Prevalence of depression and anxiety and related influencing factors in the Chinese population with noncommunicable chronic diseases: A network perspective

Abstract

BACKGROUND

The prevalence and severity of noncommunicable chronic diseases (NCDs) among Chinese residents have been increasing with mental health emerging as a critical challenge in disease management.

AIM

To examine the interactions between depression, anxiety symptoms, and related factors, and to identify key factors in the Chinese population with NCDs.

METHODS

Data from the Psychology and Behavior Investigation of Chinese Residents were used in a cross-sectional survey of 6182 individuals with NCDs. This study measured depression and anxiety symptoms as well as their influencing factors including social environments, individual behaviors and lifestyles, and subjective indicators. A network analysis approach was used for data assessment.

RESULTS

Network analysis demonstrated that several central factors (media exposure, family health, problematic internet use, suboptimal health status, intimate relationship violence, tired or little energy, and nervousness/anxious/on edge) and bridge factors (media exposure, problematic internet use, intimate partner violence, health literacy, and suboptimal health status) that significantly influenced the co-occurrence and interconnectedness of depression and anxiety symptoms. Additionally, gender, ethnicity, residency, and living status did not significantly influence the overall network strength.

CONCLUSION

Depression and anxiety are prevalent among the Chinese population with NCDs. Effective interventions should focus on managing key symptoms, promoting correct media use for health information, and fostering healthier family relationships.

Key Words: Noncommunicable chronic diseases; Depression; Anxiety; Influencing factors; Network analysis method

Core Tip: This study, leveraging data from the Psychology and Behavior Investigation of Chinese Residents and employing network analysis, revealed the intricate relationships between depression and anxiety symptoms and multidimensional influencing factors among Chinese individuals with noncommunicable chronic diseases. Media exposure, family health, problematic internet use, suboptimal health status, and intimate relationship violence were identified as key central and bridge factors. These findings suggested that interventions targeting these factors, particularly guiding the correct use of media information and fostering harmonious family dynamics, may be crucial for improving the mental health of this population.

INTRODUCTION

Noncommunicable chronic diseases (NCDs) are characterized as insidious onset, long latency, prolonged and slow progression, incurability, lack of conclusive biological cause, and unclear indications for treatment[1]. The World Health Organization (WHO) reported in 2022 that over 17 million people die from chronic diseases globally each year[2]. In China the prevalence of NCDs has increased dramatically from 1992 to 2012[3], making NCDs the leading cause of death among Chinese residents. Patients with NCDs are forced to overcome a range of challenges, such as difficult symptoms, loss of independence, and adherence to daily self-management, leading to a particularly high risk of depression and anxiety[4]. Psychiatric comorbidities, particularly anxiety and depression, are associated with a series of adverse outcomes in patients with NCDs. Specifically, patients may experience poorer self-assessed health status[5], increased absenteeism[6], and poorer prognosis[7]. In this respect an assessment of the current situation, including the prevalence of depression and anxiety among patients with NCDs and influencing factors, would assist in the subsequent care provided and the development and implementation of appropriate intervention strategies.

Scholars have long sought to identify assessment tools, prevalence, and antecedents of depression and anxiety because they believe that formulating personalized, adaptive psychological intervention programs can address these mental disorders by focusing on risk factors[8,9]. Namely, related factors were classified into socioeconomic demographic variables, social environments, personal behaviors and lifestyles, and subjective indicators based on previous research. First of all, the occurrence and development of depressive and anxiety symptoms are significantly related to socioeconomic demographic variables (e.g., age, gender, residency and living alone status)[10,11]. Variables connected with the social environments, including social status[12], neighborhood[13], family health, family communication[14], and intimate relationship violence[15], have relevance to depressive and anxiety symptoms. In addition, personal behaviors and lifestyles, such as lack of exercise[16], high sedentary time[17], problematic internet use[18], poor subjective sleep quality[19,20], and severe social media exposure[21], may contribute to the development of depression and anxiety. Previous literature has demonstrated that subjective indicators like stress[22], loneliness[23], suboptimal health status[24], neurotic personality[25], and dry eye syndrome[26] are risk factors for depression and anxiety while self-efficacy[27], social support[28], subjective well-being[29], and health literacy[30] are protective factors. As far as we know, previous studies have not considered the interaction between depression and anxiety symptoms and related factors. The previous research in this field considers depression and anxiety as independent factors that affect other variables[31] or regards them as the consequence influenced by other factors[32].

