Multimorbidity patterns and associated factors in older Chinese: results from the China health and retirement longitudinal study

Multimorbidity, the co-existence of two or more chronic conditions [1], has become a global public health concern. Multimorbidity can cause adverse health outcomes, including frailty [2], disability [3], and even mortality [4]. It also increases healthcare use and costs [5], challenging healthcare systems around the world [6].

Given its importance, there has been a significant increase in studies on multimorbidity in both developed and developing countries, investigating the prevalence, risk factors, and/or outcomes. However, the measurement of multimorbidity varied in different studies. A large number of prior studies assessed multimorbidity by either the co-existence of two or more conditions or simply counting the number of concurrent conditions [7]. As the co-occurrence of chronic conditions in a single individual is often due to their shared etiology [8, 9], understanding the clusters of multimorbidity may improve the effectiveness of the traditional clinical practice, which treats each condition individually [10, 11].

Some scholars have tried to use cluster analysis approaches, such as latent class analysis (LCA), to investigate multimorbidity patterns. LCA is an appropriate tool to identify latent classes of observed variables based on structural equation modeling. It has been applied by some studies to examine how individuals are clustered according to patterns of chronic conditions. For example, by focusing on 10 self-reported conditions from the National Health and Aging Trends Study in the United States, Nguyen and colleagues [2] identified five multimorbidity classes, including minimal, cardiovascular, osteoarticular, neuropsychiatric, and multisystem. Another study on the middle-aged and older adults in Spain identified three patterns: healthy, metabolic/stroke, and cardiorespiratory/mental/arthritis [12].

However, the clustering patterns of chronic conditions were rarely examined in China using national-level data. Among the few exceptions, some studies used China Health and Retirement Longitudinal Study (CHARLS), a nationally representative dataset in China. A study on individuals aged 50 and older in China identified 76 dyadic and 169 triadic combinations using observed-to-expected ratios [13], which are likely due to a single disease's high prevalence, and have limited practical implications due to the large number of dyads and triads. Another study revealed four multimorbidity patterns among Chinese people aged 45 and older [14]. It focused on the similarities between diseases using hierarchical cluster analysis instead of reporting the rate of different multimorbidity categories. Xu et al. used cluster analysis to identify multimorbidity patterns among the middle-aged population and used early data in 2011 [15]. Moreover, although existing studies on elderly population in China have also shown that multimorbidity was associated with older age, females, low education attainment, and unhealthy behaviors like smoking and alcohol consumption [16, 17], the risk factors for specific multimorbidity patterns have been rarely examined.

The multimorbidity concern could be a bigger problem for developing countries, as their healthcare systems may not be well prepared for the rapid transition from infectious diseases to non-communicable diseases [18]. In addition, China constitutes the largest elderly population globally, with more than 264 million (about 18.7% of the population) aged 60 and older, according to the latest national census in 2020 [19]. The rapid aging process and its accompanied high prevalence of multimorbidity in China may exacerbate the burden of multimorbidity. There are three main public medical insurance schemes in China, namely, the urban employee basic medical insurance (UEBMI) for urban employees and retirees, the urban resident basic medical insurance (URBMI) for the unemployed residents in urban areas, and the new rural cooperative medical scheme (NRCMS) for the rural population [20]. Despite the substantial increase in the coverage of social medical insurance, benefits vary dramatically across these schemes [21]. UEBMI has a higher reimbursement rate and service coverage than URBMI or NRCMS. Meanwhile, the health reform in China has not paid particular attention to the potential burden of multimorbidity [22]. Unlike developed countries which have proposed guidelines for the management of multimorbidity [23], there have been no such guidelines [24]. Therefore, research on multimorbidity patterns and their risk factors among elderly population in China is necessary, which can provide implications for clinical practice and health policy. The present study aims to identify predominant multimorbidity patterns using LCA among community-dwelling elderly population in China and analyze associated factors for certain multimorbidity patterns.

Using the latest wave of CHARLS, this study reported the prevalence of multimorbidity patterns and examined its associated factors among elderly population in China. About 65.6% of all participants had at least two chronic conditions. By applying LCA, all participants were assigned to five multimorbidity classes based on their shared disease pattern: relatively healthy, vascular, respiratory, stomach-arthritis, and multisystem morbidity. When examining the associated factors of multimorbidity patterns, we found that participants of older age and female sex were more likely to be assigned to the vascular class and multisystem morbidity class (with the relatively healthy class as reference). A positive association was shown between SES and the vascular class. Also, the probability of being assigned to the multisystem morbidity class was significantly higher for those with UEMI or other MI. Smokers were more likely to be in the respiratory class, and multisystem morbidity class. Physical activity was found to decrease the probability of being in the vascular class, respiratory class, and multimorbidity patterns.

Multimorbidity was highly prevalent (65.6%) among the studied population, which slightly differed from the prevalence reported in the literature [13, 27, 29], largely due to different age restrictions and the inclusion of chronic conditions. For instance, Chen and colleagues [16] reported a rate of 45.5% among urban residents aged 45 and above using data from the 2011 wave of CHARLS, and Zhao and colleagues [27] reported a rate of 61.9% among individuals aged 50 and older in the 2015 wave of CHARLS. Our study provided reasonable estimates of the multimorbidity rate given the increased multimorbidity with age.

