Effectiveness of Horticultural Therapy in People with Schizophrenia: A Systematic Review and Meta-Analysis

Horticultural therapy is increasingly being used in the non-pharmacological treatment of patients with schizophrenia, with previous studies demonstrating its therapeutic effects. The healing outcomes are positively correlated with the settings of the intervention. This review aimed to evaluate the effectiveness of horticultural therapy on the symptoms, rehabilitation outcomes, quality of life, and social functioning in people with schizophrenia, and the different effectiveness in hospital and non-hospital environments. This review followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA) guidelines. We researched studies through PubMed, Embase, the Cochrane Library, Science Direct, and the China National Knowledge Infrastructure. We included randomized controlled trials (RCTs) and quasi-experimental studies about horticultural therapy for people with schizophrenia, from January 2000 to December 2020, with a total of 23 studies involving 2024 people with schizophrenia included in this systematic review. This study provided evidence supporting the positive effect of horticultural therapy. This review demonstrated that non-hospital environments have a better therapeutic effect on all indicators than hospital environments. The results also demonstrated the effectiveness of horticultural therapy on symptoms, rehabilitation outcomes, quality of life, and social functioning in patients in hospital and non-hospital environments, providing further evidence-based support for landscape design.


Introduction
Schizophrenia is one of the most common severe mental disorders, being ranked among the top 20 causes of disability worldwide [1] and affecting 20 million people [2]. People with schizophrenia often share common experiences, such as hallucinations, delusions, disturbances of emotions, and distortions in behavior and language, and they face 2-3 times the risk of early death than the general population [3], qualifying the severity of this mental disorder. Schizophrenia is a debilitating disease because patients are cognitively impaired, which is often related to decreased executive functioning, eventually leading to severely impaired daily functioning and social interactions.
As schizophrenia is a chronic relapsing disease with a high recurrence rate and a high possibility of disability, the treatment of it has become one of the most challenging issues, affecting not only the everyday life of patients but also their family financial status [4]. Currently, medication is the primary treatment for schizophrenia. However, the long-term usage of antipsychotic drugs poses some risks, such as metabolic syndrome, manifested in weight gain and diabetes [5]. Recent evidence has demonstrated that non-pharmacological therapies are more desirable to alleviate symptoms of schizophrenia without producing side effects [6,7]. Horticultural therapy has received increasing attention as an effective and non-pharmacological intervention [8]. Horticultural therapy is defined by the American The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35].

Materials and Methods
In this review, we followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA) guidelines [33].

Search Strategy
Studies from January 2000 to December 2020 were searched and collected in this study. We searched PubMed, the Cochrane Library, Embase, and Science Direct using common keywords: (horticul* OR floricult* OR arboricult* OR olericult* OR agricult* OR garden* OR farm*) AND schizophrenia. The search strategy for the China National Knowledge Infrastructure was as follows: "(SU = '花园(garden)' OR SU = '园艺(horticulture)' OR SU = '农(farm)') AND SU = '精神分裂症(schizophrenia)'".

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, according to the population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35].

Materials and Methods
In this review, we followed the Preferred Reporting Items for Systematic Review Meta-Analysis Protocols (PRISMA) guidelines [33].

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, accordi the population, intervention, comparison, outcomes, and study design (PICOS). Table 1. Description of the inclusion/exclusion criteria according to population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndr Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessme Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to exp the quality of life, while the measuring tools of social functioning included the Sca Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance ( scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and we screened the studies by the titles, abstracts, and full texts according to the inclu and exclusion criteria of this review. If two independent reviewers disagreed, it wa solved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible stu For the RCTs, we evaluated the risk of bias for the included literature using the R specific bias risk assessment tool in the Cochrane handbook for systematic reviews o terventions [34], which assesses randomization procedure biases, allocation concealm and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal for the quasi-experimental studies [35].

Materials and Methods
In this review, we followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA) guidelines [33].

Search Strategy
Studies from January 2000 to December 2020 were searched and collected in this study. We searched PubMed, the Cochrane Library, Embase, and Science Direct using common keywords: (horticul* OR floricult* OR arboricult* OR olericult* OR agricult* OR garden* OR farm*) AND schizophrenia. The search strategy for the China National Knowledge Infrastructure was as follows: "(SU = '花园(garden)' OR SU = '园艺(horticulture)' OR SU = '农(farm)') AND SU = '精神分裂症(schizophrenia)'".