The network theory of mental disorders (NTMD) was proposed by Borsboom[33], and it assists in filling this gap by integrating all factors into an interrelated network. NTMD suggests that if coupled symptoms are closely related in the network structure, they tend to synchronize. The development and maintenance of a mental disorder is the causal interacting symptoms. If this connection is tight enough, it will lead to a feedback loop that contributes to the illness progression[33]. The network analysis method (NAM) is regarded as a method corresponding to the NTMD that puts symptoms into a network graph of interrelated nodes with edges, and the thickness describes the intensity of the association. Furthermore, the NAM can determine the importance of specific symptoms in promoting a disorder, namely, the most central nodes in the network according to the several centrality index (strength, betweenness, closeness)[34]. Currently, the NAM has been widely applied in research of depression and anxiety in diverse populations. For example, researchers have attempted to identify the central symptoms of depression and anxiety symptoms in Chinese patients with rheumatoid arthritis using NAM[35]. However, previous scholars have not determined the interaction between depression and anxiety and related factors from a network perspective.

This study, guided by NTMD, applied a NAM to explore depression and anxiety and their related factors. From the perspective of an interactive network, we attempted to identify the most central factors and bridge factors affecting depression and anxiety to provide a theoretical basis for the future interventions of depression and anxiety in the Chinese population with NCDs.

MATERIALS AND METHODS

Participants and procedures

The data was extracted from the Psychology and Behavior Investigation of Chinese Residents[36]. The Psychology and Behavior Investigation of Chinese Residents conducted a nationwide cross-sectional survey from June 20, 2022 to August 31, 2022. A multistage sampling design based on the seventh National Population Census data was applied to ensure the representativeness and generalizability of participants. The survey covered 148 cities from 23 provinces, 5 autonomous regions, and 4 municipalities (excluding Hong Kong, Macau, and Taiwan). Participants were invited to complete the questionnaire via the online Utilizing Questionnaire Star platform and were assigned a survey number by investigators after signing informed consent.

Interviewees met the following inclusion criteria: (1) Age ≥ 12 years; (2) Chinese identity; (3) Permanent resident of China (travel time ≤ 1 month); (4) Voluntary involvement and provided an informed consent form; and (5) Completed questionnaire independently or with the help of the researcher. The exclusion criteria for participants were: (1) Presence of confusion; (2) Presence of mental abnormalities; (3) Presence of cognitive impairment; and (4) Concurrent participation in similar research. To operationalize exclusion criteria related to cognitive impairment, two complementary approaches were adopted: (1) Documented diagnoses of cognitive disorders (e.g., dementia) retrieved from community health service records; and (2) Participant self-reports of medical history indicating physician-diagnosed cognitive impairment. Individuals failing either criterion (e.g., unverifiable medical records or incorrect responses to screening items) were excluded.

In addition, participants with limited literacy or education received standardized support from trained investigators who adhered to strict protocols. Investigators were required to read questions verbatim, avoid interpretive language, and record responses objectively. Centralized training ensured uniformity in guidance (e.g., neutral phrasing and predefined explanations for complex terms). All participants voluntarily participated in the study and are required to sign an informed consent form. For minor participants both the participant and their legal guardian provided informed consent. A total of 31480 questionnaires were distributed with 30505 valid questionnaires returned, and 21916 individuals were further screened according to age, yielding a response rate of 71.8% and a validity rate of 96.9%.

Individuals with NCDs were selected from the data according to the understanding of the top 25 causes of disability-adjusted life-years in China. Respondents diagnosed with NCDs derived from a question: “Have you been diagnosed with any of the following diseases”. A total of 19 diseases were listed: (1) Hypertension; (2) Coronary heart disease; (3) Thyroid disease; (4) Stroke; (5) Mood disorders; (6) Fatty liver; (7) Cerebral thrombosis; (8) Diabetes, chronic kidney disease; (9) Osteoporosis; (10) Alzheimer’s disease; (11) Parkinson's disease; (12) Viral hepatitis; (13) Benign tumors; (14) Chronic enteritis; (15) Chronic gastric disease; (16) Asthma; (17) Dyslipidemia; (18) Chronic obstructive pulmonary disease; and (19) Lumbar disc herniation. In the end a total of 6812 eligible subjects were obtained (Figure 1). The survey protocol was ethically reviewed and registered with the Chinese Clinical Trial Registry.

Figure 1  Flow diagram of participant selection.

Survey questionnaire

Sociodemographic information: The individual characteristic variables include age, body mass index, gender, marital status, ethnicity, residency, education level, monthly income [≤ 1000/1001-3000/3001-5000/> 5000 (Chinese yuan)], and whether living alone. Income categories were analyzed in local currency to preserve ecological validity. Approximate United State dollar conversions are provided in Table 1 for reference.

Table 1 Demographic characteristics of participants.

