Network analysis of multisite chronic pain in middle-aged and older adults in Western China

Abstract

Background

Multisite chronic pain (MCP) is a growing public health concern among aging populations worldwide, disproportionately affecting underdeveloped regions of low- and middle-income countries where pain awareness and management capacity remain limited. This study aimed to investigate MCP prevalence and co-occurrence patterns, identify key sites, and explore subgroup heterogeneity among middle-aged and older adults in Yunnan Province, western China.

Materials and Methods

A cross-sectional study was conducted using multi-stage stratified random cluster sampling to recruit 5,346 participants aged 45 years and older, and demographic characteristics and pain status across 30 body sites were collected using structured questionnaires. The Ising model was applied to eatimate pairwise associations between pain sites, and Network Comparison Test (NCT) was performed to evaluate pain pattern differences.

Results

The prevalence of MCP was 29.83%. Co-occurrence patterns were dominated by bilateral symmetry (e.g. left-right knee, weight = 4.93), followed by adjacent-region pain (e.g. right upper arm-right elbow, weight = 4.26). Conversely, contralateral adjacent sites showed negative association (e.g.,left calf-right knee, weight = −3.24). The left calf, right thigh, right knee, right calf, and left knee were identified as core pain sites, whereas the left hip, left thigh, neck, left buttock, and chest served as bridge nodes. Notably, no statistically significant differences in MCP networks were detected across gender, age, and ethnic subgroups.

Conclusions

MCP remains disproportionately prevalent and exhibits structured pain patterns. The identified network characteristics provide new insights into pain co-occurrence and may inform future research and pain management strategies in resource-limited settings.

Introduction

Chronic pain (CP), which is defined as persistent or recurrent pain over three months [1], exerts an enormous personal and economic burden worldwide, affecting approximately 30% of the world’s population [2]. About half of these individuals experience multisite chronic pain (MCP), which involves two or more body sites [3,4]. Research has demonstrated that pain originating from a single site is frequently interrelated with pain in other sites, often evolving into MCP [5–7]. MCP is associated with a wide range of adverse physical, psychological, and functional outcomes and imposes substantial healthcare and societal burdens [8–12]. Targeted interventions are critical to mitigate the public health burden of MCP.

Although previous research has revealed that pain most commonly co-occurred in mirrored contralateral sites [13], studies on the pain patterns of MCP are still limited. It remains unclear whether pain patterns differ across populations, despite the established disparities in MCP prevalence among women [14], older adults [15], ethnic minorities [16], and low-income groups [15]. Consequently, further investigation into pain co-occurrence patterns in MCP is warranted to improve our understanding of the distribution and interrelationships of pain sites.

Nationally, CHARLS data from 2018 showed a 60.02% prevalence of CP among Chinese adults aged ≥45, rising to 66.84% in western China and 65.07% among ethnic minorities [17]. Notably, 90% of older CP patients in ethnic minority areas experienced MCP [16], underscoring its high burden in western China’s ethnic minority regions. Yunnan Province, a less-developed region in western China characterized by substantial ethnic diversity, a large population residing in remote mountainous areas, limited healthcare access, and a high burden of musculoskeletal pain, provides a particularly relevant context for studying MCP among middle-aged and older adults [18,19]. Understanding MCP co-occurrence patterns in this population may help inform pain management strategies in similar underserved and low-resource settings.

Network analysis is highly suitable for the exploratory analysis of statistical association patterns in multivariate data [20] and has been increasingly applied to examine co-occurrence patterns in chronic conditions [21,22], including pain [13]. This approach provides a valuable framework for systematically examining interconnections among pain sites and visualizing the co-occurrence patterns in MCP.

This study used network analysis to examine the prevalence and co-occurrence patterns of MCP among middle-aged and older adults in a study sample from Yunnan Province, western China. Network characteristics were further compared across gender, age, and ethnic groups. The study seeks to inform the development of targeted management and intervention strategies and to provide empirical evidence for future research on pain patterns in MCP and their associations with disease.

