Cross-Sectional Study of Health Promotion and Recreation Effectiveness on Quality of Life Among Rural Older Adults

Abstract

Older adults in rural areas often face barriers to accessing formal health services. Community-based programs serve as alternative models for delivering preventive care and psychosocial support. However, the effectiveness of specific program types on well-being outcomes remains underexplored. This cross-sectional study analyzed secondary data from 1033 older adults across 44 rural communities in Taiwan. Participants were involved in 5 types of community-based programs. Subjective well-being was assessed using the WHO-5 index. Associations between participation hours and well-being were examined using ANOVA, OLS regression, and linear mixed models, with community-level clustering and individual demographics controlled. Health promotion and recreational activities were positively associated with well-being, while horticultural therapy and social participation showed negative associations. Food and agricultural education was positively associated with well-being only after controlling for community context. Neither age nor gender significantly predicted outcomes. Community context moderated several program effects. Community-based programs impact rural older adults’ well-being in diverse ways depending on program type and local implementation. Tailored, context-sensitive interventions and ongoing program evaluation are essential for optimizing care outcomes in aging rural populations.

Keywords are based on MeSH terms identified using the National Library of Medicine’s MeSH on Demand tool.

This study of 1,033 older adults in 44 rural communities found that health promotion and recreational activities improved well-being, while horticultural therapy and social participation showed negative effects. Food and agricultural education became beneficial after adjusting for community context, underscoring the importance of place-based, user-centered approaches in rural aging care.

Introduction

As global populations age rapidly, promoting the well-being of older adults has become a central priority in gerontology, public health, and social policy. In rural communities—where access to formal health and social services is often limited—community-based programs serve as critical platforms for social engagement, preventive care, and psychological support. A growing body of research has demonstrated that structured community interventions can enhance physical and mental health, reduce loneliness, and foster a sense of meaning and connection among older adults.^1-3

In Taiwan, the Ministry of Agriculture, in collaboration with local organizations, has developed a wide array of community-based wellness initiatives for rural older adults. These include food and agricultural education, horticultural therapy, health promotion activities, recreational and cultural programs, and social participation events. Such programs align with broader national efforts to support aging in place through culturally appropriate and locally embedded care strategies.

Understanding how community-based programs affect older adults’ well-being requires a framework that integrates individual functioning and environmental context. Rowe and Kahn’s Successful Aging Framework emphasizes minimizing disease, maintaining function, and active engagement. Their updated Successful Aging 2.0 model further highlights the role of structural supports, social inclusion, and intergenerational participation, shifting the focus from individual traits to environmental opportunity. ⁴ Similarly, the Ecological Model of Aging stresses the dynamic interaction between personal competence and environmental demands. ⁵ In rural areas, where infrastructure and cultural contexts vary, program effectiveness depends not only on content but also on local fit. Together, these models underscore the need to evaluate aging interventions within broader social and environmental systems, especially when designing inclusive, context-sensitive strategies for rural older adults.

While existing literature affirms the positive effects of community engagement on well-being, there remains a need to disaggregate which types of programs offer the greatest benefits and whether program dosage and contextual factors affect outcomes. For example, while health promotion and leisure activities have been widely supported by empirical evidence,^6,7 the impacts of horticultural therapy or food and agricultural education are often influenced by community dynamics, physical capacity of participants, and program implementation quality.^8,9

As global populations age, providing effective and equitable care for rural older adults is increasingly important. In areas with limited formal health infrastructure, community-based programs can complement public health systems by meeting psychosocial, preventive, and functional needs. In Taiwan, the Ministry of Agriculture, in partnership with community organizations, implemented 5 program types—food and agricultural education, horticultural therapy, health promotion, recreational and cultural activities, and social participation—under its national Green Care initiative. These categories were selected to address multiple domains of older adults’ intrinsic capacity and functional ability, in line with the WHO ICOPE framework, and to leverage rural Taiwan’s agricultural resources and cultural heritage. The goal is to integrate locally relevant activities that promote physical health, social connectedness, and cultural continuity, thereby supporting aging in place.

This study adopts a service delivery lens to evaluate how different types of community programs influence subjective well-being. Building on frameworks such as the Ecological Model of Aging and Successful Aging 2.0, we assess how program structure and community context contribute to care outcomes in rural environments. Our aim is to inform policymakers and service planners on which types of care interventions yield the greatest benefit and under what organizational conditions they succeed.

