Background

Poor diets and physical inactivity are prominent contributors to chronic diseases [1]. Dietary risks factors are thought to directly contribute 5–14% to all death and disability in high-income countries like Australia, the United Kingdom, Canada and the United States [2]. Interventions to improve diet and physical activity have become an important focus for public health and for governments, with environmental factors receiving attention [3]. Interventions that involve environmental activities such as gardening are thought to have several health benefits including physical, mental and psychosocial outcomes [4].

Conceptual models such as the one proposed by Lovell and colleagues [5] suggest several health and community benefits of participating in gardening, particularly in communal spaces. Gardening is a physically active pastime [6] and may also address food and nutrition-related factors, through fruit and vegetable production and consumption [7]. Gardening also encourages experiences in nature which may have effects independent of other health behaviors such as stress reduction [8]. Participation in community gardening activities may encourage social interactions and the development of social support networks, as well as broader community-level components such as social cohesiveness and neighborhood attachment [5]. Thus, gardening in communal spaces may be useful for chronic disease treatment and prevention by targeting multiple health behaviors, but also concurrently addressing individual-level psychosocial outcomes such as social isolation, mental health and general wellbeing. Community gardens fall under the umbrella of ‘urban agriculture’, which incorporates both domestic or home-based gardens, as well as gardens open to community members for the purposes of growing, cultivating and taking care of plants and flowers for non-commercial outcomes [4]. The current review will specifically focus on the latter type of garden, spaces open to the general public or community.

Previous publications have reviewed the evidence primarily for the effects of community gardening on food and nutrition-related outcomes. Garcia and colleagues [9] reviewed studies examining urban gardens and food and nutrition outcomes among adults, with evidence of positive outcomes on fruit and vegetable consumption, access to healthy foods, as well as improved food perceptions such as the value of organic production and cooking. Importantly, that review was limited to studies among adults, food and nutrition outcomes, and studies of home-based gardens rather than community gardens. McCormack and colleagues [10] reached similar conclusions from their review of community gardens studies conducted only in the United States, as well as methodological issues identified in the studies reviewed. Such findings were echoed by Audate et al. in their scoping review of urban agriculture and its effects on health, wellbeing, food security and social capital [11]. Recently, Kunpeuk et al. [12] conducted a meta-analysis on the health and nutrition-related outcomes associated with community gardening, which suggested a positive effect of community garden participation on body mass index (BMI).

While there have been reviews on community gardening, most past reviews only consider nutrition-related outcomes in isolation from other, broader health factors, or behavioral and psychosocial outcomes. The potential for wider neighborhood-level benefits of community gardens have been understudied. By bringing together information on multiple outcomes we hoped to establish a comprehensive view of the evidence on community gardens that is broader in scope. Therefore, the aim of this work is to systematically review the evidence on effects of community gardens for effects on the following outcomes:

  1. 1.

    Food consumption, with particular attention to vegetable and fruit intake

  2. 2.

    Health outcomes, with particular attention to physical activity

  3. 3.

    Psychosocial measures, such as (but not limited to) social isolation, mental health and wellbeing

  4. 4.

    Community sentiment, such as (but not limited to) social cohesiveness

Additionally, we aimed to collate information on the characteristics of people who use community gardens and whether the effects of community gardens on outcomes might differ according to location (urban, regional, remote) or socioeconomic position.

Materials and methods

The methods were undertaken according to a pre-written protocol which is available from the authors upon request. The review was undertaken using standard systematic review methodology following the Cochrane Collaboration methods and is reported according to the PRISMA guidelines [13].

Search strategy

We searched Medline via PubMed platform, Web of Science, PsycINFO, EBSCOhost and CAB Abstracts from inception until 4th December 2020. To capture literature across all the key outcome areas, the search strategy was deliberately broad in scope, covering databases from health, psychology and sociology. The search strategy was tailored to each database and search terms were pilot tested. MeSH terms and keywords from relevant articles were reviewed to design searches most likely to identify relevant articles. When possible, searches were limited to articles published in English and to humans, and searches were not limited by date or by setting (e.g. high and low-middle-income countries were eligible). In addition to the search strategy described above, we reviewed the reference lists of systematic reviews in this field for potentially relevant studies. The search strategies for each database are included in Supplementary Table 1.

Eligibility and PICO (Participants, Intervention, Comparator and Outcomes) criteria

Studies that make inferences about community gardens were included. Quantitative studies were prioritized for evaluation; qualitative studies were excluded from the review unless they also reported quantitative data. Evidence from randomized controlled trials (RCTs) were considered separately from observational studies and case studies were excluded. Ecological studies were eligible for inclusion as implementation of community gardens may often occur at the higher community (and not individual) level.