Materials and methods

Research design and sampling

We conducted the China-Yunnan Multi-Ethnic Health Survey (CYMEHS), a cross-sectional survey of middle-aged and older adults (≥45 years) in Yunnan from July 2022 to January 2025. Using a multi-state stratified cluster sampling method, we selected nine prefecture-level cities, presenting diverse geography, economic development, and ethnic minority distribution from Yunnan’s 16 cities. From each city, one or two counties were randomly chosen; from each county, two to three townships; and from each township, two natural villages were randomly selected. All permanent residents aged ≥45 years in each village formed the sampling cluster. The inclusion criteria for participants included individuals aged 45 years or older residing locally for at least three months, who provided informed consent and voluntarily participated in the survey. Individuals with severe mental disorders that impair cognitive or communication abilities, or those who expressed unwillingness to participate, were excluded.

The sample size was estimated by using the formula: $N=\frac{z_{\alpha }^{2}P\left(1-P\right)}{{\delta }^2}\times d\mathrm{e}\mathit{\text{ff}}$. To estimate the prevalence of MCP, we employed a value of  P= 56.6% (highest reported prevalence [23]). The relative error (ε) was set at 5%, and the corresponding absolute error (δ) was calculated as δ = P × ε = 0.566 × 0.05 = 0.028. Assuming a design effect (deff) of 3, the initial sample size was estimated to be 3,621 participants. After making an adjustment for a projected 20% non-response rate, the final required sample size was determined to be 4,346 participants.

Measurements

Socio-demographic characteristics

The structured questionnaire used in this study consisted of two components: socio-demographic characteristics and pain assessment. The investigated socio-demographic characteristics included age, gender, ethnicity, marital status, educational level, occupation, and monthly household income. Three age groups were identified: those aged 45–59, 60–74, and ≥75. Gender was classified as male or female. Based on self-reported ethnic affiliation, ethnicity was divided into Han majority and ethnic minority groups. Marital status was categorized as married, unmarried, widowed, divorced, and other. Educational level was treated as a binary variable: primary school or below, and junior high school or above. Seven categories were used to group occupations: Technicians/office workers/civil servants, farmers, laborers, retired, self-employed, unemployed, and others. The monthly household income was categorized into three groups: ≥11,000 RMB, 3,000–10,999 RMB, and <3,000 RMB. This categorization was based on official statistics from 2022, which indicated that monthly disposable income for rural households was 3,786 RMB, while that for urban households was 10,542 RMB [24].

Multisite chronic pain

In this study, MCP was precisely defined as persistent or recurrent pain that occurs in two or more distinct anatomical locations. A structured questionnaire, developed with strict reference to the Brief Pain Inventory (BPI) [25], was employed to comprehensively assess pain in this investigation. Initially, in line with the International Association for the Study of Pain’s (IASP) [1] diagnostic criteria for CP, participants were queried, ‘Do you frequently experience distress from pain lasting ≥3 months?’ Subsequently, patients diagnosed with CP were requested to clearly mark the sites of their pain on a pain map. This pain map was carefully modeled after the pain manikin from the the Pain Standard Evaluation Questionnaire (SEQ Pain) [26], as depicted in Figure 1. The pain manikin consists of a pair of lines that accurately represent the front and back contours of the body, allowing participants to precisely mark the areas where they have felt pain. Finally, pain intensity was assessed through a combined approach that incorporated the Numerical Rating Scale (NRS) and the Faces Pain Scale - Revised (FPS-R) [27]. The NRS, with a scale ranging from 0 to 10, classifies pain into four distinct grades: no pain (0), mild (1–3), moderate (4–6), and severe (7–10). The FPS-R visually represents different levels of pain intensity using six facial expressions, ranging from a smiling face to a crying face.

Figure 1.

Pain map showing 30 body sites.

Numbers correspond to body sites as follows: 1, head; 2, face; 3, neck; 4, chest; 5, upper abdomen; 6, lower abdomen; 7, back; 8, waist; 9, left buttock; 10, right buttock; 11, left shoulder; 12, left upper arm; 13, left elbow; 14, left forearm; 15, left palm; 16, right shoulder; 17, right upper arm; 18, right elbow; 19, right forearm; 20, right palm; 21, left hip; 22, left thigh; 23, left knee; 24, left calf; 25, left foot; 26, right hip; 27, right thigh; 28, right knee; 29, right calf; 30, right foot.