Method

Study Design and Participants

In this study, “rural” was defined according to the classification used by Taiwan’s Ministry of Agriculture, which identifies rural communities as administrative villages (cun/li) with a population density of fewer than 300 persons per square kilometer, a predominantly agricultural economy, and limited access to tertiary medical facilities within a 30-min travel radius. This definition differs from OECD’s rural typology, which uses regional-level population density and commuting patterns, and from the USDA’s Rural-Urban Commuting Area (RUCA) codes, which incorporate labor market commuting flows. While our definition reflects Taiwan’s unique administrative and geographic context, the criteria—low population density, agricultural economy, and limited service access—align conceptually with international rural classifications. Nonetheless, these contextual differences may limit the direct generalizability of findings to rural older adults in countries with different rural governance or infrastructure systems.

This cross-sectional study analyzed secondary data from 1033 adults aged ≥65 years in 44 rural Taiwanese communities, collected through a nationwide community aging survey. Participants had engaged in at least 1 community-based program in the past year, and their well-being was assessed using the WHO-5 index. Inclusion criteria were age ≥ 65, ≥6 months residence in the community, program participation, completed WHO-5, and ability to communicate in Mandarin or Taiwanese. Exclusion criteria were incomplete program records, missing or invalid WHO-5 data, severe cognitive impairment, bedridden or hospice status, or refusal to consent.

This cross-sectional study utilized secondary data collected from 44 rural communities in Taiwan between March 2024 and December 2024 as part of a nationwide community aging survey. This study followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) reporting guideline for cross-sectional studies. A completed STROBE checklist is provided in the Supplemental Materials. ¹⁰

Data Source

This study conducted a secondary analysis of data from Taiwan’s Ministry of Agriculture’s Green Care initiative, involving 1033 older adults in 44 rural communities. Data were collected via face-to-face structured interviews in Mandarin or Taiwanese between July and August 2024, and anonymized in the official public report.

Following the ecological model, contextual attributes of the 44 communities were incorporated. These included geographic type (18 mountainous, 12 coastal, 14 plains/valleys), population size (320-2500; median 1150), and predominant economic activity (59% crop farming, 18% fishing, 23% mixed agriculture with small-scale manufacturing or tourism). While all communities had multi-purpose activity centers, access to hospitals ranged from 10 to 20 min in plains/coastal areas to 30 to 60 min in mountainous areas. Program delivery capacity varied with facilitator availability, volunteer networks, and public transport infrastructure. Supplemental Table S1 lists each community’s code, geographic type, and economic activity, providing context for the identifiers in Figures 2 to 6.

Measures: Subjective Well-Being

Well-being was assessed using the World Health Organization Five Well-Being Index (WHO-5), a validated 5-item scale measuring psychological well-being over the past 2 weeks. Each item is rated on a 6-point Likert scale ranging from 0 (none of the time) to 5 (all of the time), producing a raw score of 0 to 25. Scores were converted to a 0 to 100 scale for ease of interpretation: (Raw Score/25) × 100. Scores below 13 suggest possible psychological distress, while scores below 8 indicate likely clinical depression.^11,12

$$AdjustedWHO-5Score=\left(RawWHO-5Score/25\right)\times 100$$

To ensure intuitive interpretation, WHO-5 scores ranged from 0 (lowest well-being) to 100 (highest), aligning with standardized quality-of-life metrics for cross-study comparability. Data were collected via structured face-to-face interviews by trained community health facilitators, as most older participants had limited literacy or mild cognitive impairments. Interviewers read questions in Mandarin or Taiwanese dialect to ensure comprehension.

Program Participation

Community-based programs were categorized into 5 domains based on their content and purpose:

• Food and Agricultural Education-Offers workshops on nutrition, sustainable farming, and preparing local produce.

• Horticultural Therapy-Provides guided gardening, plant-based crafts, and therapeutic outdoor activities.

• Health Promotion Activities-Delivers group exercise, chronic disease management, and health education sessions.

• Cultural and Recreational Activities-Features music, dance, traditional arts, and community festivals.

• Social Participation Activities-Encourages volunteering, intergenerational exchanges, and peer support groups.

The total hours each participant spent in each category over the past year were obtained from community records. For further comparison, the participation hours were divided into tertiles (Q1: low, Q2: medium, Q3: high) for each program type to examine dose-response relationships.

Statistical Analysis

Descriptive statistics were used to summarize the demographic characteristics and well-being scores of the participants. To examine the impact of program engagement on well-being, 2 primary types of statistical analyses were conducted:

• 1. ANOVA (Analysis of Variance):

• One-way ANOVA was applied to compare well-being scores across 3 levels of program participation (low, medium, high) for each of the 5 program types.

• 2. Multiple Linear Regression (OLS Regression):

• Ordinary least squares regression was used to assess the relationship between participation hours (as continuous variables) and well-being scores, with program types entered as separate predictors. The initial model included only program variables, and a second model controlled for community-level effects to address clustering and regional variations. The final model explained 53.6% of the variance in well-being (adjusted R² = .536), indicating a substantial contribution of program type and community context.

