Home- and community-based services, informal support networks, and life satisfaction among older adults in China: moving beyond filial support

Abstract

Background

With rapid population ageing and shrinking household sizes in China, Home- and Community-Based Services (HCBS) have become a critical component of elderly care. However, the interplay between formal services and informal support networks—such as intergenerational family support and horizontal social ties—remains insufficiently explored. This study examines the impact of HCBS utilization on life satisfaction among older Chinese adults and analyzes how these effects are moderated by different types of support networks.

Methods

This study utilized data from the 2023 China Longitudinal Aging Social Survey (CLASS). Ordinary Least Squares (OLS) regression was employed to estimate the baseline association between HCBS utilization and older adults’ life satisfaction. To address potential selection bias and endogeneity, propensity score matching (PSM), inverse probability weighted regression adjustment (IPWRA), and an instrumental variable (IV) approach were applied. Furthermore, we examined the moderating effects of informal support (filial support networks and social support networks). Heterogeneity analyses were conducted across urban–rural residency, income levels, activities of daily living (ADL), and age cohorts.

Results

After rigorously addressing selection bias, we found no significant main effect of HCBS utilization on life satisfaction within the full sample. However,HCBS utilization significantly improves life satisfaction exclusively among rural older residents. Moderation analyses identified distinct mechanisms: the emotional network synergistically enhanced the HCBS effect (β = 0.101∼0.136, p < 0.01), whereas the instrumental network exhibited a substitution effect, attenuating the HCBS impact (β = − 0.084∼ − 0.055, p < 0.05). Independently, the instrumental network served as the strongest protective factor for ADL-impaired (β = 0.158, p < 0.05) and low-income (β = 0.137, p < 0.01) elders. A striking geographic divergence emerged: emotional networks benefited urban residents (β = 0.037, p < 0.05) but harmed rural ones (β = − 0.046, p < 0.01), while contact networks uniquely benefited rural areas (β = 0.041, p < 0.01).

Conclusion

Aging policies must move beyond traditional assumptions of filial support, as empirical evidence uncovers that broader informal ties—particularly instrumental networks—now exert a stronger relative impact on well-being than traditional family care. Furthermore, HCBS is not a one-size-fits-all solution; its benefits are geographically masked, primarily accruing to rural and functionally capable elders. The interplay between formal and informal care features emotional synergy and instrumental substitution. Policymakers must favor precision-tailored interventions that integrate formal HCBS with robust community networks.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-07781-4.

Introduction

Driven by rapid demographic shifts, China is experiencing an unprecedented aging trajectory, with the World Health Organization (WHO) projecting its population aged 60 and over to reach 28% by 2040 [1]. This transition is characterized by a unique “getting old before getting rich” phenomenon, placing profound strain on the nation’s healthcare and social security systems. In response, the Chinese government has strategically shifted eldercare resources to the grassroots level by promoting the “9073” framework (where 90% of older adults receive home-based care, 7% rely on community support, and 3% reside in institutions). Consequently, expanding community-based care has become a national priority, evidenced by massive investments in local facilities and a continued emphasis on service accessibility in the upcoming “15th Five-Year Plan” [2].

Despite the policy emphasis on community as a critical spatial carrier for active ageing, the actual effectiveness of Home- and Community-Based Services (HCBS) in meeting diverse needs remains contested. While the Chinese government encourages both public and private sectors to deliver HCBS as a preferred model for “ageing in place” [3] practical implementation faces numerous hurdles. These include issues of availability and accessibility [4], mismatches between service supply and demand [5], and a lack of supportive resources or cultural adaptation among older adults [6]. Evaluating how HCBS utilization influences life satisfaction is therefore crucial not only for assessing current policy efficacy but also for refining the community-based care system to better enhance older adults’ well-being.

Existing literature has extensively examined the determinants of life satisfaction among older adults, highlighting factors such as intergenerational support [7], grandchild caregiving [8], volunteer participation [9], supportive environments (socioeconomic and physical factors) [10] and macro-level public financial support [11]. Some studies have specifically focused on community-level service supply [12] and supply quantity [13], fewer studies have specifically focused on the association between HCBS and health-related quality of life among older adults in China [14]. Despite existing research, two critical gaps remain. First, studies often conflate service supply with actual utilization, overlooking real-world user behaviors. Second, the interplay between formal HCBS and diverse informal networks—specifically distinguishing intergenerational family ties from broader social connections—lacks thorough analysis. Thus, how these formal and informal systems interact to affect life satisfaction, and who benefits most, remains largely unexplored.

To address these gaps, this study utilizes data from the 2023 China Longitudinal Aging Social Survey (CLASS). Drawing upon Andersen’s Behavioral Model [15], in alignment with Social Network Theory [16], we develop an integrated analytical framework to examine the interplay between HCBS utilization and multidimensional social support networks. By distinguishing between formal service use and informal network support, this study contributes to the ongoing debate on HCBS, social support, and life satisfaction.

Literature review

HCBS and life satisfaction in older adults

In the context of an aging society, life satisfaction has emerged as a crucial dimension for evaluating the quality of life among older adults. Driven by the dual challenges of accelerated demographic aging and fiscal constraints, global care paradigms are undergoing a structural transition from institutional models to Home- and Community-Based Care. Evidence suggests that HCBS significantly enhances quality of life through multiple pathways, including the provision of daily living assistance, support for social participation, and the alleviation of psychological distress [17]. Globally, HCBS have adopted diversified models, accompanied by a significant increase in the utilization of social services [18], Individuals living in service-deprived facilities tend to exhibit poorer mental health and a diminished sense of purpose in life [19]. Supportive environments, including the utilization of HCBS, enhance life satisfaction among frail older adults [20]. While Community and Home-based Elderly Care Services significantly improve life satisfaction, their impact exhibits marked heterogeneity across different service types. Specifically, compared to foundational Life Care Services and Medical Care Services, higher-order interventions such as Spiritual and Cultural Services and Rights and Legal Services exert a more pronounced positive effect on life satisfaction [21].