Type of participants

‘Participants’ refers to all community members who may freely access community gardens. Participants were not limited to any particular subgroup of the community or by any characteristic (e.g. age, gender).

Type of intervention (for RCTs) or exposure contrast (for observational)

Community gardens were conceptualized as publicly accessible spaces that are used to grow vegetables and fruit. This definition included council median strips or verges that are made accessible and permissible for food production by the public, but excluded incidental use of verges by individuals for growing vegetables or fruit for personal purposes. The definition excluded production of crops for profit and animal-based food production, such as using community spaces for animals that produce milk and eggs, or for collection of honey. It also excluded fruiting trees on government properties (e.g. botanical gardens) or gathering native and non-native foods from national parks. The motivation for developing such gardens was not considered; whether they were designed for example, for food production in response to food insecurity issues in that area, or to create social and community connection. The key concept of the definition of community gardens was that they reflected public access to spaces; therefore studies that did not involve free access to the public were excluded (e.g. gardens in schools, hospitals or jails that are not freely accessible to the public).

Type of comparator

We adopted the counterfactual approach to understanding the effects of community gardens. For RCTs, the comparator was community members who did not receive the intervention (community gardens) and for observational studies, the comparator was non-exposed controls, or a pre-exposure group for pre-/post- designs.

Types of outcome measures

Outcomes were categorized as:

  1. 1.

    Food consumption, with particular attention to vegetable and fruit intake

  2. 2.

    Health outcomes, with particular attention to physical activity

  3. 3.

    Psychosocial measures, such as (but not limited to) social isolation, mental health and wellbeing

  4. 4.

    Community sentiment, such as (but not limited to) social cohesiveness

Characteristics community garden users and differences on effects of community gardens according to location (urban, regional, remote) or socioeconomic position were also explored.

Screening

The titles and abstracts of all identified articles were examined using Rayyan software (a software program used to collate and screen papers for systematic reviews). The authors conducted the screening process and each title/abstract was viewed by two authors. Only articles that were irrelevant were excluded at this stage. The full text of the article was retrieved if either of the authors indicated that the title/abstract was eligible or unclear.

Data extraction, management and synthesis

The full text of each article was reviewed and data were extracted systematically. For RCTs, study quality was evaluated using the Cochrane Risk of Bias tool [14] and the quality of non-randomized studies was assessed using the ROBINS-I tool [15]. A narrative meta-synthesis was undertaken because a meta-analysis was not possible due to differences in study designs and outcomes.

Changes to the protocol

After commencing the searches it became apparent there were more systematic reviews on this topic than anticipated. A post hoc decision was made to include a table summarizing the main findings of each systematic review to collate the full body of literature. No assessment of the quality of each systematic review was undertaken since individual publications were being judged for quality as part of the current review.

Results

The search strategy captured 7,355 articles for screening after duplicates were removed. There were 66 papers judged as eligible for inclusion, but the full text was unable to be obtained for two papers. The flow of studies through the systematic review process is shown in Fig. 1. At least two authors extracted data from 12% of articles. Any discrepancies in data extraction or quality ratings were resolved by discussion at meetings involving all authors.

Fig. 1
figure 1

Flow of studies through the systematic review process

Full size image

Systematic reviews

The searches identified 14 systematic reviews in related topic areas. Table 1 shows the number of studies included in each review, as well as their aims and conclusions. These systematic reviews extended to areas beyond the scope of the current review (e.g. peri-urban agriculture), making only some components of these reviews directly relevant to our research aims. There were 10 to 196 articles included in these reviews. Nutrition and food security were the most commonly studied outcome (9/13 (69%)). The earliest systematic reviews indicated that various forms of community gardening had potential to improve fruit and vegetable intake and food security [16] although this view was not uniform with some suggesting gardens had little impact on food access [17]. Many systematic reviews agreed upon the poor quality of evidence [11, 12, 16, 18]. Recently, Spano et al. described that community gardening may benefit psychosocial wellbeing and this effect was more pronounced among individualist societies compared with collectivist societies [19].

Table 1 Summary of systematic reviews
Full size table

Diet and food-related outcomes

The 23 studies that reported diet and food-related outcomes are summarized in Table 2. Of these, the majority were conducted in the Unites States (16/23 (70%)), with two studies from France and one each from Canada, Japan, the Netherlands, South Africa and the United Kingdom. Most were cross-sectional surveys (15/23 (65%)), four were pre-/post reports of feasibility/pilot studies (4/23 (17%)), one quasi-experimental study, one longitudinal cohort study, and two pilot RCTs. Sample sizes varied from 20 [25] to 1000 [26].

