Unveiling intervention targets: urban-rural disparities in depression-anxiety symptoms networks among Chinese older adults living alone

Abstract

Background

Living alone is a significant risk factor for depression and anxiety among older adults in China. However, geographic disparities in the symptom-level network structure of depression and anxiety remain unexamined. This study aims to investigate urban–rural differences in the depression-anxiety network to identify central and bridge symptoms as potential intervention targets.

Methods

We analyzed data from 2,477 older adults aged ≥ 65 years living alone (urban: n = 413; rural: n = 2,064), drawn from the 2017–2018 Chinese Longitudinal Healthy Longevity Survey. Depressive and anxiety symptoms were assessed utilizing the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) and the 7-item Generalized Anxiety Disorder Scale (GAD-7), respectively. We constructed a network structure of depressive and anxiety symptoms and calculated expected influence (EI) and bridge expected influence (BEI) indices to identify central and bridge symptoms. Furthermore, a comparative analysis of the network structure across urban and rural groups was conducted.

Results

CESD3 (Feeling blue/depressed), GAD2 (Uncontrollable worry), and GAD4 (Trouble relaxing) were the most central symptoms in the network. CESD10 (Sleep disturbances) and GAD1 (Nervousness) were the key bridge symptoms. The urban–rural comparison revealed that while both groups shared CESD3 (Feeling blue/depressed) and GAD2 (Uncontrollable worry) as central symptoms, they exhibited distinct profiles: GAD3 (Worry too much) was central in the rural network, whereas GAD4 (Trouble relaxing) was central in the urban network. Bridge symptoms remained consistent across both groups. The network comparison test indicated no significant global structural difference, but significant differences were found in specific edge weights and node centralities (EI/BEI).

Conclusions

Central and bridge symptoms may represent effective intervention targets for anxiety and depressive symptoms in older adults living alone. Clinical interventions should consider geographic disparities in symptom presentation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-08141-y.

Introduction

Global population aging has become an irreversible social trend [1]. According to the United Nations, the global population aged 65 and over will surge from 727 million in 2020 to 1.5 billion in 2050, with a growth rate far exceeding that of the total population [2]. As a developing country, China is undergoing unprecedented aging due to the combined effects of the persistently low fertility rate and rising life expectancy [3]. The seventh national census shows that older adults aged 65 and over reached 13.5% (190.64 million) in 2020 and are projected to nearly double by 2050 [4]. This unprecedented pace of ageing underscores the urgent need to address the mental health challenges in this growing population.

Against this backdrop of global aging, depression and anxiety represent a major public health challenge. Depression and anxiety represent a major global mental health challenge for older adults, characterized by high prevalence, significant comorbidity, and substantial personal and societal burdens [5–7]. Meta-analytic evidence indicates global prevalence rates of 19.2% for depressive symptoms and 16.5% for anxiety symptoms among older populations [8]. These rates are even more pronounced in China, specifically 23.6% for depressive symptoms, 22.1% for anxiety symptoms, and 6.67% for their comorbidity [9–11]. In older adults, comorbid depression and anxiety causes serious psychological and physiological damage [6]. Clinically, this co-occurrence is particularly severe, associated with greater symptom severity, chronicity, poorer treatment response, and an elevated suicide risk compared to either condition in isolation [12]. Consequently, the complex presentation of comorbid depression and anxiety, which is often intertwined with somatic complaints, necessitates integrated treatment approaches and solidifies its status as a critical public health priority within global aging research.

Living alone has been identified as a significant risk factor for these debilitating conditions. As a defining feature of contemporary ageing societies, living alone places older adults at greater risk for mental health problems [13]. Research shows older adults living alone experience higher rates of loneliness and social isolation compared to those living with others [14]. This loneliness can negatively impact both their emotional and physical health, significantly harming the mental health of older adults living alone [15]. However, evidence regarding its association with anxiety and depressive symptoms in older adults remains inconsistent. Cross-national studies often report poorer mental health outcomes for those living alone. For example, research from the UK and Iran indicates older adults living alone face a higher risk of anxiety and depression [16, 17]. Conversely, findings from East Asia, particularly China, present a nuanced picture. A South Korean follow-up study found that while living alone initially increased depression risk, this effect diminished over time [18]. Similarly, an early Chinese study reported no significant correlation between living alone and depressive symptoms [19], and a recent large-scale survey found no significant difference in anxiety prevalence between older adults living alone and those living with families [20]. This inconsistency suggests that socio-cultural contexts may moderate the relationship, warranting further investigation within specific settings like China.

In China, Confucian filial piety has profoundly shaped the intergenerational cohabitation tradition, resulting in significantly higher proportions of older adults living with families compared to Western countries [21, 22]. However, accelerated modernization, industrialization, and the growing emphasis on nuclear families have shifted the family structure from a traditional co-living model to a more diversified one [23, 24]. Freed from traditional roles and obligations of intergenerational transfers, older adults increasingly opt for living alone over multi-generational households [25]. A census of older adults in China showed that only 38.8% of urban and 53.6% of rural residents preferred living with children [26]. Meanwhile, the rate of older adults living alone in China reached 14.2% in 2021, with rural areas (16.3%) significantly exceeding urban areas (12.4%) [27]. This trend stems from both active choice (e.g., autonomy) and passive circumstances (e.g., widowhood, child migration) [28].

