Patterns of social participation among older adults and their association with self-rated health: the mediating role of activities of daily living

Abstract

Background

The aging population poses challenges in various areas, particularly regarding the health of older adults, which is becoming a worldwide issue due to decreasing physical abilities. Social participation improves both mental and physical well-being, making it an important approach for promoting active aging. This research examined the patterns of social participation among the older adults and how these patterns relate to their Activity of Daily Living and their Self-rated Health.

Methods

Utilizing data from the 7th National Health Service Survey in Gansu, a potential category analysis was performed to examine the social participation patterns of older adults. To investigate how these patterns influence Self-rated Health and the Activity of Daily Living, mediation effect analysis was carried out, incorporating classical triple regression, structural equation modeling, and Bayesian structural equation modeling.

Results

In Gansu Province, the social participation patterns among older adults are categorized as Diverse Participation(9.85%), Agricultural-centered Participation (29.02%), and Low Participation (61.13%). Those engaged in diverse and agricultural participation exhibit better Activities of Daily Living and Self-rated Health compared to those with low participation. Using the Low Participation group as reference, both Diverse Participation (β = 0.038, 95% CI = 0.028–0.047) and Agricultural-centered Participation (β = 0.042, 95% CI = 0.030–0.055) showed that the association between social participation and Self-rated Health was achieved partly through Activities of Daily Living this mediators, accounting for 34.23% (Diverse Participation) and 40.00% (Agricultural-centered Participation) of the total effects respectively.

Conclusion

Activities of Daily Living mediates the relationship between social participation and Self-rated Health. Attention should be paid to the Activities of Daily Living of older adults, with particular attention to those living alone and those who are chronically ill, and to social resources to enhance the social participation of older persons.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40520-025-03131-3.

Background

The United Nations’ most recent population projections indicate that by 2050, China’s older population will reach 470 million, representing over 30% of the total population, making it the country with the largest aging population globally [1]. This demographic shift poses significant challenges to various sectors, including the economy, society, and healthcare, with the health of older adults being the most pressing issue that needs addressing and recognized as a global concern [2]. The well-being of older individuals not only impacts their quality of life but also affects societal stability and development. Currently, numerous studies focus on the health issues faced by older adults, such as chronic disease management, cognitive health, mental well-being, physical capabilities, nutrition, and social support [3–9]. The ultimate goal of these studies is to promote healthy aging, slow down the aging process, and enhance the overall health and quality of life for older adults.

Self-rated Health (SRH) serves as a crucial measure for assessing an individual’s health status, as it is closely linked not only to physical well-being but also to psychological conditions and social environmental influences [10]. SRH is widely recognized as a significant indicator of health-related quality of life among older adults. Numerous studies indicate that SRH is influenced by various factors, including socio-demographic elements (such as age, gender, education, and economic status), health behaviors (like smoking, alcohol consumption, and exercise routines), and medical conditions (including the presence and management of chronic illnesses) [11]. These factors collectively shape how individuals perceive their own health, which subsequently impacts their quality of life and overall health. By examining these determinants together, a more comprehensive understanding of the health status of older adults can be achieved, leading to potential improvements.

Researchers have identified social participation (SP) as a potentially adaptable area. SP refers to the way individuals engage in social activities and interact with others. It enhances both physical and mental well-being in older adults by offering social support, alleviating feelings of loneliness, and boosting life satisfaction [12]. Previous research has indicated that SP positively influences health outcomes in older adults, contributing to delayed mortality [13], increased healthy life expectancy [14–16], better SRH [17], improved mental health [18] and subjective well-being [19], as well as reduced cognitive decline [20]. An international study also found that high SP is a potential health-promoting factor in some low- and middle-income countries [21]. However, SP takes many different forms, and different modes of SP may have different impacts on the health of older adults. Both social and individual-level studies have found that active, organized SP (e.g., participating in community activities, volunteering) is associated with better SRH, whereas negative, passive SP (e.g., watching television, playing cards) may be associated with poorer SRH [22–24]. Additionally, the frequency of SP affects the health of older adults. A study on the association between frequency of SP and SRH showed that the probability of poor SRH decreased as the frequency of SP increased [25]. Besides, existing research indicates that family composition is the primary context that influences older people’s SP and health. Multigenerational families naturally provide opportunities for SP through intergenerational interactions and thus maintain physical and mental [26], whereas older people living alone need to actively build external social networks.

Activities of Daily Living (ADLs) serve as a crucial measure of functional independence in older adults. With advancing age, progressive physical decline typically leads to deteriorating ADLs. However, research suggests SP may mitigate this decline through multiple protective mechanisms. Studies demonstrate that older adults who are more socially active have better physical function than their inactive counterparts, such as lower rates of disability, motor decline, loss of ADLs and poor mobility [25]. Globally, SP has been associated with delayed frailty in review studies [27]. Moreover, the decline in daily living abilities of older adults not only adversely impacts their quality of life but also imposes considerable burdens on both families and society at large [28].