Table 2 Studies addressing diet and food-related outcomes included in this review
Full size table

Studies that compared community gardeners with non-gardeners generally reported higher fruit and vegetable consumption by gardeners [26, 27, 30, 31] or with higher frequency of gardening [26]. However comparisons between community gardeners and home gardeners indicated that fruit and vegetable consumption did not differ [28]. Some community gardeners grew food outside of the community garden [36].

Health outcomes

Table 3 summarizes the 16 studies reporting health-related outcomes (one of these studies was reported in two papers). Eleven (11/16 (69%)) of these studies were conducted in the United States, with two studies from Japan, 1 each from France, the Netherlands and the United Kingdom. Most were cross-sectional surveys (9/16 (56%)), in addition to 3 pre-/post- designs, one quasi-experimental, one longitudinal and two RCTs. Studies ranged in size from 13 participants at follow up [48] to 794 [29]. The diverse outcomes reported in these studies included weight-related outcomes such as BMI, overweight and obesity, self-reported outcomes such as health, physical activity, number of general practitioner visits and the number of chronic illnesses, and clinical measures of hypertension and blood glucose (HbA1c).

Table 3 Studies addressing health outcomes included in this review
Full size table

Weight-related outcomes were reported most frequently (11/16 (69%)) and the findings were mixed. Gardening was sometimes associated with lower weight-related outcomes, for example, there was less overweight and obesity among families participating in weekly gardening sessions [31]. However, studies also reported no difference in BMI, for example, in a cross-sectional survey of allotment gardeners compared with other active groups such as home gardeners and walkers [49], and a survey comparing gardeners to other local residents [50]. With respect to the five studies (5/16 (31%)) reporting blood pressure outcomes, a quasi-experimental study involving a non-randomized intervention suggested the odds of hypertension were lower [29] for gardeners compared with residents at a nearby county with no community garden, or that there were small or no differences in blood pressure [32, 48]. For other health outcomes such as physical activity, lung function, sleep and HBA1c there were too few studies to synthesize evidence or outcomes were measured inconsistently or were inadequately powered to detect changes.

Psychosocial outcomes

The 16 studies that reported psychosocial outcomes are summarized in Table 4. Seven of these studies were from the United States (44%), with two studies from Japan and the United Kingdom, and one study each from France, the Netherlands, Portugal, Singapore and Switzerland. The study designs were either cross-sectional surveys (12/16 (75%)), RCTs (2/16 (13%)), pre-/post (1/16 (6%)) or longitudinal (1/16 (6%)). Studies included between 20 [25] to 469 [38] participants. Outcomes were diverse, with community gardening associated with improvements in happiness [54], social support, social cohesion [35], mental health [50] and quality of life [55], as well as reductions in perceived stress [49]. In contrast, there were no differences observed in perceived health [38], although some outcomes such as effects on depression were not reported [33].

Table 4 Studies addressing psychosocial outcomes included in this review
Full size table

Community outcomes

Table 5 summarizes the seven studies that reported community outcomes. Three of these studies were conducted in the United States, with one each from Canada, Japan, Portugal and the Netherlands. All were cross-sectional in design and sample sizes ranged from 25 [40] to 500 [39]. Findings were generally positive for gardening and community-related outcomes. For example, gardeners had higher neighborhood attachment [61], perceptions of neighborhood aesthetics [38], measures of social cohesion [50] and civic participation [42], compared with non-gardeners, although Machida et al. did not report greater connection among neighbors among community gardeners compared with non-gardeners [39].

Table 5 Studies addressing community outcomes included in this review
Full size table

Effects of community gardens according to location or SEP

No studies were identified that directly compared effects in different locations or by socioeconomic position (SEP).

Characteristics of community gardeners

We located 24 studies that described users of community gardens (Table 6). Studies were from cross-sectional surveys (23/24 (96%)) except for one longitudinal study. Sample sizes ranging from 37 [62] to 1916 [63]. Seven studies were from the United States (8/24 (33%)), with two studies each from Canada and the Netherlands, and one each from Australia, Czechia, Denmark, Germany, Israel, Italy, Japan, Nigeria, Portugal, Spain, South Africa and Zimbabwe. Of these studies, 16 (67%) made no comparisons to non-gardeners and therefore little inference can be made from these studies but they have been tabulated for completeness. Of the eight studies (33%) that compared community gardeners against some other community group (such as non-gardeners or home gardeners), some reported that gardeners had higher educational attainment and income [42, 64] although this was not consistent as other studies reported no differences [50, 51]. Gardeners also tended to be older or were retirees [50, 51, 65].