Characteristics	Total	Depression participants	P value	Anxiety participants	P value
Age (years)	39.43 (18.85)	47.64 (20.02)	< 0.001	47.09 (19.89)	< 0.001
Body mass index (kg/m2)	21.97 (4.37)	21.69 (4.43)	< 0.001	21.63 (4.44)	< 0.001
Gender			< 0.001		< 0.001
Male	3211 (47.1)	2083 (45.2)		1719 (45.0)
Female	3601 (52.9)	2524 (54.8)		2097 (55.0)
Marital status			< 0.001		< 0.001
Unmarried	1647 (24.2)	1344 (29.2)		1140 (29.9)
Married	4569 (66.9)	2843 (61.7)		2335 (61.2)
Divorced	190 (2.8)	143 (3.1)		119 (3.1)
Widowed	415 (6.1)	277 (6.0)		222 (5.8)
Ethnicity			0.207		0.742
Han	6191 (90.9)	4173 (90.6)		3472 (91.0)
Non-Han	621 (9.1)	434 (9.4)		344 (9.0)
Residency			0.255		0.038
Urban	4676 (68.6)	3142 (68.2)		2580 (67.6)
Rural	2136 (31.4)	1465 (31.8)		1236 (32.4)
Education level			< 0.001		< 0.001
Illiterate/semi-literate	645 (9.5)	398 (8.6)		320 (8.4)
Primary school	1013 (14.9)	639 (13.9)		522 (13.7)
Middle school	2376 (34.9)	1533 (33.3)		1264 (33.1)
College or higher	2778 (40.7)	2037 (44.2)		1710 (44.8)
Monthly income			0.005		< 0.001
≤ 1000 CNY	571 (8.4)	401 (8.7)		336 (8.8)
1001-3000 CNY	1972 (28.9)	1286 (27.9)		1024 (26.8)
3001-5000 CNY	1979 (29.1)	1321 (28.7)		1117 (29.3)
> 5000 CNY	2290 (33.6)	1599 (34.7)		1339 (35.1)
Living alone			< 0.001		< 0.001
Yes	1134 (16.6)	874 (19.0)		752 (19.7)
No	5678 (83.4)	3733 (81.0)		3064 (80.3)

Income corresponds to monthly values in Chinese yuan (CNY). Approximate United State dollar (USD) equivalents (2022 average exchange rate: 1 USD = 6.72 CNY): (1) ≤ 149 USD; (2) 149-446 USD; (3) 446-744 USD; and (4) > 744 USD. CNY: Chinese yuan.

Depression symptoms: The Patient Health Questionnaire-9 (PHQ-9) was used for depression symptoms assessment. It has been proven to be a valid and reliable tool for assessing depression in various populations and has been widely applied in clinical practice[37]. The Chinese version of the PHQ-9 contains nine items and each item is scored on a four-point Likert scale ranging from 0 (not at all) to 3 (nearly every day), yielding a total score ranging from 0 to 27 with higher scores representing more severe depressive symptoms. A total score > 4 indicates the participant was suffering from depressive symptoms[38]. In the current study the Cronbach’s alpha coefficient of the PHQ-9 was 0.922.

Anxiety symptoms: Anxiety of participants was assessed by the Seven Generalized Anxiety Disorder (GAD-7)[39], which consists of seven items and each range from 0 (not at all) to 3 (almost every day) with a total score ranging from 0 to 21. Higher scores of the GAD-7 mean high levels of anxiety. A total score between 0 and 4 indicates the absence of anxiety, while a total score ≥ 5 indicates the participant was suffering from anxiety[40]. Cronbach’s alpha of the GAD-7 was 0.945 for the entire scale.

Social environments: Social environment variables include neighborhood, social status, family health, family communication, and intimate partner violence. Neighborhood was assessed by asking the subjects about the relationship between their family with their neighbors with response options ranging from 1 (lowest) to 7 (highest). Social status was measured by asking a question: “How do you perceive your family’s status in society”? with the same response options as for neighborhood. The Family Health Scale (Short Form) was used for family health assessment, which consists of ten items covering four dimensions, and each item was rated on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree) with higher scores meaning better family health. The Cronbach’s alpha for this scale was 0.826, indicating excellent reliability. Family communication of participants was examined using the ten-item Family Communication Scale (FCS-10)[41]. Each item was rated on a five-point Likert scale, with 1 indicating strongly disagree and 5 indicating strongly agree. The scores for all items were then added together to represent a total score of family communication, with higher scores indicating better quality of family communication. The Cronbach's alpha coefficient of the Family Communication Scale-10 was 0.968. Intimate partner violence was measured using a self-developed scale with five items covering physical and psychological dimensions. A higher total indicated a higher level of intimate partner violence experienced by the subject. The Cronbach’s alpha coefficient of this scale was 0.914.

Individual behaviors and lifestyles: Individual behaviors contained media exposure, sleep quality, and problematic internet use. Media exposure was measured using a self-administrated six-item scale, assessing the frequency of individuals’ media use behaviors, including social communication, self-presentation, social action, leisure and entertainment, information acquisition, and business transactions via media. Each item was rated on a five-point Likert scale ranging from 1 (never use) to 5 (always use), which yielded a total score ranging from 6 to 30, and higher scores indicated more frequent media use.

