Sex and age-specific multimorbidity profiles among working-age inpatients in China: a comparative network analysis.

Our study cohort was drawn from an existing cohort of blood donors and non-donors from Shaanxi Province, originally designed to assess the health impacts of blood donation [ 33 ]. While our study aimed to explore all possible multimorbidity patterns in the working population aged 18–59 years by analysing all eligible inpatient records from both blood donors and non-donors. Figure  1 showed the details of data source. Fig. 1 Data source and the selection flowchart of the study population Data source and the selection flowchart of the study population This study constructed an inpatient hospitalisation records database for the working-age population, derived from two cohorts: blood donor and non-blood donor cohorts in Shaanxi Province, China. The blood donor cohort was built using three major databases: 1) Shaanxi Blood Donor’s Database, which collected voluntary blood donation data from 3.38 million individuals with 6.78 million donation records between 1998 and 2018; 2) Electronic Health Records (EHR) Database, which included 31,130,122 records in Shaanxi Province, China from January 2012 to December 2018; 3) Centralized Hospital Medical Records (CHMR) Database, covering outpatient and inpatient records from 2007 to 2018 with a total of 19,483,983 hospital records. After matching participants by unique Identification Numbers (IDs) across Shaanxi Blood Donor’s Database and EHR, we excluded 1,472,935 blood donors without EHR records, 210,434 with duplicate EHR records, and 1,921 with missing diagnostic information, leaving 1,704,691 blood donors with valid EHRs. These were then matched with the CHMR Database, resulting in 418,312 blood donors with 1,640,483 outpatient records and 332,569 blood donors with 496,061 inpatient records. The non-blood donor cohort was created through a 1:1 matching process based on age, sex, and region with the blood donor cohort, resulting in 1,704,691 non-blood donors randomly selected from the provincial health records. Among them, 407,798 non-blood donors provided 1,655,725 outpatient records, and 346,097 provided 562,337 inpatient records. To capture all potential disease relationships, diagnoses were coded using ICD-10 in inpatient records. We then merged 496,061 inpatient records from 332,569 blood donors with 562,337 inpatient records from 346,097 non-blood donors, resulting in 1,058,398 records from 678,666 participants. Among the 678,666 participants, 96.34% were local residents. Diseases Diagnoses were coded using ICD-10 and initially categorised into 22 chapters (see Additional file 1, Supplementary_Table_S1) [ 32 ]. However, ICD-10 Chapters 15–22 were associated with congenital diseases, injuries, poisoning, and pregnancy (chapters 15–22 [ 34 ]). Based on previous studies’ methodology [ 35 , 36 ] and considering the study population of working-age individuals (18–59 years), including ICD-10 Chapters 15–22 would obscure the exploration of common disease relationships in the general population. Therefore, we only included diseases from ICD-10 Chapters 1–14. The details of Chapters 1–14 are presented in Table  1 . Table 1 The Content of ICD-10 1–14 Chapters Chapter Abbreviation Content Chapter1 C1 Certain infectious and parasitic diseases Chapter2 C2 Neoplasms Chapter3 C3 Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism Chapter4 C4 Endocrine, nutritional and metabolic diseases Chapter5 C5 Mental and behavioral disorders Chapter6 C6 Diseases of the nervous system Chapter7 C7 Diseases of the eye and adnexa Chapter8 C8 Diseases of the ear and mastoid process Chapter9 C9 Diseases of the circulatory system Chapter10 C10 Diseases of the respiratory system Chapter11 C11 Diseases of the digestive system Chapter12 C12 Diseases of the skin and subcutaneous tissue Chapter13 C13 Diseases of the musculoskeletal system and connective tissue Chapter14 C14 Diseases of the genitourinary system The Content of ICD-10 1–14 Chapters From the database of 1,058,398 records from 678,666 participants, our study included only the baseline hospitalisation data for each participant, defined as the earliest hospitalisation record based on admission date. Of 678,666 inpatients, 643,764 were working-age adults (18–59 years) [ 37 ]. While China’s retirement age is 60 for men and 55 for women [ 14 ], we extended it to 60 for females for the comparative analysis between sexes (see Additional file 1, Supplementary Methods_1). We excluded 250,594 records only associated with congenital diseases, injuries, poisoning, and pregnancy (ICD-10 Chapters 15–22 [ 34 ]). Analyses focused on the remaining 393,170 records, coded under ICD-10 Chapters 1–14 [ 36 ] (Fig.  1 ). We analysed multimorbidity patterns in 393,170 inpatients (203,735 males, 189,435 females) stratified by age (18–24, 25–31, 32–38, 39–45, 46–52, 53–59) and sex. Each subgroup contained between 17,193 to 42,160 males and 21,464 to 54,605 females (Fig.  1 , see Additional file 1, Supplementary_Table_S2). The dataset included 898 diseases for males and 888 for females, as defined by 3-character ICD-10 codes (see Additional file 2, Male_ICD and Female_ICD). Socio-demographic variables included age, sex, ethnicity, educational level, marital status, registration, employment status, and residential region. Multimorbidity indicates the presence of ≥ 2 diseases in the same individual, while “multimorbidity pattern” describes the co-occurrence of two diseases as a disease pair [ 38 ]. Network Type Our study constructed undirected multimorbidity networks. Node and Edge Definition Each node in the network represents a unique disease, coded using the ICD-10 classification. Edges represent multimorbidity patterns. Logistic regression calculated the odds ratio (OR) and p-value for each pattern (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The edges were selected based on the following criteria: 1) OR > 1; 2) p-value  1; Bonferroni correction [ 39 ]); 3) the multimorbidity patterns with prevalence > 1/10000 in each dataset were included (see Additional file 1, Supplementary_Figure_S1.1–1.9). Node and edge counts, and multimorbidity frequencies are shown in Additional file 1, Supplementary_Table_S4-5. Algorithm for Network Construction Logistic regression was used to calculate the odds ratio (OR) and p-value for each pattern, totalling 223,524 analyses (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The network construction employed correlation-based methods to assess relationships between diseases based on their co-occurrence. Multimorbidity Network Metrics Our network analysis assessed metrics for each node:1) Degree: The number of edges connected to each node [ 40 ]. 2) Maximal Clique Centrality (MCC): A measure of the centrality of a node in the network, based on its involvement in the largest cliques [ 40 ]. 3) Closeness Centrality (Clo_Cen): A measure of the proximity of a node to all other nodes in the network [ 41 ]. 4) Clustering Coefficient (Clu_Coe): A measure of the degree to which nodes in the network tend to cluster together [ 41 ]. 5) Betweenness Centrality (Bet_Cen): A measure of the influence of a node in terms of its position on the shortest paths between other nodes [ 41 ]. 6) Pagerank: A ranking of nodes based on their connectivity and influence within the network [ 42 ]. 