The association between multimorbidity and health-related quality of life: a cross-sectional survey among community middle-aged and elderly residents in southern China

This study was based on a cross-sectional community health survey in the Foshan city of Guangdong province in Southern China, which have characteristics similar to the national average in terms of population demographics. The samples in this survey were selected using a multistage and stratified random sampling method. The stratification according to the level of the economy and randomly selected street communities at economic levels, then randomly selected communities in street communities according to proportion. Households within residential communities were then randomly selected from the household lists. In the present study, we used the available information from people aged ≥45 years of age. Of 19,037 participants, 900 (4.73%) subjects were excluded due to incomplete or inconsistent data from questionnaires, including reluctance to provide part of their personal information or not aware of their chronic diseases. Finally, a total of 18,137 middle-aged and elderly adults were included in our analyses.

It is a generic HRQoL measure which can compare HRQoL in populations [7]. The Chinese version of EQ-5D-3 L scale has good reliability and validity [15]. The EQ-5D-3 L consists of two parts: a health description system and Visual Analogue Scale (VAS). The first part records self-assessed health status according to five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension is divided into three levels: no problem, some problems, and extreme problems. A total of 243 health state can be expressed by combining the different level from each dimension. This is then transformed into a weighted health state index score (EQ-5D index) by the Chinese time trade-off value, which is a conversion weight from health utility measurements designed based on the HRQoL preferences of populations. And the range of EQ-5D index is − 0.149 to 1 [16]. The second part used to assess self-perceived levels of health from 0 to 100, where 0 represents the worst imaginable state of health and 100 represents the best imaginable state.

The presence of chronic diseases was determined by self-report, as a yes/no response to the question stem, “has a doctor ever diagnosed that you had...” [17]. The 19 diseases investigated in this study included: hypertension, chronic pain, bone diseases, diabetes mellitus, dyslipidemia, gastroenteritis, gout, coronary disease, peripheral vascular disease, stroke, spleen and gallbladder diseases, chronic obstructive pulmonary disease, chronic kidney disease, cancer, multiple sclerosis, dementia, anxiety, depression and phthisis.. In the face-to-face interview the trained interviewer was used to verify the chronic illness of the interviewees through the diagnosis results and medication to ensure the authenticity and comprehensiveness of the interview.

The variables used for the analyses included age, gender, marital status, education level, and employment status. Marital status was divided into two categories: currently married and single (including unmarried, divorced or widowed). Education level was grouped into three categories as follows: primary school or lower, middle school, high school or above. Employment status was categorized into employed and unemployed. In this study, the definition of unemployed included all subjects who were without work [18].

First, descriptive statistics were calculated for all measures. Means and standard deviations (SD) were presented for continuous variables, while frequency and percentage were used for categorical variables. Second, we combined the groups “some problems” and “extreme problems” into one group, called “problem” group, and compared it with “no problem” group. We evaluated the proportion of having problem on five dimensions and average EQ-5D index. The univariate logistic regression and univariate Tobit regression were used to examine the differences among the category of demographic variable.

Then, we displayed 19 chronic diseases by calculating the prevalence of each disease individually. We chose the prevalent chronic medical conditions with a prevalence > 4%, and analyzed their correlations using contingency tables to generate the most prevalent multimorbidity pairs (the combination of two chronic diseases). And the logistic regression was used to analyze the association between two chronic diseases in the pairs that was controlled for demographics.

Finally, we explored the impacts of multimorbidity pairs on HRQoL. The ceiling effect, which is defined as when the scales are developed such that substantial proportions of individual obtain maximum score and the true extent of their abilities cannot be determined, is particularly evident with the EQ-5D index [19, 20]. Previous researches have shown that preference-based HRQoL index scores exhibit ceiling effects, with a significant number of respondents reporting the highest score (1.0) [9]. In these circumstances the censored least absolute deviations estimator developed by Powell and Tobit models have theoretical advantages over the ordinary least squares estimator [21‐23]. In our research, more than 80% people with “no problem” which means all of them got highest score (1.0) and Tobit regression was used as well as the marginal disutility estimates were reflected by the regression.

Within a sample of Chinese middle-aged and elderly adults, the study results showed that people with chronic conditions were associated with lower HRQoL. Based on this, we found that HRQoL declining to varying degrees was affected by different types and combinations of chronic diseases, and that this decline still persisted following correction for confounding factors. Although the prevalence of hypertension is the highest, its impact on HRQoL is not the greatest. The combination of chronic pain and bone disease has the greatest loss in HRQoL. What’s more, certain chronic diseases often occurred in pairs, especially those pairs involving hypertension or dyslipidemia.