Using LCA, our study further identified five multimorbidity patterns among older adults in China, namely, relatively healthy (49.8%), vascular (24.7%), respiratory (5.6%), stomach-arthritis (14.5%), and multisystem morbidity (5.4%). Previous studies also studied multimorbidity patterns in China but more focused on disease similarities instead of categorizing all participants into exclusive multimorbidity classes [30, 31]. They found similar disease patterns to the present analysis, such as cardiovascular/metabolic clusters [31]; however, they seldomly identified the relative healthy class or multimorbidity class due to different focuses and cluster methods. Almost half of the participants were classified into the relatively healthy class in the present study. This coincided with previous studies using the same method showing a large proportion of participants in the relatively healthy class, with 63.8% in Span [12], 60.4% in Korea [32], 71% in England [26], and 24.7% in the United States [2]. In our analysis, 24.7% were classified into the vascular class consisting of participants with an excess rate of cardiovascular diseases. Similarly, studies in developed countries identified a metabolic or cardiovascular class [2, 12, 26, 32]. The existence of a respiratory class (5.6%) characterized by the high prevalence of chronic lung disease and asthma was evident, given the shared etiology [33]. Similar to previous studies revealing a digestive-kidney-arthritis cluster in China [15], our study also identified the stomach-arthritis class, which had not been reported in previous studies in other regions. One possible reason might be that some of the existing studies did not include stomach or digestive disease in the analysis. Other explanations might be that both arthritic and digestive diseases have a high prevalence in China [14] and the medication for arthritics (i.e., non-steroidal anti-inflammatory drugs) may lead to adverse side effects for the stomach [34]. The multisystem morbidity class was revealed in most studies, indicating that chronic diseases with diverse etiologies may also be clustered together [4].

Based on previous studies on the risk factors of multimorbidity, we extended the literature for multimorbidity patterns. Consistent with previous evidence [2, 12, 26, 32], participants with an older age had higher odds of multimorbidity for all classes, which is not surprising given the deterioration of health during the natural aging process. Regarding sex differences, there were inconsistent findings regarding multimorbidity prevalence [16, 17, 35], largely due to their inclusion of different chronic conditions and covariates. The sex difference could be better understood by looking at specific multimorbidity patterns. Consistent with previous studies [2, 12, 32], we found a higher probability of belonging to the vascular class, stomach-arthritis class, and multisystem morbidity class (compared to the relatively healthy class) for women. Similarly, previous studies also demonstrated that the prevalence of cardiovascular diseases [36], arthritis [37], and multimorbidity [27] was higher among women.

In contrast to previous studies in developed countries showing that people with low SES were more likely to belong to multimorbidity classes related to cardiometabolic conditions [12, 26, 32], we documented a significantly higher probability of the vascular class for those with higher SES in China, including urban residence, higher education, higher income, and medical insurance. Similarly, previous studies in China found a higher prevalence of multimorbidity for urban residents than their rural counterparts [14, 35]. This could be explained by people in developing regions, including China, with high SES being able to afford the consumption of high-calorie foods and avoid physically demanding tasks [38, 39], which may lead to a high prevalence of chronic diseases such as diabetes, hypertension, and dyslipidemia among those in urban areas and with a high income. Another reason might be the limited access to healthcare resources for those in disadvantaged SES groups, and the underdiagnosis of chronic diseases [40, 41].

In line with the existing evidence [26], we revealed that smoking and a lack of physical activity were risk factors for multimorbidity. More specifically, ever smoking was significantly related to the respiratory class and high multisystem morbidity class, which is plausible given the detrimental effects of smoking on the respiratory system [42]. People who participated in physical activity reported lower odds of being assigned to the vascular class, respiratory class, and multisystem morbidity class. This could be explained by physical activity benefitting cardiorespiratory function [43]. We did not find any significant influences of alcohol consumption on multimorbidity patterns, given the possible health benefits of light alcohol intake [44].

The current study applied LCA to classify the participants into five latent classes and analyze the associated factors of multimorbidity patterns. Nevertheless, it has several limitations. First, most of the variables are self-reported, which may cause potential bias. For example, all 14 chronic conditions come from respondent self-reporting. Therefore, the multimorbidity prevalence was underestimated, and associated factors for multimorbidity patterns should be interpreted with caution regarding the problem of underdiagnosis. Second, the cross-sectional design of the current study made it impossible to draw confident causal conclusions. Third, other potentially influential factors such as the nutrition status were not involved due to the limitations of the secondary dataset. Our strengths include the provision of the recent picture of multimorbidity picture based on the latest data from a nationally representative survey in China, which could be generalized to the whole country, and more importantly, the examination of risk factors for different multimorbidity classes, which has great implications for both health policy and clinical practice.

In conclusion, using nationally representative data of older adults in China, we revealed a multimorbidity rate of 65.6%. LCA analysis showed that different conditions clustered together in a predictable pattern and identified five multimorbidity classes: relatively healthy, vascular, respiratory, stomach-arthritis, and multisystem morbidity. Older age, female sex, high SES, smoking, and no physical activity were risk factors for certain multimorbidity patterns. The findings have significant policy and clinical implications concerning the choice of prevention interventions and treatment strategies to target a multimorbidity pattern instead of a single condition.