7) Eigencentrality: A measure of the node’s importance based on the eigenvector centrality in the network [ 41 ]. The details of each metric can be found in Additional file 1, Supplementary_Methods_2. Network Type Our study constructed undirected multimorbidity networks. Our study constructed undirected multimorbidity networks. Node and Edge Definition Each node in the network represents a unique disease, coded using the ICD-10 classification. Edges represent multimorbidity patterns. Logistic regression calculated the odds ratio (OR) and p-value for each pattern (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The edges were selected based on the following criteria: 1) OR > 1; 2) p-value  1; Bonferroni correction [ 39 ]); 3) the multimorbidity patterns with prevalence > 1/10000 in each dataset were included (see Additional file 1, Supplementary_Figure_S1.1–1.9). Node and edge counts, and multimorbidity frequencies are shown in Additional file 1, Supplementary_Table_S4-5. Each node in the network represents a unique disease, coded using the ICD-10 classification. Edges represent multimorbidity patterns. Logistic regression calculated the odds ratio (OR) and p-value for each pattern (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The edges were selected based on the following criteria: 1) OR > 1; 2) p-value  1; Bonferroni correction [ 39 ]); 3) the multimorbidity patterns with prevalence > 1/10000 in each dataset were included (see Additional file 1, Supplementary_Figure_S1.1–1.9). Node and edge counts, and multimorbidity frequencies are shown in Additional file 1, Supplementary_Table_S4-5. Algorithm for Network Construction Logistic regression was used to calculate the odds ratio (OR) and p-value for each pattern, totalling 223,524 analyses (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The network construction employed correlation-based methods to assess relationships between diseases based on their co-occurrence. Logistic regression was used to calculate the odds ratio (OR) and p-value for each pattern, totalling 223,524 analyses (see Additional file 1, Supplementary_Figure_S1.1–1.9, Supplementary_Table_S3). The network construction employed correlation-based methods to assess relationships between diseases based on their co-occurrence. Multimorbidity Network Metrics Our network analysis assessed metrics for each node:1) Degree: The number of edges connected to each node [ 40 ]. 2) Maximal Clique Centrality (MCC): A measure of the centrality of a node in the network, based on its involvement in the largest cliques [ 40 ]. 3) Closeness Centrality (Clo_Cen): A measure of the proximity of a node to all other nodes in the network [ 41 ]. 4) Clustering Coefficient (Clu_Coe): A measure of the degree to which nodes in the network tend to cluster together [ 41 ]. 5) Betweenness Centrality (Bet_Cen): A measure of the influence of a node in terms of its position on the shortest paths between other nodes [ 41 ]. 6) Pagerank: A ranking of nodes based on their connectivity and influence within the network [ 42 ]. 7) Eigencentrality: A measure of the node’s importance based on the eigenvector centrality in the network [ 41 ]. The details of each metric can be found in Additional file 1, Supplementary_Methods_2. Our network analysis assessed metrics for each node:1) Degree: The number of edges connected to each node [ 40 ]. 2) Maximal Clique Centrality (MCC): A measure of the centrality of a node in the network, based on its involvement in the largest cliques [ 40 ]. 3) Closeness Centrality (Clo_Cen): A measure of the proximity of a node to all other nodes in the network [ 41 ]. 4) Clustering Coefficient (Clu_Coe): A measure of the degree to which nodes in the network tend to cluster together [ 41 ]. 5) Betweenness Centrality (Bet_Cen): A measure of the influence of a node in terms of its position on the shortest paths between other nodes [ 41 ]. 6) Pagerank: A ranking of nodes based on their connectivity and influence within the network [ 42 ]. 7) Eigencentrality: A measure of the node’s importance based on the eigenvector centrality in the network [ 41 ]. The details of each metric can be found in Additional file 1, Supplementary_Methods_2. Our study constructed three-level networks (details in Supplementary_Methods_3). The first-level was stratified by sex. We created Complete multimorbidity networks for males and females, encompassing all multimorbidity patterns that satisfied selection criteria (Fig.  4 a, Additional file 1, Supplementary_Figure_S1.1). The second level involved comparing the complete multimorbidity networks of males and females. This comparison identified Overlapping and Sex-specific multimorbidity networks based on shared and unique patterns, allowing for the investigation of multimorbidity patterns common to both sexes and those specific to each sex. (Fig.  4 b-c, Additional file 1, Supplementary_Figure_S1.2–1.3). The third-level networks were stratified by sex and age, which aimed to construct Age-sex-stratified multimorbidity networks in corresponding subpopulations (Additional file 1, Supplementary_Figure_S2-S7, Supplementary_Figure_S1.4–1.9). We used degree [ 40 ] (the number of edges a node has) to identify Hub diseases with the top 10 highest number of unique multimorbidity patterns. We also included multimorbidity patterns associated with hub diseases in Hub diseases’ associated network(Fig.  4 a-c, Additional file 1, Supplementary_Figure_S2-S7) . Hub diseases and the size of the node were identified by degree, with edge thickness indicating OR values and nodes color-coded by ICD-10 chapters. Details are in Additional file 2. Demographic characteristics of inpatients were examined, and multimorbidity pattern proportions were reported. Age was presented as mean ± standard deviation (mean ± SD). Analyses were conducted using R 4.1.0 and Python 2020.1.3. Cytoscape 3.7.0 and Gephi 0.9.2 were utilised for network diagram and metric analysis. Stacked bar charts, histograms, violin plots, and heatmaps were created using Microsoft Excel and GraphPad Prism 9.3.1.

The study consisted of 393,170 inpatients with a mean age of 38.39 (SD ± 10.82). Males constituted the most (51.82%), with a significantly higher age (39.36 ± 10.53) compared to females (37.35 ± 11.03) (p < 0.001) (Table  2 ). Table 2 Sociodemographic characteristics of working-age inpatients with or without multimorbidity Whole Male Female Total Without multimorbidity (< 2 diseases) (%) With multimorbidity (≥ 2 diseases) (%) uOR Total Without multimorbidity (< 2 diseases) (%) With multimorbidity (≥ 2 diseases) (%) uOR Total Without multimorbidity (< 2 diseases) (%) With multimorbidity (≥ 2 diseases) (%) uOR Total number of inpatients 393,170 209,912 (53.39) 183,258 (46.61) 203,735 99,011 (48.60) 104,724 (51.40) 189,435 110,901 (58.54) 78,534 (41.46) Sex Male 203,735 99,011 (48.60) 104,724 (51.40) Ref Female 189,435 110,901 (58.54) 78,534 (41.46) 0.67 (0.66, 0.68) *** Age (mean ± SD) 38.39  ± 10.82 36.18  ± 10.49 40.93  ± 10.64 1.04 (1.04, 1.04) *** 39.36  ± 10.53 37.55  ± 10.36 41.07  ± 10.40 1.03 (1.03, 1.03) *** 37.35  ± 11.03 34.95  ± 10.45 40.74  ± 10.94 1.05 (1.05, 1.05) *** Ethnicity Han Ethnicity 382,940 204,399 (53.38) 178,541 (46.62) Ref 199,099 96,788 (48.61) 102,311 (51.39) Ref 183,841 107,611 (58.53) 76,230 (41.47) Ref Non-Han Ethnicity 3,237 1,688 (52.15) 1,549 (47.85) 1.05 (0.98, 1.13) 1,524 697 (45.73) 827 (54.27) 0.89 (0.81, 0.99) * 1,713 991 (57.85) 722 (42.15) 0.97 (0.88, 1.07) Missing 6,993 3,825 (54.70) 3,168 (45.30) 0.95 (0.90, 0.99) * 3,112 1,526 (49.04) 1,586 (50.96) 0.88 (0.77, 0.99) * 3,881 2,299 (59.24) 1,582 (40.76) 0.94 (0.84, 1.06) Educational level Junior High and below 196,963 104,516 (53.06) 92,447 (46.94) Ref 101,319 50,236 (49.58) 51,083 (50.42) Ref 95,644 54,280 (56.75) 41,364 (43.25) Ref Senior middle school 103,811 54,134 (52.15) 49,677 (47.85) 1.04 (1.02, 1.05) *** 55,861 26,257 (47.00) 29,604 (53.00) 1.11 (1.09, 1.13) *** 47,950 27,877 (58.14) 