Conversely, other studies suggest that the magnitude of this association is contingent upon older adults’ socioeconomic status [22], physical functional capacity [23], and the availability of financial support [24]. These inconsistencies underscore the necessity of examining HCBS within a broader ecosystem of social support and contextual factors. The Social Convoy Model posits that adult children assume an increasingly pivotal role in the social support networks of aging individuals [25]. Indeed, as adults face advancing age and functional decline, the reliance on a diverse array of informal social support networks becomes increasingly critical.

Types of support network and life satisfaction

Social support networks are defined as the assistance and resources available to individuals through social networks, such as information exchange and reciprocity [26]. The hierarchical compensatory model posits that older adults’ care preferences are determined by the priority of their relationships. Specifically, spouses are the primary choice for caregiving, followed by adult children, friends, neighbors, and formal organizations [27, 28]. Fei utilized the concept of the ‘differential mode of association’ to delineate this ego-centered social network structure. In this model, individuals closer to the center, such as family members and relatives, share the strongest ties. Conversely, as the distance from the center increases, relationships become progressively looser, extending to friends, fellow townsmen, and outsiders [29].

Informal support serves as a pivotal determinant of healthy aging, encompassing both intergenerational support and social network support. In particular, intergenerational support (often framed as filial support) constitutes the cornerstone of aging in East Asian societies, with the parent–child bond lying at its core. Evidence underscores the value of these dynamics; for instance, research has identified the positive impact of caring for grandchildren on the life satisfaction of older adults in India. Consequently, fostering supportive intergenerational relationships represents a vital strategy for enhancing well-being in later life [30]. The higher the levels of emotional, psychological, and spiritual well-being, the higher the life satisfaction of older persons 60 years and older [31]. However, the impact of intergenerational relationships varies significantly. Their effects on life satisfaction can be categorized into three patterns: universal enhancement (indicating a significant overall increase), type-specific enhancement (where only specific forms of intergenerational support offer benefits), and mixed effects (where some interactions promote well-being while others may diminish it) [32].

Beyond intergenerational relationships, social support networks such as friendships and neighborhood ties are receiving increasing attention, reflecting the principle of ‘reciprocity’. Although older adults’ networks frequently comprise more kin than non-kin [33], horizontal networks (such as friends) may exert a more direct influence on subjective well-being compared to vertical family intergenerational support. Consequently, it is essential to examine the impacts of these two types of informal support separately. Diversified or friend-centered networks not only provide companionship but also serve as vital avenues for accessing health information and service resources. Generally, higher levels of social support are associated with improved health status and life satisfaction [34]. Furthermore, in Western contexts, socialized care services provide professional support that alleviates the daily care-giving burden on adult children, thereby enhancing the quality of life for older adults [35].

Chinese context

In China, the tradition of ‘rearing children for old-age security’ is deeply embedded in family structures and cultural values. Consequently, upward intergenerational support rooted in filial piety—remains a primary determinant of life satisfaction among older adults. This traditional model is based on the principle of reciprocity, following a ‘feedback mechanism’ where parents nurture the young and children, in turn, provide for the aged across generations. However, this paradigm is facing significant challenges due to modernization, specifically within the context of urbanization and declining fertility rates (the ‘low-fertility trap’). As traditional support from children diminishes, older adults are increasingly compelled to seek socialized services to meet their daily care needs”. There is no evidence of a correlation between support of general filial responsibility norms and provision of help and care 10 years later. In comprehensive welfare states like Norway, care-giving is more a matter of individual choice than a social prescription [36].

Conversely, within the framework of the ‘Active Aging’ strategy, China has proposed the construction of a multi-tiered elderly care service system that is ‘based on the home, supported by the community, and supplemented by institutions.’ This initiative establishes families and communities as the key pillars of care responsibility. Between 2016 and 2020, 203 pilot cities (including districts and prefectures) were selected in five phases to implement this system. These pilots focused on enhancing both the ‘soft environment’ of service delivery and the ‘hardware’ of physical infrastructure. Consequently, older adults have increasingly accepted not only the concept but also the utilization of these services in their daily lives [37]. China’s elderly care system is undergoing a modernization transition from a family-centered model to a multi-tiered framework. This evolving system is characterized as ‘home-based, community-supported, and institutionally supplemented,’ with socialized care serving as a secondary pillar and the integration of medical and nursing care acting as a core component.”

While studies on home- and community-based elderly care and well-being are increasing, gaps remain in our understanding of life satisfaction. Specifically, there is a lack of integrated research simultaneously addressing service utilization, intergenerational ties, and social networks. Furthermore, the relative impact of vertical versus horizontal support systems in rapidly aging populations is rarely compared. Finally, the interplay between formal and info16rmal support, whether they act as complements or substitutes in enhancing subjective well-being, requires further empirical investigation, especially in the Chinese context.

Theoretical framework and hypotheses

Based on the Behavioral Model of Service Use [16], this research categorizes factors influencing life satisfaction into predisposing, enabling, and need dimensions. Furthermore, in alignment with Social Network Theory and the evolving landscape of Chinese aged care, two significant, non-institutional support factors were integrated: intergenerational support (from adult children) and broader social network support. A key distinction is drawn between these sources: intergenerational support functions internally within the family unit, whereas social network relationships bridge diverse family backgrounds to include wider social circles (Fig. 1).

Fig. 1.

Conceptual framework

Figure 1 conceptually extends Andersen’s Behavioral Model [38] of HCBS by embedding intergenerational ties and broader social support networks into the HCBS framework. Alongside three standard categories of variables (predisposing, enabling, and need factors) used to assess the direct impact of HCBS on older adults’ life satisfaction, the model further employs filial and social support networks as moderating variables to delineate their conditional effects.

H1: The utilization of HCBS has a significant positive impact on the life satisfaction of older adults.

H2: The relationship between the utilization of HCBS and the life satisfaction of older adults will be moderated by their filial support networks (i.e. economic support, care support, mental support).

H3: The relationship between the utilization of HCBS and the life satisfaction of older adults will be moderated by their social support networks (i.e. contact network, emotional network, instrumental network).

H4: Social support networks exert a stronger moderating impact than filial support networks on the relationship between HCBS utilization and older adults’ life satisfaction.