Statistical significance was set at P < .05, and all analyses were performed using IBM SPSS Statistics (version 26.0).

For community-level visualizations (Figures 2-6), each community was assigned an alphanumeric code consisting of a county abbreviation and numeric index (eg, CH-01 = Changhaw County, community 1; NT-03 = Nantou County, community 3) to improve international readability while maintaining anonymity. The full mapping of codes to geographic type (mountainous, coastal, or plain) and predominant economic activity (crop farming, fishing, mixed agriculture, or tourism) is provided in Supplemental Table S1. To facilitate cross-figure comparisons, the vertical axis (mean WHO-5 score) was standardized across all program types, and linear trend lines were added to illustrate overall associations between program hours and community-level well-being.

Ethical Considerations

This study used anonymized, publicly available data from government and institutional reports; thus, it did not involve the collection of personally identifiable information and did not require formal ethics board review.

Results

Descriptive and inferential analyses were conducted to examine the relationship between participation in 5 types 44 rural communities (Table S1). A total of 1033 older adults participated in the study (Table 1).

Sample Characteristics and Participation Overview

Table 1.

Descriptive Statistics of Demographics, Subjective Well-Being, and Program Participation Hours Among Older Adults.

Demographic	N = 1033 (%)
Gender
Female	783 (75.8)
Male	250 (24.2)
Age
Mean	78.65
Std.	7.43
25%	72.0
50%	78
75%	84
Weighted average happiness score	Weighted (Total. 100)
Mean	76.07
Std.	7.43
25%	64.0
50%	80.0
75%	92.0
Program type and hours
Food and agricultural education	Hours/1 year
Mean	33.82
Std.	18.81
Q1 (25%)	20.0
Q2 (50%)	30.0
Q3 (75%)	40.0
Therapeutic horticulture	Hours/1 year
Mean	41.25
Std.	31.39
Q1 (25%)	21.0
Q2 (50%)	32.0
Q3 (75%)	54.0
Health promotion	Hours/1 year
Mean	30.76
Std.	21.88
Q1 (25%)	15.0
Q2 (50%)	30.0
Q3 (75%)	50.0
Recreational and cultural activities	Hours/1 year
Mean	41.44
Std.	28.18
Q1 (25%)	20.0
Q2 (50%)	38.0
Q3 (75%)	58.0
Social engagement	Hours/1 year
Mean	28.27
Std.	20.62
Q1 (25%)	16.0
Q2 (50%)	22.0
Q3 (75%)	33.0

Among the 1033 older adults included in the analysis, the majority were female (75.8%), with a mean age of 78.65 years (SD = 7.43). The weighted average WHO-5 well-being score was 76.07 (SD = 7.43), with a median score of 80, indicating generally high subjective well-being across the sample.

Program participation hours over the past year varied across the 5 types of community-based activities. On average, participants reported 33.82 h in food and agricultural education, 41.25 h in horticultural therapy, 30.76 h in health promotion activities, 41.44 h in recreational and cultural activities, and 28.27 h in social participation activities. The distribution of hours shows considerable variation, particularly in horticultural and recreational programs, suggesting differing levels of accessibility, interest, or feasibility across communities.

These descriptive findings provide foundational context for the subsequent inferential analyses, which explore the relationship between program engagement and subjective well-being in greater depth.

Descriptive Results and ANOVA

Participants were categorized into tertiles (Q1-Q3) based on hours of engagement in each program type. One-way ANOVA revealed significant differences in well-being scores across tertile groups for 4 out of the 5 program types: food and agricultural education (F = 4.67, P = .0096), horticultural therapy (F = 4.59, P = .0104), health promotion (F = 5.35, P = .0049), and recreational and cultural activities (F = 12.15, P < .001). The most pronounced differences in well-being were observed among participants engaged in recreational and cultural activities. No significant differences were found for levels of participation in social engagement programs (F = 2.10, P = .1234) (Table 2).

Table 2.

Comparison of Subjective Well-Being by Levels of Participation in 5 Program Types (One-Way ANOVA Results).

Program type	Q1	Q2	Q3	F-value	P-value
Food and Agricultural Education	73.7	77.1	78.6	4.67	.0096
Therapeutic Horticulture	74.9	73.2	78.2	4.59	.0104
Health Promotion	77.0	75.2	71.4	5.35	.0049
Recreational and Cultural Activities	75.9	70.5	79.0	12.15	.000006
Social Engagement	78.0	74.8	77.0	2.10	.1234

OLS Regression Model Without Community Controls

A multivariate ordinary least squares (OLS) regression was conducted to assess the association between program participation hours and well-being. The model was statistically significant (P < .001) and explained approximately 8.1% of the variance in well-being scores (R² = .081). Participation in health promotion (β = .149, P < .001) and recreational and cultural activities (β = .103, P < .001) were positively associated with higher well-being. In contrast, participation in horticultural therapy (β = −.078, P = .001) and social engagement (β = −.316, P < .001) were negatively associated with well-being. Participation in food and agricultural education showed no significant association with well-being (β = −.022, P = .491) (Table 3).

