A Longitudinal Analysis of the Association Between Living Arrangements and Health Among Older Adults in China

Abstract

This article used the nationally representative Chinese Longitudinal Healthy Longevity Survey to explore the associations between living arrangements and health among older adults. Living arrangements were stratified into six categories. Health was measured by self-rated health, activities of daily living (ADL) disability, and cognitive impairment. Random-effects ordered probit regressions were applied. The results indicated that coresidence had a positive effect on self-rated health compared with living alone. After introducing psychological well-being, the health differences observed in living with a spouse and living with both spouse and children were not significant. Participants with each of the living arrangement were more likely to have a higher rate of cognitive impairment and ADL disability than those living alone. Living arrangements were associated with older adults’ health. Psychological well-being was a key factor in this association, which may result from living with a spouse, and could contribute to the self-rated health of older adults.

Keywords

living arrangements health older adults random effects probit model

Geriatric health is of great importance during demographic revolutions and is very relevant in China in particular, which is facing rapid aging. The proportion of older people in China has grown substantially in the past decade. While the population aged 65 years and above accounted for more than 7% of the overall population in 2000, China has since become an aging society; in 2013, the corresponding percentage was 14.8 (X.-Q. Wang & Chen, 2014; Wu & Dang, 2013). Based on the United Nations’ medium fertility and mortality assumptions, there will be 329 million people aged 65 years and older by 2050, which will comprise nearly one third of China’s population (F. Chen & Liu, 2009). However, older adults’ health status is not favorable. Most older adults are at a high risk of experiencing health decline, whether functionally or cognitively, with a resulting substantial burden not only on elderly people themselves but also on their families and society. Therefore, efforts have increased to improve the health of older adults to promote health, delay illness, decrease burden, and extend longevity.

An increasing number of studies have demonstrated that the living arrangements of older adults are an important determinant of their health. A rapidly aging population and the declining health conditions inherent to aging lead to an urgent need for geriatric care and support. Traditionally, given the inadequate public pensions and social security systems, older Chinese adults have relied on support primarily from their families guided by the cultural traditions of filial piety and respect for one’s parents (Kincannon, He, & West, 2005). In recent decades, however, the number of children has declined rapidly partially due to the strict family planning policies in China that were established in the late 1970s, which have left older adults with fewer children to support them. In addition, the pressure from urbanization and modernization has weakened the traditional Chinese family structure, often leaving the elderly to live on their own or live with their spouse because of the increase in off-farm employment and mobility of Chinese residents (G. Liu, Dupre, Gu, Mair, & Chen, 2012). Consequently, the typical Chinese family has only one child (families in rural and minority areas may have two children) because of the one-child policy, and the average family size has declined to 3.35 according to the China Family Panel Studies (Xie & Hu, 2014). The concern for the elderly is echoed by empirical evidence that shows that older Chinese adults are increasingly living alone or only living with a spouse. Of note, the percentage of empty-nest older adults (whose children have left home as a result of transitioning from adolescence to adulthood) is rapidly increasing in China, to more than 50% (Cheng et al., 2015; L.-J. Liu, Fu, Qu, & Wang, 2014; Zhai et al., 2015).

These trends have cast some doubts on the effectiveness of the family to act as the caretaker of older adults in China. The potential loss of family support through coresidence (any situation other than living alone) could have huge effects on the health of older Chinese adults and place a profound and lasting pressure on the already overburdened health-care system in the coming decades with the start of the “getting old before getting rich” phenomenon in China (X.-Q. Wang & Chen, 2014). That is, China would be facing an aging population and an increasing old-age dependency ratio before becoming a developed country. Poor and low-income older adults still currently represent a large proportion of the population. Problems with cultivating livable environments are particularly outstanding in urban areas. Groups of older adults in rural areas who are left behind are becoming increasingly large. These emerging sociocultural and demographic trends make China an excellent site to examine the associations between the living arrangements and health of older adults. Through these investigations, we can obtain a more comprehensive understanding of potential scenarios and be able to respond more promptly in the future.

Relations Between Living Arrangements and Health

From the theoretical perspective, living arrangements, as other societal factors, can have a substantial impact on geriatric health. Based on the social–ecological framework, the living arrangement of an older adult is just one element among many in a group of factors originating within the kin group or family that affect older persons (Pollen, 2001). The household provides many forms of interaction and exchange for its members on a daily basis, such as routine assistance and personal care, and these interactions lead to social ties that intimately connect intrafamily support systems. Living with adults children may provide the best means of ensuring that the daily needs of older adults in developing countries are met (United Nations DESA, 2005). According to the convoy model, multigenerational families may help older adults successfully negotiate changes in life (Antonucci & Akiyama, 1987). An older adult’s spouse is his or her best source of mental support. Thus, household-based social ties can provide both economic capital (supply and consumption of economic resources) and emotional support (companionship, comfort, and intimacy), enhancing positive geriatric health. Moreover, the social support and social ties provided by coresidence may affect the health of older adults partly through health behaviors. Living with a spouse or with children seems to be associated with health-promoting lifestyles (Lee et al., 2005; Umberson, 1992). Therefore, living with a spouse and living with children—in fact, all forms of living arrangements—are thought to be associated with substantial consequences for older adults (United Nations DESA, 2005).

Moreover, the demand for coresidence may be heightened in societies with a dominant sense of filial obligation such as China (Samanta, Chen, & Vanneman, 2015). Living alone, if by one’s own choice or as a result of one’s child’s rural-to-urban migration for the family’s financial benefit, may not be that disadvantageous for older adults’ health. In rural China, rural-to-urban migration leads to large geographic distances between generations. The purpose of this migration is to provide financial benefit to the whole family, with older adults living alone or living with grandchildren in rural villages. In this case, an older adult living alone may not necessarily be disadvantageous. Otherwise, this living arrangement is assumed to have disadvantages for older adults’ health compared to coresidence.

There is increasing evidence, mostly in developed nations, that living arrangements can affect health in old age, such as the occurrence of poor self-rated health, disability in activities of daily living (ADL), cognitive impairment, mortality, and short-term morbidity (Hughes & Waite, 2002; L. W. Li, Zhang, & Liang, 2009; Lund et al., 2002; Martikainen, Nihtilä, & Moustgaard, 2008; Michael, Berkman, Colditz, & Kawachi, 2001; Russell & Taylor, 2009; Samanta et al., 2015). Although the mechanisms by which coresidential living contributes to better health are not completely understood, research suggests that three factors may play important roles: health behaviors, family income, and psychological well-being.

Coresidence can lead to health-enhancing behaviors (e.g., increased involvement in exercise) or change health-compromising behaviors (e.g., smoking and alcohol consumption), which in turn promote health (Lewis & Butterfield, 2005; Lewis & Rook, 1999). Furthermore, positive social control of health behaviors among married couples is associated with the practice of desired health behaviors (Tucker & Anders, 2001). Therefore, health behaviors may be a significant predictor of health.

A substantial amount of literature suggests that living with family members results in an increase in economic well-being that promotes better health outcomes (Ross, Mirowsky, & Goldsteen, 1990; L. Waite & Gallagher, 2002) because it improves the family’s conditions, partially by allowing members to specialize in a particular division of labor (Becker, 1981). In addition, the sharing of financial and social resources leads to economies of scale because many of the costs of maintaining a family are roughly fixed (e.g., housing, heating, and transportation). Moreover, the pooling of wealth also provides protection against the risk of unexpected out-of-pocket medical spending (Oppenheimer, 2000; L. J. Waite, 1995). However, this is not always the case. It is possible for older adults to live with family members because of financial constraints but not by self-choice. In this case, living with other family members may not be beneficial for the health of older adults.