Research design

Data source

Data for this study were drawn from the 2023 wave of the China Longitudinal Aging Social Survey (CLASS). Organized by the National Survey Research Center at Renmin University of China, CLASS is a continuous, large-scale national social survey targeting adults aged 60 years and older. The survey employed a stratified multi-stage probability sampling design, covering 28 provinces, municipalities, and autonomous regions across China. Compared with other datasets [38], CLASS offers more comprehensive information regarding the demand for and access to healthy aging services, and provides a more complete and representative sample of the older population.

The 2023 CLASS data initially comprised 11,670 participants. After filtering out instances with incomplete data on essential variables, the final sample for cross-sectional analysis consisted of 10,834 older adults. The sample exclusion criteria were as follows: (1) individuals under 60 years of age; and (2) observations with missing data on offspring support and covariates. A flowchart of the sample selection process is available in the Supplementary Material (Figure S1).

Additionally, the CLASS 2023 survey was ethically approved by the Academic Research Ethics Committee at Renmin University of China. Before collecting data, informed consent was secured from each participant. This study utilizes anonymized public-use data, and all methods adhered to the ethical guidelines applicable to secondary data analysis.

Measures

The dependent variable is the life satisfaction of older adults, serving as a vital indicator of their subjective perspective on their quality of life. The study measures this indicator using the specific question from the CLASS2023 questionnaire: “Overall, are you satisfied with your current life?”, Responses are categorized into five levels: “Very dissatisfied, somewhat dissatisfied, neutral, somewhat satisfied, and very satisfied,” with assigned values of 1, 2, 3, 4, and 5, respectively.

The main explanatory variable in this text is the utilization HCBS by older adults. The CLASS2023 data survey includes nine community home-based care services: home visits, elderly hotlines, medical accompaniment, assistance with daily shopping, legal aid, household chores, elderly meal tables or meal delivery, daycare centers or elderly care institutions, and psychological counseling. For each service, the interviewer asks the surveyed older adults, “Have you ever used this service?” If an older adult answers “yes” to one or more services, it is assigned a value of 1; otherwise, it is assigned a value of 0.The utilization of home- and community-based services (HCBS), was operationalized as a dichotomous variable (1 = use of any HCBS; 0 = non-use).

Informal support networks serve as moderator variables, as follows: Filial support networks: It includes three types: economic support, daily living support, and emotional support. These are measured based on the CLASS2023 survey questions: “In the past 12 months, have your children given you (or your surviving spouse living with you) money, food, or gifts? What is the total value of these items?” “In the past 12 months, how often have your children helped with household chores?” “In the past 12 months, how often have your children contacted you (including using phone, WeChat, or other means of communication)?”, The levels of support from each child are aggregated and averaged to compute the mean score, with higher scores indicating that the elderly receive more support from their children. Social support networks: to assess social support networks, the ‘friend network’ component from the Lubben Social Network Scale was utilized, while the ‘relative network’ part was omitted to avoid collinearity with the independent variable. Drawing upon in previous literature [33], the social support networks were divided into three dimensions: contact networks, emotional networks, and instrumental networks. Consistent with standard LSNS-6 friends subscale, participants rated their network size for each question in one of the 3 questions, scored from 0 to 5 (0 = none, 1 = one, 2 = two, 3 = three to four, 4 = five to eight, and 5 = nine or more). A higher score indicates a larger network, respectively.

The study incorporated the following covariates: (1) Predisposing factors: age, gender (1 = male), marital status (1 = married), educational (1 = primary level, 2 = middle level, 3 = high level), hukou(1 = urban), region (1 = eastern, 2 = central, 3 = western); (2) Enabling factors: logarithm of per capita household income (tail trimming and imputation of missing values using city-level means); (3) Need factors: score of activities of daily living (ADL) and self-rated health (5-point scale from 1 to 5,a higher score indicates better self-rated health level).

Statistical analyses

The study utilized Stata 17 for data analysis.Following descriptive summaries, baseline associations were estimated using Ordinary Least Squares (OLS) regression. To address potential endogeneity and selection bias, we address propensity score matching (PSM) to model specifications,we further employed inverse-probability-weighted regression adjustment (IPWRA) for robustness testing. As a doubly robust estimator, IPWRA simultaneously incorporates both a treatment model and an outcome mode [39]. Its primary strength lies in yielding consistent causal inferences as long as either of the two models is correctly specified, thereby effectively minimizing biases associated with structural misspecification [40]. Additionally, an instrumental variable (IV) approach, utilizing the community-level supply of HCBS as the instrument, was implemented to account for unobserved confounding and reverse causality.

Results

Descriptive statistics

Table 1 presents the basic information of this study. The older adults had a mean age of 71.3 years, with 51.6% being male and 84.1% married. Most participants possessed a primary education (57.3%). Health indicators showed moderate-to-high levels, with mean scores of 3.84 for life satisfaction, 3.57 for self-rated health, and 9.62 for ADL. In terms of support networks, older adults received the highest level of economic support from filial networks, while maintaining active engagement in social networks, particularly in contact networks.

Table 1.

Descriptive statistics of variables

Variables	N	Mean/%	Standard deviation	Minimum	Maximum
Life satisfaction	10,834	3.841	.754	1	5
HCBS	10,834	0.099	.299	0	1
Male	10,834	.516	.5	0	1
Age	10,834	71.28	5.88	60	98
Age2/100	10,834	51.15	8.63	36	96.04
Married	10,834	.841	.366	0	1
Hukou	10,834	.426	.494	0	1
Urban areas	10,834	.567	.496	0	1
Education
Primary level	6,205	57.27
Middle level	3,076	28.39
High level	1,563	14.33
ADL	10,834	9.617	1.81	9	25
Srh	10,834	3.572	.79	1	5
Income(logged)	10,834	9.564	1.12	6.908	11.264
Region
Eastern	4,268	39.39
Central	3,869	35.71
Western	2,697	24.89
Filial support networks
Economic support	10,834	4.476	1.94	1	10
Care support	10,834	2.942	1.34	1	5
Mental support	10,834	3.571	1.025	1	5
Social support networks
Contact network	10,834	2.540	1.151	0	5
Emotional network	10,834	2.204	1.080	0	5
Instrumental network	10,834	2.151	1.160	0	5

Ordinary Least Squares (OLS)

Using life satisfaction among the elderly as the dependent variable, and the utilization of HCBS as the explanatory variable, an Ordinary Least Squares (OLS) regression was conducted to explore its impact on the life satisfaction of elderly individuals in China. The baseline regression analysis results are presented in Table 2. Following the regression analysis, the Variance Inflation Factor (VIF) was assessed to check for multicollinearity in the linear regression model. The results indicated a VIF value of 1.45, which is well below the threshold of 10, suggesting that there is low correlation among the independent variables in the model, thus minimizing any significant impact on the regression analysis.

