The Impact of Spousal Chronic Health Shock on Subjective Well-Being Among Elderly in China: An Urban-Rural Dimension

Backgrounds Chronic conditions could bring not only heavy economic burden on families, but also had negative 3 emotional and mental impacts to patients and their family members. The aim of this study is to explore 4 the effect of chronic health shock of elderly people on spousal subjective well-being in China from 5 urban-rural dimension. We used two most recent databases —2011 and 2015—of China Health and Nutrition Survey, and the 8 total sample were categorized into urban sample and rural sample. Participants were defined as treatment 9 group if his/her spouse was diagnosed with chronic disease in 2015 and not diagnosed in 2011; others 10 were defined as control group. Propensity score matching was used to evaluate the average treatment 11 effect of treated(ATT)of spousal chronic health shock. Ordinary linear square(OLS) regression was also 12 deployed to explore the relationship between spousal chronic health shock and subjective well-being.

happiness between urban and rural areas [12]. This difference in system between urban and rural areas 1 resulted in the urban-rural distinction of macro living environment for the elderly. A research using panel 2 data to study changes in the delight of Chinese urban and rural residents in 2005 and 2015 showed that 3 in both years, urban residents were happier than their rural counterparts [12]. Another research using 4 nationally representative data also demonstrated that urban residents reported significantly higher 5 subjective well-being and self-reported health [13]. China's unique socioeconomic circumstances in its 6 rural and urban areas could have an important impact on the relationship between elderly couples. A 7 prior study suggested that the urban couples have higher marital quality than rural couples, due to their 8 higher cognitive ability, income level, education level; furthermore, higher marital quality results in a 9 positive effect on urban resident's mental health [14]. Therefore, it is understandable that there are 10 differences in subjective well-being between urban and rural elderly. What's more, due to the uneven 11 distribution of health resources among different areas, the prevalence of chronic diseases also differs in 12 urban and rural areas. The prevalence of chronic diseases among the elderly in urban areas is 13 significantly lower than that in rural areas [15]. Consequently, the impact of spouse's chronic health 14 shock on subjective well-being among urban and rural elderly need to explored.

15
Apparently biased estimation will be resulted because of self-selection, if we directly compare the SWB 16 of elderly who experienced spousal chronic health shock with that of elderly who didn't experienced 17 spousal chronic health shock. That is, their own characteristics (that could affect SWB) of these two 18 groups may differ and the difference may furtherly be related to which group they are in (that is to say, 19 whether they have experienced spousal chronic health shock). Therefore, to ensure the random 20 distribution of sample, propensity score matching (PSM) are widely adopted to evaluate average effect   The two most recent databases of CHNS-2011 and 2015-were used in our study, and the total sample 8 were categorized into urban sample and rural sample. Elderly was defined as those above 60 years old; 9 according to our purpose, those who were widowed or living alone were eliminated. Additionally,

13
The dependent variable in our study was self-reported subjective well-being in 2015. Subjective well-14 being predicts many life outcomes such as longevity, health, income and social skills. The SWB has 15 been measured by question "How do you rate the quality of your life at present" in CHNS. The answer 16 of this question has following options that are "very good", "good", "fair", "bad" or "very bad"; and is 17 coded from 5 to 1, respectively.

22
If the respondent's spouse answered No for all of these diseases in 2011 but answered Yes for any of 23 these diseases in 2015, then he/she was regarded that he/she has been experienced spousal chronic health 24 shock and coded 1; otherwise, coded 0. Those who experienced spousal chronic health shock were in 25 treatment group and others were in control group.

26
Control variables Except for spousal chronic health shock, some other variables that could be related to 27 subjective well-being of elderly were also concluded in our analysis [18][19][20]. Variable work was defined by question "Are you presently working" and was also allocated as 0 if the answer was no, and if the 1 answer was yes, then it was marked as 1. Smoking status was also defined by answer to the question 2 "Are you currently smoking". Drinking status has been set as 1 if the respondent drunk more than twice 3 a month and noted as 0 if the respondent didn't drink twice a month. Economic level was defined as the 4 annual household income per capita. In order to facilitate our analysis; we have taken log of annual 5 household income per capita in our study. What's more, preventive health service, state of chronic 6 disease, household size and some other variables were also taken in our analysis.  (1)

13
However, due to self-selection problem, ATE calculated by equation (1)

17
Assume that H=1 represents participants (treatment group), H=0 represents non-participants (control 18 group). Then ATT can be anticipated as: Average treatment effect on the untreated (ATU) can be defined by following equation: If the experiment was done by random assignment, then ATT=ATE and ATE could be measured by the 23 outcome difference of treatment group and control group. However, in non-experimental data, 24 researchers could get E( 0 | = 1). Appropriate strategies need to be done to solve the missing data 25 problem. Propensity score matching is a widely adopted method in cross-sectional data. The basic idea 7 control group whose covariate ≈ . Based on Ignorability Assumption, the probability that 1 individual i in treatment group is the same as that individual j in treatment group [21]. Therefore, can 2 represent 0 . Furthermore, the treatment effect of individual i can be calculated by the following In doing so, for each individual in treatment group, it can be matched by another individual in the control 6 group, so that we can measure its treatment effects. Taking the average of all the treatment effects, we 7 can get an Average Treatment Effect, which is ATE.