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, according to the population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35].

Materials and Methods
In this review, we followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA) guidelines [33].

Search Strategy
Studies from January 2000 to December 2020 were searched and collected in this study. We searched PubMed, the Cochrane Library, Embase, and Science Direct using common keywords: (horticul* OR floricult* OR arboricult* OR olericult* OR agricult* OR garden* OR farm*) AND schizophrenia. The search strategy for the China National Knowledge Infrastructure was as follows: "(SU = '花园(garden)' OR SU = '园艺(horticulture)' OR SU = '农(farm)') AND SU = '精神分裂症(schizophrenia)'".

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, according to the population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35].

Materials and Methods
In this review, we followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA) guidelines [33].

Search Strategy
Studies from January 2000 to December 2020 were searched and collected in this study. We searched PubMed, the Cochrane Library, Embase, and Science Direct using common keywords: (horticul* OR floricult* OR arboricult* OR olericult* OR agricult* OR garden* OR farm*) AND schizophrenia. The search strategy for the China National Knowledge Infrastructure was as follows: "(SU = '花园(garden)' OR SU = '园艺(horticulture)' OR SU = '农(farm)') AND SU = '精神分裂症(schizophrenia)'".

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, according to the population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35]. (schizophrenia)'".

Inclusion and Exclusion Criteria
A description of the inclusion/exclusion criteria is provided in Table 1, according to the population, intervention, comparison, outcomes, and study design (PICOS). Table 1. Description of the inclusion/exclusion criteria according to population, intervention, comparison, outcomes, and study design (PICOS). The measuring tools of the symptoms included the Positive and Negative Syndrome Scale (PANSS), the Brief Psychiatric Rating Scale (BPRS), and the Scale for Assessment of Negative Symptoms (SANS). The Inpatient Psychiatric Rehabilitation Outcome Scale (IPROS) was used to measure rehabilitation outcomes. The Schizophrenia Quality of Life Scale (SQLS) and the Generic Quality of Life Inventory-74 (GQLI-74) were used to explore the quality of life, while the measuring tools of social functioning included the Scale of Social Function in Psychosis Inpatients (SSPI), Personal and Social Performance (PSP) scale, and the Social Disability Screening Schedule (SDSS).

Selection of Articles
We imported all studies into EndNote X7. Duplicate studies were excluded, and then we screened the studies by the titles, abstracts, and full texts according to the inclusion and exclusion criteria of this review. If two independent reviewers disagreed, it was resolved through discussion or by a third reviewer.

Quality Evaluation
Two independent reviewers critically appraised the quality of the eligible studies. For the RCTs, we evaluated the risk of bias for the included literature using the RCTspecific bias risk assessment tool in the Cochrane handbook for systematic reviews of interventions [34], which assesses randomization procedure biases, allocation concealment, and selective reporting. We used the Joanna Briggs Institute (JBI) critical appraisal tools for the quasi-experimental studies [35].

Data Extraction
In terms of data extraction, we read the title and abstract. After excluding irrelevant documents, we read the full text to determine whether it should be included and then summarized the information. The data extraction mainly included: (1) basic information, including research title, first author, and publication time; (2) baseline characteristics of the research subjects, including the number, age, and sex of people included in each group and the disease diagnosis criteria of the study subjects; (3) specific details of intervention measures, including intervention form, time, and settings; (4) critical elements of bias risk assessment; (5) the outcome indicators and outcome measurement data concerned.

Statistical Analysis
We pooled the data of the individual studies using Revman5.3 software. A random effects model was used, assuming heterogeneity between the studies and their respective effect sizes. We used standardized mean differences (SMDs) and mean differences (MDs). The results were aggregated with 95% confidence intervals (CIs). A p-value < 0.05 was considered statistically significant. The standard I 2 tests were used to assess the statistical heterogeneity and we conducted a sensitivity analysis to evaluate the reliability and stability of the results.
We used subgroup analysis to investigate the effects of the different intervention environments (non-hospital vs. hospital environments) on the symptoms, rehabilitation outcomes, quality of life, and social functioning of patients.

