searches identified a total of 3442 records after the removal of duplicates. Title and abstract screening resulted in the exclusion of 3279 records that did not meet the inclusion criteria. 124 of the remaining potentially relevant articles were obtained in full text format (whilst 39 were unobtainable) and another 22 eligible articles were found by citation and manual searching. Full text screening resulted in the exclusion of 106 articles (Fig. 1). A total of 40 articles (hereafter referred to as studies) were eligible for inclusion.

Duplication and differences in reporting
Where multiple publications duplicated findings from the same trials, we have been very careful to present and synthesise the findings of these studies without duplication; for example elements of Morgan's findings were republished by Jaenke [15,30]). To confuse matters, differences were found in the reporting of the same outcomes between different studies. For example, Jaenke and Morgan both reported data from a school garden-based nutrition education program in Australia (n = 127) but they reported slightly different findings in the control group for fruit and vegetable intake [15,30]. Neither of these reported findings were statistically significant and we have presented the most recent findings [15] (Table 7). Another case of duplication and differences in reporting was in the evaluation of the pilot of the Stephanie Alexander Kitchen Garden Program in Australia [25,26]. Both studies reported children's willingness to try new foods, but with slightly different sample sizes (n = 770/n = 764) and therefore slightly different findings. We have presented the findings from the original study which also reported the largest sample size [25] (Table 9). As illustrated, in cases such as these, the data from the study reporting the largest sample size (or, if the sample sizes were equal, the most recent study) were presented in our review tables.

Quality appraisal of included studies
Where multiple studies reported quantitative data from the same trial (quantitative) or the same sample (qualitative), they were appraised collectively (Tables 5 and 6). Most of the quantitative studies were rated as weak, with six studies rated as moderate [10,15,23,[28][29][30] (Table 5). Quality criteria that were commonly rated weak (in more than half the studies) using the EPHPP system were selection bias,    No intervention Interpersonal relationships a also included for qualitative findings (see Table 3); FV fruits and vegetables FV fruits and vegetables, SBP systolic blood pressure, DBP diastolic blood pressure   UK Two samples (two stages) from 10 primary schools e.g. stage 1: senior leaders (n = 11), garden leads (n = 10), other members of teaching staff (n = 10), teaching assistants (n = 2), parent governors (n = 2), other parents (n = 2) and pupils (n = 43) To assess the impact that using a school garden had on primary pupils' learning, behaviour and health and wellbeing Open ended questionnaire; face-toface interviews Data were then categorised thematically and analysed control for confounders, and follow-up rate (withdrawals and drop-outs) (Fig. 2). While five qualitative studies were rated as strong, most of the qualitative studies were rated as weak or moderate quality [11,[24][25][26]43] (Table 6). This reflects the often unclear reporting about some aspects of quality in qualitative studies, such as theoretical perspective, adequacy of sample and selection methods, data collection methods, analysis methods, consideration of limitations and of ethical issues.

Quantitative evidence for the health and well-being impacts of school gardening Fruit and vegetable intakes
Thirteen studies reported children's fruit and vegetable intakes [9, 10, 15, 16, 21, 26-28, 30, 31, 36-38] (Table 7). Only two studies reported statistically significant increases effects [16,38]. Both used outcomes based on children's self-report, and were non-randomised studies that were rated weak in the quality appraisal.

Nutrient intakes (and other dietary outcomes)
Six studies reported children's nutrient intakes or other dietary outcomes [9,10,16,22,26,28] (Table 8). Four studies reported statistically significant changes in nutrient intake [9,10,16,28]. However, in two studies there was only one statistically significant finding from the multiple nutrient indicators measured, [9,28] one of which related to a decrease in dietary fibre in the control group, rather than improvements in the intervention group [28]. The other study showed more convincing positive effects for three nutrient indicators, but data were based on children's self-report and was selectively reported for three nutrients only [16].

