Nature-Based Interventions for Individuals with Psychiatric Disorders: A Mixed Methods Systematic Review with Random-Effects Meta-Analysis of Mental Health and Functional Outcomes

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a mixed-methods systematic review and meta-analysis examining the effectiveness and lived experience of nature-based interventions (NBIs) in individuals with psychiatric disorders. The topic is clinically relevant, particularly in light of growing interest in recovery-oriented approaches and adjunctive, non-pharmacological interventions within mental health services. A notable strength is the attempt to bring together quantitative evidence on outcomes with qualitative accounts of patient experience, which adds useful depth beyond effect sizes alone. The focus on populations receiving mental health care, rather than general or subclinical samples, further strengthens the clinical applicability of the findings.

The review is generally well structured, and the search strategy appears thorough. The results suggest that NBIs may be associated with symptom improvements, alongside favourable patient-reported experiences. These findings are of potential relevance for service development. That said, several aspects of the methodology and reporting would benefit from clarification and refinement, particularly regarding the meta-analytic approach and the integration of findings, in order to support a more transparent and balanced interpretation of the results.

Major Comments

1. Heterogeneity and clarity of the pooling approach

The included studies differ substantially in terms of psychiatric diagnoses, intervention modalities, and study designs. While the use of a random-effects model is appropriate, the manuscript would benefit from a more detailed account of how this variability was addressed within the meta-analytic framework.

Specifically, it would be helpful to clarify:

• how outcomes were selected in studies reporting multiple measures,
• how multiple timepoints were handled (e.g., whether endpoint or change scores were prioritised),
• whether any procedures were applied to avoid double-counting of participants across outcomes.

In addition, the interpretability of the pooled estimate would be strengthened by indicating whether subgroup or sensitivity analyses (e.g., by intervention type, diagnostic group, or study design) were considered. Where feasible, even exploratory analyses of this kind may be informative; if not, this should be stated explicitly as a limitation and reflected in a more cautious interpretation of the findings.

2. Integration of quantitative and qualitative findings

Both the quantitative and qualitative components are clearly presented. However, the integration between them remains somewhat limited. As currently structured, the two strands are largely reported in parallel, with relatively little synthesis.

This could be strengthened by more explicitly articulating the relationships between the findings, for example:

• identifying areas where qualitative themes (such as improvements in mood or social connection) correspond to specific quantitative domains,
• highlighting any instances of divergence between the two strands,
• including a brief integrative paragraph within the Discussion to synthesise these perspectives.

Addressing this would enhance the mixed-methods design's contribution without necessitating changes to the underlying analyses.

3. Framing of novelty

The claim that this is the “first” mixed-methods synthesis in this area appears overstated in its current form. It would be helpful to either qualify this more clearly (for example, by specifying the population, setting, or type of intervention) or to indicate how the present review differs from previous work.

Minor Comments

There are a few terminology-related and typographical/formatting issues that would benefit from clarification and correction:

• Terminology (Methods and Results sections):
  The distinction between domains such as wellbeing, quality of life, and psychosocial outcomes is not entirely clear, although these are treated as separate analytic categories. A brief clarification of how these domains were defined or operationalised would improve interpretability.
• Page 4, lines 161–171:
  The paragraph beginning “Extracted data for all studies covered…” appears partially duplicated and should be merged to avoid repetition and improve clarity.
• Page 6, lines 282–284:
  There is a formatting issue where the sentence overlaps with table text (“table S1Table 1…”), which interrupts readability and should be corrected.

 

Recommendation

This is a well-executed and clearly presented review addressing an area that is gaining increasing attention in psychiatric practice, particularly in relation to recovery-oriented and non-pharmacological approaches within mental health care.

The points raised above focus primarily on clarifying aspects of the methods and interpretation of the findings, and addressing them would further strengthen the manuscript.

Author Response

Reviewer #1  

This manuscript presents a mixed-methods systematic review and meta-analysis examining the effectiveness and lived experience of nature-based interventions (NBIs) in individuals with psychiatric disorders. The topic is clinically relevant, particularly in light of growing interest in recovery-oriented approaches and adjunctive, non-pharmacological interventions within mental health services. A notable strength is the attempt to bring together quantitative evidence on outcomes with qualitative accounts of patient experience, which adds useful depth beyond effect sizes alone. The focus on populations receiving mental health care, rather than general or subclinical samples, further strengthens the clinical applicability of the findings. 

