Unconventional multi-morbidity patterns in older adults: results from 39,172 Chinese health examination population at primary healthcare centers Authors: Kangjun Wu, Yujian Lu, Anping Liu, Yuefang Wang, Xiandong Shen Published: 2026 ABSTRACT BackgroundMulti-morbidity is prevalent among the elderly, yet specific co-occurrence patterns remain unclear. While prior studies have predominantly focused on cardiometabolic disease clusters, this study aimed to explore the proportion of recorded diagnoses of common diseases and identify unconventional multi-morbidity patterns among older adults at primary healthcare settings using electronic health records.MethodsWe conducted a cross-sectional study using data from regional health information platform in eastern China. Participants aged ≥ 65 years who underwent at least one health examination at primary healthcare setting in 2022 were included. Association rule mining integrated with the hypertext-induced topic search framework was employed to identify disease co-occurrence patterns, weighting participants and diseases based on network connectivity. Twelve common diseases were analyzed, stratified by sex and age.ResultsThere were 39,172 individuals enrolled in this study. Significant sex and age differences were found in most diseases, with hypertension, hyperlipidemia, cataract and other diseases accounting for 56.48%, 36.02%, 21.33% and 13%-20% of the overall proportion of recorded diagnoses. A total of 58 association rules met the predefined thresholds among the whole population. Beyond cardiometabolic clusters (hypertension or hyperlipidemia co-occurrence), we identified strong associations within inflammatory diseases (osteoporosis and arthritis) and between gastrointestinal conditions and sleep disorders (chronic gastritis, constipation, and sleep disorder). Patterns were consistent across sex and age subgroups.ConclusionMulti-morbidity patterns in Chinese older adults extend beyond traditional cardiovascular risks to include inflammatory and gut-sleep clusters. These findings suggest the potential value of considering cluster-based care pathways and may inform the integration of sleep screening into gastrointestinal management in primary healthcare settings.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12877-026-07387-w. Introduction Multi-morbidity refers to the physical or mental state in which an individual is diagnosed with two or more diseases simultaneously [1]. Previous studies indicated that approximately one-third of adults worldwide had multi-morbidity, and more than half of adults with chronic diseases experienced multi-morbidity, particularly in low-income countries [2–4]. Multi-morbidity was prevalent among older adults, with a gradual increase in prevalence with ageing [5]. In China, the pooled proportion of individuals aged 60 and above with multi-morbidity exceeded 35% from 2000 to 2022 [6]; however, according to two national cohort studies in 2018, the prevalence was more than 65% [7, 8], revealing that there was a trend of growth over time. Moreover, multi-morbidity patterns exhibited strong clustering characteristics among different organs or systems across ethnicities. The dominant multi-morbidity pattern was cardio-metabolic diseases in East Asia [9, 10], whereas cardiac diseases and mental disorders were more frequent in several European countries [11, 12]. Thus, specific joint measures should be adopted to delay the onset and progression of diseases in multi-morbidity prevention. Therefore, exploring the patterns of multi-morbidity could facilitate the formulation of reasonable early prevention measures or treatment approaches. Currently, most studies on multi-morbidity have been conducted through large-scale population surveys and investigations into the risk factors of chronic diseases. However, the implementation of these special surveys often required numerous personnel, materials, and time costs, and were difficult to capture real-time trends in the prevalence of diseases among the population [6]. Additionally, issues such as recall bias, inadequate representation of the target population, and difficulties in verifying diseases would arise during the field survey process. With the increasing level of healthcare digitalization, real-world studies based on electronic health records (EHRs) have proven to be more convenient and efficient than conventional surveys. Unlike latent class or network analysis which identify symmetric clusters, association rule mining (ARM) generates directional rules that offer actionable insights for sequential screening. However, traditional data mining methods often treat all patients equally, ignoring the heterogeneity in multi-morbidity burden. To address this, we employed ARM integrated with hypertext-induced topic search (HITS) weighting to prioritize patterns associated with individuals having higher disease connectivity. In this study, we obtained EHRs from the Regional Health Information Platform (RHIP), which integrates records generated from various medical and healthcare activities. We focused on the twelve most frequently diagnosed conditions