Social support and the incidence of cognitive impairment among older adults in China : Findings from the CLHLS study

Objective: Social support is thought to be associated with cognitive impairment (CI) in the older adults. However, the longitudinal relationship between the distinct dimensions of social support and incidence of cognitive impairment remains unclear. This study aims to investigate the association between social support and the incidence of cognitive impairment among the older adults in China. Method: We used longitudinal data (2005-2014) from the Chinese Longitudinal Healthy Longevity Survey (CLHLS, 2005-2014, mean follow-up years 5.32 ± 2.64). In total, 5930 participants (aged 81.7 ± 9.7 years, range 65-112 years, 48.9% male) were enrolled. Cognitive impairment was measured by the Mini-Mental State Examination (MMSE). Social supports included support from family (marital status; contact with children, siblings, relatives, friends and others; money received from children; children’s visit; siblings’ visit) and social community (social service, social insurance). We calculated subdistribution hazard ratios (SHR) of cognitive impairment by establishing Cox regression model, adjusting for residence, gender, age, education, participation in physical exercise, activities of daily living abilities, depressive symptoms, smoking and drinking. Results: Children’s visit and marital status were significantly associated with decreased incidence rate of cognitive impairment. However, social support from other resources did not consistently predict the incidence of cognitive impairment. Specifically, (1) participants who were married had a lower incidence rate of cognitive impairment compared to the widowed (SHR = 0.849, 95% confidence interval, 0.730-0.987, p = 0.033); (2) participants who had frequent visits from their children showed lower incidence of cognitive impairment compared to those who


Introduction
Cognitive impairment not only weakens older adults' ability of independent living, but also causes physical diseases and psychological pain [1], and even increases their risk of death [2]. Therefore, the preservation and promotion of cognitive function are important for improving the older population's quality of life.
Social relationships, which are contingent on access to social networks, promote participation in social activities and provide access to social support [3]. A series of studies found that social support is a strong predictor of health-related quality of life, mental health as well as functional impairment [4][5][6]. Numbers of studies have also reported a significant association between social support and cognitive decline in older adults [7][8][9][10].
Although previous studies have demonstrated that social support is associated with cognitive function, the understanding of the role of social support in cognitive function is still limited. Social support, often divided into emotional, instrumental, and informational support, refers to a person's perception of the availability of help or support from others in their social network [11]. However, most of studies did not distinguish the sources of social support, which might be unclear to investigate the specific contribution of various social supports. For example, Andrew & Rockwood (2010) used the composite "social vulnerability index" to reflect social support [7], which included emotional/instrumental/informational support from close family members, relatives, friends and someone others. However, social support provided by different people might play different roles in cognitive decline. Interactions with close family members (especially those who live with them) were likely to be the most influential factors, while relationships outside the family or some instrumental/informational support from community might have minor effect on cognition decline. According to a previous study about empty-nest older adults, formal social support (social community services) did not satisfy their demands, but informal support (daily care from close family members) was more important for older adult's mental health [12].
In addition, social relationships are diverse under different cultural background.
Previous study claimed that Chinese social network structure differs from that of Western countries, as the Chinese older adults are more likely to live with their children, and their social interaction is more family-centered [13]. English and Carstensen [14] further pointed out that, with aging, the social contact of older adults decreases, and the relations with their spouses and other family members comprise the better part of their social network. According to surveys on Chinese older adults, social support, especially emotional support from children, is one of the most important factors affecting mental health [15]. Therefore, it is worth of consideration of specific sources of social support and their different contributions to old folks' cognition in Chinese culture background.
In the present study, we focused on the relations between specific social support and cognitive impairment. We hypothesized that different sources of social supports make different contributions to cognitive decline among Chinese older adults, and support from close family members is most influential.