Psychological well-being is an important factor that affects the health outcomes of older adults, especially those living alone (L. Waite & Gallagher, 2002). Living alone is associated with experiencing different emotions, including anger, loneliness, and depression (Kim & Fredriksen-Goldsen, 2016; Russell & Taylor, 2009). Coresidence is associated with imparting psychological resources such as social ties and social support. Substantial evidence has shown that social ties and social support are both positively and causally related to psychological well-being. That is, social ties may directly affect health, and social support may buffer the harmful physical and mental health effects of exposure to stress (Cohen & Janicki-Deverts, 2009; Ertel, Glymour, & Berkman, 2009; Taylor, Friedman, & Silver, 2007; Uchino, 2004; Umberson & Montez, 2010).

Although a sizable body of literature has shown that living arrangements are significantly related to health among older adults in Western countries, the research in China shows inconsistent findings (F. N. Chen & Short, 2008; Sun, Lucas, Meng, & Zhang, 2011; Zimmer, 2005). Additionally, most studies are limited to cross-sectional data, which prevent the identification of long-term effects. Studies of older Chinese adults have found that older people who live with their children are more likely to be disabled than those who live alone (OR = 1.33, 95% CI = [1.07, 1.66]; L. W. Li et al., 2009). In a cross-sectional study of older adults in China, researchers found that those living with family members had a more than 3 times higher risk of disability than those living alone (W. Chen et al., 2015). Based on the 2005 implementation of the Chinese Longitudinal Healthy Longevity Survey (CLHLS), Sereny (2011) found that coresidence was associated with poor self-rated health. Another study, also using the CLHLS, showed a health advantage of living with a spouse (L. W. Li et al., 2009). Longitudinal data were used in a few studies, but the results of these studies had limitations, as they did not address unmeasured individual heterogeneity that might have led to bias and inconsistent results (Lockwood & McCaffrey, 2007). For example, Gu, Dupre, and Liu (2007) used data from the first three waves of the CLHLS and found that the mortality risk of the institutionalized oldest-old residents was 1.35 times greater than that of the oldest residents living in the community. However, after controlling for factors such as sociodemographics, family caregiving, and health characteristics, the mortality difference was eliminated. There are several potential reasons for the inconsistent results, such as the different approaches to operationalizing living arrangements, the statistical models, possible mechanisms in the association of living arrangements and health, and the study samples (L. W. Li et al., 2009). Hence, it remains unclear whether older people who live alone have health disadvantages or whether coresidential living arrangements are good for health. The complexity of this relationship indicates the urgent need for research in this field. Moreover, regulation of health behaviors, consumption of economic resources, and psychological well-being have been mentioned as factors explaining why living arrangements may be relevant to health (L. W. Li et al., 2009). However, few studies are dedicated to exploring these relationships.

The aims of this study were thus to address the following two questions: (a) do living arrangements affect the health outcomes of older adults aged 65 years and over in China? and (b) are the effects of living arrangements on older adults’ health explained by health behaviors, family income, or psychological well-being? Based on the literature and previous research, we hypothesize that coresidential living would be significantly associated with advantages in health among older adults in China and that this effect can be at least partially explained by health behaviors, family income, and psychological well-being. Theoretically, of the different living arrangements investigated, we expected older Chinese adults living with both their spouse and children to have the best health outcomes because of the availability of material as well as mental support. This arrangement would be followed in order by those living with a spouse (without children, may have others), with children (without spouse, may have others), and with others (no spouse or children). Older adults living alone or in an institution were considered to fare the worst in health. A relatively new and nationally representative sample of older Chinese adults was used in this study, which allowed us to identify the latest trends and reduce estimation biases. Furthermore, we aimed to estimate the associations using random-effects ordered probit models, which can improve the quality of estimations and provide reasonable conclusions after controlling for individual heterogeneity (Vaillant & Wolff, 2012). Thus, we were able to explore the relationships between living arrangements and health in older adults.

Method

Data

We used longitudinal data from the CLHLS, which is the first nationwide longitudinal survey targeting the oldest-old aged 80 years or older in a developing country. The participants were randomly sampled from half of the cities or counties in 23 provinces of China. The total population of these provinces covered 85% of the total Chinese population in 2008 (Qiu, Sautter, Liu, & Gu, 2011). The CLHLS has conducted face-to-face interviews for six waves to date, using internationally compatible questionnaires. The survey was approved and reviewed by the Duke University Health System’s institutional review board. Informed consent was obtained from the participants. The CLHLS that began implementation in 1998 was primarily aimed at the oldest-old, whereas from 2002 (the third wave) on, the CLHLS included young elders aged 65–79 years. We therefore used data from the third to the sixth waves, which were conducted in 2002, 2005, 2008/2009, and 2011/2012, respectively. The analytic sample was restricted to initial observations to minimize the potential for selection bias and confounding effects from survivors (G. Liu et al., 2012). We excluded 59 respondents whose living arrangements were missing and 44 respondents who were less than 65 years old. The final sample included 15,961 participants from the 2002 wave, of whom 8,064 (50.52% of the population in the 2002 wave) survived and were reinterviewed in 2005. Subsequently, 4,185 (26.22% of the population in the 2002 wave) participants survived and were reinterviewed in 2008. In the 2011/2012 wave, 2,485 (15.57% of the population in the 2002 wave) of the participants had survived. The mortality rate and proportion of missing data due to withdrawal were, respectively, 36.74% and 12.58% for the 2002 wave, 30.72% and 17.87% for the 2005 wave, and 47.05% and 19.68% for the 2008 wave.

Measures

Health outcomes

The health outcomes were measured by self-rated health, ADL disability, and cognitive impairment. Self-rated health represented one’s subjective health condition based on physical, mental, and social perspectives. It was assessed by a single question: “how do you think of your state of health?” Five options (very poor, poor, so-so, good, and excellent) were provided, which were scored from 1 to 5, respectively. ADL disability consists of trouble performing the repeated, most basic, and most common daily activities that allow people to survive in and adapt to their environment. This construct was measured by 6 items (dressing, bathing, using the toilet, controlling bladder and bowel movement, indoor transferring, and eating; Katz, Ford, Moskowitz, Jackson, & Jaffe, 1963). Items were scored 0 if the respondent was able to perform it independently; otherwise, they were scored 1. The ADL disability score thus ranged from 0 to 6, with higher scores implying worse functioning of ADL. The Cronbach’s α of the ADL Disability Scale indicated that the scale reliability was .84 for the 2002 wave, .89 for the 2005 wave, .91 for the 2008 wave, and .88 for the 2011/2012 wave. Cognitive impairment was defined as scoring less than 18 (Zeng et al., 2015; Zhang, 2006) on the Chinese version of the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975; Yi & Vaupel, 2002). The MMSE captures one’s orientation, registration, attention, calculation, recall, and language abilities, with a possible integrated total score of 0–30. Higher scores indicate a better cognitive function. The Cronbach’s α of the MMSE ranged from .97 for the 2002 wave (the highest) and .89 for the 2011/2012 wave (the lowest).