The review is generally well structured, and the search strategy appears thorough. The results suggest that NBIs may be associated with symptom improvements, alongside favourable patient-reported experiences. These findings are of potential relevance for service development. That said, several aspects of the methodology and reporting would benefit from clarification and refinement, particularly regarding the meta-analytic approach and the integration of findings, in order to support a more transparent and balanced interpretation of the results. 

We are grateful to the reviewer for their positive appraisal and for their constructive feedback. Our detailed responses are provided below. 

Major Comments 

R1.1. Heterogeneity and clarity of the pooling approach  

The included studies differ substantially in terms of psychiatric diagnoses, intervention modalities, and study designs. While the use of a random-effects model is appropriate, the manuscript would benefit from a more detailed account of how this variability was addressed within the meta-analytic framework. Specifically, it would be helpful to clarify: 

how outcomes were selected in studies reporting multiple measures, 

how multiple timepoints were handled (e.g., whether endpoint or change scores were prioritised), 

whether any procedures were applied to avoid double-counting of participants across outcomes. 

In addition, the interpretability of the pooled estimate would be strengthened by indicating whether subgroup or sensitivity analyses (e.g., by intervention type, diagnostic group, or study design) were considered. Where feasible, even exploratory analyses of this kind may be informative; if not, this should be stated explicitly as a limitation and reflected in a more cautious interpretation of the findings. 

We agree that greater clarity is needed regarding the pooling approach. We have expanded the Methods section to specify how outcomes were selected when studies reported multiple measures, how multiple timepoints were handled, and how we avoided double-counting participants. We now clarify that immediate post-intervention endpoints were prioritised, that each study contributed only one effect size to each pooled analysis, and that outcomes were grouped using an a priori rehabilitation-oriented framework. We also clarify that subgroup analyses by intervention type, diagnostic group, and study design were considered, but were limited by the small number of studies, diagnostic heterogeneity, and variability in intervention content. The Discussion has been revised to interpret the pooled estimate more cautiously as an exploratory synthesis across a heterogeneous evidence base. 

Quantitative analyses were conducted in Stata using the esizei and meta set commands. Extracted outcomes were organised a priori into five domains, clinical symptoms, functioning, quality of life and wellbeing, psychosocial outcomes, and physical health. For each eligible controlled study, standardised mean differences and corresponding 95% confidence intervals were calculated from reported means, standard deviations, and sample sizes. Effect sizes were coded so that positive values indicated greater improvement in the NBI group relative to the comparator condition. Where studies reported multiple measures within the same outcome domain, we selected the outcome most closely aligned with the domain definition and the study’s stated primary or most clinically relevant outcome. Where no primary outcome was specified, we prioritised validated and commonly used measures to maximise comparability across studies.   

Where studies reported outcomes at multiple timepoints, the immediate post-intervention endpoint was prioritised for the main synthesis. Follow-up outcomes were extracted where available but were not pooled with immediate post-intervention effects. Where both endpoint and change scores were available, endpoint scores were prioritised for consistency across studies.  

To avoid double-counting participants, each study contributed only one effect size to a given pooled analysis. Where studies reported multiple eligible intervention arms, comparator arms, outcome measures, or timepoints within the same analysis, a single effect size was selected according to the hierarchy described above. Multiple correlated effects from the same participants were not pooled within the same meta-analysis. 

 

R1.2 Integration of quantitative and qualitative findings  

Both the quantitative and qualitative components are clearly presented. However, the integration between them remains somewhat limited. As currently structured, the two strands are largely reported in parallel, with relatively little synthesis. This could be strengthened by more explicitly articulating the relationships between the findings, for example: 

identifying areas where qualitative themes (such as improvements in mood or social connection) correspond to specific quantitative domains, 

highlighting any instances of divergence between the two strands, 

including a brief integrative paragraph within the Discussion to synthesise these perspectives. 

Addressing this would enhance the mixed-methods design's contribution without necessitating changes to the underlying analyses. 

We have updated the results and the discussion in accordance with the reviewer’s helpful suggestion. Specifically, the paragraph “3.7 Integration” has been explored further as follows: 

When considered together, the quantitative and qualitative strands showed both broad convergences, and some divergence across intervention types and outcome domains. Consistent with our convergent segregated approach, we compared findings across strands narratively rather than through a formal convergence matrix.  