Gansu Province is located in the upper reaches of the Yellow River in northwest China and is an economically underdeveloped area in China. By the end of 2022, the province’s permanent resident population was 24.9242 million. The per capita disposable income of urban residents was 37,572 yuan, and that of rural residents was 12,165 yuan, both significantly lower than the national average (the per capita disposable income of urban residents was 48,282.9 yuan, and that of rural residents was 20,132.9 yuan) [29]. Gansu is a province where multiple ethnic groups live together. Different ethnic groups have different cultural traditions, lifestyles and concepts of elderly care. Studying the aging issue in Gansu Province can provide an in-depth understanding of the elderly care model and the needs of the elderly in a multicultural background, and offer ideas for dealing with the aging problem in multi-ethnic areas. This is of great reference significance for other multi-ethnic concentrated areas in China.

Although a large number of studies support the protective effect of SP on the ADLs and SRH, the interactions between these variables may be bidirectional, as shown in Fig. 1. However, this study focused on the SP→ADLs→SRH pathway based on the following considerations. First, in terms of theoretical mechanisms, social role theory and activity theory emphasize SP as a driver of functional capacity. Second, empirical evidence suggests asymmetric effects. Longitudinal analyses by Stuck et al. showed that the odds ratio of ADLs decline leading to a subsequent decrease in SP was only 56% of the amount of effect of SP maintaining ADLs [30]. Finally, methodologically, we reduced reverse causality bias by controlling for health confounding variables (e.g., chronic disease). Although bidirectionality could not be completely ruled out, this provided a theoretical framework for subsequent longitudinal studies. In addition, previous studies have mostly viewed SP as a single dimension or simple categorization (e.g., participation or not, frequency), failing to capture its underlying pattern heterogeneity (e.g., differences in types, combinations of social contexts). This omission may obscure the differential impacts of different patterns of participation on health, leading to a ‘one-size-fits-all’ approach to intervention strategies. This study aims to investigate the mediating role of ADLs in the relationship between SC patterns and SRH in Chinese older adults, with a view to providing a scientific basis for healthy aging in rural older adults. Therefore, we propose the following hypothesis: ADLs mediate the association between SC and SRH.

Fig. 1.

Hypothesized model

Methods

Data source and participants

Since 1993, the Chinese National Health Service Survey has been carried out every five years across the country. The 7th National Health and Services Survey (NHSS) was conducted across the country in August 2023. This study used data from China’s Seventh National Health and Services Survey (NHSS), led by the Gansu Provincial Health Commission, between August 2023 and October 2023. This survey adopted multi-stage random sampling of the whole group, following the principle of economic validity while ensuring the representativeness of the sample and assessing the whole population through the sample. First, five districts/counties were selected (Yuzhong, Jingtai, Lintan, Maiji, and Ganzhou). Second, all towns (streets) in each selected district (counties) were divided into five levels, and one town or street is randomly selected at each level. Third, two villages (residential committees) were randomly selected from each selected town or street. The detailed process is shown in Appendix Fig. 1. In this survey, the investigators used tablet computers (pads) to conduct face-to-face surveys at residential homes. A total of 7,777 residents from 3,000 households were surveyed, with 2,619 adults aged 60 and above selected as the study population.

Description of sample size and representativeness

In this study, the sample size was calculated using a prevalence sampling formula that employs cross-sectional studies. The calculation is shown in Eq. (1).

1

The prevalence of SRH being good was p = 48.9% according to previous studies [31], so in the study with SRH, π = 0.489, µα = 1.96, δ = 0.025, α = 0.05.The theoretical minimum sample size of 1536 was calculated to be required. The final number of 2,619 older adults successfully surveyed in this study (amounting to 170.6% of the theoretical sample size), which met the required sample size.

In addition, this study corrects for sampling bias through three levels of weight adjustment. The specific methods are as follows:

2

3

4

5

Calculate county/district-level weights (), township/township-level weights (), calculate household-level weights (​), and final weights as (), respectively; in formula (2), is the total number of target population in the kth district or county, is the actual sample size of the kth district or county, and is the total number of districts and counties in the province; in formula (3), is the population of older adults in the Lth township in the Kth district or county, is the actual sample size of the township, and is the total number of townships in that district and county; and in Eq. (4), is the household response rate (the proportion of surveyed households that completed the survey), and is the total number of households in the Lth township of the county. All data are from the 2020 Seventh Population Census of China. To avoid the effect of extreme values, the weights are truncated to limit them to [0.005, 0.05]. All weights were standardized by dividing by the mean value of 0.039 (sum = 2,619 persons). All data are from China’s 7th Population Census [32].

The representativeness of the sample was tested by performing a standardized difference analysis before and after weighting the key parameters. In general, standardized differences < 10% are considered to have no significant deviation before and after weighting [33]. Exhibit 1 shows that the standardized differences of the key parameters are all < 10%, which means that the sample is representative of the older population in Gansu Province.

Table 1.