Table 3.

Associations Between Program Participation and Subjective Well-Being (OLS Regression Without Community Controls).

Program type	Regression coefficient (β)	P-value
Food and Agricultural Education	−.022	.491
Therapeutic Horticulture	−.078	.001
Health Promotion	+0.149	≤.001
Recreational and Cultural Activities	+0.103	≤.001
Social Engagement	−.316	≤.001

OLS Regression Model with Community Controls

A second regression model was conducted controlling for community-level effects using dummy variables for each of the 44 communities. The model showed a substantial improvement in explanatory power (R² = .536, P < .001), indicating that contextual community factors contributed meaningfully to the relationship between program participation and well-being. After controlling for community differences, participation in health promotion (β = .159, P < .001), recreational and cultural activities (β = .249, P < .001), and food and agricultural education (β = .346, P < .001) were significantly and positively associated with well-being. The negative effect of horticultural therapy remained significant (β = −.123, P = .003), although the effect size was reduced. The effect of social engagement also remained negative (β = −.099, P = .001), suggesting that participation in this category may be influenced by unobserved factors requiring further investigation. All models used robust standard errors to correct for heteroskedasticity. The sample size for each regression model was N = 1033. Degrees of freedom (df) for the OLS models were 5 (program types) and 1027 (residuals) in the base model, and 56 (program + community dummies) and 976 (residuals) in the community-controlled model (Table 4).

Table 4.

Associations Between Program Participation and Subjective Well-Being (OLS Regression with Community Fixed Effects).

Program type	Regression coefficient (β)	P-value
Food and Agricultural Education	+.346	≤.001
Therapeutic Horticulture	−.123	.003
Health Promotion	+.159	≤.001
Recreational and Cultural Activities	+.249	≤.001
Social Engagement	−.099	.001

Note. 1. Age and gender were included as control variables in the model but were not statistically significant predictors (P > .05). Full results available upon request. 2. Models include robust standard errors. Sample size = 1033. Degrees of freedom: Model = 56, Residual = 976.

Additional Analysis Using Multilevel Modeling

To further account for the hierarchical data structure of individuals nested within 44 rural communities, we estimated a linear mixed-effects model with random intercepts for communities (Table 5). In this specification, program participation hours were measured at the community level and entered as fixed effects, while allowing the intercept to vary randomly across communities. The intraclass correlation coefficient (ICC = .542) indicated that approximately 54% of the variance in WHO-5 well-being scores was attributable to differences between communities.

Table 5.

Linear Mixed-Effects Models for WHO-5 Well-Being. Random Intercepts for Communities (44 Clusters); N = 1033. WHO-5 Scaled 0 to 100. Fixed Effects Include Program Hours (Annual), Age, and Female.

A. Fixed effects
Predictor	Estimate (β)	Std. error	95% CI lower	95% CI upper	P-value
Constant	−147.881	120.738	−384.522	88.761	.221
Food & Agricultural Education	−.038	0.123	−0.28	0.204	.756
Therapeutic Horticulture	−.044	0.106	−0.252	0.163	.676
Health Promotion	.103	0.133	−0.158	0.364	.440
Recreational & Cultural Activities	.032	0.105	−0.173	0.237	.761
Social Participation	−.15	0.135	−0.414	0.114	.265
Age (years)	−.117	0.062	−0.238	0.004	.059
Female (vs. male)	−.347	1.017	−2.34	1.646	.733
B. Random effects and model fit
Component/index			Value
Log-likelihood			−4220.383
AIC			8449.032
BIC			8498.434
Random intercept Var (τ00)			212.112
Random intercept SD			14.564
Residual Var (σ²)			179.382
Residual SD			13.393
ICC			0.542

Note. Models estimated via REML for variance components; AIC/BIC reported from ML refit. ICC = τ00/(τ00 + σ²). STROBE guideline followed; checklist provided in Supplemental Material.