Data collection and quality control

Prior to commencing the fieldwork, all interviewers underwent standardized training encompassing research ethics, informed consent procedures, questionnaire administration techniques, and confidentiality protocols. Data collection was carried out through household visits executed by trained investigators. Structured questionnaires were employed in face-to-face interviews, and participant responses were recorded in real-time. To encourage the participation of ethnic minority populations, trained interpreters provided on-site translation services in minority languages during the interviews. Once the interviews were completed, the questionnaires were promptly retrieved on-site, reviewed for comprehensiveness and precision, and tallied for final verification. In total, 5,346 questionnaires were collected. After excluding 39 invalid questionnaires due to incomplete responses or logical inconsistencies, 5,307 valid surveys were retained for analysis, resulting in an effective response rate of 99.27%. Ultimately, all data were entered into EpiData 3.2 and subjected to verification checks to guarantee data accuracy and reliability. No missing values were observed in the variables included in the final analytical dataset.

Statistical analysis

Descriptive analysis was performed using Stata software (version 18.0), with categorical data presented as frequencies (n) and percentages (%). Chi-square tests were employed to compare the proportions of participants with MCP within each socio-demographic characteristic, with the significance level set at α = 0.05.

Network analysis was conducted using R statistical software (version 4.5.0). Due to pain-site variables were binary (i.e. ‘pain’ or ‘no pain’), the Ising model was selected as an appropriate network analysis approach [28,29]. First, the ‘bootnet’ and ‘qgraph’ R packages were used to estimate and visualize the network of MCP. Specifically, this study employed the IsingFit estimator implemented in the estimateNetwork() function of the R package ‘bootnet’ [30] to estimate a Least Absolute Shrinkage and Selection Operator (LASSO) [31]-regularized Markov Random Field (MRF) network based on the Ising model [28,29]. This network quantified conditional dependencies among pain sites, representing the statistical associations between pairwise pain sites after controlling for all other regions [32]. Pseudo-likelihood estimation was used to model the conditional probability of pain presence for each site using logistic regression with LASSO penalization to induce network sparsity. The optimal penalty parameter was selected based on the Extended Bayesian Information Criterion (EBIC) [33], yielding a highly interpretable network.

Subsequently, to identify potentially important features of the network, centrality indices were computed for each node, including strength, betweenness, and closeness [30]. However, previous methodological studies have indicated that betweenness and closeness centrality may exhibit relatively low reliability when assessing node importance [34]. In contrast, strength centrality, defined as the sum of the absolute values of the edge weights connected to a node, is considered a more stable measure [35]. Therefore, strength centrality was used to identify core pain sites within the network. Similarly, to identify nodes that connect the four anatomical groups within the MCP network, bridge centrality analysis was performed. Previous research has demonstrated that bridge strength shows superior stability compared with other bridge centrality metrics [36]. Accordingly, bridge strength was estimated using the ‘networktools’ R package to identify bridge pain sites within the network.

Following this, the ‘bootnet’ R package was applied to evaluate the accuracy of the pain network structure and the stability of the estimated centrality indices, based on non-parametric bootstrap and case-dropping bootstrap methods [30,37]. Simultaneously, bootstrapped difference tests were employed to assess differences among edge weights and among node centrality indices [30]. Lastly, distinct networks were constructed by gender (male/female), age group (45–59/60–74/≥75 years), and ethnicity (Han majority/ethnic minority) in order to compare MCP networks across various subgroups. The NCT() function from the ‘NetworkComparisonTest’ R package was applied to test for differences in global network structure, overall strength, and centrality indices across gender, age, and ethnic subgroups [38]. To account for multiple comparisons conducted across the three age groups, a bonferroni correction was applied, resulting in an adjusted significance level of 0.017 (0.05/3).