Table 2.

Model of the impact of HCBS on life satisfaction among older adults

Variables	Model 1	Model 2	Model 3	Model 4	Model 5
	Filial support	Social support	HCBS	Two supports	Full model
Male	−0.009	−0.010	−0.010	−0.010	−0.010
	(0.013)	(0.013)	(0.013)	(0.013)	(0.013)
Age	−0.053**	−0.051**	−0.010	−0.049**	−0.010
	(0.018)	(0.018)	(0.013)	(0.018)	(0.013)
Age2/100	0.040***	0.039***	−0.055**	0.038**	−0.049**
	(0.012)	(0.012)	(0.018)	(0.012)	(0.018)
Education (Ref: Primary level)			0.042***		0.038**
Middle level	0.038*	0.040*	0.040*	0.038*	0.038*
	(0.017)	(0.017)	(0.017)	(0.017)	(0.017)
High level	0.084***	0.085***	0.087***	0.082***	0.082***
	(0.022)	(0.022)	(0.022)	(0.022)	(0.022)
Married	0.104***	0.109***	0.109***	0.105***	0.105***
	(0.021)	(0.021)	(0.021)	(0.021)	(0.021)
Hukou	0.064***	0.064***	0.065***	0.064***	0.064***
	(0.016)	(0.016)	(0.016)	(0.016)	(0.016)
ADL	0.009	0.009*	0.009	0.010*	0.010*
	(0.005)	(0.005)	(0.005)	(0.005)	(0.005)
Srh	0.385***	0.386***	0.385***	0.386***	0.386***
	(0.011)	(0.011)	(0.011)	(0.011)	(0.011)
Lnincome	0.026***	0.028***	0.027***	0.027***	0.027***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Region (ref: Eastern)
Central	−0.073***	−0.082***	−0.079***	−0.080***	−0.080***
	(0.018)	(0.017)	(0.017)	(0.018)	(0.018)
Western	−0.048*	−0.063***	−0.059**	−0.058**	−0.058**
	(0.020)	(0.019)	(0.019)	(0.020)	(0.020)
Filial support networks
Economic support	0.012***			0.010*	0.010*
	(0.004)			(0.004)	(0.004)
Care support	0.003			−0.001	−0.001
	(0.006)			(0.006)	(0.006)
Mental support	−0.009			−0.010	−0.010
	(0.008)			(0.008)	(0.008)
Social support networks
Contact network		0.017*		0.015	0.015
		(0.008)		(0.008)	(0.008)
Emotional network		−0.006		−0.004	−0.004
		(0.009)		(0.009)	(0.009)
Instrumental network		0.032***		0.032***	0.032***
		(0.008)		(0.008)	(0.008)
HCBS			0.006		0.006
			(0.022)		(0.022)
_cons	3.698***	3.567***	3.814***	3.485***	3.484***
	(0.655)	(0.650)	(0.650)	(0.654)	(0.654)
N	10,834
Adj R2	0.1926	0.1954	0.1916	0.1962	0.1962

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

Table 2 presents the regression results. First, no statistically significant association was found between HCBS and the life satisfaction of older adults (β = 0.006, SE = 0.022, p > 0.05). The results failed to support H1. Second, the two types of support networks demonstrated statistically significant associations across distinct dimensions. Specifically, in Model 1, economic support from filial support was positively associated with older adults’ life satisfaction (β = 0.012, p < 0.001). Model 2 reveals that the instrumental network showed the strongest positive association with life satisfaction (β = 0.032, SE = 0.008, p < 0.001), followed by a significant but comparatively smaller effect from the contact network (β = 0.017, SE = 0.008, p < 0.05). A significant U-shaped relationship emerged between age and life satisfaction. Satisfaction declines until a calculated turning point of 64.11 years, followed by an upward trajectory in later life. This trajectory aligns with the established ‘U-shaped curve of well-being’ in aging research [41].

To evaluate the relative importance of various support dimensions in predicting life satisfaction, we standardized the key predictors within the model and visualized the resulting beta coefficients using the coefplot command (Fig. 2).

Fig. 2.

Forest plot of standardized beta coefficients for different support types

Figure 2 clearly demonstrates a divergence in the efficacy of different support dimensions. The relative importance of support dimensions varies significantly. Instrumental support (β = 0.037) acts as the primary driver of satisfaction, followed by economic support (β = 0.019) and contact networks (β = 0.017). Conversely, mental, emotional, and care support yield no statistically significant independent effects, suggesting that social support play a more dominant role, reflecting the transition toward socialized eldercare. Therefore, H4 was supported.

Robustness checks

To ensure the robustness of our findings, several robustness checks were conducted. The results are presented in Table 3.

Table 3.