Food preferences
Thirteen studies reported children's food preferences, including willingness to taste and taste ratings for fruits and vegetables [15,25,26,[29][30][31][32][33][34][35][36][37][38] (Table 9). Eight of the studies reported statistically significant intervention effects showing increased preference for fruits and vegetables [15,30,[32][33][34][36][37][38]. However, these are subjective measures which are highly susceptible to social desirability bias, especially in children who may be eager to please after learning about healthy foods and growing vegetables during school gardening time. We consider food preferences as an early step on the pathway towards behaviour change, but not indicative of behaviour change in itself.

Knowledge and attitudes towards food
Ten studies reported children's knowledge and attitudes towards food [10,[30][31][32][33][34][35][36][37][38] (Table 10). Seven of the studies reported statistically significant intervention effects [10,30,33,34,[36][37][38]. For the most part these were positive effects, showing improved knowledge or attitudes towards food in the intervention groups. Interestingly, one cluster RCT (Trial 1) in the UK found that students who participated in expert-led gardening activities (intervention group) for 18 months were less likely to have positive attitudes towards fruits and vegetables, compared with students who participated in teacher-led gardening activities (comparison group), suggesting that teacher-led activities might be more effective [10]. However, this study had mixed results with other outcome measures showing no statistical difference between teacher and expert-led gardening. As stated above for children's food preferences, these measures of knowledge and attitudes are susceptible to social desirability bias and reflect possible behaviour change mechanisms rather than actual behaviour change.

Physical health and activity
Two studies reported physical health measures, both including waist circumference, body mass index (BMI), and systolic and diastolic blood pressure [22,28] ( Table 11). The only statistically significant difference was reported in a non-randomised controlled study for diastolic blood pressure, which lowered more in the intervention group (cooking, nutrition and gardening) compared to the control group [28]. However, all of the blood pressure readings in this study were within the normal range for school children (systolic: 85-120; diastolic: 50-80) so this finding may not be clinically relevant as an improvement in physical health. Another Table 3 Summary of included qualitative studies (n = 16) (Continued) Viola (2006) [52] Australia Key informants from one primary school (n = 6) and one secondary school (n = 9); students in grades 4-9; Indigenous Australians; remote rural communities   No name (nutrition education plus gardening) [35] No name (based on Junior Master Gardener) Parmer (2007) [36] No name (based on Pyramid Café/Health and Nutrition from the Garden) Ratcliffe (2011) [38] No name (garden-based learning activities)  (Table 11). While children who participated in school gardening (intervention group) reported being 'usually' less sedentary (p = 0.001) and spent more time engaged in 'moderate' physical activity (p = 0.010) compared to the control group, there was no increase in 'light' physical activity or reduction in sedentary behaviour when measured objectively using accelerometers. Note however that the accelerometer analysis was based on a selected subsample of students.

Well-being
Four studies reported children's well-being using a variety of measures, only some of which used valid and reliable scales, including quality of life, life skills and interpersonal relationships [17,24,30,39] (Table 12). Two of the four studies did not find a significant difference between intervention and control groups using their selected measures at follow up [24,30]. The other two studies did not report their child wellbeing outcomes adequately [17,39].
Overall, quantitative evidence for the impacts of school gardens is limited with some suggestion of improvements in knowledge, attitudes and preferences towards fruit and vegetables, but little objective evidence of changes in eating habits or physical health benefits.
Qualitative evidence for the health and well-being impacts of school gardening Qualitative evidence was synthesised narratively and is described below in relation to the health, wellbeing and educational impacts described by children, teachers and parents, and those factors thought to be associated with the success and sustainability of school gardens. Tables 13  and 14 show the subthemes which make up each of these broad themes and which studies contributed to each.