Basic characteristics of survey residents (N = 2,619)

Variable	Unweighted proportion(%)	Weighted proportion(%)	Standardized Difference (%)	Judgment Criteria
Male	49.64	49.75	0.23	< 10%
Chronic disease	64.76	66.76	4.22	< 10%
Social participation patterns
Low participation	61.13	64.04	6.05	< 10%
High participation	9.85	9.20	2.23	< 10%
Agricultural participation	29.02	26.76	5.04	< 10%
Number of limitations in activities of daily living	2.22	2.01	1.39	< 10%
Self-rated Health
Very poor	0.42	0.37	0.72	< 10%
Poor	1.99	2.14	1.06	< 10%
Fair	14.85	15.30	1.24	< 10%
Good	44.86	42.17	5.44	< 10%
Very good	37.88	40.03	4.41	< 10%

Theoretical foundations of social participation of older adults

Social role theory

Social role theory posits that within the framework of social identity theory, the concept of “self-categorization” pertains to the manner in which individuals perceive themselves as entities and subsequently classify or label themselves according to socially defined categories. This process is alternatively termed “identification” in sameness theory [34, 35]. The formation of roles is contingent upon an individual’s self-categorization or identification process (Stets & Burke, 2000). Within a specific social structure, roles delineate not only the behavioral expectations for the individual but also for their interaction partners, while simultaneously conveying associated social meanings. Collectively, these meanings and expectations establish a normative framework that directs individuals’ behavioral conduct in particular contexts [36]. Through social interactions, individuals progressively develop a series of stable behavioral expectation patterns, which form the foundational essence of social roles. These roles fundamentally encapsulate the organized characteristics of interpersonal relationships. Throughout the life cycle, individuals are required to adopt various social roles. It is important to highlight that when individuals encounter a loss of social identity or experience a deprivation of functional roles, they frequently exhibit adverse emotional responses, such as anxiety and depression. Such alterations in psychological state may further influence their overall physical and mental well-being [35].

Activity theory

Activity theory represents an extension of social role theory as it pertains to the aging population, positing that individuals can compensate for roles diminished by the aging process (such as occupational roles) by actively engaging in new roles (for instance, as participants in community activities) [37]. This theory asserts that, in contrast to adulthood, social roles in later life are characterized by voluntary engagement and are more aligned with individual preferences. Numerous social activities in which older adults engage, such as volunteering and participation in interest-based clubs, are categorized as non-compulsory. This self-directed social involvement has been shown to enhance their psychological well-being. The primary focus of activity theory is to underscore the essential role of SP in sustaining vitality among older adults. Empirical evidence indicates that older individuals who engage in higher levels of SP tend to experience increased life satisfaction and social adaptability [38]. Through positive social interactions, older adults can mitigate some of the negative consequences associated with aging, such as social isolation, cognitive decline, and deterioration of both physical and mental health. In the context of addressing the challenges posed by an aging population, activity theory is pivotal in fostering the concept of “successful aging.” By engaging in SP, older individuals can assume new roles and redefine their identities, thereby enhancing their capacity to adapt to the adverse effects of aging. The notion of active participation promoted by this theory not only contributes to the maintenance of quality of life for older adults but also aligns with the socialist core values of China, which advocate for “active aging.”

Measures

Social participation patterns

Latent class analysis (LCA) is a statistical method for identifying unobserved subgroups within heterogeneous populations, particularly suitable for multidimensional variables like SP patterns. Conventional studies often dichotomize SP simply as “participation/non-participation,” overlooking the complexity of activity combinations and intensities. The central assumption of latent class analysis is that a small number of mutually exclusive latent class variables can explain the probability of various responses to observed variables, and that these latent variables are locally independent of each other [39]. There are two main types of latent category analysis: exploratory LCA and validation LCA. Exploratory LCA aims at category discovery and exploration based on data, while validation LCA validates categories based on theory [40, 41]. This study employs exploratory LCA. The process of exploratory potential category analysis consists of four main steps: firstly, assuming the existence of a null model with a single category; secondly, gradually increasing the number of categories, then determining the best model based on the fitting metrics, and finally classifying and naming the model. Commonly used metrics for model fit evaluation include Peason’s chi-square test, likelihood ratio chi-square test (G2), Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), adjusted BIC (aBIC), and entropy value (Entropy) [42]. Currently, there is no uniform determination of the advantages and disadvantages of these metrics. In this study, we used the aBIC and entropy as the primary indices for evaluating model fit. In this study, based on the survey of SP patterns of older adults, the main question asked respondents what social activities they were involved in. These social activities include economic participation, family participation, public welfare participation, recreational participation, community participation, and agricultural participation. We will automatically identify potential natural subgroups based on the participation in these six social activities, providing objective probabilistic attributions that can effectively avoid the masking of true subgroups due to subjective categorization bias. In addition, conducting potential category analysis also ensures high homogeneity within categories and heterogeneity between categories, improves the statistical validity of path analysis, and effectively tests differences in mediating effects across categories.

Variables

In this study, the outcomes of the exploratory latent class analysis were utilized as independent variables. The mediating variable was operationalized as limitations in Activities of Daily Living (ADLs), assessed through six essential functions: dressing, eating, bathing, transferring in and out of bed, using the toilet, and managing bowel and bladder control. For each domain, a limitation was coded as 1 when participants reported either needing assistance or being completely unable to perform the task. The ADL limitation score was then calculated as the sum of these binary scores across all six domains (range: 0–6), representing the total number of ADL restrictions. The dependent variable, SRH, was assessed using a visual analog scale (VAS) question: “On a scale from 0 to 100, where 0 represents the worst health and 100 represents the best health, how would you rate your health today?” Responses were categorized into five levels: “very poor” (0–19), “poor” (20–39), “fair” (40–59), “good” (60–79), and “very good” (80–100). This question is commonly employed in health surveys across various countries. Control variables included socio-demographic characteristics such as gender (male, female), age (< 75, ≥ 75), education level (illiterate, primary school, secondary school, senior high school, technical school or college, bachelor’s degree or higher), marital status (unmarried, married, other), and employment status (employed, retired, unemployed). Economic status was assessed based on household income, which was categorized into five quintiles: the first quintile (Q1) representing the lowest income and the fifth quintile (Q5) representing the highest income. Additionally, health behaviors such as smoking, alcohol consumption, physical activity, and the presence of chronic diseases were also considered.