Results

The prevalence of MCP

A total of 5,307 respondents were incorporated into the final data analysis. Among them, 2,311 (43.54%) were male, and 2,996 (56.45%) were female. The age distribution was as follows: 40.95% (n = 2,173) were aged between 45 and 59 years, 45.09% (n = 2,393) were aged between 60 and 74 years, and 13.96% (n = 741) were 75 years old or above. The sample consisted of 2,553 (48.11%) individuals from the Han majority and 2,754 (51.89%) individuals from ethnic minorities. In terms of marital status, 76.90% (n = 4,081) were married. The majority of participants (73.58%) had an education level of primary school or below. Regarding occupation, farmers accounted for 63.31% of the sample (n = 3,360). With respect to monthly household income, 43.49% (n = 2,308) reported an income of less than 3,000 RMB, 50.03% (n = 2,655) had an income within the range of 3,000–10,999 RMB, and 6.48% (n = 344) had an income exceeding 11,000 RMB.

Overall, 51.46% of participants reported chronic pain, including 21.63% with single-site pain and 29.83% with multisite pain. The most commonly affected sites among MCP participants were the waist (11.74%), back (10.50%), knees (left 10.01%, right 9.29%), and neck (8.94%). The average number of painful sites was 3.50. The distribution of MCP varied across socio-demographic characteristics within the study sample (Table 1). Among participants with MCP, women accounted for 65.19% of cases, individuals aged 60–74 years represented 50.09%, ethnic minorities comprised 54.64%, and married individuals accounted for 73.22% of the MCP group. Additionally, the majority of MCP participants were farmers (68.16%), had a primary school education or below (83.58%), and reported a monthly household income below 3,000 RMB (50.47%). Regarding pain intensity, mild (49.08%) and moderate pain (47.13%) were the most commonly reported levels among participants with MCP. The distributions of participants with and without MCP differed significantly across these categories (all p < 0.05, Table 1).

Table 1.

Distribution of socio-demographic characteristics among participants with and without MCP [n (%)].

Variable	Total (n = 5307)	Not suffering from MCP (n = 3724)	Suffering from MCP (n = 1583)	χ2	p
Gender				70.08	<0.001
Male	2311(43.54)	1760(47.26)	551 (34.81)
Female	2996(56.45)	1964(52.74)	1032(65.19)
Age				71.59	<0.001
45–59	2173(40.95)	1659 (44.55)	514 (32.47)
60–74	2393(45.09)	1600(42.96)	793 (50.09)
≥75	741(13.96)	465(12.49)	276(17.44)
Ethnic groups				6.83	0.009
Han majority	2553(48.11)	1835(49.27)	718 (45.36)
ethnic minority	2754(51.89)	1889(50.73)	865(54.64)
Marital status				48.28	<0.001
Married	4081(76.90)	2922(78.46)	1159(73.22)
Not married	81(1.53)	63(1.69)	18(1.14)
Widowed	1013(19.09)	631(16.94)	382(24.13)
Divorced	129(2.43)	107(2.87)	22(1.39)
Others	3(0.06)	1(0.03)	2(0.13)
Level of education				115.90	<0.001
Primary school or below	3905(73.58)	2582(69.33)	1323(83.58)
Junior high school or above	1402(26.42)	1142(30.67)	260 (16.42)
Occupation				99.84	<0.001
Technician/office workers/civil servant	168(3.17)	144(3.87)	24(1.52)
Farmer	3360(63.31)	2281(61.25)	1079(68.16)
Worker	257(4.84)	191(5.13)	6(4.17)
Retired	640(12.06)	497(13.35)	143(9.03)
Self-employed	357(6.73)	295(7.92)	62(3.92)
Unemployed	516(9.72)	309(8.30)	207(13.08)
Others	9(0.17)	7(0.19)	2(0.13)
Monthly household income (RMB)				53.02	<0.001
<3000	2308(43.49)	1509(40.52)	799(50.47)
3000 ∼ 10999	2655(50.03)	1938(52.04)	717(45.29)
≥11000	344(6.48)	277(7.44)	67(4.23)
Intensity of pain				2200	<0.001
No pain	2754(48.50)	2574(69.12)	0(0.00)
Mild pain	1494(28.15)	717(19.25)	777(49.08)
Moderate pain	1146(21.59)	400(10.74)	746(47.13)
Severe pain	93(1.75)	33(0.89)	60(3.79)

MCP refers to multisite chronic pain.