to ensure robust statistical power and clinical relevance. This selection criterion aligns with standard ARM methodology, which requires sufficient prevalence to generate reliable association rules. Therefore, we explored the overall proportion of recorded diagnoses of twelve common diseases and multi-morbidity patterns to provide valuable insights for healthcare administration and multi-morbidity prevention. Methods Study design and population We conducted a cross-sectional study based on the electronic health records from the RHIP in eastern China. The RHIP is a comprehensive system integrating with community health records, including basic demographic characteristics, physical and laboratory examination results, cause of hospitalization, lifestyle, and history of medication. In China, individuals aged ≥ 65 are eligible for a free annual health examination at primary healthcare centers. Participants aged ≥ 65 years who underwent at least one physician-supervised physical and laboratory examination at primary healthcare centers between January 1, 2022, and December 31, 2022, were enrolled. For individuals with multiple examinations within the study period, the most comprehensive record was retained to avoid duplication. Diagnoses were not aggregated across visits; instead, the single record with the most complete diagnostic summary was selected for analysis. In addition, we excluded individuals whose examination records only included physiological and biochemical measurements such as blood pressure and glucose. This decision was made because when a record contained only these routine indicators, it was not possible to judge whether the individual had certain diseases. Therefore, this exclusion criterion also served to remove records with minimal information from the dataset ensuring that the disease status of every included participant was ascertainable (Fig. 1).Fig. 1Flow diagram for study population selection. RHIP: regional health information platform Flow diagram for study population selection. RHIP: regional health information platform Data processing Eligible records were required to include a diagnostic summary provided by the chief physician, which integrated examination findings into clinical diagnoses. Direct ICD codes were not available, so disease identification was based on these textual diagnostic conclusions. The diagnostic summary of each record was segmented into independent texts or items based on punctuation marks or special symbols (e.g., commas, semicolons). These segments were then matched to the preferred term (PT) level classifications in the Medical Dictionary for Regulatory Activities (MedDRA version 26.0, International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use) using the SAS software (version 9.4, SAS Institute Inc). Matching was performed to facilitate subsequent review via the COMPLEV function, which calculates the Levenshtein edit distance between two strings. Crucially, all text matches and diagnostic conclusions were manually validated by a licensed medical professional to ensure the accuracy of the final disease status. All text processing was conducted in Chinese. Statistical analysis The 12 most frequently diagnosed diseases in the study sample were selected for analysis, ranked by their recorded diagnosis proportions. These included: arrhythmia, arthritis, bronchitis, cataract, chronic gastritis, constipation, diabetes mellitus, hyperlipidemia, hypertension, osteoporosis, periodontitis, and sleep disorders. The proportion of individuals with a recorded diagnosis for each disease was calculated within the study sample, stratified by sex (male and female) and age groups (65–69 years and ≥ 70 years). Group differences were assessed using two-tailed Fisher exact test at a significance level of 0.05. A disease-participant matrix (0: not diagnosed; 1: diagnosed) was constructed for ARM. ARM was employed to identify disease co-occurrence patterns. A rule formatted as disease A → B indicates that individuals diagnosed with disease A (antecedent) are more likely to have disease B (consequent). Rule reliability was evaluated using three metrics:Support: The proportion of participants diagnosed with both diseases A and B.Confidence: The conditional probability of disease B given disease A.Lift: The ratio of support to the disease B proportion in the entire population (lift > 1.0 indicates a positive association). Support: The proportion of participants diagnosed with both diseases A and B. Support: The proportion of participants diagnosed with both diseases A and B. Confidence: The conditional probability of disease B given disease A. Confidence: The conditional probability of disease B given disease A. Lift: The ratio of support to the disease B proportion in the entire population (lift > 1.0 indicates a positive association). Lift: The ratio of support to the disease B proportion in the entire population (lift > 1.0 indicates a positive association). Rules were filtered using minimum thresholds (support: 1%, confidence: 50%, lift: 1.5) to prioritize clinically relevant patterns. To address limitations of traditional