Analysis of mediation

Classical mediation model(CMM)

Classical triple regression is the most common method in mediation analysis, which examines the mediation effect through three regression models [41]. Specific steps are as follows:

6

7

8

First, we tested the total effect (coefficient c) of the SP patterns (X) on SRH (Y), modeled as Eq. (6); next, we tested the effect (coefficient a) of the SP patterns (X) on ADLs (M), modeled as Eq. (7); and lastly, after controlling for ADLs, we tested the direct effect (coefficient c’) of the SP patterns on SRH and the independent ADLs effect (coefficient b), and the model equation is Eq. (8). The mediating effect was calculated by the product of coefficients method, i.e., the product of the effect of the independent variable on the mediator variable and the effect of the mediator variable on the dependent variable was determined and tested for significance using the Bootstrap method [42].

Structural equation modeling (SEM)

SEM is a more comprehensive analytical tool that is able to deal with both measurement and structural models [43]. Specific steps include:

First, the relationship between the latent and observed variables was defined and the reliability and validity of the measurement model was assessed. Then, the causal relationship between the latent variables is set. The formula of the model is:

9

In the Eq. (9), the pattern of SP patterns (X) is referred to as the ‘exogenous variable’, while Y includes both ADLs (M) and SRH (Y), which are referred to as the ‘endogenous variables’. The model parameters were estimated using maximum likelihood estimation (MLE) or other appropriate estimation methods, and the model’s goodness of fit was assessed [44]. The mediating effect was calculated by path coefficients and tested for significance using Bootstrap method.

Bayesian structural equation model (BSEM)

BSEM employ Bayesian inference methods that allow for the handling of model uncertainty. Key steps include:

10

First, a prior distribution is set for the model parameters and combined with the likelihood function of the data to generate the posterior distribution. The measurement model is Eq. (10). where y is a vector of observable random vectors, including the explicit variables of ADLs and SRH; α is a vector of measurement intercept terms; Λ is a factor loading matrix, which was used in this study to reflect the expected strong correlation between the latent variable (SRH) and the ordered categorization of the observations using the truncated normal distribution Normal ⁺(1, 0.5) as the prior; and η is a latent variable vector, which in this study was mapped to continuous latent variables as the latent variable of SRH through an ordered probit model with 5 levels of observation scores (very poor to very good), κ is the matrix of regression coefficients between the explicit variable y and the observed variable x (SP patterns), and ε is the monotonic vector with covariance matrix Ξ. The structural model is Eq. (11).

11

where ν is a vector of structural intercept terms, Β and Γ are matrices of structural coefficients, and ζ is a vector of covariance matrices Ψ [45, 46]. Bayesian inference is used to estimate the posterior distributions of the model parameters. Next, using the Bayesian inference method, this study uses Markov Chain Monte Carlo (MCMC) to estimate the posterior distributions of the model parameters. After obtaining the posterior distributions of the parameters through simulation, the posterior means and confidence intervals of the parameters can be calculated [47]. Then, methods such as Posterior Predictive Checks (PPC) are used to assess the fitness of the model to ensure the reliability of the model. Finally, the posterior distribution of the mediating effect was calculated and tested for significance, and Bayesian confidence intervals were used to assess the significance of the mediating effect.

All models adjusted for gender, age, education, marital status, employment status, economic status smoking, alcohol consumption, physical activity and chronic disease prevalence. While the classical triple regression method is highly respected for its concise and intuitive analytical framework, structural equation modeling demonstrates its unique strengths in dealing with intricate mediation models. In contrast, Bayesian structural equation modeling provides a highly flexible solution when dealing with model uncertainty. By combining these methods, we are able to provide a more comprehensive and reliable basis for our research conclusions.

Statistical analysis

First, we used descriptive analysis to describe the demographics. Second, an exploratory latent class analysis was performed with MPLUS, version 8.0. Then three estimation methods were used for mediation analysis. Mplus8.0 version was used for Bayesian SEM mediation effect analysis, and no information prior calculation was used. The SEM mediation model was fitted by Stata version 14.0, and the factor loading of each parameter and the evaluation index of model fit were output. The natural indirect effect size was estimated by the established structural equation model. Classical triple regression was performed as follows. The initial phase involved the application of ordinal logistic regression to examine the association between SP patterns and SRH. Subsequently, linear regression was employed to investigate the relationship between SP patterns and ADLs. Finally, ordinal logistic regression was fitted to analyze the relationship between SP patterns and SRH, including ADLs. The coefficient product method was used to estimate point estimates and 95% confidence interval estimates of indirect effects.