Study population
Data were obtained from the Chinese Longitudinal Healthy Longevity Survey (CLHLS). The CLHLS was initiated in 1998 and follow-up surveys were conducted in 2000, 2002, 2005, 2008, 2011 and 2014. The details of the study design and data collection of CLHLS have been fully described previously [16]. At first, the CLHLS project focused on the oldest section of the elderly population, aged 80 and older, Social support Social supports were constituted by domains of emotional support and material support from family members and the others. Social support inside family included contact with spouse/children/other relatives, Children's frequent visit (yes/no), sibling's frequent visit (yes/no), care in sick (family members' care when participants are sick), the money that participants received from children, and the money that participants gave to children. In addition, marital status (married and living together; married but separated; widowed; divorced; never married) as a mixed variable was also included in the analysis.
Contact with family members was measured using three questions: The first three persons you talk to when you need to tell something about yourself, The first three persons you ask for help when you have problems/difficulties, and The first three persons with whom you usually talk frequently in daily life. For each question, the score was rated on one of the five responses (i.e. spouse, children, children in law, friends/neighbors, and other relatives). If the first listed person was spouse, then the item "spouse" scored 3; if the second listed person was spouse, then the item "spouse" scored 2; otherwise scored 1. The item "children" and "other relatives" scored under the same rule with "spouse". Composite scores were calculated separately for each item (ranging from 3 to 9). The higher the scores the better the relationship (Table 1). Children's visits and siblings' visits were recoded according to the question "Please list the family members that you are living with (including names, gender, age, relations, alive or not, and frequently visit or not)". The money received from children (money received, for short) was measured through three questions: "how much did you receive from your son(s) or daughter(s)-in-law last year?", "how much did you receive from your daughter(s) or son(s)-in-law last year?" and "how much did you receive from your grandchild(ren) last year?" Then, the sum of money of the three answers was calculated. The money given to children (money given, for short) was measured in the same way. Care in sick was assessed by the question: "who take care of you when you are sick?", and answers were classified into 4 categories: none, spouse, children, others (friends, neighbors, nurses). Table 1 The recodification of questionnaire Daily talk Tell something Ask for help Sum score Spouse 0-3 0-3 0-3 0-9 Children 0-3 0-3 0-3 0-9 Children in law 0-3 0-3 0-3 0-9 Friends/neighbors 0-3 0-3 0-3 0-9 Other relatives 0-3 0-3 0-3 0-9 Notes: "Daily talk" refers to the question: "The first three persons to whom you usually talk frequently in daily life"; "Tell something" refers to the question: "The first three persons to whom you talk when you need to tell something of you"; "Ask for help" refer to the question: "The first three persons you asking for help when you have problems/difficulties".
Social support outside family included the social service and social insurance. Social service was assessed by 8 questions: "is personal care service available in your community?", "is home visit service available in your community?", "is psychological consulting service available in your community?", "is daily shopping service available in your community?", "is social and recreation service available in your community?", "is human rights consulting service available in your community?", "is healthcare education service available in your community?" and "is neighborhood-relation service available in your community?". All answers were classified into 3 categories: yes, no, and unknown.
Social insurance was also assessed by 8 questions: "do you have retirement wage at present?", "do you have pension at present?", "do you have private old age insurance at present?", "can you access to public free medical services at present?", "can you access to the cooperative medical scheme at present?", "do you have basic medical insurance at present?", "do you have severe disease insurance at present?" and "do you have life insurance at present?". All answers were classified into 3 categories: yes, no, and unknown.
In addition, considering the imbalance of numbers in answers of social service and social insurance (Table 2), social service and social insurance were also defined by the 8 questions together (whether the participants are available to any social insurance/social service), respectively.  [17]. The higher the score (0-30), the greater the cognitive ability of the participant. As the majority of the Chinese older adults had no formal education, several items of MMSE were simplified to make them more practical. The serial 7 subtraction was simplified to 3 subtraction, and reading and writing a sentence was replaced by verbally reporting as many names of food items as possible in one minute [18]. As over half of the participants (54%) received no formal education in the present study, we used education-based MMSE cut-off points to define cognitive impairment, that is, < 18 for participants with no formal education, < 21 for participants with 1-6 years of education, and < 25 for them with more than 6 years of education [19,20].
ADL ability was measured using the Katz Index of Activities of Daily Living scale (Cronbach's α = 0.87) [21]. Depressive symptoms were measured using three items about neuroticism ("I often feel fearful or anxious"), loneliness ("I often feel lonely or isolated"), and perceived loss of self-worth ("The older I get, the more useless I feel"). Participants answered on a five-point Likert scale, with "1" for "dose not describe me at all" and "5" for "describes me very well". It is the recommended measurement of depressive symptoms in CLHLS database book [22]. The sum of scores on the three items was treated as the indicator of depressive symptoms, with higher scores indicating worse psychological well-being (Cronbach's α = 0.64) [20].
Participation in physical exercise was measured by one question: "Did you usually exercise in the past?". Description of all variable showed in Table 3.