Compared with the OLS model with community fixed effects (Table 4), the multilevel model produced similar coefficient directions but generally smaller magnitudes and reduced statistical significance for the program variables. This attenuation is expected because, in the mixed-effects framework, community-level program hours are highly correlated within clusters, and part of the between-community variance is absorbed by the random intercepts rather than attributed to the fixed effects. For example, food and agricultural education was significantly positive in the fixed-effects model (β = .346, P < .001) but became small and non-significant in the mixed-effects model (β = −.038, P = .756). Similarly, recreational and cultural activities lost statistical significance once random intercepts were introduced. These differences highlight the impact of modeling assumptions about community effects and underscore the importance of considering both fixed- and random-effects specifications when evaluating the role of local context in program effectiveness.

Controlled Regression Model Including Age and Gender

To ensure robustness, an additional regression model was conducted that included age and gender as individual-level covariates, alongside the 5 program types and community-level fixed effects. Neither age (β = −.118, P = .057) nor gender (β = .404, P = .691) demonstrated a statistically significant association with subjective well-being. The inclusion of these variables did not substantially alter the significance or magnitude of the main effects observed for program participation. These findings suggest that the effects of program engagement on well-being are consistent across age and gender groups within this rural older adult population.

Differential Impacts of Program Participation on Well-Being

Regression coefficients (β) for the association between hours of participation in 5 types of community-based programs and subjective well-being among older adults, before and after controlling for community-level fixed effects. Programs include food and agricultural education, horticultural therapy, health promotion, recreational and cultural activities, and social participation. Positive coefficients indicate a beneficial association with well-being. After accounting for community differences, the explanatory power of the model increased substantially (R² = .536) (Figure 1), with some program effects—particularly for food and agricultural education—shifting from non-significant to significantly positive. Negative associations for horticultural therapy and social participation persisted, albeit with reduced magnitude.

Figure 1.

Impact of program participation on well-being. Bars represent regression coefficients with 95% confidence intervals, estimated without adjustment and with adjustment for age and gender.

Multilevel Modeling: Accounting for Nested Community Structure

To account for the hierarchical nature of the data—individuals nested within 44 distinct rural communities—a series of linear mixed models (LMMs) were conducted. These models allowed for both random intercepts and random slopes to capture variability in program effects across different communities. As shown in Figure 2, community-specific slopes for food and agricultural education varied moderately, indicating place-based variability in its impact on well-being.

Figure 2.

Community-level slopes for food and agricultural education. Community labels follow the format [County abbreviation] − [Index] (eg, CH-01 = Changhaw County, community 1); full mapping of labels to geographic and economic attributes is provided in Supplemental Table S1. Vertical axes are standardized to the same scale across figures for comparability.

The random intercept model confirmed that significant variance in well-being existed at the community level (Group Var = 170.49), underscoring the need to control for contextual effects beyond fixed dummy variables. The variation in slope was particularly large for horticultural therapy (Figure 3), where several communities demonstrated negative or near-zero associations.

Figure 3.

Community-level slopes for horticultural therapy. Community labels follow the format [County abbreviation] − [Index] (eg, CH-01 = Changhaw County, community 1); full mapping of labels to geographic and economic attributes is provided in Supplemental Table S1. Vertical axes are standardized to the same scale across figures for comparability.

A random slope model was further implemented to allow the effects of program participation—particularly in health promotion activities—to vary across communities. The variance in slope for health promotion was minimal (Var = 0.002), suggesting relatively consistent effects. As illustrated in Figure 4, health promotion programs showed generally positive slopes with minimal variance, suggesting robust benefits across communities. However, when all 5 program types were modeled with random slopes, substantial variability in slope was found, particularly for:

• Horticultural Therapy (Var = 14.90)

• Health Promotion (Var = 14.84)

• Social Engagement (Var = 13.29)

Figure 4.

Community-level slopes for health promotion. Community labels follow the format [County abbreviation] − [Index] (eg, CH-01 = Changhaw County, community 1); full mapping of labels to geographic and economic attributes is provided in Supplemental Table S1. Vertical axes are standardized to the same scale across figures for comparability.

This indicates that program effectiveness is not uniform, and varies significantly across communities depending on local context and implementation dynamics. Figure 5 further demonstrates that recreational and cultural activities also yielded consistent positive associations in most community contexts. In contrast, as shown in Figure 6, social engagement exhibited wide variation, with several communities experiencing negative associations with well-being.

Figure 5.

Community-level slopes for recreational and cultural activities. Community labels follow the format [County abbreviation] − [Index] (eg, CH-01 = Changhaw County, community 1); full mapping of labels to geographic and economic attributes is provided in Supplemental Table S1. Vertical axes are standardized to the same scale across figures for comparability.

Figure 6.

Community-level slopes for social engagement. Community labels follow the format [County abbreviation] − [Index] (eg, CH-01 = Changhaw County, community 1); full mapping of labels to geographic and economic attributes is provided in Supplemental Table S1. Vertical axes are standardized to the same scale across figures for comparability.