Notably, the urban-rural distribution of depression and anxiety among Chinese older adults exhibits a unique pattern, diverging from international trends. A meta-analysis showed that while urban older adults in developed countries have a higher prevalence of depressive symptoms, rural Chinese older adults show a significantly higher prevalence [29]. Similarly, anxiety symptoms among older adults are more common in rural versus urban China [30]. This divergence may be related to the urban-rural dual structure in China. Specifically, urban older adults face stressors like rapid pace, noise pollution, and inadequate social support [31, 32], whereas rural older adults are more likely to face outmigration of children, economic hardship, and limited healthcare access [4, 33, 34], which significantly increase their mental health risks. Emerging evidence suggests rural older adults living alone may experience worse depressive symptoms than their urban counterparts [34], though findings on anxiety are less consistent [20]. Given China’s pronounced urban-rural divide, it would be ideal to stratify the urban and rural samples using the same analytical framework, explicitly examining whether residence moderates on the link between living alone and depressive-anxiety symptoms.

In addition, despite recognition of the influence of living arrangements and residence on depression and anxiety in older adults, the complexity of these relationships necessitates more refined research paradigms. Traditional approaches often rely on total scale scores to assess psychological conditions [16, 17, 20], overlooking the interconnected nature and unique characteristics of individual symptoms [35]. In this context, network analysis, an innovative data-driven approach, offers novel pathways to quantify the associated structure of symptoms [36]. This method can not only estimate the interconnections between specific symptoms but also identify closely related clusters of symptoms [37]. Moreover, network analysis can pinpoint central symptoms that may drive the development or maintenance of psychopathological networks [38]. Meanwhile, bridging symptoms connect different psychiatric disorders within the symptom networks of mental disorders, and when a bridging symptom changes, it may affect the symptoms linked to it in multiple psychiatric disorders simultaneously. Intervention on core and bridging symptoms in the symptom network can help change the structure of the psychiatric disorder network and thus impede the development of psychiatric disorders [36, 39]. By identifying and emphasizing core and bridging symptoms, network analysis provides crucial insights into the underlying mechanisms of mental disorders, making it valuable for pinpointing intervention targets and enhancing intervention effects.

Although network analysis has been applied to study depression and anxiety among older adults with various vulnerabilities [9, 38, 40], including those living alone [41], the potential urban-rural variation in these network structures remains unexamined. Therefore, this study utilized data from the Chinese Longitudinal Healthy Longevity Survey to conduct a comparative network analysis of depressive and anxiety symptoms among older adults aged 65 and above who live alone, aiming to (1) identify central and bridge symptoms within the depression and anxiety symptoms network, and (2) systematically compare the network structures, including central symptoms and bridge symptoms, between urban and rural populations. Based on previous evidence that core depression symptoms such as depressed mood and core anxiety symptoms such as uncontrollable worry have been consistently identified across different populations, we hypothesized that these symptoms would also emerge as central in the network of older adults living alone. Given the distinct urban–rural differences in living environments and stress profiles, we further hypothesized that the network structures, particularly the centrality of specific symptoms, would differ between urban and rural groups. Considering that no published studies have compared the depression–anxiety symptom network between urban and rural older adults living alone, our work is largely exploratory. These findings are expected to provide evidence for optimizing intervention targets and developing geographically tailored strategies to improve mental health in this vulnerable population.

Methods

Study design and participants

Participants

This study utilized data from the seventh wave (2017–2018) of the China Longitudinal Healthy Longevity Survey (CLHLS), a collaborative research initiative conducted by the Center for Healthy Aging and Development Studies at Peking University and the National School of Development. The CLHLS is a nationally representative longitudinal study of older adults (aged ≥ 65 years) across 23 Chinese provinces, initially enrolling 15,874 participants in 2017–2018. Through systematic screening of responses to the question “Who do you currently live with?” (Question A5), we identified 2,477 older adults reporting solitary living arrangements as the analytical subsample. Participants’ current place of residence was classified based on self-reported information in the household survey. The original survey distinguished three categories: (1) city, (2) town, and (3) village. For the present analysis, “town” and “village” were combined into a single “rural” category, while “city” was retained as the “urban” category. However, 21.2% (n = 525) of this subsample exhibited missing data across critical variables, including covariates (e.g., smoking status, alcohol consumption, economic status, physical exercise, marital history, self-reported health) and outcome measures. Detailed missing variable patterns are documented in Supplementary Material. This missingness necessitated methodological adjustments to mitigate potential selection bias and maintain statistical power, as complete-case exclusion would have reduced the sample to 1,952 participants.

To address the non-random nature of missing data, we first evaluated its mechanism using Little’s Missing Completely at Random (MCAR) test. The test yielded a highly significant chi-square statistic (χ² = 2,894.82, p < 0.001), indicating systematic patterns inconsistent with random missingness. Subsequent analyses revealed associations between missing data and observed covariates such as age, rural/urban residence, and baseline functional limitations, aligning with a Missing at Random (MAR) framework. Given these findings, we adopted multiple imputation (MI), a rigorously validated method for MAR data [42], to generate 20 imputed datasets. This approach retained the full subsample (n = 2,477) while incorporating auxiliary variables to enhance imputation accuracy and preserve sample representativeness. All subsequent analyses, including descriptive statistics and network estimation, were conducted using the imputed complete dataset (n = 2,477).

Ethical considerations

The CLHLS protocol received ethical approval from the Biomedical Ethics Committee of Peking University (IRB00001052–13074). All participants provided written informed consent prior to enrollment. This secondary analysis adhered to the Declaration of Helsinki guidelines and used exclusively de-identified data.

