The baseline survey of the Henan Rural Cohort Study enrolled a total of 39,259 participants aged from 18 to 79 years between July 2015 and September 2017 in the counties of Yuzhou, Xinxiang, Tongxu, Yima, and Suiping in Henan Province by using the stratified cluster sampling method. Detail description of the cohort had been previously published [23]. In this study, participants were excluded if they (1) had cancer or renal failure (n=350; for abnormal metabolism [protein, glucose, and lipid]) [24,25] or (2) did not have complete NCD information on hypertension, dyslipidemia, type 2 diabetes mellitus (T2DM), coronary heart disease (CHD), stroke, or hyperuricemia (n=102). Ultimately, 38,807 participants were included in this study (see the study flowchart about the data process in Multimedia Appendix 1).

This study was approved by the Zhengzhou University Life Science Ethics Committee (ethics approval 2015 MEC [S128]). Informed consent was signed by all participants.

Information was collected through face-to-face interviews using a standardized questionnaire. Demographic characteristics included age (18-44, 45-59, and ≥60 years, according to the classification provided by the World Health Organization [WHO] [26]), gender, educational level (elementary school or below, junior high school, and senior high school or above), marital status (married or cohabiting and living alone), and average monthly individual income (<500 RMB [<US $72.4], 500-1000 RMB [US $72.4-144.8], and ≥1000 RMB [≥US $144.8], according to the average monthly income of Chinese rural populations [27]). Lifestyle factors included smoking status (never, former, and current), drinking status (never, former, and current), and physical activity (low, moderate, and high) [28]. Dietary status included vegetable and fruit intake (low [<500 g/day] and high [≥500 g/day]), salt diet (low and high according to dietary flavor habits), and fat diet (low [<75 g/day] and high [≥75 g/day] according to an average taking of meat from livestock and poultry). Family history of hypertension, dyslipidemia, diabetes, CHD, stroke, and gout were also recorded.

BMI was calculated as weight in kilograms divided by height in meters squared, and further categorized as underweight (<18.5 kg/m²), normal (18.5-23.9 kg/m²), overweight (24.0-27.9 kg/m²), and obesity (≥28.0 kg/m²) [29]. Systolic blood pressure and diastolic blood pressure were obtained by using an electronic sphygmomanometer (HEM-770afuzzy, Omron). Venous blood samples were obtained from individuals after at least eight hours of overnight fasting. Total cholesterol, triglyceride, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, fasting blood glucose, and serum uric acid (SUA) level were measured by a chemistry analyzer (Cobas C501, Roche Diagnostics GmbH).

In light of the lack of a standard approach to measure multimorbidity, the selection of morbidities to be included is inevitably subjective and relies on the previous studies. We have selected 6 NCDs—hypertension, dyslipidemia, T2DM, CHD, stroke, and hyperuricemia—that have been reported to be the core diseases in multimorbidity studies [30,31]. These 6 NCDs have been defined in our original survey for the cohort [23]. The definition of hypertension was average systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or self-reported hypertension diagnosed by a physician and receiving antihypertensive therapy within the past 2 weeks [32]. Dyslipidemia was defined as one of the following conditions: elevated total cholesterol level (≥6.2 mmol/L); elevated low-density lipoprotein cholesterol (≥4.1 mmol/L); low high-density lipoprotein cholesterol (<1.0 mmol/L); elevated triglyceride (≥2.3 mmol/L); or self-reported dyslipidemia diagnosed by a physician and the use of anti-dyslipidemia medications in the past 2 weeks [33]. T2DM was defined as fasting blood glucose ≥7.0 mmol/L or self-reported T2DM diagnosed by a physician and the use of antidiabetic medications in the past 2 weeks [33]. CHD and stroke were defined as self-reports of a previous diagnosis by specialist physicians according to criteria recommended by the WHO [34,35]. The definition of hyperuricemia was SUA >7.0 mg/dL (417.0 μmol/L) in men and SUA >6.0 mg/dL (357.0 μmol/L) in women. Multimorbidity was defined as the simultaneous occurrence of at least two NCDs in a single patient [36-39].