ARM (e.g., equal weighting of all diseases) [13], we integrated the equivalence class transformation (Eclat) algorithm with the HITS framework [14] for weighting every item (diseases) and transaction (participants). In this network-based approach, participants were treated as “hubs” and diseases as “authorities”. The HITS algorithm iteratively calculated weights for each participant and disease based on their connectivity within the network: participants diagnosed with multiple central diseases received higher hub weights, and diseases occurring in highly connected participants received higher authority weights. These hub weights were then integrated into the Eclat algorithm to adjust the support calculation, effectively prioritizing patterns associated with individuals having a higher multi-morbidity burden. Stratified analyses by sex and age were conducted to explore subgroup-specific patterns. All analyses were performed in R (version 4.3.2; R Development Core Team) using the arules [15] packages. Results Frequencies of recorded diagnoses for 12 diseases Hypertension was the most frequently recorded diagnosis in the study sample (56.48%), followed by hyperlipidemia (36.02%), cataract (21.33%), sleep disorders (19.95%), and bronchitis (18.59%) (Table 1). Higher frequencies of recorded diagnoses were associated with increased co-occurrence rates between diseases (Fig. 2).Table 1Proportions of recorded diagnoses for 12 diseases in older adults [n (%)]DiseasesOverallN = 39,172SexAge group (years old)MaleN = 18,287FemaleN = 20,885p65 ~ 69N = 15,156 ≥ 70N = 24,016pHypertension22,123 (56.48)10,511 (57.48)11,612 (55.60)< 0.0017497 (49.47)14,626 (60.90)< 0.001Hyperlipidemia14,109 (36.02)6042 (33.04)8067 (38.63)< 0.0015152 (33.99)8957 (37.30)< 0.001Cataract8356 (21.33)3700 (20.23)4656 (22.29)< 0.0012845 (18.77)5511 (22.95)< 0.001Sleep disorder7816 (19.95)3202 (17.51)4614 (22.09)< 0.0012589 (17.08)5227 (21.76)< 0.001Bronchitis7281 (18.59)3549 (19.41)3732 (17.87)< 0.0012648 (17.47)4633 (19.29)< 0.001Arthritis6990 (17.84)3270 (17.88)3720 (17.81)0.862569 (16.95)4421 (18.41)< 0.001Diabetes mellitus6816 (17.40)3157 (17.26)3659 (17.52)0.512230 (14.71)4586 (19.10)< 0.001Arrhythmia6029 (15.39)2605 (14.25)3424 (16.39)< 0.0011848 (12.19)4181 (17.41)< 0.001Periodontitis5999 (15.31)2868 (15.68)3131 (14.99)0.062227 (14.69)3772 (15.71)0.007Chronic gastritis5370 (13.71)2529 (13.83)2841 (13.60)0.522058 (13.58)3312 (13.79)0.56Osteoporosis5306 (13.55)2038 (11.14)3268 (15.65)< 0.0011871 (12.34)3435 (14.30)< 0.001Constipation5296 (13.52)2398 (13.11)2898 (13.88)0.0281757 (11.59)3539 (14.74)< 0.001The differences of diseases’ proportions were determined by the Fisher’s exact testFig. 2Frequency ranking of common disease diagnoses (Top 30 conditions with single and paired diseases) Table 1Proportions of recorded diagnoses for 12 diseases in older adults [n (%)]DiseasesOverallN = 39,172SexAge group (years old)MaleN = 18,287FemaleN = 20,885p65 ~ 69N = 15,156 ≥ 70N = 24,016pHypertension22,123 (56.48)10,511 (57.48)11,612 (55.60)< 0.0017497 (49.47)14,626 (60.90)< 0.001Hyperlipidemia14,109 (36.02)6042 (33.04)8067 (38.63)< 0.0015152 (33.99)8957 (37.30)< 0.001Cataract8356 (21.33)3700 (20.23)4656 (22.29)< 0.0012845 (18.77)5511 (22.95)< 0.001Sleep disorder7816 (19.95)3202 (17.51)4614 (22.09)< 0.0012589 (17.08)5227 (21.76)< 0.001Bronchitis7281 (18.59)3549 (19.41)3732 (17.87)< 0.0012648 (17.47)4633 (19.29)< 0.001Arthritis6990 (17.84)3270 (17.88)3720 (17.81)0.862569 (16.95)4421 (18.41)< 0.001Diabetes mellitus6816 (17.40)3157 (17.26)3659 (17.52)0.512230 (14.71)4586 (19.10)< 0.001Arrhythmia6029 (15.39)2605 (14.25)3424 (16.39)< 0.0011848 (12.19)4181 (17.41)< 0.001Periodontitis5999 (15.31)2868 (15.68)3131 (14.99)0.062227 (14.69)3772 (15.71)0.007Chronic gastritis5370 (13.71)2529 (13.83)2841 (13.60)0.522058 (13.58)3312 (13.79)0.56Osteoporosis5306 (13.55)2038 (11.14)3268 (15.65)< 0.0011871 (12.34)3435 (14.30)< 0.001Constipation5296 (13.52)2398 (13.11)2898 (13.88)0.0281757 (11.59)3539 (14.74)< 0.001The differences of diseases’ proportions were determined by the Fisher’s exact test Proportions of recorded diagnoses for 12 diseases in older adults [n (%)] The differences of diseases’ proportions were determined by the Fisher’s exact test Frequency ranking of common disease diagnoses (Top 30 conditions with single and paired diseases) Males exhibited significantly higher frequencies of hypertension and bronchitis. Females showed higher frequencies of hyperlipidemia, cataract, sleep disorders, arrhythmia, osteoporosis, and constipation. No significant sex differences were found for arthritis, diabetes mellitus, periodontitis, or chronic gastritis. Participants aged ≥ 70 years had significantly higher frequencies of all conditions except chronic gastritis. Association rules in the whole study sample Fifty-eight association rules met the predefined thresholds (support ≥ 1%, confidence ≥ 50%, lift ≥ 1.5), with arthritis, bronchitis, and sleep disorders serving as the most frequent consequents (Fig. 3 and eTable 1). For arthritis-centered (as a consequent) association rules, all top 5 rules (ranked by confidence) linked