Health impacts
Most studies described perceived nutritional benefits for children involved in school gardening, including greater knowledge and awareness, improved attitudes towards food such as willingness to try new foods, and healthier eating habits [13,25,40,44,45,[47][48][49][50][51][52] (Table 13). This quote illustrates how participation in school gardening created a sense of connection to the food grown, which encouraged children to be more adventurous and led to increased preference for vegetables. Staff and volunteers observed these changes in children's attitudes and behaviour and some, like this administrator, became optimistic about the potential of school gardens to generate major shifts in food culture in the long term.
We've got to start with these kids now, so that when they become the grandparents, they're modelling correctly for the kids. We're probably not going to change the values of today's elderly and today's parents, but if we begin with the kids we're going to have a chance over time to change the health and wellness of the population. -Administrator [40] These kinds of aspirations may be idealistic, but they contributed to enthusiasm and motivation to support the school gardens, the importance of which will be explored later in this review. The quote also illustrates the perception expressed in three studies [40,50,51] that starting at an "early" age (not defined) could be an important factor in the success of school gardening interventions for encouraging the development of a healthy lifestyle.
Other studies described school gardening as an opportunity for physical activity for both children and adults [13,40,[50][51][52]. Primary school teachers appreciated the  Overall  rating  1  2  2b  3  4  5  6  7  8  9  10  11  12 Ahmed (2011) [40] Block (2012) [24] Block (2009) [25] Gibbs (2013) [26] Townsend (2014) Where multiple publications reported qualitative data from the same study, they were appraised as one study Y yes, P partial, N no, CT can't tell, NA not applicable Overall quality rating: strong (11-12 ratings Y); moderate (6-10 ratings Y); weak (1-5 ratings Y) Legend for Is the sample adequate to explore the range of subjects and settings, and has it been drawn from an appropriate population? 6. Was the data collection adequately described? 7. Was data collection rigorously conducted to ensure confidence in the findings? 8. Was there evidence that the data analysis was rigorously conducted to ensure confidence in the findings? 9. Are the findings substantiated by the data? 10. Has consideration been given to any limitations of the methods or data that may have affected the results? 11. Do any claims to generalisability follow logically and theoretically from the data? 12. Have ethical issues been addressed and confidentiality respected?
The scoring system used above was adapted for the purposes of this review physical aspect of gardening for certain groups of students in particular: "the boys that you want to keep physically busy" and those "children who cannot concentrate in class" [51]. In a rural Hawai'ian elementary school, parents saw it as an opportunity "to teach work ethics, to become physically strong and healthy, and to raise awareness of how other people labour to make our lives better and easier" [13]. School gardens are positioned as part of a social, even moral, education.
It makes me feel good inside, all fresh, good… I enjoy touching the soil, the plants. You can feel them…I feel part of them…Yes, it makes me feel that I can care more about things… Being more gentle, caring more, the plants are like people. -Student, age 17 [11].
These emotions are visceral and again there is the sense of connection to nature, which is very different to the classroom experience and brings different lessonsbased on empathy and careto the children in terms of how they interact with people.
A child who struggled and had learning disabilities … and just her confidence and her ability to outshine other kids, who have strengths in other areas was just amazing and she was just really comfortable, in her element. She knew exactly what she was doing, she was in control, she was starring while she was organising the other kids. The building of confidence was just amazing. -Teacher, primary school [25].
This quote echoes those above and suggests particular benefits of mastery and empowerment for children who do not excel in the usual academic setting, such as those with learning or behavioural difficulties. The school gardens allowed them to shine in different ways and to experience success.
In some studies, children and adolescents described school gardens as peaceful places (using words like 'refuge' or 'sanctuary') where they could slow down and let go of any stresses [11-13, 24, 45]. Students who reported these kinds of benefits included some with mental health disorders like Attention Deficit Hyperactivity Disorder (ADHD) and depression [11], behavioural and emotional difficulties [12] and minority ethnic groups including recent migrants [45]. Teenage gardeners articulated reasons why they found the garden so relaxing [11]. For some, it was about being outdoors and the connection with nature, which gave them a sense of perspective. For others, the contrast of physical work allowed the brain some quiet time for reflection and this enabled them to process stress.