Results

Basic characteristics of the study subjects

A total of 2,619 subjects were included in this study. There were 1,300 (49.64%) males and 1,319 (50.36%) females in this survey. In terms of age distribution, 2,018 participants (77.05%) were aged < 75 years, while 601 (22.95%) were ≥ 75 years. The ethnic group was predominantly Han Chinese, with 2,453 residents (93.66%), and other ethnic groups accounted for 166 residents (6.34%). In terms of education level, illiteracy and elementary school accounted for the largest proportions, at 34.59% and 33.07%, respectively. Regarding employment status, the proportions of employed, retired, and unemployed individuals were 25.66%, 15.92%, and 58.42%, respectively. The majority of respondents were married (78.85%). More than half of the older adults surveyed (64.76%) reported chronic diseases. Among the 2,619 older adults, 44.68% reported good self-rated health (SRH), while 1.99% reported poor SRH. The number of limitations on activities of daily living (ADLs) ranging from 1 to 3 was 4.50%. Other features are shown in Table 1.

Table 2.

Basic characteristics of survey residents (N = 2,619)

Characteristics	Number(n)	Percent(%)
Gender
Male	1300	49.64
Female	1319	50.36
Age(years)
< 75	2018	77.05
≥ 75	601	22.95
Ethnicity
Han	2453	93.66
Minority	166	6.34
Education level
Illiteracy	906	34.59
Elementary school	866	33.07
Middle school	463	17.68
High school	293	11.19
Secondary school/Junior college	77	2.84
Bachelor degree or above	14	0.53
Employment status
Employed	672	25.66
Retired	417	15.92
unemployed	1530	58.42
Marital status
Unmarried	24	0.92
Married	2065	78.85
Other	530	20.24
Household economic status
lowest	523	19.97
Lower	524	20.01
Middle	524	20.01
Higher	524	20.01
Highest	524	20.01
Smoke
Yes	624	23.83
Quit smoking	209	7.98
Never smoked	1786	68.19
Drink in the past 12 months
Yes	321	12.26
No	2298	87.74
Frequency of physical exercise in the past month
Six or more times	1037	39.60
Three to five times	223	8.51
One to two times	125	4.77
Less than once	37	1.41
Never exercise	1197	45.70
Chronic disease
No	923	35.24
Yes	1696	64.76
Self-rated Health
Very poor	11	0.42
Poor	52	1.99
Fair	389	14.85
Good	1175	44.86
Very good	992	37.88
Number of limitations in activities of daily living
0	2443	93.28
1–3	118	4.50
> 3	58	2.22

Social participation patterns

The latent class analysis was performed on the SP of older adults, after many times of fitting to get the indicators of model fit, aBIC and Entropy are used as the criteria for model fitness decision. Specifically the model with the smallest aBIC value was preferred, Entropy > 0.8 was considered to have high categorization reliability, and in addition, combined with the principle that the probability of each category was not less than 5% [48], our study finally chose to use the 3 potential category models as the SP model of the older adults in this study. Table 2 shows the potential category probabilities and the conditional probability of taking the value of 1 for each exogenous variable in the 3 categories. In this study, category 1 was named ‘Diverse Participation’ (9.85%), which is characterized by a dual core of family (0.661) and community (0.640) participation, with multiple participation in agriculture (0.135) and public welfare (0.098); category 2 was named ‘Low Participation’ (61.13%), with conditional probabilities of all the indicators of SP tending to be close to zero (family/community/agriculture = 0.000, public welfare = 0.008), forming a significant state of social isolation; category 3 is named ‘Agricultural-centered Participation’ (accounting for 29.02% of the total), which is characterized by a significantly prominent conditional probability of agricultural labor (0.468) and a complete lack of modern-type SP (public welfare/community probability ≈ 0). This type profoundly reflects the survival strategy of ‘aging on land’ in rural Gansu. It is also consistent with the fact that rural residents accounted for a relatively large proportion of this survey, while rural older adults have a greater reliance on land as both material and spiritual dependence.

Table 3.

Social participation patterns of older adults

Explicit variable	Cluster 1 Diverse Participation	Cluster 2 Low Participation	Cluster 3 Agricultural-centered Participation
Economic participation	0.034	0.013	0.013
Family participation	0.661	0.000	0.486
Public welfare participation	0.098	0.008	0.042
Recreational participation	0.014	0.001	0.000
Community participation	0.640	0.000	0.000
Agricultural participation	0.135	0.000	0.468
Potential class probability	0.09851	0.6113	0.29019

Analysis of mediation

Classical regression mediation analysis method, structural equation model mediation analysis method and Bayesian structural equation model were used for mediating analysis. ADLs had a mediating effect on the process of SP patterns and SRH, which was statistically significant, but there were certain differences in the indirect effect estimates obtained by the three estimation methods. The comprehensive comparison results are presented in Table 3. Using Low Participation as the reference group, we estimated the indirect effects for both Diverse Participation and Agricultural-centered Participation. The classical regression method yielded mediation effect estimates with standard errors of 0.023 and 0.014, and 95% confidence interval widths of 0.091 and 0.055, respectively. In comparison, the structural equation modeling (SEM) approach produced smaller standard errors (0.005 and 0.007) and narrower 95% confidence intervals (0.019 and 0.025). The Bayesian SEM results showed intermediate precision, with standard errors of 0.011 and 0.010 and confidence interval widths of 0.041 and 0.036. Based on the comparative analysis of confidence interval widths and standard error magnitudes, the SEM-based method demonstrated superior estimation accuracy.