Improvements in clinical symptoms identified in the narrative synthesis were consistently reflected in qualitative accounts of enhanced psychological wellbeing, further supporting the “Psychological Wellbeing” and “Personal Growth” themes. Studies which showed quantitative reductions in depression (Müller et al., 2025; Pedersen et al., 2011, Pedersen et al., 2012; Walter et al., 2023), anxiety (Berget et al., 2007), and psychiatric symptom severity (Zhu et al., 2016, Oh, Park & Ahn, 2018, He et al., 2020, Kam, Siu, 2010) observed particularly in horticultural and structured outdoor activity interventions often involved the same interventions in which participants described feeling calmer (Cooley et al., 2020; Siu et al., 2020), coping better (Iancu et al., 2013; Leighton et al., 2021), and being more active in everyday life (Barley et al., 2012; Cooley et al., 2020; Siu et al., 2020). Similarly, quantitative improvements in social and functional outcomes, including self-efficacy (Berget et al., 2008), coping ability (Berget et al., 2008), and social functioning (He et al., 2020; Siu et al., 2020), aligned with qualitative themes of “Personal Growth” and “Social Relationships,” where participants reported greater confidence (Fieldhouse, 2003; Wastberg et al., 2020), renewed identity (Barley et al., 2012; Iancu et al., 2013), meaningful social participation (Iancu et al., 2013; Kam and Siu, 2010), and engagement in ordinary life activities (Cooley et al., 2020; Pedersen et al., 2012). 

The qualitative findings also helped contextualise why certain intervention types may have produced stronger quantitative effects. Themes related to emotional restoration, sensory engagement, responsibility, and connectedness to nature were particularly prominent in horticultural and outdoor activity programmes, which were also associated with the most consistent quantitative improvements. 

At the same time, some divergence between strands was observed. "Being in Nature" emerged as a consistent theme across qualitative studies, yet this dimension was absent from quantitative measurement frameworks. No included quantitative study systematically assessed sensory or material features of the natural environment as potential mechanisms of change.  In areas where quantitative findings were mixed or limited, such as quality of life and some animal-assisted or brief programmes, qualitative accounts still described changes in mood (Iancu et al., 2013; Pedersen et al., 2012), motivation (Iancu et al., 2013), and connection with others (Iancu et al., 2013; Joung et al., 2025; Pedersen et al., 2012). These discrepancies may reflect limitations in the sensitivity of standardised outcome measures to capture subjective or recovery-oriented changes that participants considered important. Looking across both types of evidence suggests some overlap between what was measured and what participants reported and also points to areas where people describe benefits that haven’t been clearly shown in quantitative results yet. 

Overall, integrating both strands suggests that NBIs may influence psychiatric rehabilitation across multiple interconnected domains, including symptom reduction, emotional regulation, social connectedness, identity, and functional recovery.  

Moreover, in the Discussion section the following paragraph has been added: 

Improvements in clinical symptoms identified in the quantitative synthesis were strongly reflected in qualitative accounts of enhanced psychological wellbeing. Similarly, quantitative improvements in social and functional outcomes, including self-efficacy, coping ability, and social functioning, aligned with qualitative results: participants reported enhancements in self-confidence, a renewed identity, social participation, and engagement in ordinary and meaningful activities. At the same time, some divergence between quantitative and qualitative findings emerged. In domains where quantitative evidence was inconsistent or limited, particularly regarding quality of life outcomes and animal-assisted interventions, participants nevertheless described meaningful improvements in mood, motivation, and social connectedness.  

Taken together with qualitative accounts, the findings suggest that NBIs may contribute to recovery via multi-domain pathways that are not consistently captured by symptom-focused measures. Indeed, the qualitative evidence complemented the quantitative findings by illuminating the lived experiences and potential mechanisms underlying observed clinical improvements, while also identifying benefits that remain insufficiently captured within existing quantitative measures. 

R1.3 Framing of novelty  

The claim that this is the “first” mixed-methods synthesis in this area appears overstated in its current form. It would be helpful to either qualify this more clearly (for example, by specifying the population, setting, or type of intervention) or to indicate how the present review differs from previous work. 

We thank the Reviewer for this comment. We have revised the Introduction to qualify the novelty claim and to clarify how the present review differs from previous work. Specifically, we now explain that previous reviews have largely focused on general populations, specific conditions, or high-level umbrella review evidence, whereas the present review focuses on individuals with diagnosed psychiatric disorders and integrates quantitative effect sizes with qualitative thematic synthesis. 

