Association of Cognitive Impairment and Elderly Mortality: Differences of two cohorts ascertained 6-years apart in China

Background: Cognitive impairment is a major contributor to mortality among the elderly. However, the trend between cognitive impairment and mortality with time is understudied. We aim to evaluate the differences in associations of cognitive impairment with all-cause mortality and explore the relationship of cognitive impairment with mortality in different age and sex groups in two cohorts ascertained 6 years apart in China. Methods: A total of 13906 and 13873 Chinese elderly aged 65 years and older were included in the 2002-2008 and 2008-2014 cohorts from the Chinese Longitudinal Healthy Longevity Survey (CLHLS). Mortality data was ascertained from interviews with family members or relatives of participants. Cognitive function, evaluated by the Mini-Mental State Examination (MMSE), were defined by different cut-offs taking educational background into account. Cox models were used to explore the relationship of cognitive impairment with mortality. Results: For the 2002-2008 and 2008-2014 cohorts, the total follow-up times were 55,277 and 53,267 person-years, and the mean (SD) age were 86.5 (11.6) and 87.2 (11.3) years, respectively. Compared to normal cognition, severe cognitive impairment was independently associated with higher mortality risk after controlling for potential confounders, with hazard ratios (HRs) of 1.48 (95% confidence interval [CI], 1.39-1.57) in 2002-2008 cohort and 1.32 (95% CI, 1.25-1.41) in 2008-2014 cohort. The trend of cognitive impairment with risk of all-cause mortality decreased from 2002-2008 to 2008-2014 cohort. The association of cognitive impairment and mortality was decreased with age in the two cohorts. Conclusions : Cognitive impairment evaluated by MMSE was associated with increased risk of mortality and the association decreased with the passage of time during the two 6-year cohorts; this advocates that periodic screening for cognitive impairment among the


Introduction
Cognitive impairment is a major risk factor for poor health in the growing population of elders worldwide 1-3 . It imposes a heavy burden on public health and is associated with shortened life expectancy 4 . The prevalence of mild cognitive impairment in China was about 20.8% in 2014; more than half of these individuals progresses to dementia within 5 years 5 . It was reported that the mortality attributable to dementia in China increased from 1.6 million in 1990 to 2.3 million in 2016 6 . Although a number of epidemiological studies have reported on a cognitive impairment-mortality relationship 3, 7 , most of them concern the association of cognitive impairment and mortality risk of elders in high-income contries [8][9][10][11][12] . In upper-middle income countries countries such as China, several studies have indicated that baseline cognitive impairment increases the risk of all-cause mortality 13,14 . However, studies are limited that include large sample sizes, national representation, different age groups in the oldest old (aged 80 and older) and sex differences.
Prior studies have explored the association of cognitive impairment and all-cause mortality risk with long-term follow-up (14 years and 20 years) 13, 15 , but have rarely studied the impacts of change in medical, demographic and social factors over time on the association between cognitive impairment and mortality. It has been reported that annual mortality among the oldest old was substantially declined between 0.2% and 1.3% in 1998-2008 compared with the participants of the same age born before ten years, but cognitive impairment increased annually between 0.7% and 2.2% in the past ten years 16 .
Therefore, it is unclear whether the impact of cognitive impairment on mortality has 4 changed with the passage of time.
The present study aims to examine and compare the relationship between cognitive impairment and mortality using two cohort studies conducted in 2002-2008 and 2008-2014. Moreover, subgroup analyses were further conducted among different sex and age groups to identify susceptible populations

Study design and participants
The Chinese Longitudinal Healthy Longevity Surveys (CLHLS) was a nationwide survey that randomly selected half of the cities and counties in 23 provinces of China, and recruited participants aged 65 years and older. The present study evaluated baseline cognitive status of two cohorts, ascertained in 2002 and 2008, using the same scale of cognitive function; each cohort was followed for 6 years (till 2008 and 2014) to quantify mortality.
A more detailed description of the CLHLS has been published elsewhere 17 .
The protection of human subjects for the CLHLS was approved by the Ethics Committees.
Informed consent was provided by all participants prior to participation.

