Relationship Between Multimorbidity, Combination and All-Cause Mortality Among Older Adults: A Retrospective Cohort Analysis

Background: Previous studies have evaluated the association of multimorbidity with higher mortality, but epidemiologic data on the association between the combination of multimorbidity and all-cause mortality risk are rare. We aimed to examine the relationship between multimorbidity (number/combination) and all-cause mortality in Chinese older adults. Methods: We conducted a population-based study of 50,100 Chinese participants. Cox regression models were used to estimate the impact of long-term conditions (LTCs) on all-cause mortality. Results: The prevalence of multimorbidity was 31.35% and all-cause mortality was 8.01% (50,100 participants). In adjusted Cox models, the hazard rations (HRs) and 95% condence intervals (CIs) of all-cause mortality risk for those with 1, 2, and ≥ 3 LTCs compared with those with no LTCs was 1.10 (1.01-1.20), 1.21 (1.10-1.33), and 1.46 (1.27-1.67), respectively (P trend <0.001). In the LTCs ≥ 2 category, the combination of chronic diseases that included hypertension, diabetes, CHD, COPD, and stroke had the greatest impact on mortality. In the stratied model by age and sex, absolute all-cause mortality was higher among the ≥ 75 age group with an increasing number of LTCs. However, the relative effect size of the increasing number of LTCs on higher mortality risk was larger among those < 75 years. Conclusions: The risk of all-cause mortality is increased with the number of multimorbidity among Chinese older adults, particularly combinations.

has been signi cantly reduced in high-income countries [17,18], whereas rates of mortality caused by multimorbidity are rising rapidly in low-and middle-income countries, posing a serious social burden [19].
Thus, a better understanding of the epidemiology of multimorbidity is necessary to develop interventions to reduce the burden of death. However, the relationship between multimorbidity (number/combination) and all-cause mortality has not been well described in the Chinese population. Therefore, we designed a retrospective cohort study to explore the relationship in Chinese older adults.

Study design and participants
We performed a retrospective cohort analysis of 103,729 participants who entered the health check-up program in Xinzheng City, Henan Province, Central China, from January 1, 2014 to November 15, 2019. A total of 84,353 (81.3%) participants were successfully followed up. We excluded participants with age < 65 at baseline (n = 31,900), or missing data for date of birth and gender at baseline (n = 2,353). A total of 50,100 individuals were included in the main analysis (Supplemental Fig. 1). The study was approved by the Ethics Committee of Zhengzhou University, and written informed consent was obtained from all participants.

Data Collection
A standardized questionnaire was used by trained research staff under stringent quality control to collect information. Marital status was categorized as living with partner and without partner. Smokers were de ned as ever smoking at least 100 cigarettes in their lifetime and classi ed as nonsmokers and previous/current smokers [20]. Alcohol consumption was divided into three categories: never, occasionally or daily. Physical activity was a categorical variable based on the self-reporting and classi ed as never, occasionally or daily. The measurements for height and weight were performed with the subjects wearing light clothing without shoes [21]. Blood pressure was measured by an automatic sphygmomanometer (Omron HEM-7125, Kyoto, Japan) after subjects had rested in a seated position for at least 5 minutes, and the mean value was recorded when the subjects were measured three times [22]. Blood samples were obtained after an overnight fast of at least 8 hours to assess levels of fasting plasma glucose (FPG) using an automatic biochemical analyzer (DIRUI CS380, Changchun, China) [22]. In this study, diabetes was de ned as fasting glucose levels ≥ 7.0 mmol/L and/or current treatment with anti-diabetes medication according to the China guideline for T2DM [23]. Hypertension was de ned as positive for participants who were considered hypertensive with systolic blood pressure (SBP) ≥ 140 mm Hg or diastolic blood pressure (DBP) ≥ 90 mm Hg or current use of antihypertensive medication [24]. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m 2 ), which was classi ed based on the WHO classi cation criteria into underweight (< 18.5 kg/m 2 ), desirable (18.5-24.9 kg/m 2 ), overweight (25.0-29.9 kg/m 2 ), obesity class I (30.0-34.9 kg/m 2 ), obesity class II (35.0-39.9 kg/m 2 ), and obesity class III (≥ 40.0 kg/m 2 ) [25]. Information on the 7 chronic conditions diagnosed by a physician or reported by the study participants was recorded: (1) hypertension, (2) diabetes, (3) coronary heart disease (CHD), (4) stroke, (5) chronic obstructive pulmonary disease (COPD), (6) cancer, and (7) mental system disease. Multimorbidity was classi ed based on number of chronic conditions into no LTCs, 1 LTC, 2 LTCs, and ≥ 3 LTCs. The outcome variable for the present study was all-cause mortality. Death certi cates were obtained as well.