It's almost like meditation, like my body is present but my mind just kind of drifts off and goes someplace else, and thinks about things…It's brainless tasks most of               the time, so it's also like zenful, so you get to listen to things…I think about stuff, so I don't have to go home and think about it right before bed, so instead I can just go to sleep and stuff. I just feel happier in a way, and more at peace. -Student, age 15 [11].
This is another example of the sensory, visceral nature of gardening activities, which may have stress-reducing or restorative effects similar to those described in Ulrich's psycho-evolutionary theory [53].
Children experienced positive and negative emotions in the school gardens and participants described how they were able to express themselves and manage their emotions more effectively in that environment [11,13,25,41,[49][50][51]. For example, following the vandalising of the school garden, a teacher said: It really offended them that these students had done this damage to their garden…So then we talk about it and say, well, it made them feel very angry that these children had destroyed part of their garden…it was a positive experience that the children learn that doing what to those children must have been a fun thing to do to go tear up our garden, didn't make us feel good. They were on the receiving end of it and so even though it was a negative experience you can make it a positive one. -Teacher [41].
In this example, the teacher had her own views about how the negative experience has been turned into a constructive learning experience, but this was not articulated by the children themselves.
The social well-being impacts of school gardening were mainly about building relationships [11-13, 25, 41, 44, 47, 49-52]. Children enjoyed interacting with their friends, teachers, gardening specialists, parents and volunteerssome of whom were people they would not normally come into contact with.
You have to work together….It's not about individualism which is promoted in the school structure in some ways, but really communicat[ion], cooperation and ownership of something. -Garden staff [40].       [43] Gardening was seen as promoting teamwork and cooperation, working together towards common goals, which may help to break down some of the social boundaries and elitism associated with traditional academic structures.
Teachers, parents and volunteers also enjoyed the opportunity to interact with children in the garden setting and described how it improved teacher-student and intergenerational relationships [25,47,51]. Volunteers described feeling valued and 'belonging' in the school gardens [43,47]. This gave them a sense of purpose because they felt they were doing something worthwhilecontributing to the children's education and giving back to the community. In this respect, the personal and social well-being impacts of school gardening are interrelated and this combination of factors creates the motivation for volunteering.
Finally, participants described how school gardening contributed to improved cultural awareness and cohesion [12,25,40,45,50]. Parents from non-Englishspeaking backgrounds who might not contribute to other school activities felt comfortable in the garden [24]. In multi-cultural communities, the combination of gardening and cooking activities provided an opportunity for children to learn about each other's cultures [45]. It was also a successful medium for developing English language skills as children relaxed and engaged in "everyday conversations" [45].
Qualitative evidence for educational impacts of school gardening The most common outcomes reported alongside health and well-being impacts were related to potential educational impacts of school gardening [11-13, 25, 40-42, 44, 45, 47-51] (Table 14). As we have included only studies also reporting health and well-being impacts, we can only present evidence for the educational impacts of school gardening from these studies, and so this is not a comprehensive synthesis of the qualitative evidence for the educational impacts of school gardening. Six studies which focussed on educational aspects were excluded during full text screening because they did not match our primary review focus on health and well-being outcomes (see Fig. 1).
Whilst none of the included studies reported academic attainment outcomes (quantitative), two qualitative studies reported children's beliefs that school gardening was having a positive impact on their school work [11,44] and these beliefs were echoed by teachers and volunteers in four more studies [13,25,50,51].
I'm able to complete my homework faster, because I'm in a better place to do other things, because I just spent an hour not worrying about my homework and my grades and my timing for anything, because there's no deadline here. -High School Student with ADHD, age 17 [11] This quote provides further evidence that it might be children with special educational needs that stand to benefit the most from school gardening. It also suggests that well-being impacts such as stress reduction may lead to academic impacts for some students, echoing the proposed mechanisms of Attention Restoration Theory, which suggests that contact with nature can restore depleted ability to concentrate [54].