While interest in NBIs has surged, the evidence base has remained heterogeneous. Previous reviews have largely focused on general populations (Coventry et al., 2021; Nguyen et al., 2023; Twohig-Bennett, Jones, 2018) or specific conditions (Jessen et al., 2025), often using narrative (Harrison et al., 2023) or scoping approaches (Wilkie and Davinson, 2021), and recently published umbrella reviews have begun synthesising this growing body of evidence at a high level (Brandt et al., 2026; Shrestha et al., 2025; Kaleta et al., 2025). However, few reviews have specifically examined outcomes in people with diagnosed psychiatric disorders- including anxiety (Jessen et al., 2025), depression (Jessen et al., 2025; Salomon, Salomon & Beeber, 2018), and stress-related conditions (Jessen et al., 2025; Paredes-Céspedes et al., 2024) - and those that do have focused almost exclusively on quantitative efficacy, with comparatively limited attention to participant experiences, implementation processes, and the meaning participants ascribe to these interventions (Cuthbert, Kellas & Page, 2021). No review to date has integrated quantitative and qualitative evidence in this population.  

We aimed to address these gaps by systematically reviewing and meta-analysing studies of NBIs for individuals with psychiatric disorders, encompassing a broad spectrum of diagnoses and intervention types, while integrating quantitative effect sizes with qualitative thematic synthesis, allowing us to identify not only if NBIs are effective across clinical and functional domains, but also how individuals with psychiatric lived experience perceive, value, and engage with these interventions. 

 

Minor Comments 

There are a few terminology-related and typographical/formatting issues that would benefit from clarification and correction: 

R1.4 Terminology (Methods and Results sections)  
The distinction between domains such as wellbeing, quality of life, and psychosocial outcomes is not entirely clear, although these are treated as separate analytic categories. A brief clarification of how these domains were defined or operationalised would improve interpretability. 

 

We thank the reviewer for highlighting this. This has been amended, and now explicitly stated in the paragraph 2.2 as follows: 

 

Outcomes domains were defined a priori during protocol development and informed by the PICO framework. These were used as guiding categories for outcome extraction and synthesis. Specifically, “Clinical Symptoms” referred to psychiatric symptom severity, including depression, anxiety, stress, and psychotic symptoms. “Quality of Life and Wellbeing” outcomes included individual evaluations of life satisfaction, mental wellbeing, self-compassion, and perceived quality of life. “Psychosocial” outcomes referred to interpersonal and self-related constructs such as self-efficacy, coping strategies, self-esteem, and social participation. “Functional outcomes” captured engagement in daily and meaningful activities and social functioning. Last, “Physical Health” outcomes reported physical health benefits.   

 

R1.5. Page 4, lines 161–171:  
The paragraph beginning “Extracted data for all studies covered…” appears partially duplicated and should be merged to avoid repetition and improve clarity. 

 

We thank the reviewer for noticing it, the duplication has been removed.  

 

R1.6. Page 6, lines 282–284: - AG 
There is a formatting issue where the sentence overlaps with table text (“table S1Table 1…”), which interrupts readability and should be corrected. 

 

We apologise for this oversight; the formatting has been fixed. 

 

R1.7. Recommendation 

This is a well-executed and clearly presented review addressing an area that is gaining increasing attention in psychiatric practice, particularly in relation to recovery-oriented and non-pharmacological approaches within mental health care. The points raised above focus primarily on clarifying aspects of the methods and interpretation of the findings, and addressing them would further strengthen the manuscript. 

We thank the reviewer for their careful reading of the manuscript and for their constructive feedback. We have addressed each of the points raised below and believe the revisions have strengthened the manuscript.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

General comments

The manuscript presents a PROSPERO-registered mixed-methods systematic review of nature-based interventions (NBIs) in adults with psychiatric disorders, integrating a random-effects meta-analysis of nine controlled studies (pooled SMD 0.71, 95% CI 0.29–1.12) with a thematic synthesis of ten qualitative and four mixed-methods studies. The topic is timely, the convergent segregated design is appropriate, and the thematic synthesis is coherent. However, the manuscript has substantive limitations.

Abstract

The abstract overstates inferential reach: given that four of five RCTs were low-quality on CASP and that diagnostic categories are markedly heterogeneous, the claim of greater improvement over comparators should be tempered. Publication bias is not assessed and should be disclosed.