Measures

Center for epidemiologic studies depression scale (CES-D10)

We assessed depressive symptoms among older adults living alone using the CES-D10 scale, which includes 10 items rated on a Likert scale (0 = “never” to 3 = “always”) [43]. Total scores range from 0 to 30, with higher scores indicating greater severity of depressive symptoms. Based on prior research [44], a CES-D10 score ≥ 10 was used to identify probable depressive symptoms [45]. The scale’s validity and reliability are well-established in the Chinese population, making it a key tool for this assessment. In our study, the scale demonstrated good internal consistency (Cronbach’s α = 0.797).

Generalized anxiety disorder (GAD-7)

We used the GAD-7 to evaluate anxiety symptoms [46]. The scale comprises seven items, each rated on a 0–3 scale (0 = “not at all” to 3 = “almost every day”). Total scores range from 0 to 21, with higher scores indicating more severe anxiety symptoms. Following prior research [36], a GAD-7 score ≥ 5 was used to identify individuals with probable anxiety symptoms. The Chinese version of the GAD-7 has been validated and demonstrates good reliability and validity [47]. In our study, the scale exhibited strong internal consistency, with a Cronbach’s α coefficient of 0.924.

Statistical analysis

All descriptive data were analyzed using SPSS 25.0 software. All network analyses were conducted using R (version 4.1.1).

Network estimation

We utilized R software along with the qgraph package to construct and analyze a network depicting the relationship between depression and anxiety symptoms among older adults living alone [48]. In this network, blue lines indicate positive correlations between symptoms, and red lines denote negative correlations. The thickness of the lines and the saturation of the colors reflect the magnitude of the correlation coefficients [35]. The nodes were categorized into two communities based on their origins: the depressive symptoms community, comprising 10 items from the CESD-10 scale, and the anxiety community, consisting of 7 items from the GAD-7 scale. To construct a regularized sparse network, we employed the Gaussian graphical model in conjunction with the graphical least absolute shrinkage and selection operator (glasso) algorithm [49]. The optimal model was selected using the Extended Bayesian Information Criterion (EBIC) with a hyperparameter γ set to 0.5 [35, 50]. We used Spearman correlation coefficients to quantify the associations between nodes [51] and applied the Fruchterman-Reingold algorithm to optimize the network layout, facilitating an intuitive visualization of the complex relationships among the symptoms [52].

Centrality and bridge centrality estimation

We used the networktools package to calculate node centrality metrics. Expected Influence (EI) was adopted to quantify each node’s importance in the network, defined as the sum of edge weights connecting the node to all other nodes, which reflects its potential to propagate effects across the network [53]. Higher EI values indicate greater centrality and influence on symptom dynamics. Thus, nodes with higher EI values are considered as core symptom nodes of the network. Additionally, we computed Bridge Expected Influence (BEI) to identify key nodes bridging different communities. BEI is defined as the sum of edge weights between a node and nodes in opposing communities, highlighting its role in inter-community connectivity [54]. Nodes with high BEI values were designated as key bridge nodes, which may be key mediators of the interaction between anxiety and depressive symptoms.

Network accuracy and stability estimation

The bootnet package was used to assess network accuracy and stability [48]. Nonparametric bootstrapping was performed to calculate 95% confidence intervals for edge weights, estimating the reliability of network structure. The case-dropping subset bootstrap method was applied to evaluate the stability of centrality metrics, with the correlation stability coefficient (CS-C) serving as the key indicator. According to recommended criteria, CS-C values should not fall below 0.25, and values above 0.50 indicate robust stability and interpretability of centrality rankings. Furthermore, the Network Comparison Test package was used for network comparisons [55]. All analyses involved 1,000 resampling iterations with a significance level set at 0.05.

Results

Descriptive statistics

The study included 2,477 participants. As shown in Table 1, the sample had a mean age of 85.56 years (SD = 9.87), with 1564 (63.14%) participants being females. Among them, 413(16.67%) participants resided in urban areas and 2,064 (83.33%) in rural areas. Depressive symptoms were identified in 500 participants (20.19%), while 340 (13.73%) exhibited anxiety symptoms. Older adults living alone in rural areas showed significantly higher prevalence rates of depression (p < 0.001) and anxiety symptoms (p < 0.05). Statistically significant differences emerged between urban and rural residents in age, sex, smoking status, exercise, and economic status (p < 0.05). Additionally, Table 2 presents the means and standard deviations of anxiety and depression symptom variables.

Table 1.