The brief version of the Pittsburgh Sleep Quality Index (PSQI) adapted from the PSQI was used to assess the subjective sleep quality of the participants over the past month[42]. The brief PSQI consisted of five questions, covering five rated items (bedtime and wake-up time are used for sleep efficiency evaluation) with a global score ranging from 0 to 15, and higher scores indicated poorer sleep quality[43].

The Problematic Internet Use Questionnaire-Short Form-6 (PIUQ-SF-6) was applied to assess problematic internet use of participants. It included three dimensions (obsession, neglect, and control), and each dimension contained two items. Each item was rated on a five-point Likert scale, with 1 meaning never and 5 meaning always, yielding a total score ranging from 6 to 30. Higher scores on the PIUQ-SF-6 denoted increased problematic internet use[44]. The Cronbach’s alpha coefficient of the PIUQ-SF-6 in the current study was 0.931.

We collected data on the lifestyles of participants, including physical exercise and sedentary habits. Physical exercise was evaluated by the frequency of engaging in vigorous-intensity activity (power sports, fast running, ball games, aerobics, fast cycling) and moderate-intensity activity (handling goods, medium-speed cycling, jogging, table tennis) within a week. Sedentary habits were assessed with a question: “How long do you generally sit each day in the past 7 days”?

Subjective indicators: Subjective indicators included stress, self-efficacy, social support, loneliness, subjective well-being, health literacy, neurotic personality, suboptimal health status, and dry eye symptoms. The Perceived Stress Scale (PSS)-4 items, extracted from the 14-item PSS, was applied to investigate perceived stress of participants[45]. The New General Self-Efficacy Scale short form is a validated self-developed instrument to measure self-efficacy of participants[46]. The brief version of the Multidimensional Scale of Perceived Social Support was used to assess perceived social support from three sources: (1) Family; (2) Friends; and (3) Significant others[47].

The Three-Item Loneliness Scale was used to measure the loneliness of individuals and included three items and a set of simplified response categories[48]. Subjective well-being was assessed using the five-item WHO Well-Being Index (WHO-5), derived from the WHO-10[49]. Health literacy was measured using the simplified version of Health Literacy Scale-Short Form[50]. Neuroticism was assessed by the neuroticism dimension of the ten-item Big Five Inventory[51]. We measured suboptimal health status with the Short-Form Suboptimal Health Status Questionnaire[52]. Dry eye symptoms of participants were measured using the Ocular Surface Disease Index-six items with a total score ranging from 0 to 24, and higher scores were associated with more severe dry eye symptoms[53].

The above instruments have shown satisfactory reliability and validity in Chinese populations, and the Cronbach’s alphas of the PSS-4, the New General Self-Efficacy Scale, the Multidimensional Scale of Perceived Social Support, the Three-Item Loneliness Scale, the WHO-5, the Health Literacy Scale-Short Form, the ten-item Big Five Inventory, the Short-Form Suboptimal Health Status Questionnaire, the Ocular Surface Disease Index-six items were 0.699, 0.929, 0.890, 0.865, 0.936, 0.940, 0.658, 0.950 and 0.932, respectively.

Coronavirus disease 2019-related variable: We assessed the influence on daily life with a broad question: “To what extent do you think coronavirus disease 2019 (COVID-19) has affected your life”? Individuals rated it on a scale from 0 to 100 with a higher score indicating a greater impact of the pandemic.

Statistical analysis

Descriptive statistics were performed using the Statistical Package for the Social Sciences 26.0. Descriptive statistics were analyzed for demographic data. Categorical variables were expressed as frequencies (n) and percentages (%). Continuous variables were expressed as means ± standard deviation. Student's t-test and the χ² test were used to identify the relationship between the independent variables, depression and anxiety symptoms. Network analysis was conducted using R 4.4.2.

To avoid potential effects of different measurement scales on the results, the scores for symptoms and influencing factors were standardized using z-scores and converted to t-scores for visualization. First, we used the qgraph package to plot the network graph and generated an undirected correlation network applying the spring layout in which highly linked nodes clustered together in the center of the network. In this study various antecedent variables and depression and anxiety symptoms were regarded as different communities.

We performed the Least Absolute Shrinkage and Selection Operator network, which employs statistical regularization techniques to limit the number of spurious edges, obtaining a sparse network structure[54]. To understand the importance of the node structure in the network, we calculated three centrality indices (strength, closeness, and betweenness). Strength represents the absolute value of the weights on the edges connected to a given node, betweenness indicates the number of times a node lies on the shortest path among nodes, and closeness refers to the average distance from a node to all other nodes. Core nodes exhibit multiple connections in the network, and deleting or changing these nodes is likely to cause significant changes in the entire network. We focused on node strength and its accuracy and stability as there is debate suggesting that closeness and betweenness tend to be unstable[55].