20,073 (41.86) 0.94 (0.92, 0.97) *** University and above 56,463 32,064 (56.79) 24,399 (43.21) 0.86 (0.84, 0.88) *** 28,859 14,007 (48.54) 14,852 (51.46) 1.04 (1.02, 1.07) ** 27,604 18,057 (65.41) 9,547 (34.59) 0.69 (0.67, 0.71) *** Missing 35,933 19,198 (53.43) 16,735 (46.57) 0.99 (0.96, 1.01) 17,696 8,511 (48.10) 9,185 (51.90) 1.06 (1.03, 1.10) *** 18,237 10,687 (58.60) 7,550 (41.40) 0.93 (0.90, 0.96) *** Marital status Unmarried 85,022 53,388 (62.79) 31,634 (37.21) Ref 42,954 24,261 (56.48) 18,693 (43.52) Ref 42,068 (22.21) 29,127 (69.24) 12,941 (30.76) Ref Married 271,938 136,869 (50.33) 135,069 (49.67) 1.67 (1.64, 1.69) *** 142,656 65,637 (46.01) 77,019 (53.99) 1.52 (1.49, 1.56) *** 129,282 (68.25) 71,232 (55.10) 58,050 (44.90) 1.83 (1.79, 1.88) *** Divorced/Widowed 4,115 1,927 (46.83) 2,188 (53.17) 1.92 (1.80, 2.04) *** 2,066 928 (44.92) 1,138 (55.08) 1.59 (1.46, 1.74) *** 2,049 (1.08) 999 (48.76) 1,050 (51.24) 2.37 (2.16, 2.59) *** Missing 32,095 17,728 (55.24) 14,367 (44.76) 1.37 (1.33, 1.40) *** 16,059 8,185 (50.97) 7,874 (49.03) 1.25 (1.20, 1.29) *** 16,036 (8.47) 9,543 (59.51) 6,493 (40.49) 1.53 (1.47, 1.59) *** Registration Native 378,794 202,234 (53.39) 176,560 (46.61) Ref 197,369 96,106 (48.69) 101,263 (51.31) Ref 181,425 (95.77) 106,050 (58.45) 75,375 (41.55) Ref Non-native 3,298 1,807 (54.79) 1,491 (45.21) 0.96 (0.89, 1.03) 1,546 699 (45.21) 847 (54.79) 1.15 (1.04, 1.27) ** 1,752 (0.92) 1,097 (62.61) 655 (37.39) 0.84 (0.76, 0,93) *** Missing 11,078 5,871 (53.00) 5,207 (47.00) 0.98 (0.95, 1.02) 4,820 2,206 (45.77) 2,614 (54.23) 1.12 (1.06, 1.19) *** 6,258 (3.30) 3,754 (59.99) 2,504 (40.01) 0.94 (0.89, 0.99) * Employment status Employed 373,356 198,887 (53.27) 174,469 (46.73) Ref 195,588 95,121 (48.63) 100,467 (51.37) Ref 177,768 (93.84) 103,766 (58.37) 74,002 (41.63) Ref Unemployed 11,176 6,086 (54.46) 5,090 (45.54) 0.95 (0.92, 0.99) * 4,652 2,167 (46.58) 2,485 (53.42) 1.09 (1.02, 1.15) ** 6,524 3,919 (60.07) 2,605 (39.93) 0.93 (0.89, 0.98) ** Missing 8,638 4,939 (57.18) 3,699 (42.82) 0.85 (0.82, 0.89) *** 3,495 1,723 (49.30) 1,772 (50.70) 0.97 (0.91, 1.04) 5,143 3,216 (62.53) 1,927 (37.47) 0.84 (0.79, 0.89) *** Residential region North Shannxi 21,938 10,752 (49.01) 11,186 (50.99) Ref 12,312 5,479 (44.50) 6,833 (55.50) Ref 9,626 5,273 (54.78) 4,353 (45.22) Ref Central Shannxi 296,199 158,186 (53.41) 138,013 (46.59) 0.84 (0.82, 0.86) *** 152,944 74,171 (48.50) 78,773 (51.50) 0.85 (0.82, 0.88) *** 143,255 84,015 (58.65) 59,240 (41.35) 0.85 (0.82, 0.89) *** South Shannxi 75,033 40,974 (54.61) 34,059 (45.39) 0.80 (0.77, 0.82) *** 38,479 19,361 (50.32) 19,118 (49.68) 0.79 (0.76, 0.82) *** 36,554 21,613 (59.13) 14,941 (40.87) 0.84 (0.80, 0.88) *** *** : P  < 0.001; **: P  < 0.01 *: P  < 0.05; uOR: unadjusted odds ratio Sociodemographic characteristics of working-age inpatients with or without multimorbidity 209,912 (53.39) 183,258 (46.61) 99,011 (48.60) 104,724 (51.40) 110,901 (58.54) 78,534 (41.46) 99,011 (48.60) 104,724 (51.40) 110,901 (58.54) 78,534 (41.46) 0.67 (0.66, 0.68) *** 38.39 ± 10.82 36.18 ± 10.49 40.93 ± 10.64 1.04 (1.04, 1.04) *** 39.36 ± 10.53 37.55 ± 10.36 41.07 ± 10.40 1.03 (1.03, 1.03) *** 37.35 ± 11.03 34.95 ± 10.45 40.74 ± 10.94 1.05 (1.05, 1.05) *** 204,399 (53.38) 178,541 (46.62) 96,788 (48.61) 102,311 (51.39) 107,611 (58.53) 76,230 (41.47) 1,688 (52.15) 1,549 (47.85) 1.05 (0.98, 1.13) 697 (45.73) 827 (54.27) 0.89 (0.81, 0.99) * 991 (57.85) 722 (42.15) 0.97 (0.88, 1.07) 3,825 (54.70) 3,168 (45.30) 0.95 (0.90, 0.99) * 1,526 (49.04) 1,586 (50.96) 0.88 (0.77, 0.99) * 2,299 (59.24) 1,582 (40.76) 0.94 (0.84, 1.06) 104,516 (53.06) 92,447 (46.94) 50,236 (49.58) 51,083 (50.42) 54,280 (56.75) 41,364 (43.25) 54,134 (52.15) 49,677 (47.85) 1.04 (1.02, 1.05) *** 26,257 (47.00) 29,604 (53.00) 1.11 (1.09, 1.13) *** 27,877 (58.14) 20,073 (41.86) 0.94 (0.92, 0.97) *** 32,064 (56.79) 24,399 (43.21) 0.86 (0.84, 0.88) *** 14,007 (48.54) 14,852 (51.46) 1.04 (1.02, 1.07) ** 18,057 (65.41) 9,547 (34.59) 0.69 (0.67, 0.71) *** 19,198 (53.43) 16,735 (46.57) 0.99 (0.96, 1.01) 8,511 (48.10) 9,185 (51.90) 1.06 (1.03, 1.10) *** 10,687 (58.60) 7,550 (41.40) 0.93 (0.90, 0.96) *** 53,388 (62.79) 31,634 (37.21) 24,261 (56.48) 18,693 (43.52) 42,068 (22.21) 29,127 (69.24) 12,941 (30.76) 136,869 (50.33) 135,069 (49.67) 1.67 (1.64, 1.69) *** 65,637 (46.01) 77,019 (53.99) 1.52 (1.49, 1.56) *** 129,282 (68.25) 71,232 (55.10) 58,050 (44.90) 1.83 (1.79, 1.88) *** 1,927 (46.83) 2,188 (53.17) 1.92 (1.80, 2.04) *** 928 (44.92) 1,138 (55.08) 1.59 (1.46, 1.74) *** 2,049 (1.08) 999 (48.76) 1,050 (51.24) 2.37 (2.16, 2.59) *** 17,728 (55.24) 14,367 (44.76) 1.37 (1.33, 1.40) *** 8,185 (50.97) 7,874 (49.03) 1.25 (1.20, 1.29) *** 16,036 (8.47) 9,543 (59.51) 6,493 (40.49) 1.53 (1.47, 1.59) *** 202,234 (53.39) 176,560 (46.61) 96,106 (48.69) 101,263 (51.31) 181,425 (95.77) 106,050 (58.45) 75,375 (41.55) 1,807 (54.79) 1,491 (45.21) 0.96 (0.89, 1.03) 699 (45.21) 847 (54.79) 1.15 (1.04, 1.27) ** 1,752 (0.92) 1,097 (62.61) 655 (37.39) 0.84 (0.76, 0,93) *** 5,871 (53.00) 5,207 (47.00) 0.98 (0.95, 1.02) 2,206 (45.77) 2,614 (54.23) 1.12 (1.06, 1.19) *** 6,258 (3.30) 3,754 (59.99) 2,504 (40.01) 0.94 (0.89, 0.99) * 198,887 (53.27) 174,469 (46.73) 95,121 (48.63) 100,467 (51.37) 103,766 (58.37) 74,002 (41.63) 6,086 (54.46) 5,090 (45.54) 0.95 (0.92, 0.99) * 2,167 (46.58) 2,485 (53.42) 1.09 (1.02, 1.15) ** 3,919 (60.07) 2,605 (39.93) 0.93 (0.89, 0.98) ** 4,939 (57.18) 3,699 (42.82) 0.85 (0.82, 0.89) *** 1,723 (49.30) 1,772 (50.70) 0.97 (0.91, 1.04) 3,216 (62.53) 1,927 (37.47) 0.84 (0.79, 0.89) *** 10,752 (49.01) 11,186 (50.99) 5,479 (44.50) 6,833 (55.50) 5,273 (54.78) 4,353 (45.22) 158,186 (53.41) 138,013 (46.59) 0.84 (0.82, 0.86) *** 74,171 (48.50) 78,773 (51.50) 0.85 (0.82, 0.88) *** 84,015 (58.65) 59,240 (41.35) 0.85 (0.82, 0.89) *** 40,974 (54.61) 34,059 (45.39) 0.80 (0.77, 0.82) *** 19,361 (50.32) 19,118 (49.68) 0.79 (0.76, 0.82) *** 21,613 (59.13) 14,941 (40.87) 0.84 (0.80, 0.88) *** *** : P  < 0.001; **: P  < 0.01 *: P  < 0.05; uOR: unadjusted odds ratio In males, the most prevalent diseases were essential hypertension (13.46%), T2DM (7.49%), liver diseases (7.42%), dyslipidemia (7.14%), and gastritis/duodenitis (6.04%) (Additional file 1, Supplementary_Table_S6). Younger males had more respiratory, infectious/parasitic, and digestive diseases, while older males had more circulatory and endocrine/nutritional/metabolic diseases (Fig.  2 a-b). Fig. 2 The proportion and per-capita disease diagnoses according to ICD-10 (Chapters 1–14) by age in male and female inpatients. The x-axis represents age (from 18–59 years, unit: years), the y-axis for a ) and c ) represents the per-capita disease diagnoses, and the y-axis for b ) and d ) represents the proportion. The colour of each bar represents the disease’s systematic chapters The proportion and per-capita disease diagnoses according to ICD-10 (Chapters 1–14) by age in male and female inpatients. The x-axis represents age (from 18–59 years, unit: years), the y-axis for a ) and c ) represents the per-capita disease diagnoses, and the y-axis for b ) and d ) represents the proportion. The colour of each bar represents the disease’s systematic chapters In females, the most prevalent diseases were anaemia (D64) (9.48%), essential hypertension (I10, 7.61%), gastritis/duodenitis (K29, 4.71%), cholelithiasis (K80, 4.53%), and pelvic inflammatory diseases (N73, 4.51%) (Additional file 1, Supplementary_Table_S6). Younger females had more blood and blood-forming and respiratory diseases, while older males had more circulatory, endocrine/nutritional/metabolic, and musculoskeletal system/connective tissue diseases (Fig.  2 c-d). ICD-10 Chapter Rankings based on the