The participants were divided into healthy individuals, individuals living with 1 NCD, those with multimorbidities of 2 NCDs, and those with multimorbidities of ≥3 NCDs. Continuous variables were expressed as mean (SD), and ANOVA was used for statistical analysis. Categorical variables were presented as counts and percentages, and the chi-square test was used for statistical analysis.

The prevalence of multimorbidity was assessed for each characteristic group, and we quantified the number of events for each NCD separately and visualized all observed combinations of NCDs with an intersection diagram. Multinomial logistic regression analysis was used to analyze the relationship between demographic factors, lifestyle, BMI, or family history of NCDs and 1 NCD or multimorbidity. When one factor was analyzed, all other factors were adjusted as covariance.

Based on the mean age at diagnosis, the temporal order of NCDs was determined, and binary logistic regression was used to explore the associations between NCD pairs. An NCD with an earlier onset was regarded as a predictor of an NCD with a later onset [40]. Following the cascade of NCD onset, binary logistic regression was used to assess the combined association of 2 conditional NCDs with a third, later NCD by creating a new combined variable with 4 groups of both conditional NCDs, in which the combined group without both conditional NCDs served as the reference. Demographic characteristics, lifestyle factors, dietary status, BMI, and family history of NCDs were adjusted in the regression models. In addition, in the regression of assessing the combined association, when 3 of the conditions were analyzed, the others were adjusted.

Results from the logistic regression models were presented as odds ratio (OR) with a 95% CI. The level of statistical significance was set at P<.05. Statistical analysis was conducted using SPSS (version 21.0; IBM Corp) and R (version 4.0.4; R Foundation for Statistical Computing).

Our study was implemented according to the Strengthening of the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [41] (see the study STROBE checklist for cross-sectional studies in Multimedia Appendix 2).

Table 1 showed the characteristics of the participants of the baseline survey from July 2015 to September 2017. The mean age was 55.6 (SD 12.2) years among the 38,807 participants. Across the NCD groups, age, gender, educational level, average monthly individual income, marital status, physical activity, smoking status, drinking status, high-salt diet, high vegetable and fruit intake, high-fat diet, and BMI were all significantly different (ANOVA and chi-square test, all P values were <.001, except high-salt diet [P=.008]). Compared with the healthy individuals, participant groups with 1, 2, or 3 or more NCDs were more likely to be older and women and have lower education and income, higher BMI, poorer marital status, a higher-salt diet, and lower physical activity level (ANOVA and chi-square test, all P<.001). They were also less likely to have a high-fat diet and adequate fruit and vegetable intake than healthy individuals (chi-square test, all P<.001).

Among all participants, 33.5% (12,990/38,807) were living with 1 NCD, among which dyslipidemia was the most common (5858/38,807, 15.1%). In contrast, 18.8% (7285/38,807) of participants were living with 2 NCDs, and the most frequent cluster of NCDs was dyslipidemia and hypertension (3153/38,807, 8.1%). Moreover, 9.3% (3614/38,807) of participants were living with 3 or more NCDs, and the cluster of dyslipidemia, hypertension, and hyperuricemia (818/38,807, 2.1%) was the most common in participants (Figure 1 and Table 2).

Overall, the prevalence of multimorbidity (2 or ≥3 NCDs) was 28.1% (10,899/38,807), and it was significantly higher in participants who were aged over 60 years (1330/3982, 36%), men (4621/15,354, 30.1%), educated in elementary school or below (5585/17,366, 32.2%), living alone (1330/3982, 33.4%), had low physical activity (4286/12,544, 34.2%), and obese (3280/6878, 47.7%; chi-square test, all P<.001; Table 2).

Overall, participants with older age (all P<.001), those who are living alone (P<.001 and P=.007), those who are overweight and obese (all P<.001), former drinkers (all P<.001), current drinkers (P=.006 and P=.047), and those with a history of family hypertension and diabetes (all P<.001) were at a significantly higher risk of multimorbidity (2 or ≥3 NCDs, respectively; multinomial logistic regression). Although women (all P<.001) and participants with moderate and high physical activity (all P<.001), high-fat diet (P=.004 and P=.008), and who are underweight (all P<.001) were less likely to report multimorbidity (2 or ≥3 NCDs, respectively), individuals who currently smoke were less likely to report multimorbidity (≥3 NCDs; multinomial logistic regression, P<.001; Table 3).