Living arrangements

As the social and psychological meaning of coresidence varies according to the demographic characteristics and cultural context of the study setting, we applied the widely used six-category definition of the variable: living alone, living with their spouse (no children, may have others), living with their children (no spouse, may have others), living with both their spouse and their children, living in an institution, and other (L. W. Li et al., 2009). The children referred to in the categories consisted of biological children, children-in-law, biological grandchildren, and grandchildren-in-law. The “other” living arrangement include situations in which participants lived with their siblings, parents, or nurses without children or spouses. Dummy variables were used to denote living arrangements and living alone was used as the reference category.

Family income

Family income was a continuous variable measured by annual household income as reported by the respondents. In addition, family income was log transformed due to the skewness of the distribution.

Health behaviors

The health behaviors assessed included drinking, smoking, and exercise. Exercise implied purposeful fitness activities, such as walking, running, playing ball, and practicing chi kung. These activities were included in the descriptions provided in the present situation and were dichotomized as yes or no.

Psychological well-being

In this study, psychological well-being in the CLHLS was considered the mood-related and personality-related concepts of affect, such as loneliness, anxiety, and happiness. Thus, the psychological well-being variable we used was constructed by 4 positive items and 3 negative items (T. Li & Zhang, 2015; Smith, Gerstorf, & Li, 2008). Positive items included optimism (“I always look on the bright side of things”), conscientiousness (“I like to keep my belongings neat and clean”), sense of personal control (“I can make my own decisions concerning my personal affairs”), and positive feelings about aging (“I am just as happy now as when I was younger”). Negative items assessed 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”). The responses were recorded on a 5-point scale (0 = never, 4 = always). The responses to negative items were reverse coded, resulting in a higher aggregated score indicating better well-being. The sum of the scores ranged from 0 to 28. The Cronbach’s α of the psychological well-being items was .91 in the 2008 wave (the highest) and .84 in the 2002 wave (the lowest).

Covariates

The covariates included chronological age, gender (male = 0 vs. female = 1), ethnicity (Han = 0 vs. non-Han = 1), education (illiteracy, primary school, middle school, high school, technical secondary school, or above), previous occupational status (agricultural worker = 0 vs. non-agricultural worker = 1), and residence (city/town = 0 vs. rural = 1). Additionally, we controlled for the number of surviving children and the number of diseases reported by the participants themselves. These diseases included hypertension, diabetes, heart disease, stroke, bronchitis or pneumonia, tuberculosis, cataract, glaucoma, prostate tumor, gastric or duodenal ulcer, Parkinson’s disease, bedsores, arthritis, dementia, epilepsy, cholecystitis, blood disease, chronic nephritis, uterine tumor, galactophore disease, hepatitis, and cancer, which are common ailments among older Chinese adults.

Information on the variables mentioned above was obtained in all four waves. Older adults were encouraged to answer the questionnaire by themselves to the greatest extent possible. Family members, neighbors, or institution workers could serve as proxies if the respondents were unable to answer because of cognitive impairment or other disease. The percentage of proxy responses was 14.16. Notably, the MMSE and psychological well-being sections were answered only by the respondents themselves (without proxies).

Statistical Analyses

Descriptive statistics were used for the sociodemographic characteristics, health behavior variables, and health variables by calculating the proportion of the distribution corresponding to each particular living arrangement. χ² tests and one-way analyses of variance were used to identify group differences in proportions and mean tests.

Random-effects ordered probit regression was used to estimate the relationship between living arrangements and self-rated health and ADL disability. Typically, self-rated health and ADL disability were measured on an ordinal scale. Cognitive impairment was measured as a dichotomous variable and was modeled by random-effects probit regression. A random-effects panel model was used because the longitudinal data contained multiple observations for a single person across waves. Introducing a random intercept for each person controlled for the intraperson variance or unobserved heterogeneity. Moreover, the panel model increased the sample size for a better estimation of the effect of living arrangements. To ensure that there were no major biases due to mortality selection, we examined the associations between living arrangements and current mortality status for the 2005, 2008, and 2011/2012 waves. The results (available upon request) showed no significant associations. The data were analyzed using Stata Version 13.0 (StataCorp). The significance level of all analyses was α = .05. Missing data were addressed by utilizing multiple imputations in the NORM program (Schafer, 1999). The percentage missing was moderate (range: 0.2–10.5%). Compared with other ways of handling missing data, multiple imputation is better for producing consistent, efficient, and normal estimates under the assumption that the data are missing at random (Little & Rubin, 2002).

We performed data analyses using two random-effects probit models in which self-rated health, ADL disability, and cognitive impairment were included as the dependent variables. In Model 1, we examined the general relationship between living arrangements and health, controlling for sociodemographic covariates (gender, age, ethnicity, residence, and number of children), two socioeconomic status covariates (education and occupation), and health status covariates (number of diseases, cognitive impairment, and ADL disability or self-rated health). Specifically, cognitive impairment and ADL disability were controlled for in the self-rated health model. Cognitive impairment but not self-rated health was controlled for in the ADL disability model. A cognitive impairment model was presented without controls for self-rated health and ADL disability. We then added health behaviors (drinking, smoking, and exercise), family income, and psychological well-being as additional covariates to discern whether these factors explained health differences by the living arrangements in Model 2. A reduction in the significance level or magnitude of the effect of living arrangements throughout the models suggested that the potential mechanism variables explained all or part of the effect of living arrangements on the health of older adults.

Results

Table 1 shows the descriptive statistics of the variables at baseline (the 2002 wave) according to living arrangements. Over half of the participants were living with their children. Approximately 16% of the older adults were living with their spouse. Those living alone accounted for approximately 14% of the sample. Approximately 12% of the participants lived with both their spouse and children. Only a few participants were living in an institution and with other individuals. There were more older women than older men. The average age of the entire sample was 86.43 years. A majority of the older adults came from rural areas, especially those living by themselves. Of note, more than 70% of those living in an institution were from a city/town. Most of the older adults (60.23%) were illiterate. The rate of cognitive impairment was lowest (6.61%) among those living with a spouse and highest (29.71%) among those living with others. Those living with both spouse and children had the best psychological well-being (19.94), while those living alone had the worst (17.69), with a total average of 18.52. When asked to rate their health, the participants tended to report having so-so or good health. Most of the older adults had no ADL disability.

Table 1.
Characteristics of Older Adults by Living Arrangements in China: 2002 Wave of Chinese Longitudinal Healthy Longevity Survey.

Characteristics Alone (2,159/13.53%) With Spouse (2,619/16.41%) With Children (8,346/52.29%) With Spouse and Children (1,923/12.04%) Institution (739/4.63%) Others (175/1.10%) Total (15,961) p Value^a