In addition, we applied the network tools package to reveal bridge nodes reflected by bridge centrality, including bridge strength (i.e. the total connectivity of a node with other disorders), bridge betweenness (i.e. the number of times a node lies on the shortest path between any two nodes of different disorders), and bridge closeness (i.e. the average distance from a node to all other nodes outside of its disorder with distance based on the inverse of the edge weights in the weighted network).

Bootstrapping was conducted with package boot net to assess the accuracy and stability of the network. We employed a case-dropping subset bootstrap method (1000 replicates, 8 cores) to calculate the correlation stability (CS) coefficient. A CS coefficient (correlation = 0.7) indicates the maximum percentage of sample cases that can be dropped from the original full cases to retain a correlation of 0.7 in at least 95% of the samples. Researchers have pointed out that the CS coefficient needs to be higher than 0.25 (acceptable level) and ideally greater than 0.5 (preferable level)[56].

To evaluate the accuracy of the edge weights in the network, we performed non-parametric bootstrapping (1000 replicates, 8 cores) to estimate the 95% confidence intervals (CIs) of the edge values. Larger CIs indicate lower precision of edge estimates while narrower CIs indicate a higher credible network. We applied bootstrapped difference tests to examine differences in the edge weights.

Finally, participants were divided into groups based on gender, age, and residency, and a network comparison test was handled using the package network comparison test to identify whether the estimations of network connectivity and centrality for various subgroups differ, and the network comparison test tested invariant network structure, invariant global strength, and invariant edge strength between two networks[57].

RESULTS

Demographic characteristics of participants

A total of 6812 individuals with NCDs were included in this cross-sectional study. The average age of participants was 39.43 ± 18.85 years with 47.1% males. Most of the participants were married (66.9%), Han Chinese (90.9%), living in urban areas (68.6%), and not living alone (83.4%). Participants who received a college education or higher accounted for 40.7% of participants. Among the 6812 individuals with NCDs, 4607 were suffering from depression (prevalence: 58.97%) and 3816 from anxiety (prevalence: 56.02%). The univariate analysis showed that participants with depression and anxiety were younger and had lower body mass index compared with those without (both P < 0.001). Depression and anxiety were more common among females, urban residents, and solitary participants (all P < 0.001). These factors were further used as a basis for network comparisons (Table 1).

Description of the network

The network structure of depression, anxiety, and their related factors in patients with NCDs was displayed in Figure 2. The strongest edge within the depression community was node P3 “trouble sleeping” and node P4 “tired or little energy” (edge weight = 0.280). The strongest edge within the anxiety community was node G5 “restless” and node G7 “afraid something awful might happen” (edge weight = 0.237). Node Ibl1 “vigorous-intensity activity a week” and node Ibl2 “moderate-intensity activity a week” (edge weight = 0.498) showed the strongest relationship within the related factors community, followed by node Ibl7 “problematic internet use” and node Se5 “intimate relationship violence” (edge weight = 0.349) and node Si2 “self-efficacy” and node Si3 “social support” (edge weight = 0.316).

Figure 2 Estimated network model of depression-anxiety and their related factors and the centrality indices in the Chinese population with noncommunicable chronic diseases. A: Estimated network model of depression-anxiety and their related factors; B: The centrality indices in the Chinese population with noncommunicable chronic diseases. Different colored nodes represent specific symptoms of anxiety and depression and related factors. Blue edges constitute positive partial correlations between variables, and red edges constitute negative partial correlations. The edge thickness represents the strength of the association between various nodes. C1: Influence on daily life; G1: Nervous, anxious, on edge; G2: Uncontrollable worry; G3: Worry about different things; G4: Trouble relaxing; G5: Restless; G6: Irritable; G7: Afraid something awful might happen; Ibl1: Vigorous-intensity activity a week; Ibl2: Moderate-intensity activity a week; Ibl3: Walking more than 10 minutes a week; Ibl4: Sitting time in a week; Ibl5 Media exposure; Ibl6: Subjective sleep quality; Ibl7: Problematic internet use; P1: Low interest or pleasure; P2: Feeling down, hopeless; P3: Trouble sleeping; P4: Tired or little energy; P5: Poor appetite/overeating; P6: Guilt; P7: Trouble concentrating; P8: Moving slowly/restless; P9: Suicidal thoughts; Se1: Neighborhood; Se2: Social status; Se3: Family health; Se4: Family communication; Se5: Intimate relationship violence; Si1: Stress; Si2: Self-efficacy; Si3: Social support; Si4: Loneliness; Si5: Subjective well-being; Si6: Health literacy; Si7: Neuroticism personality; Si8: Suboptimal health status; Si9: Ocular surface disease index.