Search Outcomes
Figure 1 explains our review process. We found 269 articles from PubMed (n = 139), the Cochrane Library (n = 1), Embase (n = 5), Science Direct (n = 73), and the China National Knowledge Infrastructure (n = 51). We removed 19 articles because of duplication, as well as 203 after reading the titles and abstracts. Of the remaining 47 articles, 12 were removed because they did not meet the inclusion criteria, five because of a lack of a control group in the intervention programs, two because they were reviews, and one because it did not have a baseline assessment. As this review was divided into two subgroups according to the settings, four other articles without mention of this information were excluded. Finally, 23 studies were included ( Figure 1).

Data Extraction
In terms of data extraction, we read the title and abstract. After excluding irrelevant documents, we read the full text to determine whether it should be included and then summarized the information. The data extraction mainly included: (1) basic information, including research title, first author, and publication time; (2) baseline characteristics of the research subjects, including the number, age, and sex of people included in each group and the disease diagnosis criteria of the study subjects; (3) specific details of intervention measures, including intervention form, time, and settings; (4) critical elements of bias risk assessment; (5) the outcome indicators and outcome measurement data concerned.

Statistical Analysis
We pooled the data of the individual studies using Revman5.3 software. A random effects model was used, assuming heterogeneity between the studies and their respective effect sizes. We used standardized mean differences (SMDs) and mean differences (MDs). The results were aggregated with 95% confidence intervals (CIs). A p-value < 0.05 was considered statistically significant. The standard I 2 tests were used to assess the statistical heterogeneity and we conducted a sensitivity analysis to evaluate the reliability and stability of the results.
We used subgroup analysis to investigate the effects of the different intervention environments (non-hospital vs. hospital environments) on the symptoms, rehabilitation outcomes, quality of life, and social functioning of patients. Figure 1 explains our review process. We found 269 articles from PubMed (n = 139), the Cochrane Library (n = 1), Embase (n = 5), Science Direct (n = 73), and the China National Knowledge Infrastructure (n = 51). We removed 19 articles because of duplication, as well as 203 after reading the titles and abstracts. Of the remaining 47 articles, 12 were removed because they did not meet the inclusion criteria, five because of a lack of a control group in the intervention programs, two because they were reviews, and one because it did not have a baseline assessment. As this review was divided into two subgroups according to the settings, four other articles without mention of this information were excluded. Finally, 23 studies were included (Figure 1).

Study Characteristics
The features of the selected studies are aggregated. The number of people ranged from 28 to 615 (2024 in total) and their ages ranged from 15 to 65 years. Most horticultural therapy activities included growing flowers or vegetables, daily maintenance, and doing handicrafts. The settings were hospitals, agricultural rehabilitation training institutions, farms, and communities, which we divided into hospital and non-hospital environments. Figure 2 shows the evaluations of each risk of bias. Allocation concealment and blinding of outcome assessment were evaluated as unclear risks, whereas blinding of participants and personnel was assessed as low risk. For incomplete outcome data, two trials contained instances of participation withdrawal. In general, most studies were evaluated as being of low-risk quality. Six quasi-experimental studies conformed to the JBI critical appraisal checklist. The detailed results are presented in Figures 2 and 3 and Tables 2 and 3.

Study Characteristics
The features of the selected studies are aggregated. The number of people ranged from 28 to 615 (2024 in total) and their ages ranged from 15 to 65 years. Most horticultural therapy activities included growing flowers or vegetables, daily maintenance, and doing handicrafts. The settings were hospitals, agricultural rehabilitation training institutions, farms, and communities, which we divided into hospital and non-hospital environments. Figure 2 shows the evaluations of each risk of bias. Allocation concealment and blinding of outcome assessment were evaluated as unclear risks, whereas blinding of participants and personnel was assessed as low risk. For incomplete outcome data, two trials contained instances of participation withdrawal. In general, most studies were evaluated as being of low-risk quality. Six quasi-experimental studies conformed to the JBI critical appraisal checklist. The detailed results are presented in Figures 2 and 3 and Table 3.