Network analysis

Network structure

As depicted in the MCP network (Figure 2), the 30 body sites were categorized into four anatomical groups: ‘head and neck’, ‘trunk’, ‘upper limbs’, and ‘lower limbs’. The network comprised 30 nodes and 62 non-zero edges. An analysis of the correlations between painful body sites indicated that the strongest pain associations occur between bilaterally corresponding body sites, followed by significant correlations between neighboring pain sites of the upper and lower limbs. The five edges with the strongest positive connections were the left knee and right knee (weight = 4.93), the left thigh and right thigh (weight = 4.87), the left foot and right foot (weight = 4.54), the left palm and right palm (weight = 4.47), and the left calf and right calf (weight = 4.28). Simultaneously, negative edges were identified in contralateral non-corresponding sites, with the strongest negative associations being observed between regions adjacent to their contralateral counterparts, specifically the left calf and right knee (weight = −3.24), the left foot and right calf (weight = −2.38), and the left knee and right thigh (weight = −2.37). The weights of all edges are provided in Supplementary Table S1.

Figure 2.

Network of MCP.

Nodes represent individual pain sites, and edges represent associations between nodes. Blue edges indicate positive relationships, whereas red edges indicate negative relationships. Edge thickness and color intensity reflect the magnitude of the edge weights.

Analysis of node centrality and bridge centrality

Centrality metrics were computed for each node. A node is considered a core pain site of the network when its centrality value is higher, which indicates greater relevance and a larger influence on pain in other sites [35]. Figure 3A presents the centrality results for the MCP network. The core sites of MCP network were determined by the significant strength centrality demonstrated by the left calf (strength = 18.63), right thigh (strength = 15.58), right knee (strength = 14.94), right calf (strength = 14.90), and left knee (strength = 12.76). The bridge centrality results are presented in Figure 3B. The left hip (bridge strength = 6.22), left thigh (bridge strength = 4.09), neck (bridge strength = 3.47), left buttock (bridge strength = 3.41), and chest (bridge strength = 3.41) exhibited high bridge strength, identifying them as bridge pain sites. For details on the strength and bridge strength of each node, see Supplementary Tables S2–S3.

Figure 3.

Node centrality and bridge centrality in the MCP network.

(A) Centrality. (B) Bridge centrality.

Evaluation of network accuracy and stability

The stability assessment of the MCP network is presented in Figure 4A. The observed close correspondence between the two lines, narrow confidence intervals, and the fact that both the red and black lines predominantly lie within these intervals suggest an accurate estimation of the network edges. Meanwhile, the correlation stability coefficient (CS-C) for strength and bridge strength were 0.75 and 0.517 respectively, both exceeding 0.5 (Figure 4B). This indicates that over 50% of the sample can be removed without significantly impacting the network structure, demonstrating that node centrality metrics were estimated stably. In addition, the bootstrap difference test for both edge weights and nodes (Supplementary Figure S1-S2) indicated that the network exhibited good stability.

Figure 4.

Evaluation of the accuracy and stability of the MCP network.

(A) Accuracy of edge weights. The black line indicates the mean edge weights estimated using the bootstrap method, while the red line represents the original sample weights. The gray area shows the 95% confidence interval (CI) for edge weights obtained from the non-parametric bootstrap.

(B) Stability of node centrality metrics.

Network comparison

Separate MCP networks were constructed according to gender, age group, and ethnicity. In the gender-stratified network comparison, no statistically significant differences were detected in the network’s global strength (male: 83.87, female: 97.80, S = 13.93, p = 0.66) or in the network structure, specifically, the distribution of edge weights (M = 1.79, p = 0.87, Supplementary Figure S3-S7). Similarly, comparisons across age-group also revealed no significant differences (p > 0.017, Supplementary Figure S8-S12). In the comparison based on ethnicity, no significant differences were noted in the network’s global strength (Han majority: 91.58, ethnic minority: 95.45, S = 3.87, p = 0.75) or the distribution of edge weights (M = 2.31, p = 0.30, Supplementary Figure S13-S17).