frequency of associated multimorbidity patterns. During the working-age period, males exhibited the highest frequencies of circulatory, endocrine/nutritional/metabolic, and digestive disease-associated patterns, while females had the highest frequencies of circulatory, endocrine/nutritional/metabolic, and genitourinary disease-associated patterns (Additional file 1, Supplementary_Figure_S8). With age, patterns-associated diseases shifted from respiratory, infectious/parasitic and genitourinary to nervous, endocrine/nutritional/metabolic, and circulatory. Females showed a similar trend, with a notable increase in musculoskeletal/connective tissue disease-associated patterns with age. Digestive-associated patterns persisted across all ages and sexes (Fig.  3 A, Additional file 1, Supplementary_Figure_S8-S11). Fig. 3 ICD-10 Chapter Rankings and the Most Common Multimorbidity Patterns among age-sex-stratified Chinese working-age inpatients. A Rankings of ICD-10 chapters based on the frequency of associated multimorbidity patterns among age-sex-stratified inpatients. Our study calculated the total frequency of multimorbidity patterns associated with each ICD-10 Chapter (Figure S10-12). Based on these frequencies, we ranked the Chapters from highest to lowest, resulting in a ranking for Chapters 1–14. These rankings formed the basis of the line chart. Using the male inpatients as an example, the horizontal axis represents six age groups: 18–24, 25–31, 32–38, 39–45, 46–52, 53–59 years. The vertical axis represents the ranking from 1 to 14. Each line is colored according to the corresponding chapter. B The top 10 most common multimorbidity patterns among age-sex-stratified inpatients (Proportion %). Our study analysed overall (18–59 years) and each age-sex-stratified network (stratified by age groups: 18–24, 25–31, 32–38, 39–45, 46–52, 53–59; and sex: Males and Females), then calculated the top ten most frequent multimorbidity patterns in each network and used these patterns to create the following multimorbidity patterns plots. The x-axis represents different age ranges (18–59; 18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years). The y-axis represents multimorbidity patterns (each node represents one disease with ICD-10 codes, and the colour of each bar represents disease systematic chapters). The numbers in each block of the heatmap represent the proportion of the frequency of the corresponding prevalent multimorbidity pattern in the population of the corresponding age-sex-stratified range to the sum of the frequencies of all prevalent multimorbidity patterns in the population. Disease name with ICD-10 codes (listed by alphabetical order): A16: Respiratory tuberculosis, not confirmed bacteriologically or histologically; A18: Tuberculosis of other organs; B37: Candidiasis; D24: Benign neoplasm of breast; D25: Leiomyoma of uterus; D64: Other anemias; E11: Type 2 diabetes mellitus; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; G63: Polyneuropathy in diseases classified elsewhere; G99: Other disorders of nervous system in diseases classified elsewhere; I10: Essential (primary) hypertension; I20: Angina pectoris; I25: Chronic ischemic heart disease; I50: Heart failure; I63: Cerebral infarction; I70: Atherosclerosis; I84: Hemorrhoids; J31: Chronic rhinitis, nasopharyngitis and pharyngitis; J32: Chronic sinusitis; J34: Other disorders of nose and nasal sinuses; K26: Duodenal ulcer; K29: Gastritis and duodenitis; K35: Acute appendicitis; K60: Fissure and fistula of anal and rectal regions; K62: Other diseases of anus and rectum; K65: Peritonitis; K71: Toxic liver disease; K76: Other diseases of liver; K80: Cholelithiasis; K81: Cholecystitis; M47: Spondylosis; N13: Obstructive and reflux uropathy; N20: Calculus of kidney and ureter; N62: Hypertrophy of breast; N70: Salpingitis and oophoritis; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases; N76: Other inflammation of vagina and vulva; N77: Vulvovaginal ulceration and inflammation in diseases classified elsewhere; N83: Noninflammatory disorders of ovary, fallopian tube and broad ligament; N97: Female infertility ICD-10 Chapter Rankings and the Most Common Multimorbidity Patterns among age-sex-stratified Chinese working-age inpatients. A Rankings of ICD-10 chapters based on the frequency of associated multimorbidity patterns among age-sex-stratified inpatients. Our study calculated the total frequency of multimorbidity patterns associated with each ICD-10 Chapter (Figure S10-12). Based on these frequencies, we ranked the Chapters from highest to lowest, resulting in a ranking for Chapters 1–14. These rankings formed the basis of the line chart. Using the male inpatients as an example, the horizontal axis represents six age groups: 18–24, 25–31, 32–38, 39–45, 46–52, 53–59 years. The vertical axis represents the ranking from 1 to 14. Each line is colored according to the corresponding chapter. B The top 10 most common multimorbidity patterns among age-sex-stratified inpatients (Proportion %). Our study analysed overall (18–59 years) and each age-sex-stratified network (stratified by age groups: 18–24, 25–31, 32–38, 39–45, 46–52, 53–59; and sex: Males and Females), then calculated the top ten most frequent multimorbidity patterns in each network and used these patterns to create the following multimorbidity patterns plots. The x-axis represents different age ranges (18–59; 18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years). The y-axis represents multimorbidity patterns (each node represents one disease with ICD-10 codes, and the colour of each bar represents disease systematic chapters). The numbers in each block of the heatmap represent the proportion of the frequency of the corresponding prevalent multimorbidity pattern in the population of the corresponding age-sex-stratified range to the sum of the frequencies of all prevalent multimorbidity patterns in the population. Disease name with ICD-10 codes (listed by alphabetical order): A16: Respiratory tuberculosis, not confirmed bacteriologically or histologically; A18: Tuberculosis of other organs; B37: Candidiasis; D24: Benign neoplasm of breast; D25: Leiomyoma of uterus; D64: Other anemias; E11: Type 2 diabetes mellitus; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; G63: Polyneuropathy in diseases classified elsewhere; G99: Other disorders of nervous system in diseases classified elsewhere; I10: Essential (primary) hypertension; I20: Angina pectoris; I25: Chronic ischemic heart disease; I50: Heart failure; I63: Cerebral infarction; I70: Atherosclerosis; I84: Hemorrhoids; J31: Chronic rhinitis, nasopharyngitis and pharyngitis; J32: Chronic sinusitis; J34: Other disorders of nose and nasal sinuses; K26: Duodenal ulcer; K29: Gastritis and duodenitis; K35: Acute appendicitis; K60: Fissure and fistula of anal and rectal regions; K62: Other diseases of anus and rectum; K65: Peritonitis; K71: Toxic liver disease; K76: Other diseases of liver; K80: Cholelithiasis; K81: Cholecystitis; M47: Spondylosis; N13: Obstructive and reflux uropathy; N20: Calculus of kidney and ureter; N62: Hypertrophy of breast; N70: Salpingitis and oophoritis; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases; N76: Other inflammation of vagina and vulva; N77: Vulvovaginal ulceration and inflammation in diseases classified elsewhere; N83: Noninflammatory disorders of ovary, fallopian tube and broad ligament; N97: Female infertility We identified the top 10 most prevalent multimorbidity patterns in the working-age population. The seven most prevalent patterns shared between both sexes include hypertension (I10)-associated patterns (with dyslipidemia, E78; T2DM, E11; chronic