Increased levels of engagement and motivation among children who participated in school gardening was noted in some studies, although it was not always clear if this was referring generally to time spent in the garden, to garden activities intended to promote academic learning, or also to classroom-based learning [12,13,24,41,50]. In one study, teachers described how they harnessed this potential by tailoring gardening activities to the needs of individual children and creating opportunities for them to demonstrate their skills and knowledge [24]. In our view this has obvious links to the well-being impacts of confidence and self-esteem. However, some children did not work well in the gardens and teachers found them difficult to 'contain' in open spaces [25]. Suggested reasons for this were that some tasks were repetitive and the rewards of gardening were not immediate, resulting in loss of engagement [25].
Some children developed a greater awareness of the environment through school gardening [11,25,41,44,45,[48][49][50]. This varied from awareness of the immediate garden environment, such as water conservation, seasonality, composting techniques and local wildlife [25,44] to consideration of global environmental issues, such as the food supply chain ("ground to plate"), sustainability, recycling and the importance of protecting the environment [44,45,48]. Gardening was seen as a positive environmental behaviour and reinforced the sense of connection with nature [11,45,48].
There were also 'educational' impacts for adult staff and volunteers in terms of learning new skills in the school gardens and related activities like cooking [25,47,50]. In some schools there were formal opportunities for volunteers to attend short courses and gain certificates, which helped to keep volunteers motivated [43].

Qualitative evidence for factors influencing the success and sustainability of school gardening programmes
Most qualitative studies discussed one or more aspects of school gardening programmes that had contributed to their success and/or challenges they had to overcome, evidence which will be useful for schools considering, implementing or managing a gardening programme [12,13,25,40,41,[44][45][46][47][48][49][50][51][52] (Table 14).
The 'experiential' or hands-on learning style was described as an effective way to teach children academic subjects in a more applied and holistic way [25, 40, 44-46, 48, 50-52].
We've done a lot of graphs, a lot of growth measurement. Planted seedlings, measured them and predicted at sixteen weeks, forecasting what size they will be. They are graphed and monitored every fortnight… And we've talked about sustainability, compost and everything just ties in…We've used maths, perimeter, and volume in the garden…Cubic metres…That would have been a really good one for the [grade] five/sixes, if they had actually bought the soil, found out the costing. A lot of things like that you think of in retrospect…There's still more scope to have time in the regular curriculum and a more consistent approach, to have more of a strategic approach. -Teacher [25].
This example from the Stephanie Alexander Kitchen Garden Program illustrates the potential for schools to integrate core curriculum subjects with fun gardening activities, and suggests a strong mechanism underpinning the potential well-being and educational impacts of school gardens. However, not all the interventions studies in this review integrated garden and curriculum activities in this way (Table 4).
Some qualitative studies indicated that some form of cooking or food preparation was integrated with the school gardening intervention, or encouragement to become involved in cooking or food preparation at school or at home, for some or all of their participants (but this was not necessarily a key component of the integrationsee Table 4) [13,24,42,45,47,48,50,52].
Doin' the cook-up with Miss…was fun. We put a recipe book together for the tuckshop as well. We did this every week so that the tuckshop would have healthy food. -Student [52].
Cooking facilitated enjoyment and a sense of achievement. Passy also suggests that the sense of celebration created by sharing cooked garden produce was also important in encouraging students to taste the food [50].