Age (year) 86.43 77.11 91.14 77.50 88.42 91.37 86.43 <.001

Gender <.001

Male 37.66 66.32 29.70 69.84 45.33 42.29 42.48

Female 62.34 33.68 70.30 30.16 54.67 57.71 57.52

Ethnicity <.001

Han 95.79 97.25 93.04 94.18 96.89 95.43 94.44

Non-Han 4.21 2.75 6.96 5.82 3.11 4.57 5.56

Residence <.001

City/town 41.69 47.35 43.42 49.51 70.09 64.57 46.03

Rural 58.31 52.65 56.58 50.49 29.91 35.43 53.97

Education <.001

Illiterate 66.47 42.54 72.39 40.40 60.62 52.57 60.23

Primary school 25.98 40.40 22.07 41.76 27.60 23.43 28.25

Middle school 3.94 8.13 2.71 7.85 4.33 4.57 4.48

High school 2.50 5.08 1.82 6.03 4.47 9.71 3.17

Technical secondary school or above 1.11 3.86 1.01 4.32 2.98 9.71 2.07

Occupation <.001

Agriculture 38.12 47.61 33.48 47.17 56.70 57.71 39.41

Non-agriculture 61.88 52.39 66.52 52.83 43.30 42.29 60.59

Number of children 5.05 4.86 5.12 4.80 3.20 4.18 4.93 <.001

Number of diseases 1.17 1.15 1.11 1.92 1.37 1.58 1.15 <.001

Health behaviors

Drinking (%) 18.53 27.34 17.82 24.96 17.32 22.86 20.37 <.001

Smoking (%) 17.51 27.72 13.46 28.45 15.56 20.57 18.33 <.001

Exercise (%) 27.74 41.31 25.82 42.64 41.41 37.14 31.50 <.001

Family income^b 0.32 0.38 0.35 0.41 0.53 0.53 0.36 <.001

Psychological well-being 17.69 19.82 17.99 19.94 18.70 18.65 18.52 <.001

Self-rated health <.001

1 2.36 1.26 2.34 1.40 4.19 5.14 2.17

2 19.59 15.73 17.06 12.22 18.27 21.14 16.70

3 37.75 33.14 35.13 38.64 32.75 30.86 35.42

4 31.50 37.04 36.28 36.20 36.81 35.43 35.77

5 8.80 12.83 9.19 11.54 7.98 7.43 9.94

Activities of daily living disability <.001

0 79.16 87.17 58.17 84.97 65.49 53.14 69.28

1 10.61 7.14 16.76 8.11 13.80 14.86 13.15

2 3.01 1.34 6.60 1.30 4.47 8.57 4.54

3 1.76 1.34 4.71 1.66 3.38 4.00 3.32

4 1.99 0.92 4.72 1.40 3.52 5.14 3.28

5 2.13 1.34 4.76 1.77 4.74 6.29 3.50

6 1.34 0.76 4.28 0.78 4.60 8.00 2.94

Cognitive impairment (%) 21.54 6.61 33.54 7.38 29.36 29.71 24.10 <.001

Note. Percentage for categorical variable and mean for continuous variable.

^aχ² tests or one-way analyses of variance.

^bUnit: ×10⁴ renminbi.

Tables 2 –4 present the results of the random-effects ordered probit model for self-rated health and ADL disability and the random-effects probit model for cognitive impairment. The figures in those tables represent the coefficients from the random-effects probit models of the different health statuses of sets of independent variables. All of the likelihood ratio tests of the models were statistically significant, reflecting the fact that the random-effects probit model could improve the quality of the estimates and provide reasonable conclusions after controlling for individual heterogeneity.

Table 2.
Coefficients From Random-Effects Ordered Probit Regression Analyses of Self-Rated Health.

Self-Rated Health

Variables Model 1 Model 2

Living arrangements

With spouse .157 (.025)* .021 (.025)

With children .206 (.021)* .137 (.021)*

With spouse and children .167 (.028)* .009 (.027)

Institution .157 (.040)* .063 (.041)

Others .127 (.074) .040 (.074)

Age (year) .007 (.001)* .008 (.001)***

Gender −.018 (.017) .034 (.017)*

Ethnicity −.150 (.031)*** −.067 (.029)*

Residence −.097 (.016)* .019 (.015)

Education .054 (.010)* −.014 (.009)

Occupation −.121 (.017)* −.048 (.017)

Number of children .003 (.003) −.004 (.003)

Number of diseases −.189 (.005)* −.175 (.005)*

Cognitive impairment −.388 (.019)* −.203 (.019)*

Activities of daily living disability −.191 (.005)* −.141 (.005)*

Health behaviors

Drinking .139 (.018)*

Smoking .027 (.019)

Exercise .171 (.015)*

Family income^a .034 (.005)*

Psychological well-being .105 (.002)*

Variance of random effect .196 (.012)* .121 (.011)*

Likelihood ratio (LR) test 434.52* 179.17*

Note. Self-rated health ranged from 1 to 5. Ref = living alone for living arrangements; Ref = male for gender; Ref = Han for ethnicity; Ref = city/town for residence; Ref = non-agriculture for occupation. The estimates of the threshold parameters are significant at the 1% level in all estimations and are not shown here.

^aUnit: ×10⁴ renminbi.

p < .05. p < .01. p < .001.

Table 3.
Coefficients From Random-Effects Ordered Probit Regression Analyses of Activities of Daily Living Disability.

Activities of Daily Living Disability

Variables Model 1 Model 2

Living arrangements

With spouse 0.270 (0.044)* 0.333 (0.045)

With children 0.480 (0.033)* 0.523 (0.034)*

With spouse and children 0.401 (0.047)* 0.483 (0.048)*

Institution 0.483 (0.058)* 0.554 (0.062)*

Others 0.687 (0.096)* 0.763 (0.099)*

Age (year) 0.055 (0.001)* 0.053 (0.001)*

Gender 0.119 (0.025)* 0.015 (0.027)

Ethnicity −0.326 (0.050)* −0.375 (0.050)*

Residence −0.148 (0.023)* −0.224 (0.024)*

Education 0.016 (0.015) 0.059 (0.015)*

Occupation −0.187 (0.025)* −0.234 (0.026)*

Number of children 0.057 (0.005)* 0.066 (0.005)*

Number of diseases 0.204 (0.007)* 0.184 (0.007)*

Cognitive impairment 1.020 (0.024)* 0.837 (0.025)*

Health behaviors

Drinking −0.247 (0.030)*

Smoking −0.198 (0.033)*

Exercise −0.535 (0.026)*

Family income^a 0.009 (0.008)

Psychological well-being −0.053 (0.003)*

Variance of random effect 0.511 (0.033)* 0.496 (0.034)*

LR test 528.56* 474.32*

Note. The activities of daily living disability score ranged from 0 to 6; Ref = the reference category; Ref = living alone for living arrangements; Ref = male for gender; Ref = Han for ethnicity; Ref = city/town for residence; Ref = non-agriculture for occupation. The estimates of the threshold parameters are significant at the 1% level in all estimations and are not shown here.

^aUnit: ×10⁴ renminbi.

p < .05. p < .01. p < .001.

Table 4.
Coefficients From Random-Effects Probit Regression Analyses of Cognitive Impairment.

Cognitive Impairment

Variables Model 1 Model 2

Living arrangements

With spouse −.093 (.048) .023 (.050)

With children .146 (.034)* .213 (.036)*

With spouse and children −.028 (.053) .109 (.055)*

Institution .302 (.063)* .396 (.068)*

Others .103 (.114) .194 (.120)

Age (year) .080 (.002)* .075 (.002)*

Gender .156 (.029)*** .067 (.031)*

Ethnicity −.140 (.052) −.217 (.053)*

Residence .116 (.026)* .001 (.027)

Education −.227 (.019) −.170 (.019)*

Occupation .082 (.029) .013 (.029)

Number of children −.013 (.006)* −.008 (.006)

Number of diseases .114 (.008)* .073 (.009)*

Health behaviors

Drinking −.129 (.034)*

Smoking −.044 (.037)

Exercise −.459 (.030)*

Family income^a −.022 (.008)

Psychological well-being −.100 (.003)*

Variance of random effect .557 (.044)* .482 (.042)*

LR test 405.80* 299.27*

Note. The response set for cognitive impairment: 1= scoring less than 18, 0 = scoring higher than 18; Ref = living alone for living arrangements; Ref = male for gender; Ref = Han for ethnicity; Ref = city/town for residence; Ref = non-agriculture for occupation. The estimates of the threshold parameters are significant at the 1% level in all estimations and are not shown here.