ischemic heart disease, I25; cerebral infarction, I63; liver diseases, K76), and chronic ischemic heart disease/heart failure (I25 + I50) (Fig.  3 B and Additional file 1, Supplementary_Table_S7). Among males-specific patterns, hyperplasia of the prostate (N40)-associated patterns were most prevalent, while female-specific patterns included those associated with cervical uteri inflammation (N72), pelvic inflammatory diseases (N73), and leiomyoma of the uterus (D25) (Additional file 1, Supplementary_Table_S8). In both sexes, the top 10 multimorbidity patterns showed similar trends with increasing age (Fig.  3 B and Additional file 1, Supplementary_Table_S9-11). Table 2 presented that 46.61% of inpatients had multimorbidity, with a higher proportion in males (51.40%) than in females (41.46%). Multimorbidity increased with age in both sexes ( P  < 0.001, Additional file 1, Supplementary_Table_S12). In males, being unemployed was associated with higher multimorbidity rates (OR = 1.09, 95%CI: 1.02–1.15, p  < 0.01) but had the opposite effect in females (OR = 0.93, 95%CI: 0.89–0.98, p  < 0.01) (Table  2 ). The complete multimorbidity network was more complex in males (322 nodes/1233 edges in males vs. 283 nodes/881 edges in females). Both sexes shared eight hub diseases, including liver disease (K76), dyslipidemia (E78), fluid/electrolyte/acid–base imbalances (E87), T2DM (E11), hypertension (I10), heart failure (I50), atherosclerosis (I70), and gastritis/duodenitis (K29) (Additional file 1, Supplementary_Table_S13). Multimorbidity patterns covered by its hub diseases’ associated network represented 66.44% (166,457/250,528) of the frequency of patterns in the complete network for males, while the corresponding percentage for females was 58.63% (74,521/127,104) (Fig.  4 a, Additional file 1, Supplementary_Table_S4). Fig. 4 a Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among complete multimorbidity networks in males and females. b Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among networks of overlapping multimorbidity patterns in males and females. c . Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among sex-specific multimorbidity networks. We presented the diseases in ICD-10 codes. Different colours denote different disease systematic chapters in the figure legend. The hub diseases shared between the male and females were highlighted in bold circles. n 1 : The number of nodes in the corresponding networks. n 2 : The number of edges in the corresponding networks. f : The total frequency of multimorbidity patterns in the corresponding networks. Node size was based on degree and the thickness of each edge was determined by OR values. The colour of the node was marked according to the disease systematic chapters. Hub disease name with ICD-10 codes (listed by alphabetical order): D25: Leiomyoma of uterus; D64: Other anaemias; E11: Type 2 diabetes mellitus; E72: Other disorders of amino-acid metabolism; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; E87: Other disorders of fluid, electrolyte and acid–base balance; I10: Essential (primary) hypertension; I50: Heart failure; I63: Cerebral infarction; I70: Atherosclerosis; J94: Other pleural conditions; J98: Other respiratory disorders; K29: Gastritis and duodenitis; K76: Other diseases of liver; N39: Other disorders of urinary system; N40: Hyperplasia of prostate; N70: Salpingitis and oophoritis; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases; N76: Other inflammation of vagina and vulva; N80: Endometriosis; N83: Noninflammatory disorders of ovary, fallopian tube and broad ligament; N84: Polyp of the female genital tract; N88: Other noninflammatory disorders of cervix uteri a Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among complete multimorbidity networks in males and females. b Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among networks of overlapping multimorbidity patterns in males and females. c . Comparisons of multimorbidity networks, Hub diseases and Hub diseases’ associated network among sex-specific multimorbidity networks. We presented the diseases in ICD-10 codes. Different colours denote different disease systematic chapters in the figure legend. The hub diseases shared between the male and females were highlighted in bold circles. n 1 : The number of nodes in the corresponding networks. n 2 : The number of edges in the corresponding networks. f : The total frequency of multimorbidity patterns in the corresponding networks. Node size was based on degree and the thickness of each edge was determined by OR values. The colour of the node was marked according to the disease systematic chapters. Hub disease name with ICD-10 codes (listed by alphabetical order): D25: Leiomyoma of uterus; D64: Other anaemias; E11: Type 2 diabetes mellitus; E72: Other disorders of amino-acid metabolism; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; E87: Other disorders of fluid, electrolyte and acid–base balance; I10: Essential (primary) hypertension; I50: Heart failure; I63: Cerebral infarction; I70: Atherosclerosis; J94: Other pleural conditions; J98: Other respiratory disorders; K29: Gastritis and duodenitis; K76: Other diseases of liver; N39: Other disorders of urinary system; N40: Hyperplasia of prostate; N70: Salpingitis and oophoritis; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases; N76: Other inflammation of vagina and vulva; N80: Endometriosis; N83: Noninflammatory disorders of ovary, fallopian tube and broad ligament; N84: Polyp of the female genital tract; N88: Other noninflammatory disorders of cervix uteri We discovered that 603 multimorbidity patterns of 221 diseases overlapped between males and females. These accounted for 83.58% (209,392/250,528) of the frequency of all patterns in males and 70.79% (89,978/127,104) in females (Fig.  4 b). Notably, hub diseases presented in both sexes in the overlapping network were associated with endocrine/nutritional/metabolic, circulatory, and digestive diseases (Additional file 1, Supplementary_Table_S13). Among males, patterns in the overlapping hub diseases’ associated network constituted 73.03% (152,915/209,392) of the frequency in the overlapping network, whereas, among females, they accounted for 68.18% (61,351/89,978) (Fig.  4 b and Additional file 1, Supplementary_Table_S4). We identified 630 unique multimorbidity patterns of 261 diseases in males and 278 unique patterns of 149 diseases in females. These accounted for 16.42% (41,136/250,528) of the frequency of all patterns in males and 29.21% (37,126/127,104) in females (Fig.  4 c). Figure  4 c displayed the hub diseases in the sex-specific network, with hyperplasia of the prostate (N40) being the top hub disease in males, while all hub diseases in females were female-specific diseases from the genitourinary system (Additional file 1, Supplementary_Table_S13). The sex-specific hub diseases’ associated network showed that in males, the percentage of frequency accounted for 52.02% (21,398/41,136), while in females they accounted for 72.92% (27,072/37,126) (Fig.  4 c, Additional file 1, Supplementary_Table_S4). Initially, within each age-sex-stratified group, males exhibited between 5,979 and 18,875 unique multimorbidity patterns across five age ranges (18–24, 25–31, 32–38, 39–45, 46–52, 53–59), while females had between 5,050 and 16,540 unique patterns (Additional file 1, Supplementary_Figure_ S1.4–1.9, Table S3). After selecting the multimorbidity patterns that met the criteria for constructing age-sex-stratified multimorbidity networks, we observed that the networks for males were consistently more complex than those for females across all age strata. Additionally, network