Study Characteristics
The features of the selected studies are aggregated. The number of people ranged from 28 to 615 (2024 in total) and their ages ranged from 15 to 65 years. Most horticultural therapy activities included growing flowers or vegetables, daily maintenance, and doing handicrafts. The settings were hospitals, agricultural rehabilitation training institutions, farms, and communities, which we divided into hospital and non-hospital environments. Figure 2 shows the evaluations of each risk of bias. Allocation concealment and blinding of outcome assessment were evaluated as unclear risks, whereas blinding of participants and personnel was assessed as low risk. For incomplete outcome data, two trials contained instances of participation withdrawal. In general, most studies were evaluated as being of low-risk quality. Six quasi-experimental studies conformed to the JBI critical appraisal checklist. The detailed results are presented in Figures 2 and 3 and Table 3.
SMDs were used because of the different scales. We used a random-effects model (p < 0.00001, I 2 = 94%) and subgroup analysis was conducted according to the intervention settings. The results, as shown in Figure 4, showed a significant difference (SMD = −2.62, 95% CI [−3.87, −1.38], p < 0.00001) in the influence of horticultural therapy in non-hospital environments on the total score of symptoms, but the result was less significant when the intervention settings were hospital environments (SMD = −0.90, 95% CI [−1.21, −0.59], p < 0.00001). We detected significant differences in the sensitivity analyses when removing Tao " and what is the "effect" (i.e., there is no confusion about which variable comes first)? Were the participants included in any similar comparisons? (2) Were the participants included in any comparisons similar? (3) Were the participants included in any comparisons receiving similar treatment/care, other than the exposure or intervention of interest? (4) Was there a control group? (5) Were there multiple measurements of the outcome both pre-and post-the intervention/exposure? (6) Was follow-up complete and, if not, were differences between groups in terms of their follow-up adequately described and analyzed? (7) Were the outcomes of participants included in any comparisons measured in the same way? (8) Were outcomes measured reliably? (9) Was an appropriate statistical analysis used?
SMDs were used because of the different scales. We used a random-effects model (p < 0.00001, I 2 = 94%) and subgroup analysis was conducted according to the intervention settings. The results, as shown in Figure 4, showed a significant difference (SMD = −2.62, 95% CI [−3.87, −1.38], p < 0.00001) in the influence of horticultural therapy in non-hospital environments on the total score of symptoms, but the result was less significant when the intervention settings were hospital environments (SMD = −0.90, 95% CI [−1.21, −0.59], p < 0.00001). We detected significant differences in the sensitivity analyses when removing Tao

Rehabilitation Outcomes
The total score of the rehabilitation outcomes was gathered from six RCTs and three quasi-experimental studies using IPROS [12][13][14]37,40,45,47,51,52]. MDs were used because of the uniform standard. We used the random-effects model because of heterogeneity (p < 0.00001, I 2 = 91%) and conducted a subgroup analysis on the basis of the intervention settings. A significantly positive difference was found in the impact of horticultural therapy. We found some differences between the two subgroups ( Figure 5). There were significant differences found in the sensitivity analyses when removing Tao (2017) [38]

Rehabilitation Outcomes
The total score of the rehabilitation outcomes was gathered from six RCTs and three quasi-experimental studies using IPROS [12][13][14]37,40,45,47,51,52]. MDs were used because of the uniform standard. We used the random-effects model because of heterogeneity (p < 0.00001, I 2 = 91%) and conducted a subgroup analysis on the basis of the intervention settings. A significantly positive difference was found in the impact of horticultural therapy. We found some differences between the two subgroups ( Figure 5). There were significant differences found in the sensitivity analyses when removing Tao (2017) [38] (SMD = −2.01, 95% CI [−2.31, −1.71], p = 0.02).

Quality of Life
The total score of the symptoms was determined from three RCTs using SQLS [45] and GQOLI-74 [15,51]. SMD was used because of the different scales. We used the random-effects model because of the existence of heterogeneity (p = 0.008, I 2 = 79%) and conducted a subgroup analysis based on the intervention settings.

Quality of Life
The total score of the symptoms was determined from three RCTs using SQLS [45] and GQOLI-74 [15,51]. SMD was used because of the different scales. We used the randomeffects model because of the existence of heterogeneity (p = 0.008, I 2 = 79%) and conducted a subgroup analysis based on the intervention settings.