Table 4.

Mediation analysis results of the three Estimation methods

Method of estimation	Social participation model	Estimated value	SE	P-value	95%CI
BSEM	Diverse Participation	0.036	0.011	0.000	(0.018,0.059)
	Agricultural-centered Participation	0.036	0.010	0.000	(0.019,0.055)
SEM	Diverse Participation	0.038	0.005	0.000	(0.028,0.047)
	Agricultural-centered Participation	0.042	0.007	0.000	(0.030,0.055)
CMM	Diverse Participation	0.083	0.023	0.000	(0.037,0.128)
	Agricultural-centered Participation	0.091	0.014	0.000	(0.065,0.120)

Structural equation modeling mediation effects analysis

After adjusting for control variables, SP is significantly correlated with ADLs. Older adults with Diverse Participation are more likely to report better ADLs compared to those with Low Participation (a1=-0.146, P < 0.01); those with Agricultural-centered Participation are relatively more likely to report better ADLs compared to those with low participation (a2=-0.161, P < 0.01).See in Table 4.

Table 5.

Factors associated with activities of daily living among the older adults in gansu, China

Variables	Reference Group	Activities of Daily Living
Social Participation
Diverse participation	Low participation	-0.146***
		(0.055)
Agricultural-centered participation		-0.161***
		(0.036)
Gender
Female	Male	-0.015
		(0.047)
Age
≥ 75	< 75	0.211***
		(0.041)
Ethnicity
Minority	Han	-0.010
		(0.068)
Education level
Elementary school	Illiteracy	-0.046
		(0.041)
Middle school		0.007
		(0.052)
High school		0.013
		(0.061)
Secondary school/Junior college		-0.042
		(0.108)
Bachelor degree or above		-0.112
		(0.225)
Employment status
Retired	Employed	0.174***
		(0.062)
unemployed		0.154***
		(0.041)
Marital status
Married	Unmarried	0.259
		(0.169)
Other		0.302*
		(0.172)
Household economic status
Lower	lowest	0.066
		(0.050)
Middle		0.027
		(0.050)
Higher		-0.047
		(0.051)
Highest		-0.034
		(0.056)
Smoke	Yes
Quit smoking		0.152***
		(0.066)
Never smoked		0.101**
		(0.050)
Drink in the past 12 months	Yes
No		0.048
		(0.053)
Frequency of physical exercise in the past month
Three to five times	Six or more times	0.028
		(0.060)
One to two times		0.188**
		(0.077)
Less than once		0.153
		(0.136)
Never exercise		0.289***
		(0.037)
Chronic disease
Yes	No	0.079***
		(0.034)
Constant		-0.492***
		(0.183)
Observations		2,619

Coefficients (standardized) are reported with standard errors in parentheses.Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1

In Table 5, the model without the mediator variable (ADLs) shows that SP patterns is significantly associated with SRH. When ADLs is included in the model, the association between SP and SRH remains statistically significant after controlling for potential confounding factors. Older adults with Diverse Participation are more likely to have good SRH (c′1 = 0.072, P < 0.05); those with Agricultural-centered Participation are relatively more likely to report better SRH levels (c′2 = 0.062, p < 0.05). At the same time, after controlling for potential confounding factors, ADLs is significantly associated with SRH (b=-0.264, p < 0.01).

Table 6.

The mediating effect of activities of daily living on the association between social participation and Self-rated health among the older adults in gansu, China

Variables	Reference Group	Self-rated Health
		Without mediators (model 1)	With mediators (model 2)
Activities of Daily Living			-0.264***
			(0.019)
Social Participation
Diverse participation	Low participation	0.111**	0.072**
		(0.055)	(0.053)
Agricultural-centered participation		0.105***	0.062**
		(0.036)	(0.035)
Gender
Female	Male	-0.022	-0.026
		(0.047)	(0.045)
Age
≥ 75	< 75	-0.186***	-0.130***
		(0.041)	(0.039)
Ethnicity
Minority	Han	0.058	0.056
		(0.067)	(0.065)
Education level
Elementary school	Illiteracy	-0.010	-0.022
		(0.041)	(0.039)
Middle school		0.019	0.021
		(0.052)	(0.050)
High school		-0.059	-0.056
		(0.061)	(0.059)
Secondary school/Junior college		-0.034	-0.045
		(0.107)	(0.103)
Bachelor degree or above		0.083	0.053
		(0.223)	(0.215)
Employment status
Retired	Employed	-0.012	0.035
		(0.061)	(0.059)
unemployed		-0.147***	-0.106***
		(0.040)	(0.039)
Marital status
Married	Unmarried	0.017	0.085
		(0.167)	(0.161)
Other		0.059	0.139
		(0.171)	(0.165)
Household economic status
Lower	lowest	0.058	0.075
		(0.050)	(0.048)
Middle		0.129***	0.136***
		(0.050)	(0.048)
Higher		0.147***	0.135***
		(0.050)	(0.048)
Highest		0.167***	0.157***
		(0.056)	(0.054)
Smoke	Yes
Quit smoking		0.003	0.043
		(0.065)	(0.063)
Never smoked		-0.126***	-0.099**
		(0.050)	(0.048)
Drink in the past 12 months	Yes
No		-0.096*	-0.083*
		(0.053)	(0.051)
Frequency of physical exercise in the past month
Three to five times	Six or more times	-0.105*	-0.098*
		(0.060)	(0.058)
One to two times		-0.179**	-0.129*
		(0.076)	(0.073)
Less than once		-0.309**	-0.268**
		(0.135)	(0.130)
Never exercise		-0.122***	-0.045
		(0.037)	(0.036)
Chronic disease
Yes	No	-0.279***	-0.258***
		(0.033)	(0.032)
Constant		4.568***	4.438***
		(0.181)	(0.175)
Observations		2,619	2,619