A cascade of NCDs was depicted by the mean age of participants at diagnosis. Among the study participants, 10.2% (3973/38,807) were diagnosed with hyperuricemia at a mean age of 54.0 years, followed by dyslipidemia (14,569/38,807, 37.5% at 55.1 years), hypertension (12,692/38,807, 32.7% at 55.4 years), CHD (1708/38,807, 4.4% at 56.1 years), T2DM (3664/38,807, 9.4% at 56.7 years), and stroke (2613/38,807, 6.7% at 59.1 years). In addition, we observed additional connections between NCD pairs (Figure 2 and Multimedia Appendix 3).

Compared with individuals without 2 conditional NCDs, individuals with 1 conditional NCD (hyperuricemia or dyslipidemia) would have significantly higher odds of hypertension (1.3-1.4, logistic regression; both P<.001). Individuals with 2 conditional NCDs (hyperuricemia and dyslipidemia) would further elevate their odds of hypertension to 2.1 (P<.001; Figure 2 and Multimedia Appendix 4).

Compared with individuals without 2 conditional NCDs, individuals with one of the conditional NCDs (hyperuricemia, dyslipidemia, or hypertension) would have significantly higher odds of CHD (1.2-1.5; the combination of hyperuricemia without dyslipidemia: P=.008; the combination of no hyperuricemia but dyslipidemia: P<.001; the combination of hyperuricemia without hypertension: P=.001; and the combination of no hyperuricemia but hypertension: P=.02). Individuals with 2 conditional NCDs (hyperuricemia and dyslipidemia combination: P<.001; or hyperuricemia and hypertension combination: P=.002) would further elevate their odds of CHD to 1.4-1.7 (Figure 2 and Multimedia Appendix 4).

Compared with individuals without 2 conditional NCDs, individuals with 1 conditional NCDs (dyslipidemia, hypertension, or CHD) would have a significantly higher odd of T2DM (1.3-2.2; all P values were <.001 except the combination of no dyslipidemia but CHD [P=.01]), and individuals with these 2 conditional NCDs (dyslipidemia and hypertension combination or dyslipidemia and CHD combination) would further elevate their odds of T2DM to 2.7-3.1 (all P<.001; Figure 2 and Multimedia Appendix 4).

Compared with individuals without 2 conditional NCDs, individuals with one of the conditional NCDs (dyslipidemia, hypertension, CHD, or T2DM) would have significantly higher odds of stroke (1.4-2.5; all P<.001). Individuals with these 2 conditional NCDs (6 NCD combinations) would further elevate their odds of a stroke to 2.3-3.5 (all P<.001; Figure 2 and Multimedia Appendix 4).

Our findings indicated a strong tendency for the coexistence and accumulation of NCDs in a rural population in Henan, China. The prevalence of multimorbidity was 28.1% (10,899/38,807) and increased with age, unfavorable lifestyles, and a higher BMI among Henan rural adults. The multimorbidity of dyslipidemia and hypertension was the most prevalent in study participants of all age groups. In addition, the onset of NCDs was associated, and a later NCD is often predictable by 1 or more preceding NCDs. The tendency of NCD accumulation indicates that compared with participants without conditional NCDs, participants with 1 existing NCD are more likely to develop a second NCD, whereas participants with 2 NCDs have an even greater likelihood to develop a third NCD.

Our study finding demonstrated a lower multimorbidity prevalence than previous studies [14]. A younger age, existing recall and reporting biases, and a fewer number of NCDs included may have contributed to the lower prevalence observed in our study. Further, consistent with previous studies [2,42,43], the prevalence of multimorbidity increases with age, and approximately one-third of older adults aged 60-79 years develop multimorbidity. Nevertheless, multimorbidity in the young and middle-aged population should not be neglected, given the increasing trend of risk factors in the young and middle-aged population [13,44]. Multimorbidity research and preventive strategies should not focus on older adults only but should instead recognize its importance throughout the whole life cycle.