^aUnit: ×10⁴ renminbi.

p < .05. p < .01. p < .001.

Table 2 presents the random-effects ordered probit regression results of self-rated health. Model 1 showed that compared with living alone, all of the living arrangements were significantly associated with better self-rated health, with the exception of living with others. When health behaviors, family income, and psychological well-being were controlled for (Model 2), only living with children was significantly associated with self-rated health (β = .137, p < .001). Drinking, exercise, family income, and higher psychological well-being were associated with better self-rated health (β = .139, .171, .034, and .105, respectively, p < .001). In addition, the following variables were associated with higher self-rated health: being older, being female, having a non-agricultural previous occupation, having fewer chronic diseases, having no cognitive impairment, and performing better in ADL disability.

Table 3 shows the probit regression results for ADL disability. In Model 1, compared to living alone, all living arrangement categories were associated with higher ADL disability (β > 0, p < .001), and the significant associations remained stable after adjusting for health behaviors, family income, and psychological well-being. Drinking, smoking, exercise, and higher psychological well-being were all negatively related to ADL disability (β = −.247, −.198, −.535, and −.053, respectively, p < .001). Furthermore, in Model 2, the likelihood of ADL disability was higher for those who were older, were Han Chinese, lived in a city/town, had more children, had a higher educational level, had a non-agricultural occupation, had more chronic diseases, and had cognitive impairment.

The probit regression results for cognitive impairment are displayed in Table 4. Model 1 showed that living with children and living in an institution were significantly associated with cognitive impairment when compared with living alone. Furthermore, these associations were still significant after controlling for health behaviors, family income, and psychological well-being (β = .213, p < .001, for living with children; β = .109, p < .05, for living with both spouse and children; and β = .396, p < .001, for living in an institution). Drinking, exercise, less family income, and higher psychological well-being were associated with a lower likelihood of cognitive impairment. Other factors associated with a higher likelihood of cognitive impairment included being female, being older, being Han, having a lower educational level, and having more chronic diseases.

Discussion

A growing body of literature based on data from developed countries indicates that living arrangements are associated with the health of older adults. This study, using four waves of the CLHLS under the framework of a random-effects probit model, extended this research to examine the living arrangements and resulting health consequences of older adults in China, a country that has been undergoing significant social and economic changes in the past several decades. Additionally, random-effects probit models were used to examine the plausible mechanisms of health behaviors, family income, and psychological well-being.

The hypothesis that coresidential living is associated with health advantages was partially supported in this study, and these associations varied by health outcomes. Coresidential living was beneficial for self-rated health, whereas it was not beneficial for ADL disability or cognitive impairment. Only approximately 14% of our participants lived alone. We found that compared with living alone, coresidential living was associated with better self-rated health, which is consistent with previous studies in Western nations (Kharicha et al., 2007; L. J. Waite & Hughes, 1999) and in China (L. W. Li et al., 2009; H. Wang, Chen, Pan, Jing, & Liu, 2013). However, people who reported coresidential living had a higher risk of ADL disability and cognitive impairment, which is also consistent with the other two studies conducted in China (L. W. Li et al., 2009; H. Wang et al., 2013). It is possible that the advantages of living alone regarding ADL disability and cognitive function reflect a self-selection process in which older adults with cognitive impairment and ADL disability are more likely to live with family members or live in an institution, especially as the observation progressed through almost 10 years. As suggested by Brown et al. (2002), poor health does trigger changes in living arrangements and both physical and mental health conditions play a role in these transitions. In addition, older adults living independently may have to address daily life activities themselves, and having a good cognitive functioning could support them in achieving their daily activities. As in a previous cohort study, cognitive impairment was more prevalent in those living with others than in those living alone (Weiler, Lubben, & Chi, 1991). In general, the similarities found in this study when compared with existing studies suggest that the associations of living alone with different health outcomes among older Chinese adults are on a par with their Western counterparts (Kharicha et al., 2007; L. W. Li et al., 2009; Lund et al., 2002).

The results of our longitudinal analyses provide evidence on the previously unexplored plausible mechanisms of family income, health behaviors, and psychological well-being in the association between living arrangements and health of older adults. Previous studies suggest that coresidential living is linked to psychological benefits, which in turn improve health (Cohen & Janicki-Deverts, 2009; Ertel et al., 2009; Taylor et al., 2007; Uchino, 2004; Umberson & Montez, 2010). In further models (data not shown) controlling for health behaviors, family income, and psychological well-being, we found that the significant coefficients of living with a spouse and living with both spouse and children became nonsignificant only when psychological well-being was added. Thus, our findings support the plausible mechanism of psychological well-being affecting the association between living with a spouse or living with both spouse and children and self-rated health. In our study, older adults living alone had lower psychological well-being scores than those living with their spouse or with both spouses and children. This indicates that living with a spouse improves older adults’ well-being psychologically, which has a further salutary effect on one’s self-rated health. However, this was not the case with ADL disability or cognitive impairment, possibly because they represent the physical and objective dimensions of health and might not be affected by psychological well-being. Notwithstanding, such findings support the claim that having a spouse is somewhat the “greatest guarantee of support in old age” (Chappell, 1991, p. 1). Thus, societal changes should be made to support older adults who do not have a spouse to enhance their psychological well-being, possibly further improving their health.

Health behaviors, family income, and psychological well-being were all associated with older adults’ health, partially explaining the associations between living arrangements and health. It is well acknowledged that exercise benefits physical and mental health, as confirmed in this study. While we widely accepted the fact that self-rated health would decline with age, our results revealed the opposite. The results were consistent with a 6-year follow-up study, in which the oldest people improved their self-rating of health as the years progressed (Idler, 1993). The potential for self-selection should be noted here—those who rated their health as good may have been more likely to live longer. Moreover, we also conducted models for self-rated health without number of diseases, ADL disability, and cognitive impairment. The results (available upon request) showed that age was negatively associated with self-rated health in Model 1 and was nonsignificantly associated with self-rated health in Model 2. The number of diseases, ADL disability, and cognitive impairment are the primary reasons that self-rated health declines with age. Controlling for these variables often reverses the direction of the relationship for age. Thus, the opposite direction of the association with age may partially have been due to the model specification rather than a real relationship. However, the counterintuitive associations between drinking, smoking, and family income and health require further research. It is possible that it is not health behaviors or family income that affect health, but rather social control of health behaviors and social resources. Further research is needed to clarify this relationship.

To more thoroughly explore the health effects of different living arrangements, we used three indicators of health as the outcomes—self-rated health, ADL disability, and cognitive impairment. These three indicators represented subjective and objective aspects of health and have frequently been used as health outcomes in research. Self-rated health was found to be a useful index that was considered subjective for older adults because it covered three important aspects of human health: physical, mental, and social (Du Toit, Pritchard, Heffernan, Simpson, & Fonn, 2002). ADL disability is an objective measurement of one’s functional status, particularly in regard to older adults and people with disabilities. Cognitive impairment represented another objective aspect of health–cognitive function. Because health is multidimensional, exploring the relationship between living arrangements and health should involve at least the subjective and objective dimensions of health.