Coefficients (standardized) are reported with standard errors in parentheses.Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1

The mediation path model is shown in Fig. 2. The path coefficients indicate that all relationships in the model are significant. After adding ADLs as a mediator, the direct effect of SP on SRH remains significant.

Fig. 2.

Mediation path model

The results of the mediation effect test show that, using the Low Participation group as a reference, the relative mediation effect value for the Diverse Participation group is 0.038, with a 95% confidence interval of [0.028, 0.047], which does not include 0, indicating that the relative mediation effect is significant. Similarly, using the Low Participation group as a reference, the relative mediation effect value for the Agricultural-centered Participation group is 0.042, with a 95% confidence interval of [0.030, 0.055], which also does not include 0, indicating that the relative mediation effect is significant (See Table 3). The results indicate that part of the association between SP patterns and SRH is realized through the mediation effect of ADLs, accounting for 34.23% (Diverse Participation) and 40.00% (Agricultural-centered Participation) of the total effects respectively (see Table 6).

Table 7.

The standardized total, direct, and mediating effects of social participation on Self-rated health with activity of daily living as mediators

Model pathways	c Total Effect	c′ Direct Effect	a*b Mediating Effect	Percentage of mediating effect
SC→ADLs→SRH
Diverse participation	0.111	0.072	0.038	34.23%
Agricultural-centered participation	0.105	0.062	0.042	40.00%

Discussion

This study conducted an exploratory profile analysis of the social participation (SP) of older adults in Gansu Province and identified three distinct patterns. More older adults (61.13%) were less likely to participate in economic, family, public welfare, entertainment, community and agriculture social activities. Only 9.85% of older adults showed high participation across all six areas. In addition, 29.02% of older adults had a high participation rate in agricultural activities. This situation is partly attributed to the economic and cultural conditions in Gansu Province, which is situated in a relatively underdeveloped region of northwest China, where the community service and support systems are lacking, potentially impacting the overall social participation of older adults [49]. The financial situation of many older adults may not allow them to participate in more social activities. In addition, due to the disparity between urban and rural areas, the unbalanced distribution of social resources may lead to differences in the participation of different older adults groups in social activities [50, 51]. The limited social resources available in rural areas can hinder the opportunities for rural seniors, who are predominantly represented in this survey, to engage in a variety of activities. Conversely, in Gansu Province, a significant number of young people are involved in migrant work, leaving many older adults as empty-nesters. For these seniors, agricultural activities often serve as their primary source of income, leading to a higher engagement in such activities. Additionally, agriculture may hold significant cultural and traditional importance in certain regions, encouraging older adults to take part in agriculture-related events [52].

The SRH of older adults in Gansu province is good

The research revealed that over 80% of older adults in Gansu Province rated their health positively, with the highest SRH observed among those engaged in multiple social activities or agricultural work. This may be attributed to the beneficial effects of social interactions and physical activity on health. These findings can be effectively interpreted through the lens of Activity Theory, which posits that maintaining purposeful engagement is crucial for successful aging. The enhanced health outcomes among socially active elders demonstrate the theory’s central premise that continued role participation preserves physical and cognitive functioning through meaningful goal-directed activities. These findings align with observations in Europe, where Santini et al. similarly found stronger health benefits among socially active individuals [53]. However, our sample finds that agricultural participation shows uniquely stronger effects. In China, agricultural participation appears to serve as a culturally significant substitute role that maintains self-identity and social integration after retirement, aligning with Activity Theory’s emphasis on environmental adaptation. Furthermore, the study indicates that older adults with higher family incomes, who do not smoke or consume alcohol, tend to have better SRH based on social demographic factors, which aligns with findings from previous research [54–57]. The income-health gradient further reflects the theory’s principle of environmental enablers, where greater financial resources facilitate access to health-promoting activities and healthcare services [58]. In a nation with a significant wealth disparity like China, poverty seems to greatly hinder a person’s capacity to cope with health crises, resulting in poorer health results [59]. According to data from China’s National Health Commission, by 2019, over 180 million older adults were affected by chronic illnesses. Additionally, smoking and alcohol consumption are separate risk factors for various chronic diseases, and older adults with these conditions often report having a poor health status [60]. Moreover, older adults who have limitations in ADLs tend to report a poorer health status, as these limitations can greatly affect their perception of health and overall quality of life [61, 62]. After controlling for various confounding factors in multiple studies, it was found that higher levels of SP and involvement in agriculture are significantly linked to improved SRH. These results collectively underscore Activity Theory’s explanatory power in demonstrating how maintained social participation, through either community activities or productive labor, can buffer against health declines in later life by preserving functional capacity, social connectedness, and psychological well-being.