Rehabilitation Outcomes
The total score of the rehabilitation outcomes was gathered from six RCTs and th quasi-experimental studies using IPROS [12][13][14]37,40,45,47,51,52]. MDs were used cause of the uniform standard. We used the random-effects model because of heteroge ity (p < 0.00001, I 2 = 91%) and conducted a subgroup analysis on the basis of the interv tion settings. A significantly positive difference was found in the impact of horticultu therapy. We found some differences between the two subgroups ( Figure 5). There w significant differences found in the sensitivity analyses when removing Tao

Quality of Life
The total score of the symptoms was determined from three RCTs using SQLS [ and GQOLI-74 [15,51]. SMD was used because of the different scales. We used the r dom-effects model because of the existence of heterogeneity (p = 0.008, I 2 = 79%) and c ducted a subgroup analysis based on the intervention settings.
We found significant differences in the results of horticultural therapy in non-hos tal environments on quality of life (SMD = 1.61, 95% CI [1.10, 2.12], p = 0.008; Figure

Social Functioning
The total score of social functioning was collected from eight RCTs and a quasiexperimental study using SSPI [36,38,39,45,47,52] and PSP [40,43,48]. SMDs were used because of the different scales. We used the random-effects model because of the existence of heterogeneity (p < 0.00001, I 2 = 98%) and conducted a subgroup analysis considering the intervention settings. Figure 7 demonstrates the significant difference (SMD = −0.19, 95% CI [−1.69, 1.30], p < 0.00001) in the effect of horticultural therapy in non-hospital environments on the score of social functioning, whereas the result was less significant (SMD = −0.03, 95%CI [−3.40, 3.33], p < 0.00001) in hospital settings. We observed some differences between the two subgroups, but no significant difference was found in the heterogeneity analysis when we removed all of the studies one by one.

Social Functioning
The total score of social functioning was collected from eight RCTs and a qu perimental study using SSPI [36,38,39,45,47,52]and PSP [40,43,48]. SMDs were us cause of the different scales. We used the random-effects model because of the exi of heterogeneity (p < 0.00001, I 2 = 98%) and conducted a subgroup analysis consi the intervention settings. Figure 7 demonstrates the significant difference (SMD = −0.19, 95% CI [−1.69, 1 < 0.00001) in the effect of horticultural therapy in non-hospital environments on the of social functioning, whereas the result was less significant (SMD = −0.03, 95%CI 3.33], p < 0.00001) in hospital settings. We observed some differences between th subgroups, but no significant difference was found in the heterogeneity analysis wh removed all of the studies one by one.

Outcomes and Processes of Horticultural Therapy
This study focused on the outcomes and processes of horticultural therapy. Th ings support the positive effect of horticultural therapy on schizophrenic patients' toms, rehabilitation outcomes, quality of life, and social functioning, as demonstra the significant difference in the scores of the experimental and control groups. This that horticultural therapy positively impacts the treatment of schizophrenic patien the effects vary in different settings (hospital vs. non-hospital environments).
Horticultural therapy can improve the symptoms of schizophrenia by signifi reducing anxiety, depression, stress, and interpersonal sensitivity [53]. To alleviate toms such as delusions and hallucinations [54], horticultural activities promote c between schizophrenic patients and real life.
In terms of rehabilitation outcomes, patients enjoy the natural environment and more connection with nature, increasing their sensitivity to plants and nature, gene more positive emotions, and promoting their emotional management ability [12,17 The results also support a positive effect on quality of life. Horticultural activit help arouse patients' interest in participating in activities, thus effectively stimulat terest in life [13].
In addition to improving quality of life, this study also clarified the effect of ho tural therapy on social functioning. The research showed that cognitive behaviora apy (CBT) can improve the social cognition, self-efficacy, and social ability of patient