Some limitations of this study should be mentioned. First and foremost, although we tried to reduce some of the selectivity effects of living arrangements on health by adding random effects to the ordered probit model, we were not fully able to avoid the self-selection issues regarding living arrangements for different health conditions. A focus on short-term health outcomes may have reduced the problem of endogeneity because short-term outcomes are unlikely to change living arrangements. However, a sensitivity analysis was performed in this study. The results showed no significant associations between living arrangements and current mortality, which confirmed the hypothesis. Additionally, we caution against inferences regarding causality although longitudinal data were used. It may be better for the results in the study to be described as supporting an association rather than causing it. To fully assess how causal processes contribute to health differences through living arrangements, future studies should adopt instrumental variable models that contain adequate measures of instrumental variables or other causality reasoning methods. In addition, although we investigated the associations between the living arrangements and health of older Chinese adults by controlling for a dozen covariates, the data limitations restricted us from including some important qualitative dimensions, such as family relationships, that could offer explanatory insight.

Despite these limitations, our study contributes to the literature by applying a random-effects probit model to a nationally representative sample from the latest four waves of the CLHLS in China. The results revealed that living arrangements are associated with the health of older adults, including their self-rated health, ADL disability, and cognitive impairment. Moreover, the mechanisms between living arrangements and health were explored to expand upon the existing literature, including health behavior, family income, and psychological well-being variables. Psychological well-being is a key factor to consider in the relationship between living arrangements and self-rated health, supporting the claim that a spouse is the most definitive guarantee of support for an older person.

As China has become the country with the largest aging population and is currently in a stage of rapid aging, the observations of living arrangements among Chinese people of old age have important implications. Family support, especially support from a spouse, is needed to maintain and improve the physical and mental health of older adults. Attention should also be paid to older adults who are living alone but who sustain a relatively good health condition, as various types of family support are needed when physical functioning begins to fail (Zimmer & Kwong, 2003). This issue warrants further attention from both researchers and policy makers, given the rapid population aging and the changing norm of living arrangements in the older Chinese population. However, the advantages of solitary living for cognitive impairment and ADL disability led us to wonder whether coresidence itself is an indicator of well-being (Sereny, 2011). If older adults willing choose to live alone and they can receive sufficient support from their society and senior institutions, solitary living or coresidence should not substantially differ. Future studies on the living arrangements of older adults should account for arrangement preferences. In addition, because gender and region (i.e., east vs. west, urban vs. rural) are important factors in the relationship between living arrangements and health (Henning-Smith, 2016), these variables may present future research directions that could provide a better understanding of this relationship.

Characteristics	Alone (2,159/13.53%)	With Spouse (2,619/16.41%)	With Children (8,346/52.29%)	With Spouse and Children (1,923/12.04%)	Institution (739/4.63%)	Others (175/1.10%)	Total (15,961)	p Value^a
Age (year)	86.43	77.11	91.14	77.50	88.42	91.37	86.43	<.001
Gender								<.001
Male	37.66	66.32	29.70	69.84	45.33	42.29	42.48
Female	62.34	33.68	70.30	30.16	54.67	57.71	57.52
Ethnicity								<.001
Han	95.79	97.25	93.04	94.18	96.89	95.43	94.44
Non-Han	4.21	2.75	6.96	5.82	3.11	4.57	5.56
Residence								<.001
City/town	41.69	47.35	43.42	49.51	70.09	64.57	46.03
Rural	58.31	52.65	56.58	50.49	29.91	35.43	53.97
Education								<.001
Illiterate	66.47	42.54	72.39	40.40	60.62	52.57	60.23
Primary school	25.98	40.40	22.07	41.76	27.60	23.43	28.25
Middle school	3.94	8.13	2.71	7.85	4.33	4.57	4.48
High school	2.50	5.08	1.82	6.03	4.47	9.71	3.17
Technical secondary school or above	1.11	3.86	1.01	4.32	2.98	9.71	2.07
Occupation								<.001
Agriculture	38.12	47.61	33.48	47.17	56.70	57.71	39.41
Non-agriculture	61.88	52.39	66.52	52.83	43.30	42.29	60.59
Number of children	5.05	4.86	5.12	4.80	3.20	4.18	4.93	<.001
Number of diseases	1.17	1.15	1.11	1.92	1.37	1.58	1.15	<.001
Health behaviors
Drinking (%)	18.53	27.34	17.82	24.96	17.32	22.86	20.37	<.001
Smoking (%)	17.51	27.72	13.46	28.45	15.56	20.57	18.33	<.001
Exercise (%)	27.74	41.31	25.82	42.64	41.41	37.14	31.50	<.001
Family income^b	0.32	0.38	0.35	0.41	0.53	0.53	0.36	<.001
Psychological well-being	17.69	19.82	17.99	19.94	18.70	18.65	18.52	<.001
Self-rated health								<.001
1	2.36	1.26	2.34	1.40	4.19	5.14	2.17
2	19.59	15.73	17.06	12.22	18.27	21.14	16.70
3	37.75	33.14	35.13	38.64	32.75	30.86	35.42
4	31.50	37.04	36.28	36.20	36.81	35.43	35.77
5	8.80	12.83	9.19	11.54	7.98	7.43	9.94
Activities of daily living disability								<.001
0	79.16	87.17	58.17	84.97	65.49	53.14	69.28
1	10.61	7.14	16.76	8.11	13.80	14.86	13.15
2	3.01	1.34	6.60	1.30	4.47	8.57	4.54
3	1.76	1.34	4.71	1.66	3.38	4.00	3.32
4	1.99	0.92	4.72	1.40	3.52	5.14	3.28
5	2.13	1.34	4.76	1.77	4.74	6.29	3.50
6	1.34	0.76	4.28	0.78	4.60	8.00	2.94
Cognitive impairment (%)	21.54	6.61	33.54	7.38	29.36	29.71	24.10	<.001

	Self-Rated Health
Living arrangements
With spouse	.157 (.025)***	.021 (.025)
With children	.206 (.021)***	.137 (.021)***
With spouse and children	.167 (.028)***	.009 (.027)
Institution	.157 (.040)***	.063 (.041)
Others	.127 (.074)	.040 (.074)
Age (year)	.007 (.001)***	.008 (.001)***
Gender	−.018 (.017)	.034 (.017)*
Ethnicity	−.150 (.031)***	−.067 (.029)*
Residence	−.097 (.016)***	.019 (.015)
Education	.054 (.010)***	−.014 (.009)
Occupation	−.121 (.017)***	−.048 (.017)**
Number of children	.003 (.003)	−.004 (.003)
Number of diseases	−.189 (.005)***	−.175 (.005)***
Cognitive impairment	−.388 (.019)***	−.203 (.019)***
Activities of daily living disability	−.191 (.005)***	−.141 (.005)***
Health behaviors
Drinking		.139 (.018)***
Smoking		.027 (.019)
Exercise		.171 (.015)***
Family income^a		.034 (.005)***
Psychological well-being		.105 (.002)***
Variance of random effect	.196 (.012)***	.121 (.011)***
Likelihood ratio (LR) test	434.52***	179.17***

	Activities of Daily Living Disability
Living arrangements
With spouse	0.270 (0.044)***	0.333 (0.045)**
With children	0.480 (0.033)***	0.523 (0.034)***
With spouse and children	0.401 (0.047)***	0.483 (0.048)***
Institution	0.483 (0.058)***	0.554 (0.062)***
Others	0.687 (0.096)***	0.763 (0.099)***
Age (year)	0.055 (0.001)***	0.053 (0.001)***
Gender	0.119 (0.025)***	0.015 (0.027)
Ethnicity	−0.326 (0.050)***	−0.375 (0.050)***
Residence	−0.148 (0.023)***	−0.224 (0.024)***
Education	0.016 (0.015)	0.059 (0.015)***
Occupation	−0.187 (0.025)***	−0.234 (0.026)***
Number of children	0.057 (0.005)***	0.066 (0.005)***
Number of diseases	0.204 (0.007)***	0.184 (0.007)***
Cognitive impairment	1.020 (0.024)***	0.837 (0.025)***
Health behaviors
Drinking		−0.247 (0.030)***
Smoking		−0.198 (0.033)***
Exercise		−0.535 (0.026)***
Family income^a		0.009 (0.008)
Psychological well-being		−0.053 (0.003)***
Variance of random effect	0.511 (0.033)***	0.496 (0.034)***
LR test	528.56***	474.32***