Robustness check

Variables	Model 1	Model 2	Model 3	Model 4	Mode5
	Changing the estimation model		Replacing the independent variable		Limit the sample population
Male	−0.025	−0.040	−0.010	−0.025	−0.012
	(0.039)	(0.048)	(0.013)	(0.039)	(0.014)
Age	−0.118*	−0.160*	−0.048**	−0.115*	−0.119**
	(0.054)	(0.067)	(0.018)	(0.053)	(0.040)
Age2/100	0.092*	0.119**	0.037**	0.090*	0.089**
	(0.037)	(0.046)	(0.012)	(0.036)	(0.029)
Education (Ref: Primary level)
Middle level	0.124*	0.224***	0.037*	0.121*	0.040*
	(0.048)	(0.059)	(0.017)	(0.049)	(0.017)
High level	0.269***	0.539***	0.077***	0.258***	0.082***
	(0.066)	(0.093)	(0.022)	(0.064)	(0.023)
Married	0.285***	0.305***	0.105***	0.287***	0.113***
	(0.056)	(0.065)	(0.021)	(0.059)	(0.022)
Hukou	0.193***	0.345***	0.063***	0.192***	0.062***
	(0.045)	(0.055)	(0.016)	(0.047)	(0.016)
ADL	0.033**	0.039**	0.010*	0.034**	0.002
	(0.012)	(0.013)	(0.005)	(0.013)	(0.005)
Srh	1.154***	1.045***	0.387***	1.157***	0.390***
	(0.029)	(0.034)	(0.011)	(0.034)	(0.009)
Lnincome	0.071***	0.094***	0.025***	0.066***	0.025***
	(0.019)	(0.022)	(0.006)	(0.019)	(0.007)
Region (ref: Eastern)
Central	−0.263***	−0.295***	−0.078***	−0.259***	−0.082***
	(0.050)	(0.061)	(0.018)	(0.052)	(0.018)
Western	−0.203***	−0.045	−0.054**	−0.195**	−0.042*
	(0.057)	(0.069)	(0.020)	(0.059)	(0.020)
Filial support networks
Economic support	0.030**	0.062***	0.010**	0.031**	0.011**
	(0.011)	(0.013)	(0.004)	(0.011)	(0.004)
Care support	−0.003	−0.040	−0.001	−0.004	−0.003
	(0.018)	(0.022)	(0.006)	(0.018)	(0.007)
Mental support	−0.031	0.043	−0.010	−0.031	−0.009
	(0.023)	(0.028)	(0.008)	(0.023)	(0.008)
Social support networks
Contact network	0.028	0.027	0.016*	0.030	0.011
	(0.022)	(0.027)	(0.008)	(0.022)	(0.008)
Emotional network	−0.024	−0.043	−0.005	−0.026	−0.003
	(0.026)	(0.031)	(0.009)	(0.026)	(0.009)
Instrumental network	0.104***	0.094***	0.031***	0.103***	0.037***
	(0.024)	(0.028)	(0.008)	(0.025)	(0.008)
HCBS	0.018	0.150			0.013
	(0.065)	(0.084)			(0.024)
HCBS number			0.018**	0.044**
			(0.006)	(0.017)
_cons		0.582	3.447***		5.909***
		(2.507)	(0.655)		(1.412)
cut1	−3.763			−3.671
	(2.014)			(1.961)
cut2	−1.624			−1.531
	(2.010)			(1.963)
cut3	0.838			0.933
	(2.009)			(1.964)
cut4	3.868			3.964*
	(2.010)			(1.964)
N	10,834				9747
Adj R2/Pseudo R2	0.0978	0.1376	0.1967	0.0980	0.1989

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

Changing the estimation model

The continuous variable life satisfaction was evaluated using ordered logistic regression for its original ordinal scale, and binary logistic regression for its dichotomized form (coded as 1 for scores 4–5, and 0 for scores 1–3) with the results presented in column 1–2 of Table 3. As shown in the regression results, the regression results revealed no statistically significant association between HCBS and life satisfaction.

Replacing the independent variable

To address the potential limitation of using a binary measurement for HCBS, we replaced the binary variable (HCBS) with a continuous count variable representing the total number of HCBS (HCBS number, ranging from 0 to 9), the results presented in column3(ols) and column4(ologit) of Table 3. While the binary indicator was not statistically significant, the robustness check revealed that the number of HCBS used is positively and significantly associated with older adults’ life satisfaction. This indicates a clear dose–response relationship: the life satisfaction of older adults is driven by the intensity of service utilization rather than mere service initiation.

Excluding sample group

In advanced age groups, particularly those over 80 years old, the reliance on HCBS and social support tends to increase due to the decline in physical abilities. This heightened dependency may impede the accurate evaluation of the effectiveness of these services. To mitigate potential bias in the analysis, individuals aged 80 and above were excluded from the sample, as presented in the five column of Table 3. For older adults under the age of 80, the use of HCBS has no significantly enhances their life satisfaction. This result demonstrates that the research findings are robust and not driven by outlier samples.

Endogenous analysis

To address selection bias, PSM and IPWRA are used to conduct balance tests on the sample. SM and IPWRA were applied. Given that the pooled sample results were non-significant and likely obscured China’s substantial urban–rural disparities, we performed subgroup analyses for urban and rural older adults separately (Table 4).

Table 4.

Estimation results using PSM and IPWRA

		Panel A									Panel B: IPWRA
		Radius caliper			K nearest-neighbor			kernel
	N	ATT	S.E	T	ATT	S.E	T	ATT	S.E	T	ATET	Robust S.E	Z
Full sample	10,834	0.028	0.026	0.31	0.012	0.030	0.39	0.007	0.025	0.29	−0.012	0.024	−0.50
Urban areas	6140	0.009	0.033	0.28	−0.017	0.037	−0.45	−0.027	0.033	−0.79	−0.052*	0.031	−1.68
Rural areas	4694	0.126***	0.042	2.98	0.104**	0.050	2.10	0.150***	0.042	3.58	0.100**	0.042	2.40

K-nearest neighbor matching, K = 4, and in radius matching, the radius is 0.01

^* p < 0.05, ** p < 0.01, *** p < 0.001

Table 4 presents the results of the sensitivity analysis. To ensure the reliability of our findings and account for observable selection bias, we employed three Propensity Score Matching (PSM) algorithms (radius, K-nearest neighbor, and kernel matching) alongside the doubly robust Inverse Probability Weighted Regression Adjustment (IPWRA).

A striking rural–urban heterogeneity persists across all matching methods. In rural areas, the positive effects remain highly robust; the Average Treatment Effect on the Treated (ATT/ATET) ranges from 0.100 to 0.150.This firmly confirms that the utilization of home- and community-based services provides substantial benefits to rural older adults. Conversely, the estimated effects for the full sample and the urban subsample are statistically insignificant across almost all models. Overall, estimating the models with alternative methods does not alter our core conclusions, indicating high robustness.