complexity increased with age. Specifically, for males, the number of nodes ranged from 266 to 299, and the number of edges ranged from 521 to 898; for females, the number of nodes ranged from 186 to 292, and the number of edges ranged from 248 to 732(Additional file 1, Supplementary_Figure_S2-7, Supplementary_Figure_S12, and Supplementary_Table_S5). In males, Fig.  5 showed that gastritis/duodenitis (K29), dyslipidemia (E78); hypertension (I10), and liver diseases (K76) were hub diseases across the whole working age. Fluid/electrolyte/acid–base imbalances (E87) and heart failure (I50) were hub diseases across five age ranges from 25–59 years. The hub diseases changed from infectious, respiratory, genitourinary, and digestive diseases to endocrine/nutritional/metabolic and circulatory diseases by age in both sexes (Fig.  5 , Additional file 1, Supplementary_Table_S14). The hub diseases’ associated multimorbidity networks, as a percentage of the corresponding complete multimorbidity networks increased from 34.98% at age 18–24 to 75.98% at 53–59 (Additional file 1, Supplementary_Table_S5). Fig. 5 Hub diseases among age-sex-stratified populations. We constructed each age-sex-stratified complete multimorbidity network (Stratified by age groups: 18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years; and sex: male and females). We then used the network metric degree (the number of edges a node has) to obtain hub diseases with the top 10 highest number of unique multimorbidity patterns (edges). (1): The hub diseases’ network among age-sex-stratified populations. We constructed the hub diseases’ network of multimorbidity patterns all composed of hub diseases for each age-sex-stratified population. (2): Comparisons of hub diseases among various age-sex-stratified networks. The x-axis represents different age ranges (18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years). The y-axis represents diseases (each node represents one disease with ICD-10 codes, and the colour of each bar represents disease systematic chapters). A yellow patch means the disease is a hub disease in that age group. The colour of the node was marked according to the disease systematic chapters. Hub disease name with ICD-10 codes (listed by alphabetical order): A16: Respiratory tuberculosis, not confirmed bacteriologically or histologically; A18: Tuberculosis of other organs; A19: Miliary tuberculosis; B24: Unspecified human immunodeficiency virus [HIV] disease; D25: Leiomyoma of uterus; D64: Other anemias; D70: Agranulocytosis; E11: Type 2 diabetes mellitus; E72: Other disorders of amino-acid metabolism; E77: Disorders of glycoprotein metabolism; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; E87: Other disorders of fluid, electrolyte and acid–base balance; H52: Disorders of refraction and accommodation; I10: Essential (primary) hypertension; I50: Heart failure; I63: Cerebral infarction; I65: Occlusion and stenosis of precerebral arteries, not resulting in cerebral infarction; I70: Atherosclerosis; J31: Chronic rhinitis, nasopharyngitis and pharyngitis; J98: Other respiratory disorders; K29: Gastritis and duodenitis; K74: Fibrosis and cirrhosis of liver; K76: Other diseases of liver; K80: Cholelithiasis; K92: Other diseases of digestive system; M32: Systemic lupus erythematosus; M47: Spondylosis; N03: Chronic nephritic syndrome; N18: Chronic kidney disease; N28: Other disorders of kidney and ureter, not elsewhere classified; N40: Hyperplasia of prostate; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases Hub diseases among age-sex-stratified populations. We constructed each age-sex-stratified complete multimorbidity network (Stratified by age groups: 18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years; and sex: male and females). We then used the network metric degree (the number of edges a node has) to obtain hub diseases with the top 10 highest number of unique multimorbidity patterns (edges). (1): The hub diseases’ network among age-sex-stratified populations. We constructed the hub diseases’ network of multimorbidity patterns all composed of hub diseases for each age-sex-stratified population. (2): Comparisons of hub diseases among various age-sex-stratified networks. The x-axis represents different age ranges (18–24; 25–31; 32–38; 39–45; 46–52; 53–59 years). The y-axis represents diseases (each node represents one disease with ICD-10 codes, and the colour of each bar represents disease systematic chapters). A yellow patch means the disease is a hub disease in that age group. The colour of the node was marked according to the disease systematic chapters. Hub disease name with ICD-10 codes (listed by alphabetical order): A16: Respiratory tuberculosis, not confirmed bacteriologically or histologically; A18: Tuberculosis of other organs; A19: Miliary tuberculosis; B24: Unspecified human immunodeficiency virus [HIV] disease; D25: Leiomyoma of uterus; D64: Other anemias; D70: Agranulocytosis; E11: Type 2 diabetes mellitus; E72: Other disorders of amino-acid metabolism; E77: Disorders of glycoprotein metabolism; E78: Dyslipidemia; E79: Disorders of purine and pyrimidine metabolism; E87: Other disorders of fluid, electrolyte and acid–base balance; H52: Disorders of refraction and accommodation; I10: Essential (primary) hypertension; I50: Heart failure; I63: Cerebral infarction; I65: Occlusion and stenosis of precerebral arteries, not resulting in cerebral infarction; I70: Atherosclerosis; J31: Chronic rhinitis, nasopharyngitis and pharyngitis; J98: Other respiratory disorders; K29: Gastritis and duodenitis; K74: Fibrosis and cirrhosis of liver; K76: Other diseases of liver; K80: Cholelithiasis; K92: Other diseases of digestive system; M32: Systemic lupus erythematosus; M47: Spondylosis; N03: Chronic nephritic syndrome; N18: Chronic kidney disease; N28: Other disorders of kidney and ureter, not elsewhere classified; N40: Hyperplasia of prostate; N72: Inflammatory disease of cervix uteri; N73: Other female pelvic inflammatory diseases In females, gastritis/duodenitis (K29), and cervical uteri inflammation (N72) were hub diseases across working ages. Liver diseases (K76), fluid/electrolyte/acid–base imbalances (E87), dyslipidemia (E78); essential hypertension (I10), heart failure (I50), female pelvic inflammatory diseases (N73) across most of the age ranges. The hub diseases changed from infectious, respiratory, digestive, blood and blood-forming, musculoskeletal/connective tissue, and genitourinary diseases to endocrine/nutritional/metabolic, and circulatory diseases by age (Fig.  5 , Additional file 1, Supplementary_Table_S15). The hub diseases’ associated multimorbidity networks, as a percentage of the corresponding complete multimorbidity networks increased from 43.33% at age 18–24 to 73.95% at 53–59 (Additional file 1, Supplementary_Table_S5). Additional file 1, Supplementary_Figure_S13-S14 compared the distribution of network metrics for hub diseases and all diseases in multiple multimorbidity networks. Hub diseases always ranked ahead of other diseases in all network metrics, indicating hub diseases are more likely to appear in inpatients with multimorbidity.

Our study examined multimorbidity in working-age inpatients in China and compared multimorbidity patterns by sex and age. Males demonstrated more complex multimorbidity networks than females. Age-associated shifts in multimorbidity patterns and hub diseases were observed, transitioning from respiratory, infectious/parasitic to nervous, endocrine/nutritional/metabolic and circulatory diseases. Hub diseases’ associated network constituted the majority of multimorbidity frequencies in the complete network.