Outcomes and Processes of Horticultural Therapy
This study focused on the outcomes and processes of horticultural therapy. The findings support the positive effect of horticultural therapy on schizophrenic patients' symptoms, rehabilitation outcomes, quality of life, and social functioning, as demonstrated by the significant difference in the scores of the experimental and control groups. This shows that horticultural therapy positively impacts the treatment of schizophrenic patients, but the effects vary in different settings (hospital vs. non-hospital environments).
Horticultural therapy can improve the symptoms of schizophrenia by significantly reducing anxiety, depression, stress, and interpersonal sensitivity [53]. To alleviate symptoms such as delusions and hallucinations [54], horticultural activities promote contact between schizophrenic patients and real life.
In terms of rehabilitation outcomes, patients enjoy the natural environment and have more connection with nature, increasing their sensitivity to plants and nature, generating more positive emotions, and promoting their emotional management ability [12,17].
The results also support a positive effect on quality of life. Horticultural activities can help arouse patients' interest in participating in activities, thus effectively stimulating interest in life [13].
In addition to improving quality of life, this study also clarified the effect of horticultural therapy on social functioning. The research showed that cognitive behavioral therapy (CBT) can improve the social cognition, self-efficacy, and social ability of patients with chronic schizophrenia [18]. Horticultural therapy can be used with CBT to strengthen the sense of accomplishment, responsibility, and belonging [19].
Previous studies focused on the subgroup analysis of the characteristics of activities and populations [23][24][25][26], not on the environment. This study fills this gap and demonstrates that non-hospital environments have a better therapeutic effect on all indicators than hospital environments. The reasons for this result are as follows: (1) there is less chance of a natural experience in hospital environments, whereas non-hospital environments (e.g., farms) immerse people in the sense of beauty and selflessness. Non-hospital environments also have better microclimates, which are beneficial to the healing process, implying a better therapeutic effect. This finding is also consistent with those of some previous studies [55][56][57][58][59][60][61][62][63][64] that greenspace may have a more pronounced effect on individuals with mental illness [65]. A comfortable environment also increases patients' motivation to participate in activities to reap physical benefits. (2) The types of horticultural therapy activities in hospitals are limited and mainly focus on planting flowers and vegetables and making bonsai; in non-hospital settings, patients can participate in a larger number of activities, such as cultivating plants and picking fruits. More specifically, patients can fully experience the whole growing process throughout the year in non-hospital environments: fertilizing, sowing, watering, weeding, planting, and harvesting. (3) The duration of activities in hospitals was shown to be three (six studies), six (four studies), and 12 months (three studies), whereas the activities in non-hospital settings tended to have a longer follow-up: 6 (five studies), 10 (one study), 12 (one study), and 24 (one study) months. The intervention duration in non-hospital settings was found to be 4-16 h per week, whereas that in hospital settings ranged from 0.5 to 10 h per week. Overall, the treatment duration in most non-hospital environments was longer than in hospital environments, which could also have produced differences in results.

Contributions and Limitations of the Study
The main contributions of this study are as follows. First, this study provides valid evidence supporting the positive effect of horticultural therapy. Our results support a promising avenue of research with relevant application implications. Schizophrenia caregivers (including hospitals and rehabilitation facilities) should provide patients with as many opportunities as possible to participate in horticultural therapy. Therefore, horticultural therapy should be considered an essential tool to treat schizophrenia in future adjuvant therapies for schizophrenic patients. Second, we discussed the differences in the treatment effects in two different environments. We found that non-hospital settings have better healing outcomes, guiding future design and activity organization. The establishment of more professional healing farms or landscapes could be considered to improve the effectiveness of complementary horticultural therapies.
Designing landscapes for horticultural therapy in psychiatric hospitals can make horticultural therapy activities a commonly accepted treatment for patients. The process and the outcome of therapy can provide a further evidence-based reference for future design. We conducted a meta-analysis of horticultural therapy in the auxiliary treatment of schizophrenia. From the analysis, the conclusions provide a basis for evidence-based design to help create a new medical environment based on scientific research data. Thus, patients could receive more optimized treatment, and medical staff could maximize their efficiency and relieve stress in these environments. Evidence-based designs provide a theoretical and empirical foundation for the renovation of the hospital environment and provide a method to promote horticultural therapy.
This study has several limitations. First, the intervention settings were hospitals, agricultural rehabilitation training institutions, farms, and communities. Given the wide range of environments, we only classified these environments into hospital and nonhospital settings instead of more specific environmental subgroups. Second, the studies could be divided into subgroups according to different types of activities to explore which activities are more useful for the recovery of patients with schizophrenia in a future study.

Conclusions
This meta-analysis showed that horticultural therapy yields positive outcomes in terms of symptoms, rehabilitation outcomes, quality of life, and social functioning of schizophrenic patients. In terms of the environment, different settings can influence treatment; non-hospital environments were shown to have a better therapeutic effect. The result herein can provide a basis and guidance for the future evidence-based landscape design of the treatment of schizophrenia.
Further high-quality studies are needed to explore the substantial therapeutic effect of horticultural therapy. Additional studies on horticultural therapy need to explore more details about the intensity of horticultural therapy activities and the characteristics of the settings in which the activities occur. More research from other countries on horticultural therapy and schizophrenia is needed to contribute to the generalizability of these results.
Author Contributions: Data curation, S.L.; writing-original draft preparation, S.L. and Y.Z.; writing-review and editing, S.L., Y.Z., and J.L.; supervision, F.X. and Z.W. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.