To establish the validity of the causal estimates presented in Table 4, we rigorously evaluated the underlying assumptions for both the PSM and IPWRA models across the full, urban, and rural samples. For the PSM analyses, successful matching was visually and statistically confirmed: standardized bias plots (Supplementary Figs. 2–4) and sample balance tests (Supplementary Tables S2–S4) indicated substantial reductions in covariate imbalance. Correspondingly, for the IPWRA models, the overlap and common support assumptions were fully satisfied (Supplementary Figs. 5–7), supported by comprehensive covariate balance checks (Supplementary Tables S5–S7). All detailed diagnostic results are provided in the supplementary material.

Secondly, to address reverse causality, the two-stage least squares (2SLS) method employing instrumental variables is utilized (Table 5).

Table 5.

Estimation results of the instrumental variables method

	Full sample		Urban area		Rural area
	First stage	Second stage	First stage	Second stage	First stage	Second stage
Com_provision	0.009***		0.004**		0.017***
	(6.01)		(2.05)		(6.08)
HCBS		0.282		−1.173		1.644***
		(0.74)		(−1.01)		(3.21)
_cons	−0.143**	1.475***	−0.073	1.452***	−0.169**	1.632***
	(−2.56)	(10.16)	(−0.93)	(6.97)	(−2.18)	(6.30)
N	10,834		6140		4694
Control variables	YES
KP rk Wald F-stat		36.08		4.22		36.99

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

Considering the potential reverse causality between older adults’ utilization of HCBS services and their life satisfaction—that is, older adults with higher life satisfaction may be more likely to use HCBS services—such bidirectional causality, along with omitted variables, may give rise to endogeneity concerns. To address potential endogeneity, we employed the community-level provision of HCBS as an instrumental variable (Com-provision). Specifically, the original survey asked whether the community offered nine distinct categories of HCBS (1 = yes,2 = no,3 = don’t know). We binarized the “yes” responses and aggregated them to calculate a community-level mean score. The validity of utilizing community-level supply as an instrument for individual service utilization has been well-documented in the Chinese context [42]. Furthermore, the low residential mobility of Chinese older adults—deeply rooted in the traditional culture of ‘aging in place’—ensures the strict exogeneity of their community environment. The detailed results are reported in Table 5.

Table 5 presents the results of the instrumental variable two-stage least squares (IV-2SLS) regressions for the full sample, as well as the urban and rural subsamples. In the first-stage models, the instrumental variable (com_provision) was highly significant for both the full and rural samples. Notably, the Kleibergen-Paap (KP) rk Wald F-statistics substantially exceeded the conventional rule-of-thumb threshold of 10, thereby alleviating concerns regarding weak instruments. Conversely, the F-statistic for the urban subsample was markedly lower at 4.22. This statistical discrepancy likely reflects deep-seated structural disparities in the allocation of community infrastructure between urban and rural areas in China.

Because the first-stage F-statistic in the urban sample did not meet the conventional threshold of 10, we applied the Anderson-Rubin (AR) test to address weak instrument concerns. The AR test results (χ2 = 1.25, p = 0.263) indicate that, after adjusting for the weak instrument problem, the impact of HCBS on life satisfaction is still non-significant among urban residents.

Analysis of moderating effect

The impact of HCBS utilization on life satisfaction exhibits evident rural–urban heterogeneity. It yields a significant positive effect solely among rural older adults, whereas no significant associations are found in the overall or urban samples. This raises the question of whether the effect of HCBS on life satisfaction varies by different types of support. To examine this, interaction terms are constructed between HCBS utilization and two forms of support. To mitigate selection bias and ensure robust causal inference, all subsequent moderation and subgroup analyses were conducted using the doubly robust IPWRA approach. Prior to generating the interaction terms, continuous moderators were mean-centered to ensure estimation stability. To avoid omitted variable bias and account for potential correlations among different dimensions of family support, we included all six interaction terms simultaneously in three models (Table 6).

Table 6.

The moderating role of filial support and social network support

	Model 1	Model 2	Model 3
	Full sample	Urban areas	Rural areas
HCBS	0.009	−0.043	0.177***
	(0.022)	(0.027)	(0.045)
Economic support	0.003	0.001	0.000
	(0.004)	(0.005)	(0.009)
HCBS × Economic support	−0.005	−0.006	0.021
	(0.014)	(0.016)	(0.026)
Care support	−0.004	−0.003	−0.014
	(0.009)	(0.011)	(0.014)
HCBS × Care support	−0.007	−0.022	0.037
	(0.021)	(0.027)	(0.033)
Mental support	−0.012	−0.007	−0.027
	(0.011)	(0.015)	(0.017)
HCBS × Mental support	−0.025	−0.032	0.007
	(0.027)	(0.036)	(0.042)
Contact network	0.020*	0.012	0.037**
	(0.011)	(0.014)	(0.018)
HCBS × Contact network	−0.026	−0.008	−0.044
	(0.023)	(0.028)	(0.040)
Emotional network	−0.004	0.023	−0.058***
	(0.012)	(0.016)	(0.019)
HCBS × Emotional network	0.115***	0.101***	0.136***
	(0.031)	(0.039)	(0.050)
Instrumental network	0.019	−0.000	0.066***
	(0.013)	(0.016)	(0.020)
HCBS × Instrumental network	−0.061**	−0.055*	−0.084*
	(0.027)	(0.033)	(0.045)
Controlled	Yes
N	10,834	6140	4694
Adj R2	0.2114	0.2382	0.1942

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

As shown in Table 6, the interactions between HCBS and filial support indicators (economic, care, and mental support) were not statistically significant. However, emotional network demonstrated a significant positive moderation across models (β = 0.101∼0.136, p < 0.01), suggesting a synergistic effect. Conversely, instrumental network negatively moderated the HCBS effect (β = − 0.084∼ − 0.055, p < 0.05), indicating an attenuating effect that remained consistent in both urban and rural strata. These findings provide support for H2 and H3. These significant interactions are visually depicted in the Supplementary Material (Figure S8-S10).