Analysis
Cox models were constructed to estimate the subdistribution hazard ratio (SHR) and the 95% confidence interval of cognitive impairment associated with social support.  Table 4. According to the result above, the composite scores of contacts with spouse/children/other relatives were excluded in the following analysis (as children's visits predicted the incidence of cognitive impairment while contact with family did not. The composite scores might cannot distinguish the difference of social support inside family).
Secondly, all types of social supports were examined separately (i.e. marital status, child visit, sibling visit, money give, money receive, care in sick, social insurance, social service), controlling for demographic variables (Table 5a, model 1). In addition, when children's visits and sibling visits were examined, children alive and sibling alive were additional adjusted in the model respectively; when money giving and money received were examined, children alive was also controlled; when care in sickness was examined, children alive and marital status were additional adjusted in the model. As social insurance and social service were measured by 8 yes/no questions respectively, we conducted Cox analysis both for the composite measurement (whether the participants are available to any social insurance/social service; Table 5a) and for each question (Table 5b).
Thirdly, demographic variables like gender, age, residence, education, depression, ADL, drinking, smoking, exercise were included in regression equation (Table 6,   model 2). Finally, all variables were included in Cox regression model together (Table 6, model 3).

Results
Age, education, depressive symptoms, ADL, and contact with family were continuous variables in analysis, but for the clarity of results presentation, they were classified into several categories (for details of demographics, see  Regarding social supports, marital status had a significant impact on the incidence of cognitive impairment. Model 3 in Table 6 showed that participants who were married had a 15.1% decreased risk of developing cognitive impairment compared to those who were widowed (SHR = 0.849, 95% confidence interval 0.730-0.987, p = 0.033). In addition, participants who were married but lived separately (the number of this group is small, 145) also showed a lower risk of developing cognitive impairment compared to the widowed (SHR, 0.389, 95% confidence interval, 0.224-0.677, p < 0.001).
Children's visits had a consistently impact on the incidence of cognitive impairment in model 1, model 2 and model 3 (Tables 5a and 6 confidence interval,0.755-0.984, p = 0.028; Table 4), but neither social insurance or social service was significant in further univariate analysis (controlling for all demographic variable) and multivariable analysis. In addition, the analysis based on single question about social insurance showed that, participants with server disease insurance and life insurance were exposed to lower incidence of cognitive impairment (Table 5b).

Results
Age, education, depressive symptoms, ADL, and contact with family were continuous variables in analysis, but for the clarity of results presentation, they were classified into several categories (for details of demographics, see Table 3).   Table 4), but neither social insurance or social service was significant in further univariate analysis (controlling for all demographic variable) and multivariable analysis. In addition, the analysis based on single question about social insurance showed that, participants with server disease insurance and life insurance were exposed to lower incidence of cognitive impairment (Table 5b).