Discussion

This study employed network analysis to investigate pain co-occurrence patterns among individuals with MCP, providing additional insights into the complex interrelationships among pain sites. The prevalence of MCP among middle-aged and older adults from Yunnan Province was 29.83% in this study sample, which is comparable to estimates reported in the United Kingdom (33%) [39] and Guiyang (32.8%), China [40]. Although it is slightly lower than the 34.97% reported among older adults aged 60 years and above in Shanghai, China in 2023 [41], this finding further highlights the considerable burden of MCP across different contexts. The socio-demographic profile of individuals with MCP in our study sample was also consistent with previous epidemiological evidence [14,15,42]. Specifically, women, older adults, ethnic minority groups, farmers, individuals with lower educational attainment, and those from lower-income households were more frequently represented among participants with MCP. Consistent with previous reports [43], the waist was the most commonly affected pain site, followed by the back and knees. In addition, most individuals with MCP reported mild-to-moderate pain intensity, a pattern that aligns with findings from earlier studies [7].

Based on the network analysis, the most prominent pain co-occurrence patterns in MCP were observed between corresponding bilateral sites in the upper and lower limbs, followed by adjacent body sites. This pattern is consistent with previous studies of pain co-occurrence [5,13]. However, some differences were observed when compared with other populations. A longitudinal study using data from the Netherlands Twin Registry (NTR) reported strong tetrachoric correlations primarily between adjacent pain sites and did not identify the bilateral symmetry observed in our study [6]. One possible explanation is that the NTR study assessed pain across only eight body regions (head, neck, back, abdomen, joints, chest, teeth, and face), which may have limited its ability to detect bilateral pain patterns in the limbs. Similarly, in a Danish study, the strongest associations were observed between the left and right hips and between the left and right buttocks [13], whereas the most pronounced associations in our study occurred between corresponding bilateral limb regions. Although the reasons for these differences remain uncertain, variations in age composition between study populations may have contributed. Participants in our study were substantially older (mean age = 62.69 years) than those in the Danish cohort (mean age = 45.6 years) and may therefore have had a greater burden of age-related musculoskeletal conditions [44].

Several hypotheses may help explain the pain co-occurrence patterns observed in our study. For example, previous studies have suggested that repetitive occupational activities and altered pain-processing mechanisms may be associated with multisite pain. Middle-aged and older adults in sample population are often engaged in physically demanding agricultural work, which could potentially contribute to pain occurring in symmetrical or anatomically adjacent regions. Furthermore, studies on central sensitization (CS) and experimental pain models have reported the occurrence of contralateral pain hypersensitivity [45,46], which may be relevant to some of the co-occurrence patterns observed in the present network. However, these interpretations should be regarded as tentative, as occupational exposures and physiological pain mechanisms were not directly assessed in the present study. Future studies incorporating detailed occupational assessments and mechanistic measures are needed to evaluate these hypotheses.

Notably, the MCP network constructed on the basis of the Ising model revealed negative associations between contralateral non-corresponding sites, with the strongest negative edges observed between adjacent contralateral regions. This pattern may indicate inhibitory conditional associations between pain sites. Similar findings have also been reported in a Danish study [13]. Several explanations may account for these negative associations. One possibility involves nociception-activated inhibitory processes, such as diffuse noxious inhibitory controls (DNIC) [47], also referred to as conditioned pain modulation (CPM). These mechanisms have been proposed to inhibit pain perception in one body region following noxious stimulation in another and have been supported by previous experimental studies [48,49]. However, this interpretation remains speculative because the present study did not directly assess pain modulation mechanisms. Moreover, negative edges may arise from factors beyond physiological inhibition, including statistical properties of the model such as statistical suppression, conditioning artifacts, or insufficient sample sizes. Therefore, the observed negative associations should be interpreted cautiously. Nevertheless, these findings provide additional insight into the complexity of pain co-occurrence in MCP and may help generate hypotheses for future research.