The characteristics of study participants stratified by residence

Variables	Total (N= 2477)	Urban (N = 413)	Rural (N= 2064)	t / χ2	p
	n (%)/M(SD)	n (%)/M(SD)	n (%)/M(SD)
Age	85.56 (9.87)	84.06(9.88)	85.87 (9.84)	-3.41	< 0.05
Sex				4.15	< 0.05
Male	914(36.86)	134(32.45)	779(37.74)
Female	1564 (63.14)	279(67.55)	1285(62.26)
Smoking	360 (14.53)	41(9.93)	319 (15.46)	8.47	< 0.05
Drinking	351 (14.17)	49(11.86)	302(14.63)	2.17	0.14
Exercise	760(30.68)	217(52.54)	543(26.30)	111.36	< 0.001
Economic status				57.79	< 0.001
Wealth	355 (14.33)	101(24.46)	254(12.31)
Ordinary	1752 (70.73)	284(68.77)	1468(71.12)
Poverty	370(14.94)	28(6.77)	342(16.57)
Marital status				0.12	0.73
Married	211(8.52)	37(8.96)	174(8.43)
Widowed/separated/single	2266 (91.48)	376(91.04)	1890(91.57)
Self-reported health				3.44	0.18
Good	1103(44.53)	201(48.67)	902(43.70)
Ordinary	1014(40.94)	157(38.01)	857(41.52)
Bad	360(14.53)	55(13.32)	305(14.78)
Depressive symptoms(≥ 10)	500(20.19)	51(12.35)	449(21.75)	18.90	< 0.001
Anxiety symptoms(≥ 5)	340(13.73)	42(10.17)	298(14.44)	5.30	< 0.05

Table 2.

The means, standard deviations of variables of anxiety and depression symptoms

Variable	Total(N = 2477)		Urban(N = 413)		Rural(N = 2064)
	M	SD	M	SD	M	SD
GAD1	0.34	0.64	0.26	0.56	0.36	0.65
GAD2	0.68	0.85	0.59	0.84	0.69	0.84
GAD3	0.34	0.63	0.25	0.55	0.36	0.64
GAD4	0.87	0.91	0.60	0.79	0.92	0.92
GAD5	0.86	0.94	0.78	0.96	0.88	0.93
GAD6	0.26	0.56	0.23	0.54	0.27	0.57
GAD7	1.14	1.04	1.05	1.05	1.16	1.03
CESD1	0.67	0.86	0.54	0.76	0.70	0.87
CESD2	0.26	0.58	0.18	0.50	0.27	0.60
CESD3	0.73	0.83	0.67	0.80	0.75	0.83
CESD4	0.36	0.60	0.31	0.57	0.37	0.61
CESD5	0.26	0.55	0.20	0.47	0.27	0.56
CESD6	0.30	0.59	0.24	0.49	0.31	0.61
CESD7	0.21	0.50	0.16	0.43	0.22	0.52
CESD8	0.18	0.48	0.11	0.39	0.20	0.50
CESD9	0.18	0.46	0.14	0.45	0.19	0.47
CESD10	0.15	0.44	0.11	0.38	0.16	0.45

M mean, SD standard deviation

Network structure

The network structure of depression and anxiety symptoms among older adults living alone is illustrated in Fig. 1A. Logically, the network could have contained a maximum of 136 edges. There were 77 edges with non-zero edge weights (ranging from − 0.04 to 0.35) in the network, and positive edges constituted the vast majority of all edges (98.53%), as depicted in the correlation matrix shown in Supplementary Materials. Within the depressive symptom community, the strongest associations were observed between CESD5 (Hopelessness) and CESD7 (Lack of happiness), followed by CESD1 (Feeling bothered) and CESD3 (Feeling blue/depressed), as well as CESD2 (Difficulty with concentrating) and CESD4 (Everything was an effort), and CESD8 (Loneliness) and CESD9 (Inability to get going). In the anxiety symptom community, the strongest associations were found between GAD4 (Trouble relaxing) and GAD5 (Restlessness), followed by GAD2 (Uncontrollable worry) and GAD4 (Trouble relaxing), as well as GAD5 (Restlessness) and GAD6 (Easily annoyed/irritated). The cross-community edges were generally weaker than the within-community edges. Nevertheless, some relatively strong cross-community edges were identified. Specifically, GAD1 (Nervousness or anxiety) was positively associated with CESD10 (Sleep disturbances), followed by the association between GAD3 (Worry too much) and CESD4 (Everything was an effort), as well as the association between GAD4 (Trouble relaxing) and CESD8 (Loneliness).

Fig. 1.

Network model of depression and anxiety symptoms and the EIs and BEIs of the nodes in the network among older adults living alone. A Network model of depression and anxiety symptoms. The blue edge represents a positive partial correlation; the red edge represents a negative partial correlation. A thicker and more saturated edge indicates a stronger association. B The EI indices of the nodes in the network (z-score). C The BEI indices of the nodes in the network (z-score). Note: GAD1 = Nervousness or anxiety; GAD2 = Uncontrollable worry; GAD3 = Worry too much; GAD4 = Trouble relaxing; GAD5 = Restlessness; GAD6 = Easily annoyed/irritated; GAD7 = Afraid something terrible might happen; CESD1 = Feeling bothered; CESD2 = Difficulty with concentrating; CESD3 = Feeling blue/depressed; CESD4 = Everything was an effort; CESD5 = Hopelessness; CESD6 = Felt fearful; CESD7 = Lack of happiness; CESD8 = Loneliness; CESD9 = Inability to get going; CESD10 = Sleep disturbances

Centrality and bridge centrality

Figure 1(B) indicates that GAD4 (Trouble relaxing), GAD2 (Uncontrollable worry), and CESD3 (Feeling blue/depressed) exhibited the highest level of EI, while Fig. 1(C) reveals that within their respective communities, GAD1 (Nervousness or anxiety) and CESD10 (Sleep disturbances) had the highest BEI values.