Discussion
In the present study, the association between social support and cognitive impairment was investigated by using a large population sample representative of Chinese older adults during a 9-year follow up. We found that emotional support from children (children's visit) and marital status were consistently associated with lower incidence of cognitive impairment in current analyses, while material support and support outside family were unstable variables associated with cognitive impairment among older adults. With aging, the social contact of older adults becomes relatively decreased, and the relations with spouse and family members are the major forms of social network [14]. Thus, we hypothesized that family relation is one of the major factors that influence the cognition of older adults. found that being widowed was associated with a lower incidence of cognitive impairment than being married. Although, the result also showed that participants who were married but lived separately were less likely to develop cognitive impairment than those who were widowed, the number of the group is relatively small (145 participants in "separated", 3277 in "widowed", 2457 in "married"), and this result should be interpreted with caution. We used multiple recoding measures to assess social supports provided by family members, friends and neighbors (i.e. composite scores of contact with spouse/children/siblings/relatives/friends/neighbors, and children's/siblings' frequent visits). Result showed that composite scores were not associated with cognitive impairment, but children's frequent visits (yes/no) consistently predict the incidence of cognitive impairment. This might be that the composite variables did not reflect the availability of social support from family members, but the closeness of the relationships. Whereas children's visits reflected the direct contact between parents and children and it might be more appropriate in current analysis.
Interestingly, that participants who gave money to their children were exposed to lower risk of cognitive impairment than those who did not; while participants who receive money from their children showed no difference from those did not. First, the result might indicate that material support from children was not a vital factor associated with cognitive decline. Consistent to a previous study, Ellwardt et al., (2013) found that instrumental support did not buffer cognitive decline [9]. Second, in our study, material support provided by these older parents might reflect the emotional contact with their children and might not be just about material support.
In general, the old parents in China lean on their children for financial support more or less. Therefore, the money or material support provided by parents is likely to be their care and love for their children rather than the simple financial support.
Numbers of studies confirmed the current result, with the finding that both the provision and acceptance of social support play an important role in cognitive function in older adult [29][30][31][32].
According to Berkman's theoretical model, social support refers to a person's perception of the availability of help of support from others in their social network [11], which do not emphasize distinguishing the specific resources of social support.
However, different resources of social support probably play different role in cognitive impairment. For example, a survey on Chinese older adults reported that emotional support from children is one of the most important factors affecting mental health [15]. Zhu, Hu, & Efird (2012) also found that, comparing to support from friends and important others, support from family was the most important indicator that predicted older adults' cognitive function [33].
However, previous studies in America reported opposite results. Brown et al., (2009) and Ficker et al., (2002) found that it was friends' support rather than family's support that had a bigger impact on cognitive function of older adults [34,35]. Zhu et al., (2012) though these contradictory results could be explained from the perspective of cultural differences [33].
In Chinese family culture, the social networks of older adults are more familycentered, which empathizes the contact between older parents and other family members. In addition, traditional Chinese culture advocates filial piety, which is the reflection of the blood tie between parents and children in families. The traditional filial morality has importance functions to promote the personal morals and Chinese children are expected to take good care of and respect parents when they are old.
For many old Chinese parents, children are their important spiritual pillar and the contact with children makes a great contribution to their happiness. The essence of filial piety is love, which implies gratefulness, respect, generosity, happiness, selflessness. Numbers of studies have found that filial piety were closely associated with subjective happiness, depression and life satisfaction [36][37][38]. The current result is consistent with the expectation, with the evidence that only children's visits but not sibling's visits predict the cognitive decline of older adults. Neither social service, social insurance nor material support from children could consistently predict older adults' cognitive decline, which confirmed the irreplaceable role of emotional support from children.
There are some limitations in the present study.

Conclusion
In the present study, we found that emotional support from children (children's visits) and marital status were consistently associated with lower incidence of cognitive impairment in current analyses, while material support and support outside family were unstable variables associated with cognitive impairment among older adults.   Figure 1 The flowchart of the study sample from 2005 to 2014. Notes: "lost" means the data was lost