Accuracy and stability of the network

Figure 3 illustrates that the edge weights in the current sample were consistent with those in the bootstrapped samples, and the bootstrapped 95%CIs were narrow, indicating a good accuracy of the network model. Figure 4 showed that the CS coefficients for strength (traditional centrality) and bridge strength (bridge centrality) were both 0.75, reaching the critical cutoff point (0.5).

Figure 3 Accuracy of the network edges by non-parametric bootstrapping. The gray area represents the bootstrap 95% confidence intervals. Narrower confidence intervals indicate reliable accuracy.

Figure 4 Stability of the network. The stability of central and bridge strength by case-dropping bootstrap.

Centrality of the network

Centrality indices of all nodes were illustrated in Figure 2. According to three traditional centrality indices, node Se3 (strength_(rank) = 1, betweenness_(rank) = 4, closeness_(rank) = 6), node Ibl5 (strength_(rank) = 2, betweenness_(rank) = 1, closeness_(rank) = 4), node Ibl7 (strength_(rank) = 3, betweenness_(rank) = 6, closeness_(rank) = 2), node Si8 (strength_(rank) = 5, betweenness_(rank) = 2, closeness_(rank) = 1), node Se5 (strength_(rank) = 7, betweenness_(rank) = 5, closeness_(rank) = 3), and node P4 (strength_(rank) = 9, betweenness_(rank) = 7, closeness_(rank) = 11) exhibited the most significant position in the network. Node G1 had relatively high strength (strength_(rank) = 8) despite its low closeness and betweenness. In addition, the centrality differs test for strength indicated that node Se3 and node Ibl5 were statistically stronger than other nodes in the network (Supplementary Figure 1). Therefore, the central factors within the network were node Ibl5 “media exposure”, node Se3 “family health”, node Ibl7 “problematic internet use”, node Si8 “sub-health”, node Se5 “intimate relationship violence”, node P4 “tired or little energy”, and node G1 “nervous, anxious, on edge”.

Bridge centrality of the network

Figure 5 displayed the results of the bridge centrality of the network. For personal behavior and lifestyle community, node Ibl5 (bridge strength_(rank) = 1, bridge betweenness_(rank) = 1, bridge closeness_(rank) = 2) and node Ibl7 (bridge strength_(rank) = 4, bridge betweenness_(rank) = 4, bridge closeness_(rank) = 1) exhibited the highest bridge centrality. For the social environment community, node Se5 (bridge strength_(rank) = 2, bridge betweenness_(rank) = 6, bridge closeness_(rank) = 3) had the highest bridge centrality. For the subjective indicators community, node Si6 (bridge strength_(rank) = 5, bridge betweenness_(rank) = 3, bridge closeness_(rank) = 5) and node Si8 (bridge strength_(rank) = 6, bridge betweenness_(rank) = 2, bridge closeness_(rank) = 4) showed the highest bridge centrality compared with other nodes in the community. For the depression and anxiety community, node P9 (bridge strength_(rank) = 9, bridge betweenness_(rank) = 5, bridge closeness_(rank) = 6) had the highest bridge centrality. More detailed information on the centrality differs test for bridge strength was summarized in Supplementary Figure 2.

Figure 5 Bridge centrality indices of depression, anxiety, and their related factors. The bridge centralities were ranked in order, and the z-scores (not raw score) are represented.

The bridge centrality indices highlighted the significance of node Ibl5 “media exposure”, node Ibl7 “problematic internet use”, node Se5 “intimate partner violence”, node Si6 “health literacy”, node Si8 “suboptimal health status”, and node P9 “suicidal ideation”, indicating critical impacts on the interactions among different communities.

Network comparison of different subgroups

We compared the network model of depression-anxiety and their related factors across gender, ethnicity, residency, and living alone status. There were no statistical differences for the network global strength between males and females (17.125 vs 17.855, P = 0.168), Han and non-Han (18.248 vs 16.282, P = 0.574), urban area and rural area (17.271 vs 16.673, P = 0.703), and living alone and not living alone (16.764 vs 17.631, P = 1.000).

We further conducted disease-specific subgroup analyses focusing on cardiovascular diseases (CVDs) (hypertension, coronary heart disease, and stroke, n = 2381) and diabetes mellitus (n = 712) to address potential disease heterogeneity (Supplementary Figure 3). Crucially, node Ibl5 maintained its central role in both subgroups, demonstrating consistently high centrality [CVDs: (1) Strength_(rank) = 1; (2) Betweenness_(rank) = 1; and (3) Closeness_(rank) = 2; diabetes: (1) Strength_(rank) = 4; (2) Betweenness_(rank) = 3; and (3) Closeness_(rank) = 5] and bridge centrality [CVDs: (1) Bridge strength_(rank) = 1; (2) Bridge betweenness_(rank) = 1; and (3) Bridge closeness_(rank) = 2; diabetes: (1) Bridge strength_(rank) = 3; (2) Bridge betweenness_(rank) = 1; and (3) Bridge closeness_(rank) = 3]. The above results further emphasized the importance of node Ibl5 “media exposure”, indicating critical impacts on the interactions among different communities.