Network accuracy and stability

The 95% confidence intervals (CIs) derived from the bootstrapping method were narrow, suggesting that the estimation of edge weights was accurate and stable. The results of the bootstrapped difference test for the edge weights indicated that the weights of the strongest edges stood out significantly above those of most other edges. The CS-C of EI and BEI were both 0.75, suggesting that the estimations of EI and BEI were both adequately stable. The bootstrapped difference test results for node EIs showed that the EI values of the three central nodes were significantly higher than those of the other nodes, and the findings of the bootstrapped difference test for node BEIs indicated that the BEI values of the bridge nodes were significantly elevated compared to non-bridge nodes. All results are detailed in the supplementary materials.

Network comparison

To further investigate geographic disparities in the network structure of depressive and anxiety symptoms among older adults living alone, we constructed separate rural and urban networks and performed comparative network analyses.

The network structure between rural and urban older adults living alone

Figure 2A and B illustrate the depression and anxiety symptoms networks for rural and urban older adults living alone. Logically, each subnetwork could have contained a maximum of 136 edges. In the rural network, there were 91 edges with non-zero edge weights (ranging from − 0.01 to 0.33), accounting for 66.91% of all possible edges, and positive connection edges made up 98.57% of all connections. The two strongest connection edges in the anxiety and depression network of rural older adults living alone are the connections between CESD1 (Feeling bothered) and CESD3 (Feeling blue/depressed), and between GAD1 (Nervousness or anxiety) and GAD2 (Uncontrollable worry). In the urban network, there were 71 edges with non-zero edge weights (ranging from 0.01 to 0.42), constituting 52.21% of the total possible edges, and all connection edges represent positive correlations. In the anxiety-depression network for urban older adults living alone, the two strongest connection edges are the connection between CESD5 (Hopelessness) and CESD7 (Lack of happiness), and the one between GAD2 (Uncontrollable worry) and GAD3 (Worry too much). Of note, when compared to the rural network, the urban network displays a sparser structure with fewer connection edges. Detailed connection edge weight distributions are provided in the Supplementary Material.

Fig. 2.

Network model of depression and anxiety symptoms and the EIs and BEIs of the nodes in the network among rural and urban older adults living alone. A Network model of depression and anxiety symptoms in rural older adults. B Network model of depression and anxiety symptoms in urban older adults. The blue edge represents a positive partial correlation; the red edge represents a negative partial correlation. A thicker and more saturated edge indicates a stronger association. C The EI indices of the nodes in the rural and urban networks (z-score). D The BEI indices of the nodes in the rural and urban networks (z-score). Note: GAD1 = Nervousness or anxiety; GAD2 = Uncontrollable worry; GAD3 = Worry too much; GAD4 = Trouble relaxing; GAD5 = Restlessness; GAD6 = Easily annoyed/irritated; GAD7 = Afraid something terrible might happen; CESD1 = Feeling bothered; CESD2 = Difficulty with concentrating; CESD3 = Feeling blue/depressed; CESD4 = Everything was an effort; CESD5 = Hopelessness; CESD6 = Felt fearful; CESD7 = Lack of happiness; CESD8 = Loneliness; CESD9 = Inability to get going; CESD10 = Sleep disturbances

Centrality and bridge centrality between rural and urban older adults living alone

The comparative analysis of network structure revealed both shared and distinct features between rural and urban older adults living alone. As illustrated in Fig. 2C, the rural network was characterized by CESD3 (Feeling blue/depressed), GAD3 (Worry too much), and GAD2 (Uncontrollable worry) as central symptoms with the highest EI values. In contrast, the urban network identified CESD3 (Feeling blue/depressed), GAD4 (Trouble relaxing), and GAD2 (Uncontrollable worry) as its central symptoms. This indicates that while feeling depressed (CESD3) and uncontrollable worry (GAD2) were common central symptoms across both groups, their symptom profiles were distinct: rural networks featured excessive worry (GAD3) as an additional central node, whereas urban networks featured trouble relaxing (GAD4).

Regarding bridge symptoms connecting the anxiety and depression clusters, nervousness (GAD1) consistently demonstrated the highest BEI values within the anxiety cluster across both rural and urban populations. Similarly, within the depression cluster, sleep disturbances (CESD10) exhibited consistently high BEI values in both groups (see Fig. 2D). This pattern suggests the invariant role of these two symptoms as key bridges in the depression and anxiety symptoms network, regardless of geographic location.

Network stability and accuracy between rural and urban older adults living alone

Bootstrap analyses (1,000 bootstrap samples) confirmed satisfactory stability for both rural and urban networks. Network robustness tests revealed that the bootstrapped 95% CIs were relatively narrow, suggesting that the edge weights were accurately estimated in both rural and urban networks (see Supplementary Material). The results of the bootstrapped difference test for edge weights are also detailed in the Supplementary Material. The CS-C highlighted geographic divergence: rural networks demonstrated good stability for both EI (CS-C = 0.75) and BEI (CS-C = 0.75), indicating consistent symptom connectivity patterns. In contrast, urban networks exhibited reduced BEI stability (CS-C = 0.52) despite adequate EI stability (CS-C = 0.67), potentially reflecting greater environmental complexity or variability in symptom interactions among urban residents (see Supplementary Materials for further details).