	Cognitive Impairment
Living arrangements
With spouse	−.093 (.048)	.023 (.050)
With children	.146 (.034)***	.213 (.036)***
With spouse and children	−.028 (.053)	.109 (.055)*
Institution	.302 (.063)***	.396 (.068)***
Others	.103 (.114)	.194 (.120)
Age (year)	.080 (.002)***	.075 (.002)***
Gender	.156 (.029)***	.067 (.031)*
Ethnicity	−.140 (.052)**	−.217 (.053)***
Residence	.116 (.026)***	.001 (.027)
Education	−.227 (.019)**	−.170 (.019)***
Occupation	.082 (.029)**	.013 (.029)
Number of children	−.013 (.006)*	−.008 (.006)
Number of diseases	.114 (.008)***	.073 (.009)***
Health behaviors
Drinking		−.129 (.034)***
Smoking		−.044 (.037)
Exercise		−.459 (.030)***
Family income^a		−.022 (.008)**
Psychological well-being		−.100 (.003)***
Variance of random effect	.557 (.044)***	.482 (.042)***
LR test	405.80***	299.27***

Footnotes

Authors’ Note

Data used for this research were provided by the study entitled “Chinese Longitudinal Healthy Longevity Survey” (CLHLS) managed by the Center for Healthy Aging and Development Studies, Peking University. CLHLS is supported by funds from the U.S. National Institutes on Aging (NIA), China Natural Science Foundation, China Social Science Foundation, and the United Nations Population Fund.

Acknowledgment

We would like to thank the Center for Healthy Aging and Development Studies, Peking University, for supporting this database.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by the National Natural Science Foundation of China (grant number: 81573257 and 81602941) and the Natural Science Foundation of Fujian Province (grant number: 2016J0101).

References

Antonucci

T. C.

Akiyama

(1987). Social networks in adult life and a preliminary examination of the convoy model. Journal of Gerontology, 42, 519–527. doi:10.1093/geronj/42.5.519

Becker

G. S.

(1981). Altruism in the family and selfishness in the market place. Economica, 48, 1–15.

Brown

J. W.

Liang

Krause

Akiyama

Sugisawa

Fukaya

(2002). Transitions in living arrangements among elders in Japan does health make a difference? The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 57, S209–S220.

Chappell

N. L.

(1991). Living arrangements and sources of caregiving. Journal of Gerontology, 46, S1–S8.

Chen

Liu

(2009). Population aging in China. In Uhlenberg

(Ed.), International Handbook of Population Aging (pp. 157–172). New York, NY: Springer Press.

Chen

F. N.

Short

S. E.

(2008). Household context and subjective well-being among the oldest old in China. Journal of Family Issues, 29, 1379–1403.

Chen

Fang

Mao

Hao

Chen

Yuan

… Hong

Y. A.

(2015). Assessment of disability among the elderly in Xiamen of China: A representative sample survey of 14,292 older adults. PLoS One, 10, e0131014.

Cheng

Jin

Sun

Tang

Zhang

Chen

… Huang

(2015). Disparities in prevalence and risk indicators of loneliness between rural empty nest and non-empty nest older adults in Chizhou, China. Geriatrics & Gerontology International, 15, 356–364.

Cohen

Janicki-Deverts

(2009). Can we improve our physical health by altering our social networks? Perspectives on Psychological Science, 4, 375–378.

10.

Du Toit

Pritchard

Heffernan

Simpson

Fonn

(2002). A comparison of three different scales for rating contact lens handling. Optometry & Vision Science, 79, 313–320.

11.

Ertel

K. A.

Glymour

M. M.

Berkman

L. F.

(2009). Social networks and health: A life course perspective integrating observational and experimental evidence. Journal of Social and Personal Relationships, 26, 73–92.

12.

Folstein

M. F.

Folstein

S. E.

McHugh

P. R.

(1975). “Mini-mental state”: A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, 189–198.

13.

Dupre

M. E.

Liu

(2007). Characteristics of the institutionalized and community-residing oldest-old in China. Social Science & Medicine, 64, 871–883.

14.

Henning-Smith

(2016). Quality of life and psychological distress among older adults the role of living arrangements. Journal of Applied Gerontology, 35, 39–61.

15.

Hughes

M. E.

Waite

L. J.

(2002). Health in household context: Living arrangements and health in late middle age. Journal of Health and Social Behavior, 43, 1–21.

16.

Idler

E. L.

(1993). Age differences in self–assessments of health: Age changes, cohort differences, or survivorship? Journal of Gerontology, 48, S289–S300.

17.

Katz

Ford

A. B.

Moskowitz

R. W.

Jackson

B. A.

Jaffe

M. W.

(1963). Studies of illness in the aged: the index of ADL: A standardized measure of biological and psychosocial function. Journal of the American Medical Association, 185, 914–919.

18.

Kharicha

Iliffe

Harari

Swift

Gillmann

Stuck

A. E.

(2007). Health risk appraisal in older people 1: Are older people living alone an ‘at-risk’ group? British Journal of General Practice, 57, 271–276.

19.

Kim

H. J.

Fredriksen-Goldsen

K. I.

(2016). Living arrangement and loneliness among lesbian, gay, and bisexual older adults. The Gerontologist, 56, 548–558.

20.

Kincannon

C. L.

West

L. A.

(2005). Demography of aging in China and the United States and the economic well-being of their older populations. Journal of Cross-cultural Gerontology, 20, 243–255.

21.

Lee

Cho

Grodstein

Kawachi

F. B.

Colditz

G. A.

(2005). Effects of marital transitions on changes in dietary and other health behaviours in US women. International Journal of Epidemiology, 34, 69–78.

22.

Lewis

M. A.

Butterfield

R. M.

(2005). Antecedents and reactions to health-related social control. Personality and Social Psychology Bulletin, 31, 416–427.

23.

Lewis

M. A.

Rook

K. S.

(1999). Social control in personal relationships: Impact on health behaviors and psychological distress. Health Psychology, 18, 63–71.

24.

L. W.

Zhang

Liang

(2009). Health among the oldest-old in China: Which living arrangements make a difference? Social Science & Medicine, 68, 220–227.

25.

Zhang

(2015). Social network types and the health of older adults: Exploring reciprocal associations. Social Science & Medicine, 130, 59–68.

26.

Little

R. J.

Rubin

D. B.

(2002). Statistical analysis with missing data. New York, NY: Wiley Press.

27.

Liu

Dupre

M. E.

Mair

C. A.

Chen

(2012). Psychological well-being of the institutionalized and community-residing oldest old in China: The role of children. Social Science & Medicine, 75, 1874–1882.

28.

Liu

L.-J.

Y.-F.

Wang

(2014). Home health care needs and willingness to pay for home health care among the empty-nest elderly in Shanghai, China. International Journal of Gerontology, 8, 31–36.

29.