osteoporosis to arthritis. The strongest association was observed for the rule bronchitis + osteoporosis → arthritis. Meanwhile, inflammatory conditions (arthritis, chronic gastritis, osteoporosis, and periodontitis) were preserved frequently in rules involving bronchitis. For sleep disorder-centered rules, constipation was a frequent antecedent, with the highest confidence rule being chronic gastritis + constipation → sleep disorder (confidence: 62.21%, lift: 2.1). Specifically, the rule hypertension + constipation → sleep disorder had the highest support (12.27%).Fig. 3Association rules in the whole study sample. Note: Association rules with different consequents are enclosed by dashed-line boxes, and their corresponding disease categories are displayed directly above them. For each subplot, left and right sectors represent two antecedents in each rule. The size of the opening angle of the left sector represents the support (sup) of the association rules, while that of the right sector represents the confidence (cof). The radius of the sectors corresponds to the magnitude of the lift, and the color of the sectors represents different disease categories. All data used to generate this visualization have been normalized. Specific examples of the opening angles and radius sizes are located in the lower-right corner of the figure Association rules in the whole study sample. Note: Association rules with different consequents are enclosed by dashed-line boxes, and their corresponding disease categories are displayed directly above them. For each subplot, left and right sectors represent two antecedents in each rule. The size of the opening angle of the left sector represents the support (sup) of the association rules, while that of the right sector represents the confidence (cof). The radius of the sectors corresponds to the magnitude of the lift, and the color of the sectors represents different disease categories. All data used to generate this visualization have been normalized. Specific examples of the opening angles and radius sizes are located in the lower-right corner of the figure Sex-specific association rules Association rule mining identified distinct multi-morbidity patterns between sexes, with 80 rules generated in males (eFigure 1) and 31 in females (eFigure 2). A notable commonality across both groups was the high confidence of rules linking chronic gastritis + osteoporosis → arthritis, and chronic gastritis + constipation → bronchitis. In males, osteoporosis frequently appeared in rules involving arthritis and bronchitis, whereas female-specific rules for bronchitis predominantly involved chronic gastritis and periodontitis. Consistent with the whole population-level trends, constipation emerged as a main antecedent for sleep disorder-centered rules in both sexes. Age-stratified association rules Overall, 76 and 41 significant association rules were identified from the population aged 65 ~ 69 and ≥ 70 years (eFigure 3 and eFigure 4). Both age groups shared common antecedents: osteoporosis, chronic gastritis, and constipation were frequently associated with arthritis, bronchitis, and sleep disorder. However, there were two extra rules in the 65 ~ 69-year cohort, constipation + chronic gastritis → sleep disorder and constipation + periodontitis → bronchitis. Discussion Our study explored the proportion of recorded diagnoses of twelve common diseases and the mode of multi-morbidity based on individual data extracted from a regional health information platform. Hypertension and hyperlipidemia were the two most prevalent diseases, accounting for the overall proportions of 56.48% and 36.02% among the elders. The proportion of recorded diagnoses of both diseases was consistent with results from several national surveys [16, 17]. The high co-prevalence of these cardiometabolic conditions underscores the need for integrated management protocols in primary care settings rather than siloed disease-specific approaches. A meta-analysis showed that the prevalence of any cataract was more than 30% among Chinese population aged ≥ 65 years old [18], but there was only about 20% of the recorded diagnoses rate in this study. This discrepancy likely reflects underdiagnosis in routine health examinations, suggesting that ophthalmologic screening should be strengthened during annual elderly health checkups. Among the diseases with sex differences, except for hypertension and bronchitis, all other diseases had a lower proportion in men than that in women. In general, there were sex difference among behavior, lifestyle and experience which led to various health outcomes or susceptibility [19]. A population study showed that women being homemaker had higher cardiovascular disease (CVD) risk than other employment types, and proper household work could lower CVD incidence [20]. However, there were non-consistent results in men. Meanwhile, sex hormones and gut micro-biome also played essential roles in these differences [21, 22]. These findings suggest that sex-specific health education and preventive strategies may be warranted in community health programs. Regarding the