Network comparison test (NCT) between rural and urban older adults living alone

The R package Network Comparison Test was used [55]. Network invariance testing revealed no significant structural differences between rural and urban groups (M = 0.858, p = 0.656), with comparable global network strength (rural: 7.40 vs. urban: 7.23, p = 0.238) and total expected influence (S = 2.049, p = 0.580). However, three edges showed significant weight differences (p < 0.05) between the groups: rural networks demonstrated stronger associations between GAD3 (Worry too much) and CESD4 (Everything was an effort), as well as between CESD5 (Hopelessness) and CESD7 (Lack of happiness), whereas urban networks featured a stronger connection between GAD4 (Trouble relaxing) and CESD9 (Inability to get going). Furthermore, significant differences were observed in nodes’ EI and BEI across the regions. Permutation tests identified significant differences in the centrality of one node and the bridge centrality of three nodes (p < 0.05). Specifically, GAD3 (Worry too much) exhibited higher EI in rural areas, CESD6 (Felt fearful) demonstrated elevated BEI in rural contexts, while GAD4 (Trouble relaxing) and CESD2 (Difficulty concentrating) showed stronger BEI in urban networks. Collectively, these results indicate that although the overall network structure and global strength did not differ significantly between urban and rural older adults living alone, partial, symptom-level differences were present in specific edge weights and node centrality values.

Discussion

This study constructed and compared urban-rural symptom networks of depression and anxiety symptoms among Chinese older adults living alone. Results indicated that CESD3 (Feeling blue/depressed) and GAD2 (Uncontrollable worry) functioned as core symptoms in both urban and rural networks, serving as pivotal activation nodes within the depression and anxiety symptom network. Urban-rural subnetwork comparisons revealed distinct geographical patterns. Specifically, GAD4 (Trouble relaxing) exhibited significantly higher EI values in urban networks whereas GAD3 (Worry too much) showed elevated EI values in rural networks. Crucially, GAD1 (Nervousness or anxiety) and CESD10 (Sleep disturbances) consistently demonstrated the highest BEI values within their respective symptom communities across both urban and rural networks, suggesting their potential as candidate bridge targets for future cross-regional intervention research. Notably, although specific edge weights and node centrality values differed between urban and rural groups, the overall network structure and global strength did not differ significantly, indicating partial, symptom-level differences rather than broad structural divergence.

According to network theory, nodes within the same symptom community (e.g., depressive symptoms) interact closely, which in turn form stable pathological networks to maintain the state of mental disorders [36]. The present study found that the strongest connections in both rural and urban subnetworks were focused within their respective symptom communities, rather than between depression and anxiety symptoms across communities. This is consistent with the results of many previous network analyses exploring the correlation between the two disorders, which have found that the strongest edges are between symptoms of the same disorder [38, 56, 57]. It suggests that although anxiety and depression are related, each has a relatively independent symptom subnetwork. Based on EI values, CESD3 (Feeling blue/depressed) and GAD2 (Uncontrollable worry) were identified as central to the anxiety and depressive symptoms in both urban and rural older adults living alone, a finding in line with previous research. Zhang et al. [38] conducted network analyses of older adults with disabilities, Zhang et al. [56] focused on the network of older adults with diabetes, and Beard et al. [57] studied a large psychiatric sample in the US, all of which confirmed the centrality of “Feeling blue/depressed” in the network of depression and anxiety symptoms. The centrality of CESD3 (Feeling blue/depressed) may be closely related to the living patterns of older adults living alone. Studies indicate that they face a significantly higher risk of depression compared to those living with family, primarily because they must handle independent physical care and life planning [58]. The absence of support due to social isolation may exacerbate their sense of isolation and disrupt the balance of their need for belonging and love [14]. One possible neurobiological interpretation is that this emotional deprivation continuously amplifies the experience of sadness by activating the limbic system [59], which may strengthen the central driving role of CESD3 (Feeling blue/depressed) within the symptomatic network. GAD2 (Uncontrollable worry), defined as persistent concern about potential harm or risk, is a core symptom of anxiety [60], and is commonly observed across various anxiety and mood disorders [61]. Research during the COVID-19 pandemic further highlighted that GAD2 (Uncontrollable worry) is a key anxiety symptom closely linked to tension [62]. It is plausible that older adults living alone are more prone to worry and tension than other elderly groups due to limited interaction with their children [41].

In addition, urban-rural subgroup analyses further revealed differentiated core symptoms. In the urban network, GAD4 (Trouble relaxing) emerged as the third most important core symptom among urban older adults living alone, which may be attributed to chronic stress resulting from high-frequency stimulation, fast-paced living, and social detachment in urban environments [31, 32]. Research has demonstrated that traffic noise and air pollution in urban areas are associated with high levels of autonomic nervous system activation [63], which may significantly hinder individuals’ ability to relax. Moreover, China’s rapid urbanization has exacerbated these environmental problems [64], and housing tensions and resource shortages triggered by rapid urban expansion may further intensify interpersonal competition and psychological isolation [64], which could collectively contribute to persistent chronic tension among urban older adults living alone. In contrast, in the rural network, GAD3 (Worry too much), along with GAD2 (Uncontrollable worry) and CESD3 (Feeling blue/depressed), constitute the core symptoms in the network of older adults living alone. In the DSM-5 [65] and ICD-10 [66], “worrying too much” is the hallmark symptom required for a diagnosis of generalized anxiety disorder. In this sample, worrying may stem from anxiety about survival issues, as studies have shown that rural older adults living alone have significant worries about basic survival issues such as declining farming ability and delayed diagnosis and treatment of illnesses due to weak financial support and low access to healthcare resources [4, 33]. Empirical studies have also revealed the vulnerability of China’s rural elderly population; a survey covering 6,483 Chinese seniors found that 64.64% of rural seniors suffered from one or more chronic diseases and were more vulnerable to catastrophic healthcare expenditures [67], and data from 2018 also showed that the incidence of catastrophic healthcare expenditures was significantly higher in rural China (27.0%) than in urban China (17.4%) [68]. At the same time, rural areas lag far behind urban areas in terms of healthcare coverage, ease of access to healthcare, and level of healthcare services [68]. Thus, it is plausible that this severe imbalance between high health risks and lack of care resources may make rural older adults living alone more likely to fall into excessive worrying due to disease burden and financial hardship and to have less control over their worries, thereby maintaining high levels of psychopathological symptoms [69]. Notably, the core symptom status of CESD3 (Feeling blue/depressed) and GAD2 (Uncontrollable worry) remained stable across both urban and rural networks of older adults living alone, suggesting that affective deprivation and cognitive uncertainty may represent common psychological risk mechanisms across geographic regions. The geographical specificity of GAD4 (Trouble relaxing) and GAD3 (Worry too much) points to the possible role of environmental resources in shaping the expression of core symptoms, namely, psychological tension under chronic stress in the urban older adults living alone and realistic worry under existential stress in the rural older adults living alone.