We conducted a study using data from 393,170 working-age inpatients (18–59 years) in Shaanxi, China, collected from 1998 to 2018. We aimed to identify multimorbidity profiles by sex and age and explore multimorbidity patterns using network analysis. Our study demonstrated being unemployed was associated with increased multimorbidity risk in males but the opposite in females. In males, several factors may contribute to this finding. First, poor mental health was more likely among unemployed males than otherwise [ 43 ]. The experience of unemployment may be particularly stressful for males, as it entails the loss of financial support and life goals. The feeling of social isolation from a working environment heightens susceptibility to mental disorders [ 44 ], which in turn increases the likelihood of physical morbidities [ 45 ]. Second, previous research indicates that unemployed males exhibit higher rates of smoking, excessive drinking, and substance abuse than otherwise; these unhealthy lifestyles are risk factors for multimorbidity [ 46 ]. Third, studies showed that males had a low awareness of disease prevention [ 47 ]. Employed males access better health education and prevention in work environments compared to those unemployed. On the contrary, under conventional values, Chinese females often assume family duties, including childbirth and family care [ 48 ]. Employed females often face the stress of juggling between work and family responsibilities. Elevated physical and psychological strain may contribute to an increased risk of multimorbidity. Our study consistently identified more spondylosis-associated patterns in females, often associated with pregnancy, childcare, and work pressures. Our study revealed higher multimorbidity and network complexity in males than females. This finding is consistent with several studies [ 49 – 51 ] but conflicting with others [ 27 , 45 ]. Three factors explain this. First, age seems to be an important confounder in sex differences. Bivariate analyses might erroneously suggest a higher likelihood of multimorbidity in males due to higher average age. Second, unhealthy behaviors, including smoking, binge drinking, and physical inactivity, are more common among males [ 49 ]. Third, males may exhibit lower health awareness and seek less preventive primary care than females for routine check-ups and prevention. They typically seek medical help mainly upon experiencing symptoms, often at a more advanced disease stage, heightening complication risks [ 47 ]. Our study shows a significant association between multimorbidity and working-age males, indicating their susceptibility to multimorbidity may affect overall labor productivity. Targeted strategies such as prevention education, early diagnosis, and treatment for males could delay multimorbidity onset, enhancing health and productivity. Our study shows similar age-associated multimorbidity patterns in older individuals of both sexes, primarily linked to endocrine/nutritional/metabolic and circulatory diseases. Previous studies have documented that circulatory and endocrine/nutritional/metabolic disease-associated patterns are predominant in China [ 51 ], and primary healthcare alone cannot fully address the increasing burden of these patterns in its ageing population [ 52 ]. The workplace has been proposed as a platform to increase healthcare accessibility and early intervention for chronic diseases [ 22 ]. Occupational health promotion can be approached from various angles. First, workplaces should provide a conducive environment to reduce exposure to hazardous materials like asbestos, X-rays, noise, and smoke, which can trigger acute and chronic diseases. Second, workplaces should promote work-life balance through reasonable work hours, breaks, and annual leave to avert burnout and stress-linked health problems. Third, workplaces should enhance mental and physical health support by offering counselling, mental health campaigns, exercise facilities, free classes, and promoting nutritious choices. Fourth, workplaces could co-fund health insurance and collaborate with hospitals for regular check-ups, and screenings for early chronic disease symptoms. Ensuring fair employment-associated health insurance from the Urban Employee Basic Medical Insurance Scheme and the new Urban–Rural Resident Medical Insurance Scheme can improve the medical affordability of the workers [ 53 ]. Increasing the health insurance coverage and reimbursement ratios, and reducing out-of-pocket expenses, are crucial steps to achieve so. Multimorbidity patterns covered by hub diseases’ associated network comprise 66% and 58% of the multimorbidity network for males and females, highlighting the importance of hub diseases. Targeted surveillance and prevention of hub diseases may reduce multimorbidity patterns, improve healthcare utilisation [ 38 ], and aid policymakers in developing multimorbidity guidelines. Further, our study identifies sex-specific hub diseases and multimorbidity patterns, suggesting targeted measures to reduce multimorbidity. For example, implementing sex-specific prevention and screening strategies tailored to diseases such as genitourinary diseases for females and hyperplasia of the prostate for males would optimise resource allocation. This study has limitations. First, undirected graphs hinder causal inference, and relying on baseline inpatient records may overlook temporal patterns. Second, misdiagnosis and missed diagnosis may affect multimorbidity complexity. These limitations equally affect both sexes and minimally impact sex comparisons. Third, our database excludes data on the working population under 18 years. However, this small group’s low proportion insignificantly impacts our research findings. Fourth, future extensions of the legal working age range may limit the generalizability of our findings to future working populations. Nevertheless, our study has included most of China’s working age ranges and diseases. Fifth, this study is geographically limited to Shaanxi, China, potentially leading to geographical biases. Sixth, the inclusion of blood donors, who may exhibit a healthy donor effect, could underestimate disease prevalence, introduce sampling bias, and reduce the study’s representativeness of the general population, affecting external validity. However, because hospitalization dates, in the final analysis, do not align with blood donation dates, the health status of donors at the time of donation does not directly influence their baseline health status during hospitalization, minimizing the impact of the healthy donor effect. Seventh, our final dataset included predominantly local residents (96.34%), with only a small proportion of migrant populations, and by including only individuals with EHR records, we excluded those not covered by the health record management system, which introduces a potential selection bias that may significantly affect the generalizability of the findings. Future studies should improve sample representativeness. Eighth, this study did not account for occupational factors, such as the impact of physical versus non-physical labor on multimorbidity patterns. This potential heterogeneity may affect the results and warrants further investigation in future research.

Multimorbidity, the coexistence of multiple diseases in an individual, affects 33.1% of adults worldwide [ 1 ]. Multimorbidity prevalence has rapidly increased across developed and developing countries, with a UK study predicting a rise from 54% in 2015 to 67.8% in 2035 among the elderly [ 2 ], and a study in Guangzhou, China, observing an increase from 12.69% in 2009 to 35.13% in 2018 among adults [ 3 ]. Multimorbidity causes disability, decreased quality of life, increased treatment complications and premature mortality [ 4 ]. It poses significant challenges to countries’ primary care as it increases the demand for medical and pharmacotherapeutic resources [ 5 ]. The World Health Organisation (WHO) prioritises reducing the multimorbidity burden in the Multimorbidity Report published in 2016 [ 6 ]. WHO also established the Multimorbidity Working Group, which aims to develop a research program that builds on current knowledge of the etiology of single morbid conditions. The goal is to generate novel scientific evidence on the determinants of multimorbidity by leveraging the existing infrastructure of a prospective study and a network of researchers [ 7 ]. Additionally, the Healthy China 2030 strategy recognises its challenges in promoting healthy ageing [ 8 ]. China, the world’s second-largest economy and population [ 9 , 10 ], has substantial influence on the global workforce. Despite the two-child policy implemented in 2015, China’s fertility rate still declined, from 18.83 million births in 2016 to 9.02 million in 2023 [ 11 ]. China is now moving toward an ageing society due to declining fertility rates and increasing life expectancy [ 12 ]. The previous study projected that the proportion of the elderly population (aged > 65) will double from 11.7% to 23.9% during 2020–2050 [ 13 ]. Projected trends show a shrinking working-age population and a rising economic burden from the elderly [ 14 ]. Chinese policymakers suggest emulating delayed retirement from developed countries as a remedy [ 14 , 15 ]. Early onset of multimorbidity in the working-age population substantially reduces work participation and productivity and increases absenteeism [ 14 ], medical costs, healthcare demand, and treatment complications [ 16 ]. Reviews reported an increasing trend of early onset of dementia and type-2 diabetes mellitus (T2DM) in working-age adults [ 17 , 18 ]. The United Kingdom (UK) reported that the working-age population bears > 50% disease burden of multimorbidity of the whole population [ 19 ]. China will likely face similar challenges with increasing multimorbidity in the coming decades [ 20 ]. Preventing early-onset diseases and reducing multimorbidity in working-age populations are crucial for a healthy and productive workforce. Limited research on multimorbidity in working-age populations has been published (India [ 21 ], Korea [ 22 ], Germany [ 23 ], Canada [ 24 ], Dutch [ 25 ], UK [ 26 ]). Although these studies are timely, they often focused on limited aspects of multimorbidity and examined limited diseases, without exploring the full spectrum of diseases. Research on working populations reported varying multimorbidity prevalence rates (2.4% in India [ 21 ], 3.0% in Germany [ 23 ], and 16.4% in the UK [ 26 ]). Scarce evidence exists on multimorbidity in the workforce in East Asian countries. Exploring sex differences in multimorbidity is crucial for developing sex-specific interventions and treatment guidelines for working-age populations [ 27 – 29 ]. Precision medicine recognises the importance of identifying sex-specific multimorbidity, influenced by genetics, hormones, physiology, behaviour, and sociocultural factors [ 27 ]. For instance, older Chinese females have substantially less access to health resources than their male counterparts, potentially leading to a greater multimorbidity prevalence [ 30 ]. Different sexes may differ in susceptibility to multimorbidity, as females with arthritis or osteoporosis have a higher likelihood of comorbidity [ 29 ], while males with depression are more likely to do so [ 31 ]. However, comprehensive studies comparing multimorbidity between sexes, particularly in developing countries, are limited [ 27 , 28 ]. Our study explored multimorbidity in 393,170 working-age inpatients in Shaanxi, China (1998–2018). It compared multimorbidity patterns by sex and age, using the International Classification of Diseases, 10th revision (ICD-10) [ 32 ] for diagnoses. These results will offer insights into multimorbidity management and prevention strategies across sexes and ages for China’s working-age population.