Analysis of heterogeneity

To further investigate potential heterogeneity in the effects of HCBS, we conducted subgroup analyses stratified by socioeconomic status such as household registration type (urban, rural), personal income (high = at or above the median, low = below the median), and functional vulnerability such as ADL scores (assessed via a standard 9-item scale ranging from 9 to 27, where higher scores indicate worse functional capacity; participants with a baseline score of 9, indicating no need for assistance, were coded as 1, and those scoring ≥ 10 were coded as 0), age group(based on the non-linear threshold at age 64 from Table 2 and the WHO age classification criteria [43]), the sample was divided into three age groups using cut-offs of 65 and 74 years). Detailed multivariable regression models, including all interaction terms, are presented in Table 7. Additionally, to clearly depict the differential impact across subgroups, the interaction between HCBS and two supports is visualized using a forest plot (Supplementary Fig. S11).

Table 7.

The heterogeneity analysis by subgroups

	Hukou		Income(log)		ADL(Need for assistance)		Age group
	Urban	Rural	Low	High	No	Yes	60–64	65–74	> 75
HCBS	−0.040	0.118***	0.041	−0.004	0.049**	−0.052	−0.022	0.030	−0.017
	(0.030)	(0.035)	(0.042)	(0.029)	(0.024)	(0.052)	(0.056)	(0.031)	(0.056)
Economical support	−0.001	0.004	0.022***	−0.006	0.006	−0.010	0.014	0.024***	0.020**
	(0.006)	(0.008)	(0.008)	(0.006)	(0.005)	(0.012)	(0.011)	(0.006)	(0.010)
	(0.018)	(0.022)	(0.013)	(0.011)	(0.009)	(0.025)	(0.025)	(0.012)	(0.015)
Care support	−0.004	−0.012	−0.003	−0.021	−0.016	0.003	−0.003	−0.031*	−0.027
	(0.013)	(0.012)	(0.016)	(0.014)	(0.011)	(0.029)	(0.031)	(0.016)	(0.023)
Mental support	−0.013	−0.016	−0.036**	0.007	−0.018	0.032	−0.092**	−0.008	0.030
	(0.017)	(0.014)	(0.018)	(0.017)	(0.013)	(0.028)	(0.042)	(0.016)	(0.028)
	(0.040)	(0.035)	(0.017)	(0.017)	(0.014)	(0.029)	(0.038)	(0.017)	(0.028)
Contact network	0.002	0.041***	−0.004	−0.010	0.003	−0.018	0.044	−0.013	−0.012
	(0.015)	(0.016)	(0.024)	(0.016)	(0.015)	(0.031)	(0.027)	(0.019)	(0.037)
Emotional network	0.037**	−0.046***	0.005	−0.044	0.000	−0.076	0.031	−0.000	−0.048
	(0.018)	(0.017)	(0.042)	(0.035)	(0.029)	(0.068)	(0.067)	(0.038)	(0.060)
Instrumental network	−0.003	0.046**	0.137***	0.109***	0.082**	0.158**	0.054	0.178***	0.140*
	(0.018)	(0.018)	(0.051)	(0.038)	(0.035)	(0.065)	(0.075)	(0.043)	(0.074)
N	4611	6223	5328	5506	8367	2467	1135	7098	2601
Controlled	Yes
Adj R2	0.2534	0.1898	0.2190	0.2118	0.1775	0.2851	0.2047	0.2283	0.2243

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001

Table 7 presents the heterogeneity of the main effects of HCBS on life satisfaction across different subgroups, categorized by hukou (urban, rural), income level, ADL status (need for assistance), and age group.

Table 7 details the subgroup analyses evaluating the heterogeneous associations between HCBS utilization and life satisfaction. Stratification by residential status revealed that HCBS utilization was positively associated with life satisfaction among rural older adults (β = 0.118, p < 0.01), whereas no significant association was observed among urban residents. When assessed by functional capacity, the positive effect of HCBS remained significant only for older adults without limitations in activities of daily living (ADL) (β = 0.049, p < 0.05).Conversely, this effect was attenuated and non-significant for those requiring ADL assistance. Furthermore, the analysis did not detect statistically significant heterogeneity across income levels or age cohorts (60–64, 65–74, and > 75 years), suggesting a relatively consistent impact across these demographic strata.

Furthermore, instrumental network was the most robust protective factor, especially for ADL-impaired (β = 0.158, p < 0.05) and low-income (β = 0.137, p < 0.01) elders. Economical support selectively benefited low-income (β = 0.022, p < 0.01) and older (≥ 65 years) cohorts. Geographically, emotional network positively impacted urban residents (β = 0.037, p < 0.05) but negatively affected rural ones (β = − 0.046, p < 0.01), whereas contact network only benefited rural areas (β = 0.041, p < 0.01).Finally, some supports showed adverse effects: mental support negatively impacted low-income (β = − 0.036, p < 0.05) and youngest-old (60–64 years:β = − 0.092, p < 0.05) groups, while care support marginally decreased satisfaction for those aged 65–74 (β = − 0.031, p < 0.1).

Discussion

Based on the CLASS 2023 data, this study comprehensively examines the effects of HCBS satisfaction. In addition, we examined how HCBS use interacts with different types of support networks among older adults, and how these interactions influence their life satisfaction. Our findings provide a novel perspective on how older adults in China enhance their health and well-being within the context of modern kinship structures, by navigating the relationships between different types of informal support networks and formal support services. The following are several key findings derived from our analysis:

Firstly, HCBS utilization showed no significant main effect on life satisfaction in the full sample. However, subsequent analyses revealed a robust pattern: it significantly improves satisfaction exclusively among rural older adults.This disparity likely stems from differences in baseline care resources. Urban older adults generally have access to a broader array of formal and informal support, higher pensions, and more developed commercial care markets, which may diminish the added benefit of basic HCBS. Conversely, rural areas are experiencing a severe decline in traditional family care due to the massive out-migration of younger populations. In this resource-depleted context, government-subsidized HCBS serves as a vital safety net. Furthermore, our interaction analyses reveal a nuanced interplay of synergy and substitution. Traditional filial support showed no significant interaction with HCBS, operating through independent, parallel pathways. Conversely, emotional networks positively moderated HCBS. This synergy suggests that by alleviating daily practical burdens, HCBS enables older adults to derive greater psychological fulfillment from emotional ties. However, instrumental networks negatively moderated HCBS across both urban and rural strata. This robust substitution effect indicates that redundant practical assistance yields diminishing returns and risks overcare, inadvertently undermining functional autonomy and underscoring the critical need for targeted, needs-based service allocation.