Assessment of cognitive impairment and mortality
Cognitive impairment was evaluated using the Mini-Mental State Examination (MMSE), a widely used cognitive test 18 . The total MMSE score ranges from 0 to 30 within 6 dimensions: orientation, registration, attention, language, memory, and visual construction skills. Three methods were used to define cognitive impairment: (1) ≥24 and <23 were used to define cognitive normal (reference) and cognitive impairment 18, 19 ; (2) <18 was used to define cognitive impairment for participants who didn't receive any formal education, <21 for participants who received 6 years of education or less, and <25 for participants who received more than 6 years of education 20, 21 ; (3) ≥24, 18-23, and <18 were used to define cognitive normal (reference), mild cognitive impairment and Mortality date was ascertained from interviews with family members or relatives of participants. The cause-specific mortality was not involved in this study because (1) many of the elderly died at home rather than in medical institutions where cause of mortality might be recorded, and (2) mortality surveillance systems are uncertain in many survey fields.

Assessment of potential confounding variables
A number of variables were collected through a face-to-face standardized questionnaire, including demographic characteristics, economic status, lifestyles, health conditions and medical services.
Marital status was classified into unmarried and married. Education level was classified as no formal education, elementary school graduate (1-6 years of education), and high school graduate ( 6 years of education). Region was defined as: urban, rural and suburban. Exercise was categorized into yes or no. Housework and reading were divided into 3 categories: never, sometimes, and often. Binary variables were defined to assess current status of smoking, drinking, depression and disability in six activities of daily living (ADL) including dressing, bathing, using the toilet, getting in/out of a bed or chair and feeding. Participants with systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg were considered hypertensive. Self-reported history of heart disease and stroke were also collected. According to the survey, we assessed the level of medical care by whether the participant was receiving adequate medical care at present? (Yes or No)" and on the basis of the payor of the medical costs (public medicare or not)". We 6 assessed economic status by asking, "are all financial sources enough for your life?", "do you have a pension?" and "how is your life compared with other local people? (richer, similar and poorer)".

Statistical analysis
We divided the elderly into 3 groups by the conventional MMSE cut-off points for 2002-2008 and 2008-2014. Mean and standard deviation were summarized for continuous variables, frequency and percentage for categorical variables. Comparisons between the elderly were conducted using the chi-square test for categorical variables, Kruskal-Wallis test for continuous variables.
Kaplan-Meier analysis was used to draw the survival curves stratified by MMSE score, compared by the log-rank test. We used the Cox proportional hazards models and the important confounders were identified by previous studies 13,15 . Less than 1.3% for all independent variables had missing values, and multiple-imputation approach were adopted to reduce the influence of missing values on the models 22 .
Hazards ratio (HR) and 95% confidence intervals (CIs) were estimated with the construction of Cox proportional hazards models: the crude model was model 1; age and sex were adjusted in model 2; marital status, living alone, exercise, alcohol consumption, and smoking status were further adjusted in model 3; ADL and depression were further adjusted in model 4; and medical care and economic status were further adjusted in model 5. For subgroup analyses, the elderly were stratified by age group (65-79, 80-89, 90-99, and ≥100) and sex (male and female) in model 5. Interactions of baseline cognitive impairment with age group and sex on mortality risk were explored.
The following sensitivity analyses were conducted to check the robustness of the primary results: (1) removed the participants lost to follow-up to examine possible attrition bias; 7 (2) excluded survival time less than one year due to the possibility that disease status in the last year of life could have affected the risk effects; (3) additionally adjusting for place of residence, dietary habits, hypertension, self-reported heart disease and stroke.
Data analysis was conducted using R version 3.3.1. All statistical tests were 2-sided, and statistical significance was judged by P-values 0.05.  (Table 1).

Relationship between cognitive impairment and all-cause mortality
Appendix Figure 1 shows the Kaplan-Meier survival curve by different categories of MMSEbased cognitive impairment. Significant differences were revealed by the log-rank test in the 3 groups (cognitive normal, mild cognitive impairment and serious cognitive impairment; P<0.001) among the two cohorts.   (Table 3).