Statistical analysis
All quantitative variables are described using the median (interquartile range [IQR]) for skewed distribution, and qualitative variables are expressed as frequency (%). Cox regression models were used to estimate hazard ratios (HRs) and 95% con dence intervals (CIs) for risk of all-cause mortality by groups of LTC categories. With no LTCs as the reference group, we constructed three multivariateadjusted Cox models: model 1 was adjusted for age, sex and marital status; model 2 was adjusted for model 1 plus smoking status, alcohol status and physical activity; and model 3 was further adjusted for BMI. In order to understand the pattern and effect of multimorbidity, we further analyzed the common combination of multimorbidity in older adults. A test for multiplicative interaction showed that age (< 75/ ≥75 years) and sex at baseline modi ed the association of LTCs and risk of all-cause mortality (both pinteraction < 0.001). Therefore, for subgroup analyses, we strati ed participants by age and sex at baseline. Finally, sensitivity analysis was performed using LTCs de ned by the medical examination diagnostic record at study baseline and not including self-reported history. Statistical analyses were performed using SAS 9.1 (SAS Institute, Cary, NC, USA), the forest plot was performed by GraphPad Prism 8, and the level of signi cance was considered at P < 0.05 (two-tailed).
During 205,022.95 person-years of follow-up, the mortality rates (per 1,000 person-years) were 15.92, 19.83, 21.18, and 24.80 in the no LTCs,1 LTC, 2 LTCs, and ≥ 3 LTCs groups, respectively. The number of LTCs at baseline had a signi cant association with all-cause mortality observed in the three models (   LTCs, long-term conditions; One asterisk Per 1000 person-years; Two asterisk adjusted for age, sex, marital status, smoking status, alcohol status, body mass index, and physical activity levels at baseline; CHD, coronary heart disease; COPD, chronic obstructive pulmonary disease. The analysis of all-cause mortality for multimorbidity strati ed by age and sex is shown in Fig. 1. In the multivariable-adjusted models, the absolute event rate for all-cause mortality was higher for the age group ≥ 75 years, but the relative effect sizes for mortality risk with increasing number of LTCs were higher for the age group < 75 years. Absolute mortality was highest in the age group ≥ 75 years with ≥ 3 LTCs (6.48% for males and 3.71% for females). However, participants in the age group < 75 years with ≥ 3 LTCs had the highest relative risk of all-cause mortality (HR 1.85, 95% CI 1.40-2.44 for males and HR 1.84, 95% CI 1.34-2.54 for females) compared to participants with no LTCs in the same age group. The observed effect size was similar for both female and male. In addition, we found that being older, living without partner, and being underweight had a higher risk of mortality. In contrast, participants who were female, overweight, class I obesity, and physically active had a signi cantly lower adjusted risk of allcause mortality (Supplemental Table 1). The sensitivity analysis yielded similar ndings as our main results, and the risk of death increased with the increase in LTC count when the LTC count was de ned only by the physical examination records and did not include the self-reported conditions (Supplemental Table 2).