The ADLs of older adults in Gansu province are good

The research revealed that only 30 older adults were identified as having limitations in all six areas of ADLs, while 2,443 had no limitations in any of the six areas. This suggests that most older adults are capable of performing ADLs effectively, with only a small number facing limitations across the board. Those who engaged more in social activities and agricultural activities demonstrated better performance in ADLs compared to those with lower participation levels. Hiroyuki and colleagues reported similar findings in a Japanese sample, where both men and women with higher social participation were less likely to be physically inactive [63]. From the perspective of Activity Theory, social engagement provides older adults with meaningful motives (e.g., social interaction) and structured opportunities (e.g., group activities) to participate in physical activities, thereby fulfilling their psychological and social needs while promoting an active lifestyle. Furthermore, the study indicated that certain sociodemographic factors, such as being widowed or having a different marital status, being unemployed, and having chronic illnesses, were associated with poorer ADLs performance. This may be attributed to the fact that being widowed or in non-traditional marital situations can lead to decreased social support and emotional loneliness, which negatively impacts ADLs [64]. Older adults who are unemployed or inactive exhibit poorer ADLs. From the perspective of Activity Theory, unemployment reduces engagement in structured activities, diminishing physical/cognitive stimulation and accelerating functional decline. Furthermore, the loss of employment diminishes social roles that previously provided purpose, identity, and motivation to maintain functional independence. Related studies have also shown that unemployment may lead to financial difficulties, increased psychological stress, etc., which may affect the ADLs [65]. Older adults with chronic diseases have poor ADLs. Chronic diseases usually lead to a decline in physical function, affecting the ability to perform ADLs and the quality of life of older adults [66]. After adjusting for a range of study confounders, SP patterns as Diverse Participation and Agricultural-centered Participation were significantly associated with better ADLs.

This study is the first to examine the mediating effect of ADL on the relationship between SP and SRH. Our results suggest that ADLs play a mediating role in the connection between SP patterns and SRH, with mediation effects of 34.23% (Diverse Participation) and 40.00% (Agricultural-centered Participation), respectively. Research suggests that older adults who engage more in SP are more likely to report better SRH due to enhanced ADLs. Based on the research findings, we propose the following comprehensive policy recommendations: First, the assessment of older adults’ ADLs should be integrated with fall prevention measures, with particular focus on solitary, widowed, and chronically ill older populations. Drawing upon Azizan et al.‘s research [67]on the application of sensor technology in fall prevention, community health service centers should introduce simple motion monitoring devices to enable dual screening of both ADLs and fall risk. Second, we recommend that civil affairs departments collaborate with communities and village committees to increase investment in older adults activity facilities. Diverse community activities (such as farming exchanges, cultural and recreational events) should be actively organized, and a participation reward point system should be established to enhance the SP motivation of older adults. Finally, financial departments should create special elderly care subsidy funds, with priority support for older adults without child care and with low incomes, to establish an appropriate older adults care security system and improve the social security system for older adults.

This study has several limitations. First of all, this study was based on cross-sectional data, so it cannot test the causal relationship between SP and SRH, nor can it strictly rule out bidirectional causality whereby poorer SRH lead to reduced SP. In addition, the data of this study comes from Gansu Province, which is not nationally representative for China. Subsequent studies could be extended to all of China using data from the China Health and Retirement Longitudinal Study (CHARLS). Finally, the survey was conducted after the COVID-19 pandemic ended and did not consider the impact of COVID-19 on the results.

Conclusion

This study aimed to investigate the relationship between SP patterns and SRH among older adults in Gansu Province, China, as well as the role of ADLs in this relationship. The results of the study showed that SC has a significant directly affect SRH, and the association between SC and SRH was achieved partly through ADLs this mediators, accounting for 34.23% (high participation) and 40.00% (agricultural participation) of the total effects respectively. As society ages, it is important to implement policies that enhance social services and support systems for older adults to improve their SP, ADLs, and SRH.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

Jiaxian Zhou, YiNan Yang, Xiaobin Hu and Jie Lu conceived and designed the project. Jiaxian Zhou, Ningwen Mao, Shuyi LI, Xi Chen, Dan Wang, Yanxia Zhang and Xiaoru Shi acquired and interpreted the data. Jiaxian Zhou and YiNan Yang analyzed data and wrote the manuscript. Jie Lu, Xiaobin Hu, Jianmiao Li, Xin Gao, Shengxin Tao and Xuhong Pu revised and gave some comments on the manuscript. All authors have read and approved the final manuscript.

Funding

This research was funded by the Gansu Provincial Health and Family Planning Statistical Information Center Project (20231011-(23)0667); Gansu Provincial Higher Education Industry Support Plan Project (2023CYZC-06);Study on total health expenditure accounting of 2023 in Gansu province ((23)0621).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethical Committee of the Department of Public Health, Lanzhou University, China. Written informed consent was obtained from each participant taking part in this study. We confirmed that all methods were performed in accordance with the relevant guidelines and regulations.

Competing interests

The authors declare no competing interests.