In the MCP network, the left calf, right thigh, right knee, right calf, and left knee exhibited the highest strength centrality, indicating a higher likelihood of pain co-occurrence (or inhibition) with other sites, and identifying them as core nodes within the network. The prominence of these sites may partly reflect the substantial burden of chronic diseases among middle-aged and older adults in rural Yunnan. Previous research has reported that the prevalence of chronic diseases among middle-aged and older adults aged 45 years and above in Yunnan Province was 57.91%, with rheumatoid arthritis (14.06%), diabetes (7.82%), gout (3.73%), dyslipidemia (3.14%), and osteoporosis (1.81%) being particularly common[18]. These conditions, as well as their associated vascular and neuropathic complications, frequently involve the lower extremities, which may help explain the central role of lower-limb pain sites observed in the present network. In addition, this study further revealed that the left hip, left thigh, neck, left buttock, and chest serve as bridge nodes. Previous network studies have suggested that bridge symptoms may facilitate interactions between symptom clusters and may therefore influence the overall organization of symptom networks [32,50]. Given their prominent positions within the network, both the core pain sites and bridge nodes identified in the present study may warrant particular clinical attention and represent potentially important targets for future assessment, management, and intervention strategies in individuals with MCP.

This study has several limitations. First, self-reported pain data may be influenced by recall and information bias. Second, the cross-sectional design precludes inferences regarding temporal or causal relationships within the MCP network. The observed associations reflect pain co-occurrence at a single time point and cannot determine whether central or bridge nodes contribute to the development or progression of MCP. Third, because non-response was extremely rare and not systematically documented during data collection, sampling weights could not be constructed retrospectively. Consequently, prevalence estimates and socio-demographic comparisons were based on unweighted data and may not fully represent the underlying population of middle-aged and older adults in Yunnan Province. Additionally, the study population primarily comprised middle-aged and older adults engaged in agricultural labor, which may limit the external validity of the findings. Furthermore, subgroup differences were not detected in the NCT, possibly due to limited statistical power to detect subtle differences in local network characteristics, particularly in smaller subgroups. Finally, pain sites were analyzed as binary variables without accounting for differences in pain intensity, which could have influenced the estimated network structure and resulted in an incomplete representation of pain co-occurrence patterns. Future studies should address these limitations by employing longitudinal designs, recruiting more diverse populations, improving the documentation of sampling procedures and non-response information to facilitate weighted analyses and population-level inference, and incorporating pain intensity measures into network analyses.

Conclusion

Among middle-aged and older adults in Yunnan Province, China, MCP exhibited a structured rather than random pattern of pain co-occurrence, characterized by bilateral symmetry, anatomical proximity, and negative associations between contralateral adjacent sites. Core pain sites were primarily concentrated in the lower extremities (left calf, right thigh, right knee, right calf, and left knee), whereas bridge nodes were mainly located in the trunk and proximal lower-limb regions (left hip, left thigh, neck, left buttock, and chest). These findings provide novel insights into the spatial organization of pain co-occurrence in MCP and may help inform future research and more targeted pain assessment and management strategies, particularly in resource-limited settings.

Funding Statement

This study was supported by Yunnan Fundamental Research Projects (202401AT070178); Yunnan High-level Talent Introduction Plan (YNQR-QNRC-2018-140); Yunnan Philosophy and Social Science Innovation Team (2024CX08); First-Class Discipline Team of Kunming Medical University (2024XKTDTS16); Science Research Fund Project of Yunnan Provincial Department of Education (2026Y0399).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethics approval and consent to participate

CYMEHS was approved by the Ethic Committee of Kunming Medical University (KMMU2021MEC095). At the onset of the research, the researchers elucidated the project to the participants. The information encompassed the objectives of the study, potential merits and demerits of participation, the expected benefits of the research, principles of privacy and confidentiality, and a statement of voluntary participation. Participants were additionally informed that they had the right to withdraw from the study at any time. The study was conducted in accordance with the Declaration of Helsinki, and all participants signed informed consent forms.

Data availability statement

The datasets generated and analyzed during the current study are not publicly available due to privacy considerations. Participants were informed at the time of data collection that their data would not be made publicly available to ensure confidentiality. However, the data are available from the corresponding author upon reasonable request, subject to compliance with applicable privacy regulations and consent agreements.