Lockwood

McCaffrey

D. F.

(2007). Controlling for individual heterogeneity in longitudinal models, with applications to student achievement. Electronic Journal of Statistics, 1, 223–252.

30.

Lund

Due

Modvig

Holstein

B. E.

Damsgaard

M. T.

Andersen

P. K.

(2002). Cohabitation and marital status as predictors of mortality—An eight year follow-up study. Social Science & Medicine, 55, 673–679.

31.

Martikainen

Nihtilä

Moustgaard

(2008). The effects of socioeconomic status and health on transitions in living arrangements and mortality: A longitudinal analysis of elderly Finnish men and women from 1997 to 2002. The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 63, S99–S109.

32.

Michael

Y. L.

Berkman

L. F.

Colditz

G. A.

Kawachi

(2001). Living arrangements, social integration, and change in functional health status. American Journal of Epidemiology, 153, 123–131. doi:10.1093/aje/153.2.123

33.

Oppenheimer

V. K.

(2000). The continuing importance of men’s economic position in marriage formation. In Waite

L. J.

Bachrach

Hindin

Thomson

Thornton

(Eds.), The ties that bind: Perspectives on marriage and cohabitation (pp. 283–301). New York, NY: Gruyter.

34.

Pollen

(2001). Living arrangements of older persons. Living Arrangements of Older Persons: Critical Issues and Policy Responses, 42, 54.

35.

Qiu

Sautter

Liu

(2011). Age and gender differences in linkages of sleep with subsequent mortality and health among very old Chinese. Sleep Medicine, 12, 1008–1017.

36.

Ross

C. E.

Mirowsky

Goldsteen

(1990). The impact of the family on health—The decade in review. Journal of Marriage and the Family, 52, 1059–1078.

37.

Russell

Taylor

(2009). Living alone and depressive symptoms: The influence of gender, physical disability, and social support among Hispanic and non-Hispanic older adults. The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 64, 95–104.

38.

Samanta

Chen

Vanneman

(2015). Living arrangements and health of older adults in India. The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 70, 937–947.

39.

Schafer

(1999). NORM: Multiple imputation of incomplete multivariate data under a normal model [computer software]. University Park: Department of Statistics, Pennsylvania State University.

40.

Sereny

(2011). Living arrangements of older adults in China: The interplay among preferences, realities, and health. Research on Aging, 33, 172–204.

41.

Smith

Gerstorf

(2008). Psychological resources for well-being among octogenarians, nonagenarians, and centenarians: Differential effects of age and selective mortality. In Zeng

Poston

J. D. L.

Vlosky

D. A.

(Eds.), Healthy longevity in China: Demographic, socioeconomic, and psychological dimensions (pp. 329–346). Dordrecht, The Netherlands: Springer Press.

42.

Sun

Lucas

Meng

Zhang

(2011). Associations between living arrangements and health-related quality of life of urban elderly people: A study from China. Quality of Life Research, 20, 359–369.

43.

Taylor

S. E.

Friedman

Silver

(2007). Social support. In Friedman

H. S.

Silver

R. C.

(Eds.), Foundations of health psychology (pp. 145–171). New York, NY: Oxford University Press.

44.

Tucker

J. S.

Anders

S. L.

(2001). Social control of health behaviors in marriage. Journal of Applied Social Psychology, 31, 467–485.

45.

Uchino

B. N.

(2004). Social support and physical health: Understanding the health consequences of relationships. New Haven, CT: Yale University Press.

46.

Umberson

(1992). Gender, marital status and the social control of health behavior. Social Science & Medicine, 34, 907–917.

47.

Umberson

Montez

J. K.

(2010). Social relationships and health a flashpoint for health policy. Journal of Health and Social Behavior, 51, S54–S66.

48.

United Nations (Department of Economic and Social Affairs [DESA]). (2005). Living arrangements of older persons around the world. Department of Economic and Social Affairs (DESA). New York, NY: Population Division.

49.

Vaillant

Wolff

F.-C.

(2012). On the reliability of self-reported health: Evidence from Albanian data. Journal of Epidemiology and Global Health, 2, 83–98.

50.

Waite

Gallagher

(2002). The case for marriage: Why married people are happier, healthier and better off financially. New York, NY: Broadway Books.

51.

Waite

L. J.

(1995). Does marriage matter? Demography, 32, 483–507.

52.

Waite

L. J.

Hughes

M. E.

(1999). At risk on the cusp of old age: Living arrangements and functional status among black, white and Hispanic adults. The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 54, S136–S144.

53.

Wang

Chen

Pan

Jing

Liu

(2013). Associations and impact factors between living arrangements and functional disability among older Chinese adults. PLoS One, 8, e53879.

54.

Wang

X.-Q.

Chen

P.-J.

(2014). Population ageing challenges health care in China. Lancet, 383, 870.

55.

Weiler

P. G.

Lubben

J. E.

Chi

(1991). Cognitive impairment and hospital use. American Journal of Public Health, 81, 1153–1157.

56.

Dang

(2013). China report of the development on aging cause. Beijing, China: Social Sciences Academic Press.

57.

Xie

(2014). An introduction to the China family panel studies (CFPS). Chinese Sociological Review, 47, 3–29.

58.

Vaupel

J. W.

(2002). Functional capacity and self–evaluation of health and life of oldest old in China. Journal of Social Issues, 58, 733–748.

59.

Zeng

Chen

Ruan

Feng

Nie

… Gu

(2015). GxE interactions between FOXO Genotypes and tea drinking are significantly associated with cognitive disability at advanced ages in China. The Journals of Gerontology. Series A: Biological Sciences and Medical Sciences, 70, 426–433.

60.

Zhai

Shen

Xiao

Fan

… Lin

(2015). Association of empty nest with depressive symptom in a Chinese elderly population: A cross-sectional study. Journal of Affective Disorders, 187, 218–223.

61.

Zhang

(2006). Gender differentials in cognitive impairment and decline of the oldest old in China. The Journals of Gerontology. Series B: Psychological Sciences and Social Sciences, 61, S107–S115.

62.

Zimmer

(2005). Health and living arrangement transitions among China’s oldest-old. Research on Aging, 27, 526–555.

63.

Zimmer

Kwong

(2003). Family size and support of older adults in urban and rural China: Current effects and future implications. Demography, 40, 23–44.

	Self-Rated Health
Variables	Model 1	Model 2
Living arrangements
With spouse	.157 (.025)***	.021 (.025)
With children	.206 (.021)***	.137 (.021)***
With spouse and children	.167 (.028)***	.009 (.027)
Institution	.157 (.040)***	.063 (.041)
Others	.127 (.074)	.040 (.074)
Age (year)	.007 (.001)***	.008 (.001)***
Gender	−.018 (.017)	.034 (.017)*
Ethnicity	−.150 (.031)***	−.067 (.029)*
Residence	−.097 (.016)***	.019 (.015)
Education	.054 (.010)***	−.014 (.009)
Occupation	−.121 (.017)***	−.048 (.017)**
Number of children	.003 (.003)	−.004 (.003)
Number of diseases	−.189 (.005)***	−.175 (.005)***
Cognitive impairment	−.388 (.019)***	−.203 (.019)***
Activities of daily living disability	−.191 (.005)***	−.141 (.005)***
Health behaviors
Drinking		.139 (.018)***
Smoking		.027 (.019)
Exercise		.171 (.015)***
Family income^a		.034 (.005)***
Psychological well-being		.105 (.002)***
Variance of random effect	.196 (.012)***	.121 (.011)***
Likelihood ratio (LR) test	434.52***	179.17***