Figures 2 to 6 illustrates the variation in estimated slopes for each program type across communities. Positive slopes indicate beneficial associations with well-being, while negative slopes suggest potential contextual or implementation-related barriers. This variation highlights the importance of tailoring program strategies to specific community contexts.

These multilevel analyses offer a more robust estimation of program impacts by modeling the real-world nested data structure. Compared to OLS models with fixed community controls, LMM provides more accurate standard errors and deeper insight into community-level heterogeneity in program effectiveness.

Notably, the greatest variability in slopes was observed for therapeutic horticulture and social engagement programs. For instance, community-level β coefficients for horticultural therapy ranged from approximately −0.45 to +0.35, with over one-third of the communities showing negative or near-zero associations with well-being. This finding indicates that while horticultural therapy may be beneficial in certain settings, it could be ineffective or even stressful in others—possibly due to factors such as task intensity, physical accessibility, or lack of therapeutic supervision.

Conversely, health promotion and recreational activities exhibited more consistently positive slopes across most communities, with narrower variance. These patterns suggest that while some programs exhibit robust effects regardless of locale, others are highly context-dependent and may require tailored design and implementation to ensure efficacy. The heterogeneity observed reinforces the value of community-level customization and iterative feedback mechanisms in program planning.

Discussion

Our findings indicate that community-based care programs vary significantly in their impact on rural older adults’ well-being. Health promotion and recreational activities consistently demonstrated positive effects, reinforcing their role as high-value components in decentralized aging care systems. In contrast, the negative associations observed with horticultural therapy and social participation suggest that program effectiveness depends on alignment with participants’ functional capacities and local context. The emergence of positive effects for food and agricultural education after adjusting for community factors further highlights the importance of place-based program design. These results underscore the need for tailored service models that integrate user-centered design, contextual sensitivity, and ongoing evaluation to optimize rural care delivery.

Positive Impacts of Health Promotion and Recreational Activities

Participation in health promotion and recreational or cultural activities showed a consistently strong positive association with older adults’ subjective well-being across both bivariate and multivariate analyses. These findings align with a growing body of literature highlighting the physical, psychological, and social benefits of structured health education, exercise, and culturally engaging programs for aging populations.^13,14 Notably, a dose–response relationship was observed in recreational activity participation, suggesting that both the type and duration of engagement significantly influence well-being outcomes. ¹⁵

In rural areas, where access to formal mental health services is often limited, community-based programs play a critical role in fostering psychological resilience and social connectedness. ¹⁶ The design and delivery format of these programs—whether in-person or digital—also appear to matter. Evidence from Japan supports this, showing that structured, socially interactive programing can effectively enhance happiness and reduce isolation among older adults. ⁶ Further supporting these findings, prior research has shown that mobility and participation in cultural events are associated with greater life satisfaction among older adults in East Asia. ² Together, these findings underscore the importance of accessible, engaging, and well-structured community interventions—particularly in rural settings—as powerful tools to support healthy and fulfilling aging.

Contextual Complexity of Food and Agricultural Education

The effects of different community-based programs on older adults’ well-being revealed a nuanced picture shaped by both program type and contextual factors. Food and agricultural education, for instance, demonstrated a complex pattern across analytic models. While bivariate analysis showed a modest yet significant positive association with well-being, this relationship was not evident in the initial regression model. However, after adjusting for community-level clustering, the positive association re-emerged, underscoring the importance of contextual influences such as program delivery methods, local food systems, and levels of social cohesion.

These findings align with prior research suggesting that place-based, culturally grounded learning experiences—particularly in rural settings—enhance older adults’ identity, autonomy, and social connectedness.^9,14 Food and agricultural education often incorporates intergenerational components, which may foster a sense of purpose, continuity, and belonging among older adults, especially when tied to local traditions and community practices.

Unexpected Negative Associations in Social and Horticultural Programs

The negative associations observed for horticultural therapy and social participation programs diverge from conventional assumptions that all community engagement enhances well-being. These counterintuitive results warrant careful theoretical and contextual interpretation. From a role theory perspective,^17,18 such programs may unintentionally impose implicit expectations—such as caregiving duties, coordination roles, or emotional labor—that generate psychological stress, particularly among frail older adults. This phenomenon may be exacerbated in rural settings, where community cohesion is high but role flexibility is limited, leading to participation driven by obligation rather than intrinsic motivation.