8
The propensity score is the probability for individual i to be in the treatment group with given . It can 9 be stated as the following equation.

11
In this study, following a standard propensity score matching procedure, we have firstly chosen

18
ATT of matched sample could be expressed as this: 20 Where 1 = ∑ , which represents observations in treatment group. Similarly, ATU could be 21 expressed as this: 22 Where 0 = ∑ (1 − ), which represents observations in control group. ATE could be expressed as 24 following: Where represents the total sample size.          . What's more, there are more and more empty-related to care, for the chronically ailing spouse. However, being a caregiver also bring several burdens 1 to elderly. Caregivers need to take on many different types of tasks. For example, caregivers need to 2 assist their ailing spouse with enough daily activities, monitor their diets and provide care after a 3 debilitating illness. A large body of literature has documented that caregivers tend to report worse 4 physical health and higher rates of depression and anxiety, especially among the elderly[32-35]. As a 5 consequence, a potential explanation for our result that spousal chronic health shock has a negative 6 effect on elderly's SWB is that elderly whose spouse suffered from chronic health shock may need to 7 provide care for their spouse and being a caregiver brings both physical and mental burden to their daily 8 life and consequently, their attitude and satisfaction to current life decreases. Thirdly, looking after 9 mutually in couple's plays a crucial part in life of old age. Elderly couples support each other by 10 accompanying in hardship of life, comforting and protecting each other. If one suddenly suffers from 11 chronic disease, his/her daily life would be affected, so would his/her spouse. Not only spousal chronic 12 condition brings economic burden, but also social support and family support will also be damaged.

17
The results of our study from urban and rural dimensions also have been suggested the different role 18 spousal chronic health shock played in SWB of elderly. Our results proved that spousal chronic health 19 shock significantly worsens the SWB of urban elderly, but the effect for rural elderly isn't statistically  Chronic diseases are becoming more and more popular in China, especially among elderly. Our study 9 found that spousal chronic health shock would have a significant negative effect on elderly's subjective 10 wellbeing in urban China. We can explain this relationship in three ways. First, the onset of chronic 11 diseases has brought heavy economic burden to a family and SWB of elderly is relatively associated 12 with economic level. Second, spousal suffering from chronic disease usually means elderly need to 13 become a caregiver and this will bring living burden to him/her. Third, for urban elderly couples with 14 closer marital relationship, if one partner suffers from chronic illness, the other may be lonely and lack 15 of emotional support. In rural China, there was no effect has been found between spousal chronic health 16 shock and elderly's SWB. It is understandable because rural residents have lower educational level and 17 thus they are less sensitive to subjective well-being. Furthermore, rural elderly usually have more ways 18 to participate in labor activities and more family members and friends, which could compensate SWB 19 reduction caused by spousal chronic health shock. This study firstly evaluated the effect of spousal 20 chronic health shock of SWB of elderly from rural and urban dimensions in China. Our results suggest 21 that more attention need to be payed to the elderly whose spouse are suffering from chronic illness, 22 especially in urban area.

23
There are also several limitations in our analysis. We proposed several possible explanations. However, 24 the influence mechanism of this effect needs to be further explored in future studies. In our study we

11
 Availability of data and materials:

12
The datasets generated and analysed during the current study were derived from the China Health       2 confidence interval. The coefficient isn't statistically significant, which is there is no effect found here.

3
Additional Files 4 Additional file 1.docx: Table A1, bias reduction results by nearest neighbor matching; 5 Additional file 2.docx: Figure A1, kernel density distribution -urban sample; Figure A2, kernel 6 density distribution -rural sample. 7 Figure 1 OLS regression after PSM -urban sample, shows the OLS regression result after matching treatment and control group in urban sample. The far right column shows the coe cient and its 95% con dence interval. With controlling other variables, the negative effect of spousal chronic health shock is statistically signi cant.

Figure 2
OLS regression after PSM -rural sample, shows the OLS regression result after matching treatment and control group in rural sample. The far right column shows the coe cient and its 95% con dence interval.
The coe cient isn't statistically signi cant, which is there is no effect found here.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.