Introduction

The rationale around stress reduction, attention restoration and social connection is well-built but too narrow on neurobiological mechanisms. In a review aimed at informing rehabilitation practice, at least two convergent biological pathways should be briefly acknowledged — circadian entrainment and HPA-axis/stress regulation — which anchor the plausibility of benefits from outdoor NBIs to contemporary psychiatric pathophysiology. Appropriate citations are Monteleone P, et al. Circadian rhythms and treatment implications in depression. Prog Neuropsychopharmacol Biol Psychiatry. 2011;35(7):1569-74. doi:10.1016/j.pnpbp.2010.07.028 (for the light/circadian mechanism linking outdoor exposure to depressive pathophysiology) and Flattened cortisol awakening response in chronic patients with schizophrenia onset after cannabis exposure. Psychiatry Res. 2014;215(2):263-7. doi:10.1016/j.psychres.2013.12.016 (for HPA dysregulation in schizophrenia, relevant given that 43% of participants had this diagnosis). The authors should also more sharply articulate what their review adds beyond the 2025–2026 umbrella reviews they cite.

Methods

Several issues require attention. The use of CASP for RCTs is suboptimal: RoB-2 is the accepted standard and should be adopted. No publication-bias assessment is reported despite a pooled analysis; this should be added, with the caveat that nine studies is below the usual threshold. Section 2.2 ("Data extraction") is duplicated almost verbatim and must be consolidated. The exploratory meta-regression (three moderators, nine studies) is underpowered and this should be acknowledged. The absence of patient and public involvement warrants at least brief reflection, given the qualitative strand's emphasis on lived experience.

Results

The forest plot (Figure 1) labels studies generically as "Study 1"–"Study 9" rather than by author/year; this must be corrected, as it currently prevents readers from identifying which studies drive the pooled effect. A sensitivity analysis excluding the two outliers (effect sizes 2.39 and 2.04) would clarify robustness. The CASP finding that four of five RCTs were low-quality should be brought forward into the main text, since it materially affects interpretation of the meta-analysis.

Discussion

The discussion integrates the two strands reasonably but remains largely descriptive. The mechanistic paragraph should move beyond the Ulrich and Kaplan frameworks and cite the circadian and HPA references noted above. A further forward-looking anchor is Barlattani T, Glymphatic system and psychiatric disorders: need for a new paradigm? Front Psychiatry. 2025;16:1642605. doi:10.3389/fpsyt.2025.1642605, to be cited in one sentence acknowledging that improved sleep and reduced stress — both repeatedly reported in the qualitative strand — may engage neurobiological pathways (including glymphatic clearance) increasingly recognized as relevant to psychiatric disorders, situating NBIs within a biologically informed rehabilitation framework. The limitations section should be expanded to include low RCT quality, the small pooled sample, diagnostic heterogeneity, and absence of publication-bias assessment.

Conclusions

Soften in line with the certainty of evidence: "NBIs show promise" is appropriate; stronger formulations are not, and an explicit call for adequately powered RCTs with standardized recovery outcomes should precede any scale-up recommendation.

Author Response

Reviewer #2  

General comments 

The manuscript presents a PROSPERO-registered mixed-methods systematic review of nature-based interventions (NBIs) in adults with psychiatric disorders, integrating a random-effects meta-analysis of nine controlled studies (pooled SMD 0.71, 95% CI 0.29–1.12) with a thematic synthesis of ten qualitative and four mixed-methods studies. The topic is timely, the convergent segregated design is appropriate, and the thematic synthesis is coherent. However, the manuscript has substantive limitations. 

We are grateful to the reviewer for their appraisal and for their helpful feedback. Our detailed responses are provided below. 

R2.1. Abstract  

The abstract overstates inferential reach: given that four of five RCTs were low-quality on CASP and that diagnostic categories are markedly heterogeneous, the claim of greater improvement over comparators should be tempered. Publication bias is not assessed and should be disclosed. 

We thank the reviewer for this helpful suggestion. We have revised the abstract to temper the inferential language. We agree that transparency regarding publication bias assessment is important and, following PRISMA 2020 guidelines, we have ensured that disclosure of publication bias assessment is clearly addressed in paragraph 2.4 as follows: 

Formal assessment of publication bias was not conducted because fewer than 10 studies were included in the meta-analysis. Instead, potential publication bias was considered narratively, including whether small studies reported disproportionately large or significant effects and whether unpublished or grey literature was searched.   