Continuity theory further suggests that well-being is maintained when older adults engage in activities aligned with their established routines, preferences, and capacities. ¹⁹ When horticultural or social programs demand unfamiliar roles or physical effort beyond an individual’s capability—such as strenuous gardening tasks or complex social dynamics—they may disrupt this continuity and induce distress rather than comfort. Although these programs are often designed to be therapeutic or socially enriching, their implementation may lack age-appropriate adaptations or sufficient emotional support, especially when conducted in non-senior-friendly environments. ⁸

These findings are consistent with recent frameworks like the WHO’s Healthy Aging model, ¹² which emphasize functional ability and environmental fit as critical determinants of well-being. A mismatch between program design and participants’ functional or psychosocial needs may explain the adverse outcomes. In line with this, Aerts et al ²⁰ and Chen et al ⁷ have noted that older adults may experience stress when community activities involve emotionally or physically taxing roles.

Additionally, selection bias may partially account for these findings, as older adults experiencing loneliness or psychological distress may be more likely to participate in such programs, skewing outcomes. However, the persistence of negative associations even after controlling for age, gender, and community variation suggests underlying design or contextual issues. While social participation is broadly recognized as a vital contributor to successful aging1, its benefits may depend on delivery within familiar, supportive, and appropriately structured environments, as community attachment and psychological traits can mediate its effects on well-being. ²¹ The absence of such reinforcing elements may explain the diminished well-being observed in this study’s context.

Collectively, these results underscore that the effectiveness of community-based interventions hinges not merely on participation frequency or intent, but on program quality, contextual alignment, and individual-environment fit. Especially in rural areas, programs must be co-designed with older adults, culturally and functionally tailored, and implemented with sufficient flexibility to accommodate diverse capacities and expectations.

Program Design and Demographic Moderators

Together, these findings suggest that the effectiveness of community-based interventions for older adults is not uniform across all program types. Instead, it depends heavily on thoughtful design, cultural relevance, environmental suitability, and sensitivity to participants’ physical and emotional capacities. For interventions to be truly impactful, especially in rural settings, they must be locally adapted, intergenerationally inclusive, and responsive to the lived realities of older adults. Future qualitative research should further explore how older individuals experience these programs and identify the contextual conditions that optimize their well-being outcomes.

Community Context, Demographics, and Policy Implications

The community-level effects identified in our models likely reflect a combination of geographic, economic, and infrastructural differences among the 44 participating rural communities. For instance, mountainous agricultural communities tended to focus more heavily on food and agricultural education, leveraging local crop production and farming traditions. In contrast, coastal fishing villages often emphasized recreational and cultural activities linked to maritime heritage. Variations in program outcomes may also be shaped by differences in transportation accessibility, availability of trained facilitators, and strength of local volunteer networks. These contextual attributes—though not directly modeled as independent variables—represent important underlying factors influencing program engagement and well-being outcomes.

The impact of food and agricultural education on older adults’ well-being revealed a complex pattern across analytic models. While bivariate analysis showed a modest but significant positive association, this effect disappeared in the initial regression model. Interestingly, after adjusting for community-level clustering, a significant positive relationship re-emerged, highlighting the critical role of contextual factors such as how the program is delivered, the presence of robust local food systems, and the degree of social cohesion within the community.

These findings suggest that the effectiveness of food and agricultural education is not solely dependent on participation, but also on the surrounding cultural and environmental context. Prior research supports this view, emphasizing that place-based, culturally grounded learning experiences can strengthen older adults’ sense of identity, autonomy, and social connectedness.^9,14 In rural settings, where food traditions and community ties are often deeply rooted, such programs may be particularly impactful.

Moreover, the intergenerational component commonly embedded in food and agricultural education may further contribute to a sense of purpose and continuity among older adults, enhancing their subjective well-being. Collectively, these results underscore the importance of culturally responsive and locally integrated program designs that leverage community assets to foster meaningful engagement and psychological well-being in later life.The final model, which included community-level controls, explained 53.6% of the variance in well-being scores—a substantial improvement over the base model. This finding highlights the importance of accounting for inter-community variability in research on aging. Factors such as geographic location, availability of resources, facilitator expertise, and community engagement culture may significantly shape how programs impact individuals.

Furthermore, in our regression analyses, age and gender were included as control variables (covariates) rather than as primary independent variables of interest. As such, the absence of statistical significance for these covariates does not contradict prior studies that examined age or gender as focal predictors with different model specifications and control structures. Instead, our findings indicate that, after accounting for program participation and community-level effects, age and gender did not add explanatory power to the variation in well-being. This suggests that, in rural contexts, community engagement and participation in meaningful programs may exert stronger influence on well-being than demographic characteristics alone.

Additionally, while not directly measured in our study, the role of family social capital may also shape how older adults engage with their communities and how they benefit from program participation. Prior research has highlighted the influence of family relationships and caregiving roles on older adults’ social involvement and well-being outcomes. ²² It is possible that unobserved family dynamics in our sample influenced both program participation and its perceived impact, underscoring the need for future studies to incorporate familial and intergenerational factors when evaluating community-based interventions.