We have a whole bunch of young adults who know how to go to the shop or the market and pick up some vegetables and make themselves something delicious out of it…imagine uni students nourished on seasonal vegetables instead of two-minute noodles. -Volunteer [47]. This example, from the Stephanie Alexander Kitchen Garden Program, illustrates how combining experience of gardening and cooking also gave children some of the life skills needed to live healthily [47]. It also supports the earlier quote from Ahmed et al. (2011) which proposes the potential long term impacts of changing attitudes to food through gardening. School gardens were also described by teachers, parents and volunteers as being supportive and inclusive environments; a 'level playing field' where all children could participate equally [12,13,24,47,50]. This characteristic may help to explain well-being benefits if children felt comfortable to be themselves and participate without any sense of pressure or competitiveness. One study described how achievements in the garden were celebrated in school newsletters or assemblies, which contributed to feelings of pride, confidence and motivation among children who were previously unruly [50].
Considering the socio-demographic and cultural characteristics of the school community in the design of school gardening programmes could ensure that they effectively engage both children and the wider community [12,25,40,45,52]. This connection with cultural heritage and local foods was particularly important in multi-ethnic or native/indigenous communities.
I bring in a variety of people throughout the year to help with various facets of our garden … Last week we had someone come in and show us how to prune our fruit trees and so they get all different kinds of role models…some people are perhaps more patient, some people are less patient....they realize that not all men are the same, not all women are the same, and they get to see people who are not teachers. -Teacher.
This quote relates to gardening specialists, but support from volunteers (including parents, grandparents and other members of the community) and local organisations/businesses was also valued by teachers for similar reasonsincreasing capacity, diversity of skills, materials and resources. However, some schools experienced difficulty recruiting enough volunteers [25,46,50].
The four of us especially have all realized they need some support for this kitchen garden program, being that our funding runs out at the end of this year. So we ran this bloody fair and that was six months of my life and that's what I gave up to ensure that my children still have this program in their school.
-Parent volunteer [43]. This is clearly an example of overdependence on volunteers, which became a source of resentment and threatened the sustainability of the garden. This suggests that schools need to consider the balance between making use of volunteers and keeping them motivated. The same applies to teachers (and other school staff) and two studies highlighted concerns about increased pressure on workloads and fitting school gardening into an already overcrowded timetable [25,50]. One innovative way of motivating teachers was to offer continuing professional development (CPD) opportunities, benchmarks to work towards and cash prizes [50]. A study of the California Instructional School Garden Program found that when schools formed a garden committee (including administrators, teachers, parent/community volunteers and garden coordinators) it helped to define roles, share responsibilities and reduce the risk of overburdening any one person [46].
Finally, some schools experienced financial challenges, such as securing ongoing funding and resources for the school gardens [13,25,46,50,51]. Schools had found various solutions included fundraising events, donations from local businesses and grant applications [25].
There is a lot of sharing that goes on within the gardening community, and I think it's important to reach out beyond the school gardens and contact people in community gardens and local gardening clubs. I have also contacted all of the retailers in this areaall the big box stores, the local nursery stores if they have damaged goodsif they have goods that are unsalable in any way, if they're just old seeds, I'll take them. -Member of staff [46]. This example demonstrates the importance of developing links with the wider community to increase the visibility and sustainability of the school garden.

Discussion
In this mixed methods review, we have systematically and transparently identified, selected, appraised and synthesised the best available evidence on the health and well-being impacts of school gardens. We have used the highest quality international evidence available, although much of this, particularly the quantitative evidence, was judged to be weak in our quality appraisal (Tables 5 and 6).
We found some quantitative evidence for nutritional impacts of school gardening, such as increased preference for, and consumption of, fruits and vegetables. However, many of the included studies relied upon self-reported outcomes, likely to be affected by social desirability bias, especially in children in school settings. It is notoriously difficult to measure food consumption accurately, with different measures having different challengesthe CADET tool, for example, has been found to overestimate fruit intake and underestimate vegetable intake [55]. It was not possible to conduct meta-analyses due to study design and data limitations. Measurement scales, or methods of applying the same measures, were too heterogeneous to allow outcomes to be pooled. In addition, interventions ranged in length from 10 weeks to 18 months.