Based on BEI values, GAD1 (Nervousness or anxiety) and CESD10 (Sleep disturbances) were identified as key bridging symptoms connecting depression and anxiety symptoms. Their high stability across urban and rural areas suggests that they may function as co-drivers across symptom clusters, sustaining the pathology. This finding appears generalizable given the cross-geographical consistency of the symptom network. Specifically, the bridging role of GAD1 (Nervousness or anxiety), a core symptom of anxiety disorders as defined by the American Psychiatric Association [65], has been demonstrated in previous studies. Recent network analyses of diabetic and widowed older adults indicate that GAD1 (Nervousness or anxiety) serves as a key bridging node in the anxiety-depression comorbidity network [40, 56]. Older adults living alone may be more susceptible to this type of emotional distress due to the lack of parent-child interaction [41], and studies have shown that the prevalence of anxiety or depressive symptoms is higher in older adults living alone compared to those not living alone [70]. In addition, CESD10 (Sleep disturbances) consistently exhibited the highest BEI in the depressive symptom community across both geographical groups, which is in line with previous research concluding that sleep problems are a significant risk factor for mental illness [71]. Research suggests a bidirectional interaction between sleep disorders and depressive symptoms [72]. Loneliness triggered by the solitary state may exacerbate sleep disorders, which in turn may aggravate depressive symptoms [73]. Moreover, pathological features such as negative cognitive bias and excessive emotional arousal accompanying depressive symptoms may disrupt normal sleep structure, reduce effective sleep duration, and induce or worsen sleep disorders [72]. In the elderly group living alone, this potential bidirectional relationship may be more likely to form a vicious circle due to the decline in their physical functions, weakened psychological adjustment ability, and the absence of timely emotional support and life care from the outside world [74]. It is also possible that the synergistic pathological effects of GAD1 (Nervousness or anxiety) and CESD10 (Sleep disturbances) are related to the neuromodulatory mechanisms of the dopamine system. It has been demonstrated that dopamine plays a crucial role in both sleep regulation and mood regulation [75], providing one possible neurobiological pathway for the pathological association between sleep disorders and depressive symptoms [76]. Although bridging symptom intervention strategies based on mental disorder network theory are of theoretical interest [36], it must be emphasized that the present cross-sectional findings provide only preliminary evidence. Whether interventions targeting GAD1 (Nervousness or anxiety) and CESD10 (Sleep disturbances) can effectively disrupt the vicious cycle of anxiety and depressive symptoms requires rigorous empirical validation through longitudinal and experimental studies. Currently, cognitive behavioral therapy (CBT) and its sleep intervention subtype therapy (CBT-I) have proven effective in improving anxiety, depression, and comorbid states [77, 78]. Notably, a recent study demonstrated that digital cognitive behavioral therapy for insomnia (dCBT-I) was effective in alleviating insomnia in patients with concomitant anxiety symptoms [79], offering a potential direction for future intervention research.

We also observed significant differences in edge weights and EI and BEI values of individual nodes between urban and rural networks, yet there were no differences in overall network structure, which suggests that geographic context may shape specific symptom-level associations rather than the global network architecture of mental health among older adults living alone. Our results indicate that the edges between GAD3 (Worry too much) and CESD4 (Everything was an effort), as well as between CESD3 (Feeling blue/depressed) and CESD7 (Lack of happiness), are significantly stronger in rural older adults living alone compared to the urban network; this may be closely related to the real-life dilemmas faced by rural older adults, such as a lack of healthcare resources, prominent economic stress, and forced separation from children [32, 33]. Research has shown that rural older adults living alone are more likely to worry excessively when confronted with illness and financial difficulties [80], and such excessive worrying may continuously deplete their psychological resources, keeping their minds in a state of chronic tension, which could not only significantly reduce the efficiency and motivation of their daily activities but also make otherwise easy tasks seem difficult, consequently creating a subjective experience of “everything is an effort” and exacerbating their negative emotional state [13]. At the same time, the absence of their children may significantly deepen their sense of isolation, directly undermining their subjective well-being and further reinforcing feelings of sadness and depression [81].Moreover, we found that the strong link between GAD4 (Trouble relaxing) and CESD9 (Inability to get going) is more pronounced in urban older adults living alone, suggesting a possible association between the psychological stress linked to fast-paced urban living, complex environmental stimuli, and social alienation [31, 32]; one tentative interpretation is that this stress-induced tension may affect cognitive control in the dorsolateral prefrontal cortex, manifesting as distraction and difficulties in behavioral initiation, which in turn could affect their mobility [82, 83]. Furthermore, we found that GAD3 (Worry too much) had higher EI values in rural areas, and CESD6 (Felt fearful) had higher BEI values in rural areas, whereas GAD4 (Trouble relaxing) and CESD2 (Difficulty with concentrating) had stronger BEI values in the urban network; these findings provide preliminary evidence that there may be differences in the symptom-level drivers and symptom evolution pathways of mental health problems in urban and rural older adults living alone, suggesting that geographic specificity warrants consideration in future intervention research.