DISCUSSION

In this study conducted through a nationwide survey, we identified the prevalence of depression and anxiety among the Chinese population with NCDs and analyzed their related factors from a network perspective to provide a theoretical basis for healthcare to identify, prevent, and treat depression and anxiety.

Poor management of NCDs leads to negative outcomes such as reduced quality of life[58], increased healthcare utilization[59], and financial deterioration[60], making patients more susceptible to depression and anxiety. The prevalence estimates for depression (9.3%-23.0%) and anxiety (2.9%-8.8%) widely vary[61] with morbidity increasing dose-dependently with the number of NCDs. A survey based on 47 Low-income and middle-income countries globally showed that the prevalence of depression and anxiety in patients with a single NCD was 5.6% and 10.3%, respectively, and increased significantly with the number of NCDs (11.5% and 17.0%)[62]. Notably, our results suggested that the prevalence of depression in the Chinese population with NCDs was 58.97%, and the prevalence of anxiety was 56.02%, which is higher than previous study results.

Differences in population composition, assessment tools, survey timing, and the types of NCDs included in surveys may account for the variations in the prevalence of depression and anxiety. The sampling procedure was performed rigorously and the PHQ-9 and GAD-7 have been proven to be reliable and effective measures for depression and anxiety, ensuring a representative sample and reliability of the study results. Therefore, we speculated that the dynamic zero-tolerance quarantine policy maintained by China during the COVID-19 pandemic, a unique epidemic restrictions background, severely affected the mental health of Chinese residents[8], leading patients with NCDs to face more challenges in obtaining health support, including social distancing, transportation barriers, treatment burdens, and discontinuous care.

Studies have shown that the prevalence of depression and anxiety varies with the type and severity of comorbid NCDs. Individuals with inflammatory diseases such as asthma, angina, and chronic back pain were at particular risk of anxiety (odds ratio = 1.37-2.16)[62] and depression (odds ratio = 1.38-2.24)[63], suggesting that mental disorders may be related to chronic low-grade inflammation, cell-mediated immunity activation, and a compensatory anti-inflammatory reflex system characterized by negative immune regulatory processes[64]. The risk of depression and anxiety in patients with inflammation-related chronic diseases highlighted the importance of screening for mental disorders in patients suffering from cumulative chronic disease effects.

This study discovered that “tired or little energy”, and ”nervous, anxious, on edge” were core symptoms in the network, which is similar to the findings of Yang et al[65] in patients with breast cancer, indicating that these symptoms play an important role in activating and maintaining the psychopathological network of depression and anxiety. However, under similar survey conditions to our study, Peng et al[66] found that “restlessness” exhibited the highest centrality in the depression and anxiety network of Chinese nurses, reflecting the uniqueness of the Chinese population with NCDs.

Generally, patients with normal depression experienced an obvious energy decline[67], and patients with NCDs may face a greater degree of energy deficiency, possibly reflecting the heterogeneity of depression. However, it may also be a comorbid symptom of both somatic and psychiatric disorders. NCDs require long-term, sometimes lifelong, disease care. The primary care physicians are usually the first point of contact for the patients, and the relationship established between healthcare professionals and patients is the foundation for health monitoring, providing the necessary education on related pathologies and treatment adherence. However, pandemic restrictions and the saturation of the healthcare system during successive waves of infection led to social isolation and a lack of communication with healthcare professionals, resulting in discontinuity of care. When the most basic needs are unmet, persistent worry or anxiety is inevitable[68]. According to the fifth version of the Diagnostic and Statistical Manual of Mental Disorders, “persistent worry or anxiety” served as a core symptom for diagnosing generalized anxiety disorder, and this feature remained prominent when anxiety was included in a network.

Family health, media exposure, problematic internet use, suboptimal health status, and intimate relationship violence were core factors in the network. Family health was described as a resource at the family unit level, developed from the intersection of the health of each family member and their interactions and abilities with the family’s physical, social, emotional, economic, and medical resources[69]. Our results were generally consistent with previous studies, suggesting that a healthy family was essential in preventing physical and mental problems by establishing stronger, more harmonious intimate relationships and more effective family dynamics, regulating each other’s behaviors and providing information and encouragement to act more healthily[70,71].

Regarding the core role of social media exposure and problematic internet use, researchers have argued that excessive exposure to social media may replace activities that could be more beneficial to individuals, such as forming more important interpersonal relationships, achieving real goals, or even just valuable moments of reflection[72]. Additionally, patients’ depletion of resources due to illness led them to replenish resources through internet use, and overindulgence in the internet for obtaining new resources may generate negative emotions.