R2.2. Introduction  

The rationale around stress reduction, attention restoration and social connection is well-built but too narrow on neurobiological mechanisms. In a review aimed at informing rehabilitation practice, at least two convergent biological pathways should be briefly acknowledged — circadian entrainment and HPA-axis/stress regulation — which anchor the plausibility of benefits from outdoor NBIs to contemporary psychiatric pathophysiology. Appropriate citations are Monteleone P, et al. Circadian rhythms and treatment implications in depression. Prog Neuropsychopharmacol Biol Psychiatry. 2011;35(7):1569-74. doi:10.1016/j.pnpbp.2010.07.028 (for the light/circadian mechanism linking outdoor exposure to depressive pathophysiology) and Flattened cortisol awakening response in chronic patients with schizophrenia onset after cannabis exposure. Psychiatry Res. 2014;215(2):263-7. doi:10.1016/j.psychres.2013.12.016 (for HPA dysregulation in schizophrenia, relevant given that 43% of participants had this diagnosis). The authors should also more sharply articulate what their review adds beyond the 2025–2026 umbrella reviews they cite. 

We thank the reviewer for this suggestion. In response, we have expanded the introduction in order to acknowledge the two additional biological pathways as follows: 

There are also plausible neurobiological pathways through which outdoor NBIs may support psychiatric rehabilitation exposure to natural daylight, particularly in the morning, may strengthen circadian entrainment and stabilise sleep–wake rhythms, which are frequently disrupted in depression, bipolar disorder, and psychosis and are increasingly recognised as relevant to psychiatric pathophysiology and treatment (Monteleone et al., 2011). Outdoor activity may also provide regular behavioural timing cues through movement, routine, and social engagement, further supporting circadian regulation and sleep. In parallel, NBIs may influence HPA-axis and stress-regulatory systems, given evidence that chronic psychiatric disorders, including schizophrenia, can be associated with altered cortisol dynamics and stress responsivity (e.g., flattened cortisol awakening responses) (Aas et al., 2014). Together, these pathways suggest that NBIs may plausibly affect both psychological recovery processes and biological systems implicated in mental health. 

With respect to the positioning of this review relative to recent umbrella reviews, we have also revised the Introduction to more explicitly articulate what the present review adds: 

While interest in NBIs has surged, the evidence base has remained heterogeneous. Previous reviews have largely focused on general populations (Coventry et al., 2021; Nguyen et al., 2023; Twohig-Bennett, Jones, 2018) or specific conditions (Jessen et al., 2025), often using narrative (Harrison et al., 2023) or scoping approaches (Wilkie and Davinson, 2021), and recently published umbrella reviews have begun synthesising this growing body of evidence at a high level (Brandt et al., 2026; Shrestha et al., 2025; Kaleta et al., 2025). However, few reviews have specifically examined outcomes in people with diagnosed psychiatric disorders- including anxiety (Jessen et al., 2025), depression (Jessen et al., 2025; Salomon, Salomon & Beeber, 2018), and stress-related conditions (Jessen et al., 2025; Paredes-Céspedes et al., 2024) - and those that do have focused almost exclusively on quantitative efficacy, with comparatively limited attention to participant experiences, implementation processes, and the meaning participants ascribe to these interventions (Cuthbert, Kellas & Page, 2021). No review to date has integrated quantitative and qualitative evidence in this population.   

We aimed to address these gaps by systematically reviewing and meta-analysing studies of NBIs for individuals with psychiatric disorders, encompassing a broad spectrum of diagnoses and intervention types, while integrating quantitative effect sizes with qualitative thematic synthesis, allowing us to identify not only if NBIs are effective across clinical and functional domains, but also how individuals with psychiatric lived experience perceive, value, and engage with these interventions.   

 

R2.3. Methods 

R2.3.1 Several issues require attention. The use of CASP for RCTs is suboptimal: RoB-2 is the accepted standard and should be adopted.  

We thank the reviewer for this important suggestion. We acknowledge that RoB-2 can be regarded as the preferred tool for risk-of-bias assessment in RCTs. However, because this review employed a mixed-methods design incorporating diverse study methodologies, we selected the CASP appraisal tools to enable a consistent and integrated approach across study types. We have clarified this rationale in the methods section: 

Given the mixed-methods nature of this review and the inclusion of studies with diverse designs, the methodological quality of all included studies was assessed using the Critical Appraisal Skills Programme (CASP) checklists for Randomised Controlled Trials, Cohort Studies, and  Qualitative Research (Long, French & Brooks, 2020) .   