Table 6 Characteristics of individuals or households who use community gardens
Full size table

Quality of included studies

The quality assessment of the two RCTs have been included as Supplementary Table 2, and the quality assessments for other study designs are in Supplementary Table 3. Of the 34 non-randomized studies included in the review, only two were rated overall as having a low risk of bias. The most common problems were poor or no adjustment for confounding and the potential for selection bias. Deviation from any intended intervention was frequently unclear due to inadequate reporting, as was reporting of missing data.

Discussion and conclusions

The results of this systematic review describe quantitative evidence from 53 studies (54 papers). The outcome with the largest amount of quantitative information was for fruit and vegetable intake, overall diet, nutrients or nutrition knowledge (k = 23 studies). Sixteen studies included health related outcomes, such as physical activity, BMI or blood pressure, and sixteen reported a diverse range of psychosocial outcomes such as happiness, stress and quality of life. Fewer studies reported community-related outcomes of gardeners (k = 7). Importantly, there were few studies located that were conducted in developing countries; the vast majority of studies reviewed here were from developed countries, particularly the United States.

Quite unexpectedly, 14 systematic reviews that had not been identified during the preliminary searches of databases were located. This is testament to how difficult this literature is to capture due to the varying terminology, breadth of outcomes examined, and places where this type of work has been published. Many of the other systematic reviews focus on specific content areas or a particular definition of gardening such as peri-urban agriculture, which has a different scope to our review. Where there was crossover, findings of the current review are somewhat similar to past reports though the current review is more up-to-date and suggests that ongoing (poor) quality of publications is proving difficult to shift.

For dietary outcomes, the results of the current review suggest that users of community gardens consume slightly more fruit and vegetables than non-users of community gardens, with little difference between findings of studies of low, moderate or serious risk of bias. Overall, the quality of the evidence is low with many studies at risk of selection bias and poor adjustment for confounding.

Harvesting fruit and vegetables from community gardens is typically seasonal and this may have influenced data collection, with few studies stating that had been taken into consideration. Of note is one publication indicating that community gardeners purchased more fruit and vegetables than other members of the community [41]. This might indicate that community gardeners are more interested in consuming fruits and vegetables than non-gardeners. Teasing apart the effects of community gardens from the effects of people who choose to use them is particularly challenging. The trial by Heilmayr and Friedman is neatly designed to tease apart the mechanism by which community gardens are purported to have effects, by using different comparison groups that focus on social contact, physical activity or outdoor exposure [34]. While this is a clever design for understanding the mechanisms, the RCT was underpowered and no effects on diet, activity or psychosocial outcomes were noted. Counter to expectations was that food security was consistently higher among community gardeners, as one study suggested highly food insecure participants were less likely to be involved in community gardens [42]. It is plausible to hypothesize that the community gardens may have made participants less food insecure or that gardens are not viewed by people experiencing food insecurity as a possible solution.

With respect to health outcomes, a wide variety of measures were reported in the included studies. It was common for articles to not report whether the more frequent measures such as BMI were self-reported or measured. Self-reported measures of BMI are often lower than measured BMI, and measured BMI is preferable particularly for pre/post designs, which might be vulnerable to outcome reporting bias. Nevertheless, studies indicated that community gardeners perceived themselves as having good to excellent health and as having lower odds of hypertension and overweight/obesity than non-gardeners. However the evidence was not consistent as one study [49] found no differences in physical activity, BMI, blood pressure and lung function between community gardeners and people in other active pursuits (such as home gardeners, walkers). This issue in particular, points to a need for careful consideration of who is being compared in each analysis, as well as the problem of self-selecting into active pursuits, such as community gardening, by healthy people.

Of the psychosocial outcomes, it is important to keep in mind the context. For example, psychosocial outcomes of community gardening from highly impoverished areas in low-income countries are not generalizable to high-income countries and vice versa. However, studies involving immigrants, refugees or culturally and linguistically diverse communities may be relevant. Among many potential benefits, the growth of culturally relevant produce may support resettlement. Gardeners tended to have more social contact and higher indicators of wellbeing than comparators, but again the body of evidence is both small and low in quality.