Secondly, the heterogeneity findings provide further insights. Geographically, HCBS significantly improved life satisfaction only among rural older adults. This divergence aligns with the resource substitution hypothesis: urban elders possess diverse formal and informal care alternatives, minimizing the marginal utility of basic HCBS. Conversely, severe youth out-migration has left rural areas with profound family care deficits, making publicly provided HCBS a critical compensatory lifeline rather than a mere supplement. There is no evidence of an association between support for general filial responsibility norms and helping or caregiving behaviors ten years later in Norwegian. In contrast, perceived relationship quality is significantly associated with subsequent helping and caregiving behaviors [44].

Finally, although upward intergenerational support remains important, our findings indicate that social network support may exert a stronger influence on life satisfaction. Compared with vertical family support, horizontal social ties—such as friends and neighbors —may contribute more directly to emotional well-being and social integration. This may be because children’s excessive involvement in intergenerational relationships can generate relational strain and emotional distancing, thereby prompting older adults to seek external support. Broader social engagement may help older adults maintain autonomy, reduce loneliness, and reinforce their sense of belonging, which are key components of life satisfaction. Diversified or friend-centered social networks not only provide a broader range of informal support but also offer greater access to information and knowledge regarding service utilization [45].

One notable finding is that while the coefficient for HCBS in the full-sample regression is positive, it remains statistically insignificant. This non-significance is primarily driven by a dilution effect from the urban sample. It underscores that in resource-constrained rural settings, the actual uptake of public services exerts a profound “compensatory effect” for vulnerable populations. Consequently, the insignificant “average effect” observed in the overall sample effectively masks an urgent, high-yield opportunity for targeted policy interventions in rural areas.In addition, a weak complementary relationship is observed between informal care and nursing services, and this relationship is not significantly influenced by the severity of disability among older adults [46]. These findings highlight the importance of taking older adults’ overall support environment into comprehensive consideration when seeking to enhance their health and well-being.

Conclusions

This study examined the association between HCBS utilization and life satisfaction among older Chinese adults, while accounting for intergenerational and social network support. HCBS utilization demonstrated no significant main effect on the overall life satisfaction of older adults. However, utilizing PSM and IPWRA we uncovered significant “efficacy heterogeneity and resource mismatch” across different subpopulations. Notably, rural areas have emerged as the core arena where HCBS exerts a crucial “safety-net” function. The impact of broader social support networks substantially surpassed that of traditional filial support. Specifically, emotional networks synergistically enhanced the HCBS effect, whereas instrumental networks exhibited a substitution effect.

Policy implications

These findings offer policy implications that extend beyond China and contribute to broader debates on global ageing governance and long-term care reform in rapidly ageing societies. First, policymakers should abandon the “one-size-fits-all” approach. Future funding for HCBS should be disproportionately directed to rural areas to strengthen their foundational “safety-net” function. In urban settings, policies should halt inefficient, generalized expansion and instead pivot to highly specialized, skilled care (e.g., medical nursing, rehabilitation) specifically targeting older adults with severe functional impairments (ADL decline).

Second, Given the significant substitution effect of instrumental networks, local governments must implement rigorous pre-service assessments. HCBS should only provide instrumental assistance (e.g., cleaning, meal delivery) when informal support is absent. Duplicating practical help wastes public resources and risks “overcare,” which inadvertently accelerates the loss of functional autonomy in older adults.

Third, community interventions should be designed to complement, rather than replace, family functions. Since emotional networks synergistically enhance HCBS, community services should focus on alleviating the heavy, practical burdens of daily caregiving. By taking over exhaustive physical tasks, HCBS can “free up” families and informal networks to focus on providing high-quality emotional and psychological support, thereby maximizing the older adults’ subjective well-being.

Limitations and future directions

This study has several strengths. It uses nationally representative survey data and adopts a multi-dimensional framework that simultaneously considers formal services, intergenerational support, and social network support. The analysis also explores heterogeneity and moderation effects, providing a more nuanced understanding of elderly well-being. However, several limitations should be noted. Firstly, although methods such as PSM, IPWRA and IV were employed, the cross-sectional design still cannot fully eliminate biases arising from omitted variables, reverse causality, and measurement error. Secondly, HCBS utilization was measured based on self-reported information, which may introduce reporting bias. Thirdly, unobserved factors such as service quality and local policy variation were not fully captured. Future longitudinal research is needed to examine causal mechanisms and long-term effects of HCBS use.

Abbreviations

HCBS

Utilization of home- and community-based services

Srh

Self-rated health

ADL

Activities of Daily Living

Com-provision

Community supplies nine categories of home- and community-based services

Authors’ contributions

Conceptualization, JX and JC; methodology, JX and LY; software, JX; formal analysis – review & editing, JC; writing—original draft preparation, JX; writing—review and editing, JC and LY; supervision, LY. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the Guangdong Philosophy and Social Sciences Planning (Grant: GD25YTY04) and Young Doctoral Research Project of Guangzhou Sport University (Grant: 5250180628).

Data availability

The datasets analyzed for this study can be found in the China Longitudinal Aging Social Survey [http://class.ruc.edu.cn/index.htm].

Declarations

Ethics approval and consent to participate

This study utilized data from the Chinese Longitudinal Aging Social Survey (CLASS), which was approved by the Ethical Review Committee of Renmin University of China (RER201902). All participants in the original survey provided informed consent. This study, based on publicly available data, did not involve new human subject research and thus required no additional ethics approval. The research was conducted in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.