association rule mining results, our findings reveal three clinically meaningful multi-morbidity clusters that extend beyond traditional cardiovascular patterns. Highly prevalent diseases tended to occur together. In this study, co-occurrence of hypertension and hyperlipidemia was the most frequent combination which was consistent with the pooled prevalence in China [6]. This co-morbidity may be associated with similar physiological change and environmental exposure. However, our analysis identified additional clusters that warrant equal clinical attention. First, we observed a cluster within inflammatory diseases among the whole study sample. Strong association existed in individuals between osteoporosis and arthritis. In China, the pooled proportion of osteoporosis in patients with rheumatoid arthritis (RA) was 35%−50% [23], whereas plenty of studies considered osteoporosis as a common complication for RA [24]. In the view of gene polymorphism, RANK 575CC genotype and RANKL 290 C allele were related with osteoporosis in RA patients [25]. However, a Mendelian study indicated that no causal effect was found between osteoporosis and RA [26]. This association might be affected by individual behavior and medical treatment. We propose that chronic low-grade inflammation may serve as a shared underlying factor associated with skeletal and joint conditions in the elderly, consistent with the “inflammaging” concept. Clinically, this suggests that patients diagnosed with either osteoporosis or arthritis should be proactively screened for the other condition, and anti-inflammatory lifestyle interventions may benefit both. Second, bronchitis also had strong associations with other inflammatory diseases. As one of common conditions of chronic obstructive pulmonary Disease (COPD), the prevalence of chronic bronchitis ranged from 27% to 35% in patients with COPD [27]. Krishnan et al. also found a higher prevalence of osteoporosis in patients with COPD, most of whom were predisposed to being exacerbation [28]. For ageing-related diseases, there were more prevalent and detectable in those with COPD by complex gene-environment mechanism, furthermore, specific treatment for some diseases could change the susceptibility of multi-morbidity [29]. The respiratory-skeletal inflammatory cluster indicates that pulmonary function testing and bone density assessment should be considered complementary in elderly patients with chronic respiratory conditions. Third, we observed that chronic gastritis-constipation was most closely associated with sleep disorder under the sort by association rules’ confidence. There were bi-directional effects between sleep and gastrointestinal diseases (GD) [30]. On the one hand, sleep disorder is associated with disrupted circadian rhythm, which may affect gastrointestinal system functioning. On the other hand, GD may be associated with sleep disruption through nocturnal symptoms. Several population-based studies revealed that inappropriate sleep duration and quality were associated with gastrointestinal symptoms such as constipation, diarrhea and irritable bowel syndrome in both patients with diabetes and general population [31–33]. In addition, constipation not only directly reduced sleep quality but affected it by mental conditions [34]. This bi-directional gut-sleep relationship revealed that community physicians should incorporate sleep quality assessment into routine gastrointestinal care, and conversely, address digestive symptoms when managing sleep disorders in older adults. Simple interventions such as sleep hygiene education and dietary fiber modification may simultaneously improve both conditions. In this study, although there were disparities among actual combinations between various diseases, multi-morbidity modes were similar within whether both sex or age groups. In reality, previous studies showed no significant sex- and age-differences of multi-morbidity prevalence in Chinese population, whereas multi-morbidity was more prevalent in the female among various regions (urban and rural) or country [6, 35]. This suggested pattern of multi-morbidity might be determined by ethnicity and environment. The consistency of multi-morbidity patterns across sex and age subgroups in our study suggests that cluster-based care pathways could be universally applied in primary healthcare centers without extensive stratification. Totally, we obtained some patterns of multi-morbidity from association rules except for hypertension-hyperlipidemia; however, it is essential to recognize inner relationships or potential mechanisms. Notably, the substantial difference in the number of association rules among sex and age subgroups warrants clarification. This variation reflects multiple factors. First, parameter thresholds inherently affect the number of rules retained. Subgroups with lower disease prevalence may generate fewer rules meeting the minimum thresholds. Second, genuine differences in disease prevalence existed across sex and age subgroups. Third, differences in sample size between subgroups also contributed to this variation. These factors should be considered