We also found substantial qualitative evidence on a wide range of health and well-being impacts but these were rarely supported by the quantitative evidence, either because these outcomes were not measured, or because few studies identified significant impacts. It is not clear why these perceptions in the qualitative evidence synthesis are supported by few findings in the quantitative evidence base; whether this should be treated as evidence of no effect or whether, especially due to limitations in the quantitative study designs, it should be seen as no evidence of effect. It should also be noted that only three studies used mixed methods to evaluate impacts, and the quantitative and qualitative evidence comes largely from different school gardening interventions.
Our qualitative synthesis provides contextual information about which aspects of school gardening may be important, how health and well-being impacts may be related to educational impacts, and what factors are important for the success and sustainability of school gardening programmes. This qualitative evidence provides plausible suggestions for how school gardens could lead to health and well-being improvements which may help to influence better study design and the elements of school gardens that have the potential to be beneficial.
Based on the qualitative synthesis, we have developed a conceptual model (Fig. 3) to visually represent some of the possible mechanisms and pathways through which gardening could lead to health and well-being impacts. These are our interpretations of the evidence we have synthesised and, to some extent, the model has been left open to further interpretations, without the use of lines joining up specific pathways. It should be read from left to right to consider how the physical and social aspects of school gardening, coupled with factors influencing success and sustainability, might lead to health and wellbeing (and other) outcomes. The bottom arrow suggests a feedback loop mediated by feelings of enjoyment, engagement and motivation. This model builds on the 'social-ecological conceptual model' presented in a previous review, which depicted potential short-term (proximal) and long-term (distal) effects of school gardens and the interconnections between individual, family, school and community-level effects [3]. Our model also suggests the potential for broader intermediate and long term impacts, although we have focused on more immediate, individual-level health and wellbeing effects as determined by our original review questions. Such long term effects are supported in the broader literature, particularly those suggested that understanding and appreciation of the natural world in childhood may lead to environmental responsibility in adulthood as well as support broader perceptions of wellbeing and quality of life [56,57].
Although much of the evidence from individual qualitative studies is context specific, we have observed several cross-cutting themes that we believe to be transferable between studies. Firstly, school gardening can be integrated with the wider curriculum to maximise opportunities for learning: from nutrition education, to practical growing and food preparation skills, to core curriculum subjects taught in fun ways. To achieve this it is important for teachers to be involved in developing and delivering school gardening activities, with support from other stakeholders in the school and community. Secondly, school gardens appear to have particular benefits for children who have complex needs (behavioural, emotional, or educational) and do not thrive in an academic environment. The evidence suggests that these children may be able to express themselves better in the garden, leading to feelings of calmness, self-esteem and success. Gardening may therefore be described as physical, social and visceral; distinct and complementary to the individual and cerebral nature of classroom education. Thirdly, we noticed a two-way flow whereby the perceived benefits associated with school gardening meant that children were motivated to continue gardening and adults (teachers, parents and volunteers) were motivated to continue to support the school gardening programmes. This feedback loop contributes to the ongoing success and sustainability of school gardening programmes, as indicated in our conceptual model (Fig. 3).

Strengths and limitations of the review
By combining quantitative and qualitative synthesis methods, this review has highlighted the divergence between these research methods and the need for greater synergy. The qualitative research suggests that health and well-being impacts may be felt by those children who struggle in a classroom setting, but quantitative studies did not examine this subgroup. It is possible that average population outcomes obscure impact among these children, or that the outcomes are less relevant to them. Although improvements in eating habits and physical activity were reported in the qualitative research, these were poorly supported by the quantitative studies. Again, it is unclear whether this is due to lack of effect, or deficiencies in the study designs. The qualitative research suggests holistic effects that may be difficult to quantify, as well as suggesting that impacts may be felt in the medium to long term, whereas included studies report only short term follow up.