Sensitivity analysis
Among two cohorts, there was almost no change in the association between cognitive impairment and all-cause mortality after excluding participants lost to follow-up or with survival time less than one year. The association was still robust after further adjustment for potential confounders (Appendix Table 1, 2). Previous studies on sex differences in the association of cognitive impairment with all-cause mortality have been inconsistent 13, 30 ; this might be attributed to regional differences and insufficient sample size for analysis. An et al. indicated that males had a higher risk than females, which might be attributed to an undesirable lifestyle among men, such as smoking and drinking 15 . Conversely, Kirsten found that women perform worse than men with respect to lifelong subnormal cognitive functioning or emotional disorders 18 . However, there are also studies that reported no sex-specific differences 13 .
Similarly, we did not find sex differences between baseline cognitive impairment and allcause mortality. It is necessary for further analyses to explore possible different patterns of mortality among sexs with cognitive impairment.
Many population-based studies have reported that cognitive impairment was strongly associated with subsequent mortality in the elderly 18, 31, 32 . Our study was consistent with previous findings and further demonstrated that the association of cognitive impairment with all-cause mortality was more pronounced among younger elderly in two Chinese cohorts. One possible reason that cognitive impairment of younger elderly poses a greater risk of mortality may be due to their ability to develop cognitive impairment faster compared to the oldest old, thus triggering a greater risk of mortality. The varying association of cognitive impairment and all-cause mortality in different age groups might be due to the survival bias that the oldest old represent hardy survivors who have successfully adapted to cognitive impairment 26 . Another possible explanation is that the oldest old have a higher risk of mortality, a common competing risk for cognitive impairment, thus causing loss to follow up bias and confusing the association 13 .
Several strengths are worth mentioning in our findings. We included the representative sample of deaths among the Chinese elderly population based on two 6-year cohorts.
Moreover, our age-specific analyses including 65-79, 80-89, 90-99 and ≥100 age groups can help us obtain a more comprehensive understanding of the impact of cognitive impairment on risk of mortality in elders. In addition, our sensitivity analyses suggested that the findings of this study were robust.
Some limitations of this study should be acknowledged. Firstly, cognitive impairment was measured using the MMSE (not based on clinical diagnosis). Secondly, despite the effort of adjusting for a number of confounders such as demographic characteristics, lifestyle factors, heath conditions, health service and economic status, we could not completely eliminate the risk of confounding bias due to unobserved differences in personal characteristics. Thirdly, the association of rapid cognitive decline and mortality might differ in the elderly whose cognitive status did not decline or declined slowly. In our study, we only focused on baseline cognitive impairment and did not assess whether cognitive decline over time was associated with elevated risk of mortality.

Conclusions
The data from this population-based longitudinal study revealed that cognitive impairment was significantly associated with increased risk of all-cause mortality, and the association of cognitive impairment with mortality decreased over a time period during the two 6-year cohorts. Thus, prevention and management of cognitive impairment have substantial benefits for mortality. We advocates that periodic screening for cognitive impairment among the elderly is warranted.

Declarations
Informed consent was obtained from all participants and/or their relatives, and the study was approved by the Ethics Committee of Peking University (IRB00001052-13074).

Disclosure Statement
The authors declare that they have no conflict of interest.

Author Contributions
Jun Duan and Yue-Bin Lv conducted the data analysis, drafted and revised the manuscript.
Xiaoming Shi, Hong Su and Yi Zeng designed the survey, reviewed, and revised this manuscript. Xiang Gao and Virginia Byers Kraus helped to conduct data analysis and critically revised this manuscript, Jin-Hui Zhou helped to implement the survey, and review the manuscript.

Funding Sources
This work was supported by the Chinese Longitudinal Healthy Longevity Survey, which provided the data analysed in this article, is jointly supported by National Natural Sciences Foundation of China (81273160 and 81573247).