Our results also emphasized the prominent role of suboptimal health status in the network. It is worth mentioning that this survey was conducted in the late stages of the COVID-19 pandemic, and people tended to be in a suboptimal health status because of COVID-19 sequela with chronic fatigue, a series of physical symptoms, and mental problems persisting for more than 3 months[24], especially those with preexisting chronic pathologies[73].

Building upon the NTMD, Jones et al[74] expanded the NTMD. They indicated that nodes in the network should not be limited to symptoms but also include biological, cognitive, or other individual-level processes that might contribute to a disorder. Additionally, Jones et al[75] proposed bridge centrality (bridge strength, bridge betweenness, and bridge closeness) in networks composed of various disorders or communities to assess the importance of nodes. Our findings demonstrated that media exposure, problematic internet use, intimate relationship violence, health literacy, suboptimal health status, suicidal thoughts, and irritability exhibited high bridge centrality. They were generally similar to traditional centrality results above in terms of factor importance (e.g., they all emphasized the core role of media exposure, problematic internet use, intimate relationship and suboptimal health status). The high bridge centrality of suicidal thoughts suggests that it plays a crucial role in the progression and maintenance of mental health issues in NCDs. This finding aligns with previous research, which has highlighted the association between the severity of anxiety symptoms and suicidal thoughts and behaviors among patients with mood disorders[76]. Addressing suicidal thoughts early may help prevent the escalation of other interconnected mental health issues in this population.

Interestingly, the variables associated with social media (media exposure, problematic internet use) were more connected with the entire network. We speculated that social media and the internet were regarded as the main sources of information about the epidemic due to the social isolation policy[77]. There were serious concerns about misinformation and unverified rumors spreading rapidly through social media, causing distrust and posing additional challenges to public health efforts to combat the epidemic. Social media also overemphasized risks, and repeated exposure may increase panic and fear. Compared with traditional media, social media not only provides information but also allows personal sharing and emotional expression. During infectious disease pandemics negative emotions are more likely to spread on social media[78]. Most importantly, mass media information may alter individuals’ perceptions of disease susceptibility and severity. For patients with NCDs high concerns about diseases during a global pandemic may lead to feeling more threatened by the disease, resulting in depression and anxiety[79].

In our results the persistent centrality of media exposure observed consistently across various disease subgroups highlighted its broad influence on psychological well-being. This robust pattern suggests that interventions targeting media usage could provide generalized mental health improvements, irrespective of specific disease classifications. Our findings emphasized the importance of interventions to guide patients with NCDs in using social media correctly to obtain health information, carefully consider the authenticity and quality of health information, and handle information objectively.

Strengths and limitations

This study used a large sample size, including a nationwide sample of participants, manifesting the representativity and reliability of depression and anxiety prevalence among Chinese individuals with NCDs. Additionally, our research contained multidimensional factors, making the results more comprehensive and persuasive. Lastly, we innovatively applied network analysis, surpassing traditional regression methods, to visualize and quantify complex relationships between variables, identifying central symptoms and bridge factors to inform targeted interventions. Media exposure emerged as a unique bridge factor, reflecting China’s digital health information ecology, emphasizing the need for localized interventions.

However, several limitations of the current study deserve mention. Firstly, although participants with diagnosed cognitive impairment were excluded, the screening relied primarily on self-reported conditions rather than standardized clinical assessments. Importantly, we did not explicitly exclude individuals with preexisting mood disorders (e.g., depression or anxiety) or early-stage Alzheimer’s disease, potentially leading to underdetection of mild cognitive deficits (particularly among elderly respondents with neurodegenerative conditions) and a higher baseline prevalence of depression/anxiety in these subgroups that might influence overall estimates. Secondly, although we determined the importance of core and bridge factors based on NTMD and empirical research, the cross-sectional design limited the ability to infer causality. Therefore, further experimental or longitudinal studies are strongly warranted to validate whether interventions targeting these core or critical bridge factors will be more effective than others. Thirdly, the self-reported data may introduce information bias and social desirability effects. Objective measurements are desirable for future research. Finally, the interactions between factors in the network change over time and could not be addressed in our study. Consequently, future research should consider using longitudinal data-based NAMs or recursive network methods based on time series data.

CONCLUSION

The study investigated the prevalence of anxiety and depressive symptoms in individuals with NCDs and used NAM to explore the interactions (social environments, individual behaviors and lifestyles, and subjective indicators) between anxiety and depression symptoms and their associated factors. From a perspective of the interaction network, we applied network methods to explore the factors influencing depression and anxiety in patients with NCDs, identifying media exposure, problematic internet use, intimate relationship violence, and suboptimal health status as both core and bridge factors. Future interventions on depression and anxiety of patients with NCDs should focus on these specific and critical factors, which may be more efficient than current modalities.