Moreover, we acknowledged the potential limitation that CASP provides a broader critical appraisal framework rather than the more specialised domain-based bias assessment offered by RoB-2 in the discussion: 

Moreover, a potential limitation of this review is the use of CASP tools for quality appraisal across all included study designs. While this enabled a consistent and integrated approach across heterogeneous methodologies, CASP provides a broader critical appraisal framework rather than a domain-based risk-of-bias assessment specific to randomised controlled trials, such as the Cochrane RoB-2 tool. 

R2.3.2 No publication-bias assessment is reported despite a pooled analysis; this should be added, with the caveat that nine studies is below the usual threshold.  

A statement clarifying that publication bias was not formally assessed has been added to paragraph 2.4: 

Formal assessment of publication bias was not conducted because fewer than 10 studies were included in the meta-analysis. Instead, potential publication bias was considered narratively, including whether small studies reported disproportionately large or significant effects and whether unpublished or grey literature was searched.   

R2.3.3 Section 2.2 ("Data extraction") is duplicated almost verbatim and must be consolidated.  

We thank the reviewer for noticing it, the duplication has been removed. 

 

R2.3.4 The exploratory meta-regression (three moderators, nine studies) is underpowered and this should be acknowledged.  

We thank the Editor for this helpful comment, we acknowledged that as follows: 

Across nine studies​ (He et al., 2020, Oh, Park & Ahn, 2018, Zhu et al., 2016, Vujcic Trkulja et al., 2021, Siu, Kam & Mok, 2020, Keenan et al., 2021, Kam, Siu, 2010, Pedersen et al., 2012, Barton, Griffin & Pretty, 2012)​, a random-effects meta-analysis showed that the intervention was associated with improvements in clinical symptoms than control conditions (pooled effect size 0·71 [95% CI 0·29–1·12]; p=0·0009; see Figure 1). Study-specific effects ranged from 0·04 to 2·39, and between-study heterogeneity was moderate (τ²=0·18; I²=48·6%; Q(8)=16·30; p=0·038). All included studies were controlled trials, comprising both randomised and non-randomised designs. In an exploratory random-effects meta-regression including NBI type and trial design, neither exercise-focused (β=0·76 [95% CI –1·63 to 3·14]; p=0·54) nor horticultural programmes (β=0·46 [95% CI –1·53 to 2·45]; p=0·65) differed significantly from carefarming, and randomised controlled trials did not differ significantly from non-randomised controlled studies (β=0·31 [95% CI –1·03 to 1·66]; p=0·65). As a set, these moderators did not explain between-study variation (Wald χ²(3)=0·45; p=0·93; R²=0%), and substantial residual heterogeneity remained (residual τ²=0·56; I²=75·7%; Q_res(5)=16·16; p=0·006). However, this meta-regression included only nine studies and three moderators, and was therefore underpowered; the findings should be interpreted as exploratory and hypothesis-generating rather than confirmatory. 

 

R2.3.5 The absence of patient and public involvement warrants at least brief reflection, given the qualitative strand's emphasis on lived experience. 

We thank the Editor for this comment, which has helped strengthen the manuscript. The following paragraph has been included in section 1.7: 

 

Public and stakeholder input was provided by HF, a non-academic climate and sustainability contributor with experience in community mobilisation, clean air advocacy, green space access, and public climate engagement. Her contribution supported the framing of the review in relation to environmental justice, community participation, and the public relevance of nature-based approaches. No patient or service-user contributors with lived experience of psychiatric disorders were involved, which is a limitation given the qualitative synthesis’ focus on lived experience. Future work should involve people with psychiatric lived experience in shaping review questions, interpreting themes, and identifying outcomes relevant to recovery-oriented practice 

 

R2.4. Results 

R2.4.1 The forest plot (Figure 1) labels studies generically as "Study 1"–"Study 9" rather than by author/year; this must be corrected, as it currently prevents readers from identifying which studies drive the pooled effect.  

We thank the reviewer for their useful suggestion, the year details have been reported.   

 

 

R2.4.2 A sensitivity analysis excluding the two outliers (effect sizes 2.39 and 2.04) would clarify robustness. The CASP finding that four of five RCTs were low-quality should be brought forward into the main text, since it materially affects interpretation of the meta-analysis.  

We thank the reviewer for this helpful suggestion. We conducted a sensitivity analysis excluding the two studies with the largest effect sizes (Keenan et al., 2021; Kam & Siu, 2010). The pooled effect was reduced from 0.71 to 0.48, but remained statistically significant, while heterogeneity decreased substantially from I²=48.6% to I²=10.0%. We have added these results to the Results section and reflected them in the Discussion, noting that the overall direction of effect was robust, although the magnitude of the pooled estimate and degree of heterogeneity were influenced by these high-effect studies. 