when interpreting subgroup-specific multi-morbidity patterns, and future studies with balanced subgroup samples could further validate these findings. This study offers several distinct contributions to the field of multi-morbidity research. First, we utilized electronic health records from RHIP that captures real-world clinical diagnoses from primary healthcare centers, rather than relying on survey-based self-reports which are subject to recall bias. Second, we applied a HITS-weighted ARM framework that accounts for heterogeneity in multi-morbidity burden, refraining from a methodological limitation of traditional ARM approaches. Third, our focus on community-dwelling older adults aged ≥ 65 years in eastern China provides region-specific evidence that complements existing multi-morbidity studies from other geographic areas and healthcare settings. Together, these methodological and population characteristics distinguish our work from prior clustering analyses and enhance the clinical applicability of our findings for primary care settings. However, this study had some limitations. First, regarding data quality, we excluded individuals whose examination records only included physiological and biochemical measurements to ensure that the disease status of every included participant was ascertainable. However, this criterion may have excluded individuals with undiagnosed conditions or early-stage diseases who exhibited abnormal clinical indicators but had not yet received a formal diagnosis from the chief physician. This selection criterion might result in an underestimation of disease prevalence and could overlook potential early-stage multi-morbidity clusters that precede formal diagnosis. Additionally, relying on the physician’s diagnostic summary rather than standardized ICD codes introduced potential variability in disease classification depending on the physician’s documentation habits. Second, the cross-sectional design precludes the determination of temporality, meaning we cannot ascertain the sequence of disease onset or infer causal relationships between comorbidities. Third, while ARM effectively identifies co-occurrence patterns, it does not imply causality, and the observed rules should be interpreted as statistical associations rather than pathological pathways. Fourth, the selection of thresholds for support, confidence, and lift involves a degree of subjectivity; although based on prior literature, different cutoffs may yield varying numbers of rules. Fifth, although national basic public health services program mandates free annual health examinations for all residents aged ≥ 65 years in China, participation is voluntary and we do not have access to the exact participation rate for our study region. This introduces potential healthcare engagement bias, as individuals who attend routine health examinations may differ from those who do not. Certain groups may be underrepresented, including institutionalized elderly, those with severe mental or physical illnesses who cannot attend examinations, and the migrant population who may not be registered in the RHIP. Among those not included, some were mentally ill individuals being at a higher risk of physical outcomes such as infection and hospitalization [36, 37]. Sixth, by focusing on the 12 most prevalent diseases, we may have overlooked less common but clinically meaningful multi-morbidity clusters. Future studies with larger sample sizes or longer observation periods could explore broader disease sets to capture rarer co-occurrence patterns. These limitations affect the generalizability of our multi-morbidity patterns to the broader elderly population. Moreover, in the view of disparities between present and previous studies, there were differences in not only inclusive population but also lifestyle and behavior. The time-dependent determinants such as interval of various health examinations and onset of symptoms, would obviously influence the proportion of multi-morbidity. In conclusion, we found that besides the high co-occurrence of hypertension and hyperlipidemia, the inflammatory diseases of skeletal and respiratory system as well as the co-occurrence of sleep disorders and gastrointestinal diseases are the main multi-morbidity patterns, regardless of sex and age groups. Based on these findings, we recommend integrating sleep screening into gastrointestinal disease management protocols and implementing coordinated screening for osteoporosis in patients with arthritis and chronic respiratory conditions. Additionally, we advocate for shifting from single-disease guidelines to cluster-based care pathways in primary healthcare centers while prioritizing patient education on the inter-connectivity of these conditions to improve adherence to holistic lifestyle interventions. Therefore, in addition to focusing on traditional high-risk factors for cardiovascular diseases, it is important that we prioritize enhancing the intervention strategies for other illnesses, including mental disorders and digestive system diseases. Supplementary Information Supplementary Material 1. Supplementary Material 1.