Discussion
This cohort study demonstrates that multimorbidity has a positive association with all-cause mortality.
The results from sensitivity analyses were robust. In the LTCs ≥ 2 category, the chronic disease combination that included hypertension, diabetes, CHD, COPD, and stroke had the greatest impact on mortality. In the strati ed model by age and sex, absolute all-cause mortality was higher in the age group ≥ 75 years, but the relative magnitude of the effect on mortality risk was greater among those < 75 years.
The higher all-cause mortality risk among participants with 1, 2, and ≥ 3 LTCs found in this study is consistent with previous studies that have focused on older adults [26,27]. The study of Martinez-Gomez et al. observed signi cant upward trends with 1 LTC (HR 1.26), 2 LTC (HR 1.78), and ≥ 3 LTC (HR 2.27) than those without LTC [26]. Similarly, Nunes et al. conducted a meta-analysis of 26 studies and observed similar effect sizes [10]. And we also found that LTCs combinations were most strongly associated with all-cause mortality relative to the number of multimorbidity alone, whereas previous studies mainly focused on the role of a single disease in death, such as diabetes [28]. Moreover, the present study found that hypertension was the most frequent occurrence, with presence in 11 of the 16 different combinations, suggesting that there may be a potential link between hypertension and a variety of chronic diseases. However, a large sample of participants aged 40 to 69 years study based on 36 chronic conditions showed that conditions such as hypertension, diabetes, and asthma were at the center of common multimorbidity [16]. These ndings suggest there is a need for further research to explore the links between these diseases and their possible interactions, which is of great signi cance for the guidance and treatment of multimorbidity, and also provides a theoretical basis for the formulation of health management measures and allocation of medical resources for the elderly population to some extent.
A prospective population-based cohort study of England people aged 37 to 73 years shown that participants in the younger age group 37-49 years with ≥ 4 LTCs had the highest relative risk of all-cause mortality [11]. And we found that the association on risk of all-cause mortality with an increasing number of LTCs was particularly evident among younger age groups (< 75 vs ≥ 75 years old), which not only lls in the gap of the age of participants, but also is basically consistent with the trend of previous studies. This study found that female had a higher prevalence of multimorbidity than male, which was consistent with previous ndings [29,30]. However, the observed magnitude of mortality effect size was similar for both female and male. Potential explanations for the phenomenon are that females are generally more sensitized to their health [30], more likely to report more conditions [31], and more likely to engage in preventive health behavior [32]. These ndings suggest the need for early intervention to manage and prevent chronic diseases and to reduce the prevalence of multiple diseases as much as possible.
Studies have shown that the causes of death from multimorbidity can be attributed to 4 major underlying risk factors: smoking, alcohol consumption, underweight, and physical inactivity [11,33]. While previous studies found that smoking was a risk factor for all-cause mortality [11], our results found no statistically signi cant difference, which is possibly explained by the quitter bias (people may stop smoking because they are in poor health and may be advised not to continue). The results of the risk study on alcohol consumption were basically consistent with the results of Ortolá et al., which showed that there was no statistically signi cant difference in mortality between light-to-moderate alcohol consumption and no alcohol consumption among people over 60 years old [34]. As well, we found the greater all-cause mortality risk among persons with underweight (BMI < 18.5 kg/m 2 ) in this study, which agreed with previous studies [21,35,36]. In addition, this study suggests that physical activity is associated with a lower mortality risk, and the underlying mechanism behind this nding may be that physical activity delays disease progression by preventing many chronic diseases, including diabetes, cardiovascular and respiratory diseases, and some types of cancer [37]. Therefore, once any of these chronic conditions is diagnosed, physical activity is often incorporated into treatment plans to improve quality of life and survival [26,37,38].
Our study has several strengths. First, the determination of chronic diseases was relatively accurate and comprehensive, including the self-reported condition of participants and the diagnosis made by the physician based on the professional comprehensive examination. Second, our study was also novel in that we explored combinations of multimorbidity that were associated with the highest risk of death. In addition, it is more convincing to assess the relationship between multimorbidity and all-cause mortality based on a large sample size. Finally, the main results remained robust after conducting sensitivity analyses. However, the study has some limitations. Since we included only seven chronic diseases registered at baseline, some other diseases associated with older people such as hyperlipidemia and arthritis could not be taken into account, which may underestimate the prevalence of many diseases in this study, and there was no way to estimate the severity of conditions as well. Another restriction is that our participants were people over 65 years, which should caution generalizing our ndings to younger age groups. In addition, although we adjusted for various covariates, there is still a possibility of residual confounding, such as diet factors. Last, recall bias is unavoidable in self-reported information.

Conclusion
Our ndings suggest an increased risk of all-cause mortality with an elevated number of multimorbidity, particularly LTC combinations, which provides scienti c basis for the treatment, prevention and control of the multimorbidity, and has important public health signi cance in guiding the rational allocation of medical and health resources.

Declarations
Ethics approval and consent to participate The study was approved by the Ethics Committee of Zhengzhou University, and written informed consent was obtained from all participants.

Consent for Publication
Not applicable.
Availability of data and material The datasets used and/or analyzed in this study were obtained from third parties and cannot be publicly available to con dentiality requirements.
Competing interests