Additional file 1: Supplementary_Table_S1. The Content of ICD-10 1–22 chapters. Supplementary_Table_S2. The number of inpatients of the subpopulation. Supplementary_Table_S3. The table of the initial number of multimorbidity patterns before exclusion criteria was performed in each inpatient group. Supplementary_Table_S4. Number of nodes, edges and frequency of multimorbidity patterns in the multimorbidity networks stratified by sex. Supplementary_Table_S5. Number of nodes, edges, and frequency of multimorbidity patterns in the multimorbidity networks stratified by age and sex. Supplementary_Table_S6. The top 10 most prevalent diseases in the male and female population. Supplementary_Table_S7. The top 10 most prevalent multimorbidity patterns in different sexes among complete multimorbidity networks. Supplementary_Table_S8. The top 10 most prevalent multimorbidity patterns in the Sex-specific multimorbidity networks. Supplementary_Table_S9. The proportion of the top 10 most prevalent multimorbidity patterns of each age-sex stratified group. Supplementary_Table_S10. The frequency and odds ratio (OR) of the top 10 most prevalent multimorbidity patterns in the 6 subpopulations of male inpatients. (25 patterns). Supplementary_Table_S11. The frequency and odds ratio (OR) of the top 10 most prevalent multimorbidity patterns in the 6 subpopulations of female inpatients. (31 patterns). Supplementary_Table_S12. The proportion of inpatients with or without multimorbidity stratified by age and sex. Supplementary_Table_S13. The summary of the hub diseases of different sex among complete, overlapping and sex-specific multimorbidity networks. Supplementary_Table_S14. The summary of hub diseases in the 6 subpopulations of the male population. (24 nodes). Supplementary_Table_S15. The summary of hub diseases in the 6 subpopulations of female inpatients. (22 nodes). Supplementary_Figure_S1.1 The selection flowchart of the complete multimorbidity networks in the male and female population. Supplementary_Figure_S1.2 The selection flowchart of the overlapping multimorbidity networks in the male and female population. Supplementary_Figure_S1.3 The selection flowchart of the sex-specific multimorbidity networks. Supplementary_Figure_S1.4 The selection flowchart of the complete multimorbidity networks in 18–24 years male and female population. Supplementary_Figure_S1.5 The selection flowchart of the complete multimorbidity networks in 25–31 years male and female population. Supplementary_Figure_S1.6 The selection flowchart of the complete multimorbidity networks in 32–38 years male and female population. Supplementary_Figure_S1.7 The selection flowchart of the complete multimorbidity networks in 39–45 years male and female population. Supplementary_Figure_S1.8 The selection flowchart of the complete multimorbidity networks in 46–52 years male and female population. Supplementary_Figure_S1.9 The selection flowchart of the complete multimorbidity networks in 53–59 years male and female population. Supplementary_Figure_S2. The complete multimorbidity network of 18–24 years age range. Supplementary_Figure_S3. The complete multimorbidity network of 25–31 years age range. Supplementary_Figure_S4. The complete multimorbidity network of 32–38 years age range. Supplementary_Figure_S5. The complete multimorbidity network of 39–45 years age range. Supplementary_Figure_S6. The complete multimorbidity network of 46–52 years age range. Supplementary_Figure_S7. The complete multimorbidity network of 53–59 years age range. Supplementary_Figure_S8. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in males and females. Supplementary_Figure_S9. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in males by various age groups. Supplementary_Figure_S10. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in females by various age groups. Supplementary_Figure_S11. ICD-10 Chapter (Chapter 1–14) rankings based on the number of unique multimorbidity patterns and multimorbidity frequencies in males and females by various age groups. Supplementary_Figure_S12. The trend of the number of nodes and edges across age-sex stratified subpopulations. Supplementary_Figure_S13. The distribution of network predictors of the complete, overlapping and sex-specific multimorbidity networks. Supplementary_Figure_S14. The distribution of various predictors for the age-sex stratified multimorbidity networks. Supplementary_Method_1 Working-Age Definition. Supplementary_Method_2 The explanation of network metrics. Additional file 2: Male_ICD: The diseases’ ICD-10 codes and corresponding chapters and proportion in the male inpatients. Female_ICD: The diseases’ ICD-10 codes and corresponding chapters and proportion in the female inpatients. The other sheets provide detailed information on the nodes and edges in all the network graphs involved in the study. The meaning and explanation of each sheet can be found in the “Explanation” section of Additional File 2. Additional file 1: Supplementary_Table_S1. The Content of ICD-10 1–22 chapters. Supplementary_Table_S2. The number of inpatients of the subpopulation. Supplementary_Table_S3. The table of the initial number of multimorbidity patterns before exclusion criteria was performed in each inpatient group. Supplementary_Table_S4. Number of nodes, edges and frequency of multimorbidity patterns in the multimorbidity networks stratified by sex. Supplementary_Table_S5. Number of nodes, edges, and frequency of multimorbidity patterns in the multimorbidity networks stratified by age and sex. Supplementary_Table_S6. The top 10 most prevalent diseases in the male and female population. Supplementary_Table_S7. The top 10 most prevalent multimorbidity patterns in different sexes among complete multimorbidity networks. Supplementary_Table_S8. The top 10 most prevalent multimorbidity patterns in the Sex-specific multimorbidity networks. Supplementary_Table_S9. The proportion of the top 10 most prevalent multimorbidity patterns of each age-sex stratified group. Supplementary_Table_S10. The frequency and odds ratio (OR) of the top 10 most prevalent multimorbidity patterns in the 6 subpopulations of male inpatients. (25 patterns). Supplementary_Table_S11. The frequency and odds ratio (OR) of the top 10 most prevalent multimorbidity patterns in the 6 subpopulations of female inpatients. (31 patterns). Supplementary_Table_S12. The proportion of inpatients with or without multimorbidity stratified by age and sex. Supplementary_Table_S13. The summary of the hub diseases of different sex among complete, overlapping and sex-specific multimorbidity networks. Supplementary_Table_S14. The summary of hub diseases in the 6 subpopulations of the male population. (24 nodes). Supplementary_Table_S15. The summary of hub diseases in the 6 subpopulations of female inpatients. (22 nodes). Supplementary_Figure_S1.1 The selection flowchart of the complete multimorbidity networks in the male and female population. Supplementary_Figure_S1.2 The selection flowchart of the overlapping multimorbidity networks in the male and female population. Supplementary_Figure_S1.3 The selection flowchart of the sex-specific multimorbidity networks. Supplementary_Figure_S1.4 The selection flowchart of the complete multimorbidity networks in 18–24 years male and female population. Supplementary_Figure_S1.5 The selection flowchart of the complete multimorbidity networks in 25–31 years male and female population. Supplementary_Figure_S1.6 The selection flowchart of the complete multimorbidity networks in 32–38 years male and female population. Supplementary_Figure_S1.7 The selection flowchart of the complete multimorbidity networks in 39–45 years male and female population. Supplementary_Figure_S1.8 The selection flowchart of the complete multimorbidity networks in 46–52 years male and female population. Supplementary_Figure_S1.9 The selection flowchart of the complete multimorbidity networks in 53–59 years male and female population. Supplementary_Figure_S2. The complete multimorbidity network of 18–24 years age range. Supplementary_Figure_S3. The complete multimorbidity network of 25–31 years age range. Supplementary_Figure_S4. The complete multimorbidity network of 32–38 years age range. Supplementary_Figure_S5. The complete multimorbidity network of 39–45 years age range. Supplementary_Figure_S6. The complete multimorbidity network of 46–52 years age range. Supplementary_Figure_S7. The complete multimorbidity network of 53–59 years age range. Supplementary_Figure_S8. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in males and females. Supplementary_Figure_S9. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in males by various age groups. Supplementary_Figure_S10. The number of unique multimorbidity patterns and total frequency of multimorbidity patterns related to each Chapter (Chapter 1–14) in females by various age groups. Supplementary_Figure_S11. ICD-10 Chapter (Chapter 1–14) rankings based on the number of unique multimorbidity patterns and multimorbidity frequencies in males and females by various age groups. Supplementary_Figure_S12. The trend of the number of nodes and edges across age-sex stratified subpopulations. Supplementary_Figure_S13. The distribution of network predictors of the complete, overlapping and sex-specific multimorbidity networks. Supplementary_Figure_S14. The distribution of various predictors for the age-sex stratified multimorbidity networks. Supplementary_Method_1 Working-Age Definition. Supplementary_Method_2 The explanation of network metrics. Additional file 2: Male_ICD: The diseases’ ICD-10 codes and corresponding chapters and proportion in the male inpatients. Female_ICD: The diseases’ ICD-10 codes and corresponding chapters and proportion in the female inpatients. The other sheets provide detailed information on the nodes and edges in all the network graphs involved in the study. The meaning and explanation of each sheet can be found in the “Explanation” section of Additional File 2.