The significance of community attachment as a contributing factor to well-being is also supported in the literature. For example, Seo and Lee ²³ found that while both community participation and attachment positively relate to well-being, the effect of attachment is especially salient in rural areas. Such findings reinforce the relevance of community integration strategies for rural aging policies.

Study Limitations

This study has several limitations. First, although we applied the ecological model as a guiding framework, our analysis was constrained by the availability of community-level variables. We were able to include factors such as geographic type, population size, predominant economic activity, access to health facilities, and program delivery capacity. However, important contextual indicators—such as average education level, household income, prevalent lifestyle habits, program-specific local government funding, the number of institutions delivering each program, and the frequency of program provision—were unavailable in the secondary dataset. The absence of these data limited our ability to fully capture the mechanisms underlying community-level effects.

Second, “rural” was defined according to the Taiwanese Ministry of Agriculture criteria, which may differ from definitions used in other countries or frameworks. Consequently, the generalizability of our findings to rural populations in different settings should be approached with caution.

Third, the cross-sectional design restricts causal inference and precludes the assessment of within-person changes in well-being over time. Future research should adopt longitudinal or quasi-experimental designs (eg, pre-post surveys or difference-in-differences approaches) to more accurately estimate program effects. Potential self-selection bias may also exist, as more motivated or socially active individuals might have been more likely to participate in the programs.

Finally, while we accounted for clustering at the community level, unmeasured contextual factors—such as leadership quality, volunteer engagement, or infrastructure—may still have influenced outcomes. Incorporating additional individual-level measures (eg, physical or cognitive status, education, income, and lifestyle habits) and operational metrics in future evaluations could strengthen the contextual analysis and provide deeper insights into program effectiveness.

Policy and Practice Implications

These results have several implications for rural aging policy and practice. First, investments in health promotion and culturally meaningful leisure activities should be prioritized in rural community centers. Second, the design and implementation of horticultural and social participation programs should be critically reviewed to ensure they are not inadvertently generating distress or excluding vulnerable populations. Finally, it is essential to consider community context when evaluating program outcomes, as local dynamics may significantly mediate program effectiveness.

Conclusion

This study demonstrates that community-based programs can improve the well-being of rural older adults, though effects vary by program type and local context. Health promotion and recreational or cultural activities showed consistent benefits, while horticultural therapy and social participation require better alignment with participant needs. Food and agricultural education had a positive effect only after accounting for community-level factors, underscoring the importance of local implementation quality, cultural relevance, and social cohesion.

Key recommendations include: (1) prioritize health promotion and culturally meaningful recreational activities; (2) strengthen quality assurance for horticultural and social programs; (3) use community-led co-design to ensure cultural fit; (4) integrate intergenerational, place-based strategies in food and agricultural education; and (5) conduct localized needs assessments to tailor delivery to each community’s capacity. These strategies can enhance the impact, equity, and sustainability of rural aging interventions, supporting aging in place through inclusive, community-driven care.

Supplemental Material

sj-docx-1-inq-10.1177_00469580251382758 – Supplemental material for Cross-Sectional Study of Health Promotion and Recreation Effectiveness on Quality of Life Among Rural Older Adults

Supplemental material, sj-docx-1-inq-10.1177_00469580251382758 for Cross-Sectional Study of Health Promotion and Recreation Effectiveness on Quality of Life Among Rural Older Adults by Kai-Lin Liang and Yi-Chun Hung in INQUIRY: The Journal of Health Care Organization, Provision, and Financing

We would like to thank the community organizations and facilitators involved in the Green Care initiative for their support in data collection and program implementation. We also acknowledge the administrative assistance provided by the Ministry of Agriculture’s regional offices.

ORCID iD

Kai-Lin Liang https://orcid.org/0009-0006-1804-7327

Ethical Considerations

This study involved secondary analysis of anonymized program data collected by community organizations as part of routine government-funded health promotion programs. No identifiable personal information was collected, and no experimental interventions were conducted. All data were aggregated at the community level and used solely for research purposes. As such, this study does not fall under the scope of human subjects research requiring Institutional Review Board (IRB) approval.

Author Contributions

Kai-Lin Liang (85%): Conceptualization, Methodology, Formal Analysis, Writing – Original Draft, Supervision. Yi-Chun Hung (15%): Data Curation, Visualization.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data supporting the findings of this study are derived from anonymized secondary sources provided by Taiwan’s Ministry of Agriculture. These data are not publicly available but can be obtained upon reasonable request from the corresponding author, subject to institutional and governmental approval.

Supplemental Material

Supplemental material for this article is available online.