We have also addressed the reviewer’s second point by bringing the CASP quality finding into the main Discussion. We now explicitly state that confidence in the meta-analytic estimate is limited by study quality, as four of the five RCTs included in the quantitative synthesis were rated as low quality using CASP. This has been used to support a more cautious interpretation of the pooled effect. 

Results: 

Finally, a sensitivity analysis excluding the two studies with the largest effect sizes (Keenan et al., 2021; Kam & Siu, 2010) attenuated the pooled effect, but the association remained statistically significant (SMD 0.48 [95% CI 0.17–0.79]; p=0.0025). Heterogeneity was substantially reduced (τ²=0.018; I²=10.0%; Q(6)=6.74; p=0.345), suggesting that these two studies contributed meaningfully to between-study heterogeneity, but did not fully account for the overall direction of effect. 

Discussion: 

Overall, the pooled estimate from controlled studies favoured NBIs over comparator conditions, and exploratory analyses did not indicate clear differences in effect size by trial design. A sensitivity analysis excluding the two studies with the largest effect sizes attenuated the pooled estimate from 0.71 to 0.48, although the association remained statistically significant (SMD 0.48 [95% CI 0.17–0.79]; p=0.0025). Heterogeneity was substantially reduced (I²=10.0%; τ²=0.018), suggesting that these two studies contributed meaningfully to between-study variability. However, confidence in the meta-analytic estimate is limited by study quality, as four of the five RCTs included in the quantitative synthesis were rated as low quality using CASP. The pooled effect should therefore be interpreted as promising but cautious, rather than definitive  

R2.5. Discussion  

The discussion integrates the two strands reasonably but remains largely descriptive. The mechanistic paragraph should move beyond the Ulrich and Kaplan frameworks and cite the circadian and HPA references noted above. A further forward-looking anchor is Barlattani T, Glymphatic system and psychiatric disorders: need for a new paradigm? Front Psychiatry. 2025;16:1642605. doi:10.3389/fpsyt.2025.1642605, to be cited in one sentence acknowledging that improved sleep and reduced stress — both repeatedly reported in the qualitative strand — may engage neurobiological pathways (including glymphatic clearance) increasingly recognized as relevant to psychiatric disorders, situating NBIs within a biologically informed rehabilitation framework. The limitations section should be expanded to include low RCT quality, the small pooled sample, diagnostic heterogeneity, and absence of publication-bias assessment. 

We thank the reviewer for their suggestion; the discussion was integrated accordingly: 

However, the potential mechanisms are likely broader than these classic psychological frameworks. Outdoor NBIs may support circadian entrainment through daylight exposure, routine, and daytime activity, which is relevant given the role of circadian rhythm disruption in depression and other psychiatric disorders (Monteleone et al., 2011). They may also influence HPA-axis and stress-regulatory systems, given evidence of altered cortisol dynamics in chronic psychiatric disorders, including schizophrenia (Aas et al., 2014). More speculatively, improvements in sleep and reductions in stress - both repeatedly described in the qualitative synthesis - may engage neurobiological pathways, including glymphatic clearance, that are increasingly recognised as relevant to psychiatric disorders (Barlattani, 2025). These converging psychological and biological pathways position NBIs not simply as adjunctive wellbeing activities, but as potentially biologically informed rehabilitation approaches that may affect symptoms, functioning, identity, social participation, and physiological regulation.   

6. Conclusions

Soften in line with the certainty of evidence: "NBIs show promise" is appropriate; stronger formulations are not, and an explicit call for adequately powered RCTs with standardized recovery outcomes should precede any scale-up recommendation. 

We thank the Reviewer for this comment. We have revised the concluding paragraph as follows: 

Overall, NBIs show promise as recovery-oriented, low-intensity interventions within mental healthcare, particularly when delivered in structured and sustained formats. Our findings point to the potential utility of integrating NBIs within flexible, person-centred psychiatric rehabilitation pathways, especially where access to safe natural spaces can be ensured and interventions can be aligned with individual preferences and goals. However, given the heterogeneity and methodological limitations of the current evidence, further high-quality research is needed to clarify which intervention approaches are most effective for specific psychiatric populations and contexts.   

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

No further questions

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for your valuable work. Only one minor comment. The new references added has not been included in the reference list. Thank you again for your efforts.