This study has several limitations that warrant consideration. First, the cross-sectional design is a fundamental constraint. Cross-sectional network analysis can only identify conditional associations between symptoms and cannot establish causal relationships, temporal dynamics, or symptom propagation pathways. The central and bridge symptoms identified in this study should therefore be regarded as potential candidates for future longitudinal investigation rather than confirmed intervention targets. Second, regarding assessment methods, numerous studies have demonstrated that while the CES-D10 and GAD-7 self-assessment scales are well-validated and widely utilized [84–90], the self-reporting approach is prone to recall bias and the social desirability effect. These factors may particularly affect the accuracy of symptom assessment among older adults living alone, who often experience cognitive decline and face a high burden of physical illness [91]. To comprehensively address this limitation, future research should integrate multiple assessment tools, such as clinical interviews, psychiatrist diagnoses, and physiological measurements (e.g., cortisol levels and heart rate variability). Third, the study did not directly assess important contextual variables such as social support, physical health status, and healthcare access, which may influence both the experience and the network structure of depression and anxiety symptoms. The absence of these measures limits our ability to interpret the urban–rural differences observed. Fourth, in terms of sample representativeness, the exclusive focus on Chinese older adults living alone necessitates caution when generalizing findings to other regions or populations with differing living arrangements. Additionally, the urban and rural sample sizes were highly imbalanced (urban: n = 413; rural: n = 2,064), which may reduce the stability of the urban network estimates, as reflected in the lower bridge expected influence stability coefficient (CS-C = 0.52) observed in the urban network. This imbalance may also have reduced the statistical power of the network comparison test, potentially contributing to the non-significant overall structural difference. Subsequent research should broaden sampling diversity and strive for more balanced subgroup sample sizes to enhance generalizability. Finally, group-level network analyses neglect individual heterogeneity, and future studies employing individual-level network modeling are needed to advance understanding of mental health dynamics in this population.

Despite these limitations, this study has several strengths. To our knowledge, it represents the first network analysis to systematically compare the depression and anxiety symptoms structure between urban and rural Chinese older adults living alone. The use of a nationwide dataset enhances the representativeness of our findings. Our analysis identified CESD3 (Feeling blue/depressed), GAD2 (Uncontrollable worry), and GAD4 (Trouble relaxing) as the most central symptoms in the depression-anxiety network. The urban-rural comparison further revealed that although both groups shared CESD3 (Feeling blue/depressed) and GAD2 (Uncontrollable worry) as central symptoms, they exhibited distinct profiles: GAD3 (Worry too much) was central in the rural network, whereas GAD4 (Trouble relaxing) was central in the urban network. Meanwhile, CESD10 (Sleep disturbances) and GAD1 (Nervousness or anxiety) were identified as key bridge symptoms connecting the depression and anxiety clusters, remaining consistent across both groups. These findings, while preliminary given the cross-sectional design, provide a theoretical foundation for future research to examine whether targeting these core and bridge symptoms may help disrupt the comorbidity cycle. The urban-rural differences in central symptoms further suggest the potential value of developing geographically tailored interventions. For example, it is conceivable that interventions for urban older adults might prioritize addressing trouble relaxing, while rural interventions might focus on managing excessive worry. However, such tailored approaches remain speculative at this stage and require validation through prospective studies. If confirmed, they could maximize the effectiveness of mental health interventions and promote healthy aging in this vulnerable population.

Abbreviations

CES-D10

Center for epidemiologic studies depression scale

GAD-7

The Generalized anxiety Disorder-7

LASSO

Least absolute shrinkage and selection operator

EBIC

Extended bayesian information criterion

EI

Expected influence

BEI

Bridge expected influence

CS-C

Correlation stability coefficient

Authors’ contributions

Y.H. participated in study design, writing and revising the manuscript. K.S. performed the statistical analysis in the writing and revising the manuscript. W.C. contributed to data interpretation and validated the analytical findings. M.M. conducted the investigation, created visualizations, and assisted in revising the manuscript. N.C. provided valuable input for revising the manuscript, along with supervision, resources, and project administration. W.L. supervised the entire study operation and the drafting process, and provided critical inputs for drafting and revising the manuscript. All authors read and approved the final manuscript.

Funding

This study received no funding support from any grant program.

Data availability

Publicly available datasets were analyzed in this study. The datasets utilized can be accessed here: https://opendata.pku.edu.cn/dataset.xhtml?persistentId=doi:10.18170/DVN/WBO7LK .

Declarations

Ethics approval and consent to participate

The studies involving human subjects were approved by the Biomedical Ethics Committee at Peking University (IRB00001052–13074). These studies were conducted in accordance with both local legislation and institutional requirements. The participants provided their written informed consent to participate in this research.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.