Of the seven studies reporting on indicators of community sentiment, gardeners rated neighborhood aesthetics and neighborhood attachment more highly than other members of their communities, and their civic participation is higher. Importantly, the current review did not distinguish between community gardens developed for the purposes of creating positive community sentiment or connection, and those gardens developed for the purposes of alleviating food security concerns. The differences in motivation for developing and participating in community gardens may well be important to consider as suggested in a review by Guitart and colleagues [80] and in empirical research from Trendov [81] and Bende and Nagy [82]. Despite motivational differences in community gardening, the current review suggests that the effect on social interactions and community connection appears to exist regardless. Once again, whether this finding is a result of community gardening or because people seeking social interactions self-select into gardening cannot be clearly delineated from the literature due to poor control of confounding and possible selection bias.

The aim to collect information on characteristics of community gardeners was made difficult by the majority of studies not comparing gardeners to either non-gardeners or to the general population. It would appear that community gardeners were generally older members of the community, with a higher proportion of retirees, and with more years of formal education. However, the samples included in individual studies are entirely dependent on the eligibility criteria (and research questions) of individual studies.

Limitations

The limitations of the current review fall into two areas, those that arise as limitations of the studies included in the review and those that are limitations of the review processes itself. With respect to limitations of studies included in the review, there were no high quality well-powered RCTs and most of the evidence from observational research was rated as having a high risk of bias. The lack of randomized trials in this area is not surprising as it is difficult to randomize individuals to involvement (or not) in community gardens. Non-compliance within intervention and control groups would be problematic as some individuals in the treatment group would not interested in gardening, and some individuals in the control group would want to be gardening. This reflects the ‘problem’ of selection bias through self-selecting into desired activity (common to observational studies reviewed here). Other possibilities that could help elucidate the effects of community gardens could involve randomizing individuals as part of a prescription or treatment for health conditions, or randomizing entire communities to the implementation of a community garden though this would involve large commitments by councils and residents. Thus, the small amount of evidence from ‘gold standard’ RCTs will likely continue, and more attention should be paid to improving the quality of the observational evidence. Many studies had poor or no adjustment for confounding. Furthermore, careful attention needs to be paid to what is being compared in each study. For example, a comparison of food security outcomes from more advantaged community gardeners versus individuals accessing food banks could lead to over-estimates of the beneficial effects of gardens [42]. Even though such comparisons may be adjusted for confounders, it is unlikely that individuals are exchangeable on all other factors, and residual confounding is likely to be present. Thus, the evidence from individual observational studies are probably overly optimistic effects across all outcomes. Such challenges with research in the community gardens setting and the poor quality of evidence is unsurprising given the diversity of likely motivations for developing and participating in community gardens, the length of time needed to develop such gardens and then see any health or behavioral changes resulting from participation and the unique nature of each community garden and of the users themselves. Future research should not be dissuaded from investigating the benefits of community gardens, rather as much as practical, attention paid to the issues such as selection bias, adjustment for confounding and exchangeability.

Another potential limitation of the included studies is around external validity or in deciding whether the evidence from this review is applicable to other settings. Studies from low- or middle-income studies may not be directly applicable to high-income countries, and vice versa. However, there may also be external validity problems with the high-income country settings (where there is more evidence). For example, studies conducted in highly disadvantaged rural areas of the United States are not likely to be applicable to affluent areas of Europe or Asia, or to high-density living. Hence, the benefits observed in one setting may not be transferable to others.

Potential limitations from the systematic review process are predominantly around the inclusion of relevant literature and the scope of the outcomes. Databases that would have outcomes to inform the review were deliberately searched but no grey literature was searched and it is possible that potentially relevant studies were missed. Finding all sources of grey literature would be unrealistic for an academic review of this nature. If the results of grey and unpublished literature differed from the published literature, the current paper may have a potentially biased view of evidence. No formal tests of the potential for (positive) publication bias, were undertaken as the outcomes of studies were too disparate. As mentioned earlier, the literature in this field is published across many areas and there are many different terms used to reflect conceptualizations of ‘community gardens’. This became apparent during the search and screening processes. Potentially negative outcomes such as community gardens conducted in areas of poor or contaminated soil quality were also not considered. Although the search strategy located such articles, these studies were out of scope.

Conclusions

In conclusion, the results of the studies included in this review indicate that community gardeners tend to consume more fruit and vegetables, are healthier and participate in civic settings more frequently than non-gardeners. However, the observational evidence that involves selected populations have poor (often no) adjustment for confounding, are at risk of bias. Thus, although the evidence is positive for all outcomes, the potential for bias is sufficiently high that the findings are likely to be overly optimistic effects of community gardens.