Whilst the qualitative studies provided the greatest insights, we recognise that by combining multiple studies set in different contexts, some of the meaning and depth of findings from individual studies will have been lost. Whilst we added our own interpretations to those of the authors, we were limited by the original study designs and implicit biases. For example, the primary studies did not consider the perspectives of children who did not participate in school gardening and reasons for this.
School gardens can be seen as operating in line with the WHO Health Promoting School's Framework which aims to take a holistic approach to health promotion in schools [58]. Our focus was health and well-being impacts and we did not include studies that focused only on educational impacts of school gardens. Future reviews could consider this broader remit which may be important for school policy makers.
We had hoped to comment on whether there were different impacts of school gardening interventions on health and wellbeing for different age groups. However, the majority of studies focus on younger children, in pre-school [21] or primary/elementary school [9, 15-17, 24-31, 33, 35, 37, 39] with only a small number of participants of middle school [22,38] or junior high age [17] (Table 2). We therefore conclude that there is insufficient evidence to answer this question.
We restricted this review to OECD countries because developing countries have very different baseline health (and nutrition) characteristics and needs. Most of the evidence comes from the UK, Australia and the USA and it is unclear how transferable the findings are beyond these locations.

Implications and recommendations for future research
The quality appraisal of both quantitative and qualitative studies included in this review highlighted weaknesses in study design and reporting, despite using strict criteria to exclude the weakest study designs. We would recommend that future studies apply the quality criteria used in systematic reviews at the design stage to improve the robustness of the findings and facilitate meta-analysis. More convincing quantitative evidence is needed to promote school gardening programmes as public health interventions. Greater use of objective measures would provide more robust evidence and consistency in measures used across studies would allow meta-analysis in future reviews.
Our findings have some resonance with theories identified in a recent systematic review of how the school environment impacts on student health [59]. For example, the 'theory of human functioning and school organisation' suggests that the ways in which schools implement formal and informal modes of teaching, and develop relationships between staff and students, influence students' commitment and engagement to learning [60]. The 'social development model' suggests that prosocial activities can increase students' commitment to school [61]. Activities that support social and emotional learning have the potential to reduce stress and improve behaviour, both of which may ultimately improve school performance [62]. Although gardening was not one of the mainstream activities tested in the meta-analysis by Durlak et al., it has the potential to be. Considering in more detail how school gardens are anticipated to impact on the school experience, student health and wellbeing and school outcomes, including through the development of logic models or theories of change, garden programmes' impact could be enhanced.
A set of theories in the broader literature suggest that school gardening may have longer term impacts than those addressed in the studies included in this review. Positive and repeated contact with elements of the natural world in childhood has been suggested to relate to pro-environmental behaviours and beliefs in adulthood [57]. Pro-environmental behaviours and 'connectedness to nature' have been found to be related to dimension of wellbeing [63]. Furthermore the type and frequency of childhood exposure to natural environments is thought to influence adult use of such spaces [64]. There is a growing body of evidence which has shown robust associations between use of natural environments (for leisure, physical activity and so on) and a range of positive health outcomes [65]. There is therefore interest in identifying ways in which children can be provided with regular and meaningful opportunities to experience the natural environment.
Future studies on school gardening could usefully make use of theory-led methods, such as realist synthesis or evaluation [66,67], to develop evidence-based causal explanations of how and why school gardens work, for which groups of students, in which types of schools.

Conclusions
Despite their popularity, there is currently limited quantitative evidence that school gardens can have health and well-being benefits for students, and the evidence that does exist is based on self-reported outcome measures.
The qualitative evidence suggests that participants in school gardening programmes (including children and adults) may experience and perceive a range health and well-being impacts. Further high quality evidence is needed to facilitate subgroup analysis of health benefits and the extent of well-being benefits.
School gardens are complex interventions, yet few studies articulated a logic model to show how it was believed that school gardens might have an impact on health and wellbeing. More appropriate study design, and more consistency in the way food intake is measured, is required.