The finding, that pain/discomfort was generally the most concerned domain, was line with the previous reports in China [24] and Western developed countries [25]. This suggested that, despite the diversity of populations, the trend remains the same. It could be because the pain/discomfort was the domain which put the greatest limitations on people’s lives, and it was the most important and least bearable condition for them [4]. The differences of HRQoL in age, gender, education level, and marital status in this study were matched with the published literatures [7, 9]. With the age increased, health problems increasing was a common phenomenon, which also affected on HRQoL [26, 27]. Furthermore, the decrease in HRQoL could be caused by marriage problems and it could increase gradually with age for the singles. Because compared with those who are single or divorced, married people have improved mental health due to the social relationship with the spouse [28]. And in our study, female had higher rates of chronic pain and bone disease, the difference by gender was significant. Some studies also showed that females perceived more severe physical damage and poorer health than males [29‐31]. HRQoL was lower in females than in males. People with high levels of education can better understand and adhere to the treatment of the disease, and those with lower education levels had more problems related to self-management on a daily basis [32]. HRQoL increased with the improvement of educational level [33]. Thus older single women, those with lower levels of education often reported poorer HRQoL.

In our study, 48.3% adults aged 45 years and above suffered from chronic diseases and 52.5% reported the presence of only one or two chronic condition. Only a few have three or more chronic diseases. The prevalence of multimorbidity increased with age from 12.8% among the middle-aged (45–64 years old) to 33.8% in older adults (over 65 years old), which was lower than that of 20.5 and 45.5% reported in other studies [8, 34]. But in different studies, estimates of the prevalence of multimorbidity varied widely, depending on age of study population, number of involved chronic diseases, measurements, and sampling frame [34, 35]. Although the prevalence of multimorbidity was lower than other studies, the effect of multimorbidity on HRQoL was still significant in the study.

Hypertension was found to be the single most prevalent disease, in line with the actual situation in China [36]. In our study, nearly half of the people who reported a hypertension experienced only one condition, which was different from the study about China of Wang et al. [37]. Perhaps it is because the population selected in this study included the middle-aged, which had a lower rate of multimorbidity compared with older people. Nevertheless hypertension was still the most common in chronic pairs in the study. Then we selected seven prevalent chronic diseases with prevalence rates higher than 4.0% to examine the associations between each other. Almost all of the seven selected prevalent chronic diseases tend to co-occur more often than expected. We detected that six pairs of chronic diseases (hypertension and diabetes mellitus, hypertension and dyslipidemia, dyslipidemia and gout, chronic pain and bone disease, chronic pain and gastroenteritis, and diabetes mellitus and dyslipidemia) were highly correlated. This was consistent with a Japanese study which demonstrated that people with either hypertension or diabetes mellitus had a higher risk of having both conditions [38], and a French study also observed that patients with gout often have comorbid conditions of cardiovascular disease, like dyslipidemia [39]. One study revealed that older adults with diabetes mellitus were significantly more likely to have certain chronic conditions simultaneously, including dyslipidemia [40], which was in agreements with our finding regarding the co-occurrence of diabetes mellitus and dyslipidemia. These findings imply that the primary health care system should provide targeted prevention and treatment program for these co-occur chronic disease.

The associations between chronic diseases and the effect of multimorbidity on the HRQoL made us further explore the impacts of different disease pairs on the HRQoL. Among those chronic diseases with high prevalence, the effects on HRQoL ranged from chronic pain (− 0.036) to hypertension (− 0.006). Previous studies showed that hypertension seemed as a clinically silent disease, impaired the HRQoL, but some other variables had the greater impacts on HRQoL [41]. Chronic pain caused the body or nerves system to continuous or recurrent pain, which had a direct effect on the dimensions of usual activities and pain/discomfort. And there was strong correlation between pain severity and HRQoL [42]. The effects of disease pairs on quality of life were not simply the sum of the effects of two diseases, but had different effects in the combination of different diseases. In the common disease pair, co-occurrence of chronic pain and bone disease (− 0.039) had the greatest impact on HRQoL. But there were some disease pairs leading to a less decline in the quality of life compared with the single disease. That maybe because HRQOL was related to individual’s goals, expectations, and concerns of their life. This implies that the concept of HRQoL is subjective and multidimensional, including positive and negative elements. And with the increasing of diseases, people appeared to become more accustomed to their diseases and to have adapting to them [4]. This allowed them to subjectively reduce the severity of the disease, which accordingly mitigated the impact on HRQoL. Although most literatures agreed that the HRQoL decreases with the increase of the number of chronic diseases [7, 9, 43]. But a single disease which had a great impact on the HRQoL was more serious than a combination of two diseases which had a smaller impact. Therefore, more attention should be paid to diseases and disease combinations that have a greater impact on HRQoL.

There are several limitations in this self-reported study. First, the list of chronic diseases used in this study may have encouraged participants to over-report. And the severity of the chronic illness was not taken into account in this study, which may have an impact on the HRQoL. Second, the community-based survey excluded respondents who were too ill to respond and those in institutions. Third, this study used cross-sectional survey data to analyze the correlations between HRQoL and the associated factors, rather than the causation. Last, we only explored the effects of chronic disease pairs on HRQoL, and the multimorbidity patterns with three or more chronic disease were not taken into account.