Our study found that unfavorable lifestyles and a higher BMI were associated with a higher risk of multimorbidity. Our finding is supported by evidence in other settings. A multinational cohort study found that a prediagnostic healthy lifestyle, such as a healthy diet, lower BMI, and higher physical activity, could reduce the risk of multimorbidity [45]. A pooled study from 16 cohort studies demonstrates that a high BMI significantly contributes to the risk of cardiometabolic multimorbidity [46]. Further, a systematic review identifies that physical activity may improve immunity and reduce systemic inflammation and thereby ameliorate multimorbidity [47]. Overall, many modifiable risk factors associated with multimorbidity, including drinking, overweight and obesity, and physical inactivity, could be targets of behavioral prevention and interventions. Differing from previous studies [48-50], our study found that a high-fat diet, current smoking status, and being underweight may reduce the risk of multimorbidity. Given our study is a cross-sectional study with certain biases, the power to explain the causative effects of risk factors on diseases is limited. Further investigation is necessary to confirm this particular finding. Our findings support the WHO’s recommendation to implement evidence-based strategies to reduce poor use of alcohol, encourage physical activity, and maintain a healthy weight.

Our study suggested a tendency for NCD accumulation among the participants. The clustering of dyslipidemia and hypertension was the most prevalent multimorbidity observed in our study, consistent with previous studies [30,51]. A Japanese study further describes hypertension as the most common comorbidity among individuals with dyslipidemia [52]. In addition, the temporal sequence reflects the order of NCDs onset and suggests a potential causal association between the occurrence of these NCDs and the development of other comorbidities [40]. The accumulation of NCDs reflects that participants who lived with 1 conditional NCD are more likely to develop a second NCD, whereas individuals with 2 conditional NCDs are more likely to develop a third NCD than otherwise. Similar results have also been reported in a large-series prospective study: participants with the multimorbidity of T2DM and hypertension have a higher risk of CHD or stroke than participants with hypertension or T2DM [53]. This observation might be partly explained by obesity and the variations of metabolites that have accumulating side effects on health over time [54,55]. Kivimaki et al [54] found that obesity-related disease predicted or was predicted by 1 or more other obesity-related diseases, leading to the accumulation of NCDs. A study of 11,000 participants to examine metabolic pathways associated with 27 NCDs shows that two-thirds of metabolites are associated with more than 1 NCD. In addition, 420 metabolites shared between multimorbidity have been found, revealing several key common pathways in NCDs [55]. Our study suggests that taking reasonable health management and therapies of early NCDs can reduce the incidence or slow down the progression of multimorbidities effectively and ultimately improve the quality of life of the population [56,57].

Our findings have important implications for research and public health policy for multimorbidity. This study indicates that multimorbidity is an important health concern in rural Henan. In light of the WHO Sustainable Development Goals for preventing and controlling NCDs [58], existing health care needs to be improved to cope with the burden of multimorbidity. First, the prevalence of multimorbidity is higher in those older than 45 years. A life cycle approach to multimorbidity and to the challenges it poses to public health is vitally important [59]. Second, there is a clear need to increase the awareness of the importance of a healthy lifestyle in the rural population in Henan. Measures of propagating publicity and education on relevant health knowledge, improving health literacy, and changing unfavorable lifestyles should be implemented. Third, disease-specific guidelines are insufficient to effectively manage patients with multimorbidity, and new detailed guidelines for multimorbidity need to be developed [6]. Given the tendency of the coexistence and accumulation of NCDs, preventing multimorbidity will become difficult if the NCDs are treated separately. Integrated strategies for multimorbidity should be developed, such as monitoring at the individual level with electronic health records and increasing the screening management of high-risk populations.

This study has several limitations. First, the causal relationship between NCDs may not be reliable on account of the data being from the baseline survey of a prospective cohort study in Henan Province. Therefore, prospective studies are needed to confirm the results of this study. Second, we define hypertension, dyslipidemia, hyperuricemia, and T2DM by self-report with laboratory measurements, whereas CHD and stroke are defined only by self-report, which might underestimate their prevalence. Finally, although several potential confounders are controlled, there may be other unknown confounders affecting the results of this study.

In conclusion, our study indicates that multimorbidity is an important health issue in rural Henan. Older age, unfavorable lifestyles, and a higher BMI are associated with a higher risk of multimorbidity. In addition, our findings highlight a plausible explanation for the coexistence and accumulation of NCDs. Care for multimorbidity is complex, and concerted efforts and early interventions to prevent the occurrence of multimorbidity are essential to improve health in the rural population in Henan, China.