Data source and study sample
The data used in this study were obtained from the 2010 EASS, which is an East Asian version of the European Social Survey. The EASS has been viewed as unique, a result of substantial efforts by East Asian countries to establish an internationally comparable database on social issues. The EASS was conducted on the basis of a vigorously controlled study protocol, including standard procedures for translating the measures into different languages and for collecting and controlling the data, thus enabling direct cross-national comparisons in a coordinated research setting. In the EASS, China and Korea shared a common module in a General Social Survey-type questionnaire, and each country conducted a survey of a nationally representative sample selected using a multistage stratified random sampling method.
In particular, the 2010 EASS included a health module concerning health status (self-rated health, chronic disease, etc.), health-related behaviors (smoking, drinking, exercise, etc.), caregiving and receiving care, health and social security insurance, social support and trust, epidemiology, family care need and care management, and so forth. The survey was administered face-to-face at respondents’ homes by trained interviewers from June to December, 2010. Valid response rates of 73.0 and 63.0% were obtained for China and Korea, respectively. Participants’ ages ranged from 18 to 98 years. The survey methodology has been described in an online report (http://www.eassda.org/).
Of the 5327 individuals aged ≥20 years, we excluded 347 (6.5%) due to missing values; however, we found no significant differences between the datasets before and after the exclusion (p < 0.827 for gender; p < 0.867 for age). Finally, we analyzed data from 4980 individuals, 3629 of whom were in China and 1351 in Korea.
The EASS data archive provides publicly available data from anonymous respondents. Verbal informed consent was obtained from all participants due to the limited time for survey interviews, and waivers of written consent were authorized by an ethics committee. Ethical approval for this study was granted by the institutional review board of the Graduate School of Public Health, Yonsei University, Seoul, Korea.
Measures and variables
The dependent variable was individuals’ self-rated health. Respondents were prompted to answer the question, “How would you rate your health?” on a 5-point scale (excellent, very good, good, fair, and poor). The responses were grouped into the following three categories because of the sparseness of observations in each category: “poor” (“poor” or “fair”), “good” (“good”), and “excellent” (“very good” or “excellent”).
The independent variables of interest included socioeconomic factors (education, employment, household income, and self-assessed social class) and religious affiliation. Respondents’ education levels were classified into the following four categories: elementary school/lower (no formal qualification or elementary school), junior high school, senior high school, and college/higher (junior college, university, or graduate school completed). The highest education level reported by each individual was used as the indicator of education. Employment status was dichotomized into “employed” and “not employed” (having no current work income). Household income, which was a continuous variable, was divided by the square root of household size in order to adjust for household size [22], and categorized into income quartiles. To prevent any risk of bias, we included the 12.9% of respondents who did not report household income in the “missing income” category. Considering self-assessed social class, the following question was posed in the survey: “In our society, there are groups that tend to be toward the top and groups that tend to be toward the bottom. Below is a scale that runs from bottom to top. Where would you put yourself on this scale?” Available choices were on a 10-point scale from 1 (lowest) to 10 (highest). We converted the 10-point scale into a 5-point scale and merged the two highest categories because of the sparsity of cases in these groups. As a result, we obtained the following four categories of the variable of self-assessed social class: lowest, low, high, and highest. Individuals were also categorized on the basis of religious affiliation, namely, Christian, Buddhist, and others (Islam, Hinduism, atheism, agnosticism, and other religions). The Christian group included individuals who were affiliated with Roman Catholicism, Protestantism, Christian Orthodoxy, and other Christian religions.
As potential confounders, we considered various factors regarding (1) demographics, (2) health-related risks and the healthcare system, and (3) social capital. Demographic factors included gender, age, and marital status. Respondents were divided into six age groups, namely, 20–29, 30–39, 40–49, 50–59, 60–69, and ≥ 70 years. For the marital status variable, individuals were divided into three categories, married, never married, and formerly married (widowed, divorced, or separated). We removed unmarried individuals cohabiting with their partners because this category included only 52 individuals (0.5%).
Health-related risks and healthcare system factors include chronic disease, current smoking habits, drinking frequency, body mass index (BMI), physical exercise, health insurance, and unmet medical needs. The following variables were dichotomized: chronic disease (“having a chronic disease” or not), current smoking habits (“smoking a few times a year or more” or not), physical exercise (“doing physical exercise for at least 20 minutes a few times a year or more frequently” or not), health insurance (“having health insurance” or not), and unmet medical needs (“having healthcare needs in the past 12 months but did not receive care” or not). Drinking frequency was categorized into two groups, frequent (drinking daily or several times a week) and none or infrequent (drinking less than several times a month and not drinking). BMI was categorized as underweight (BMI < 18.5), normal weight (18.5 ≤ BMI < 25), overweight (25 ≤ BMI < 30), and obese (BMI > 30).
Social capital factors include generalized trust, emotional support, and instrumental support. Generalized trust was measured by the degree to which respondents agreed with the statement, “Generally, you can trust other people.” The response options were as follows: “highly trust,” “trust,” “do not trust,” and “do not trust at all.” This category was dichotomized, with the first two alternatives combined as “highly trust” and the latter two as indicating low trust. Emotional support was assessed with the question, “During the past 12 months, did people listen to your personal problems or concerns when you needed it?” The answers were grouped into three, namely, “yes,” “no,” and “do not have such needs.” Instrumental support was assessed with the question, “During the past 12 months, did people take care of your household chores (housework, childcare, and nursing care) when you needed it?” The answers were grouped into three, namely, “yes,” “no,” and “do not have such needs.”
Analytic procedures
A three-step analysis was performed. First, the demographic factors of Chinese and Korean respondents were compared using χ2 tests. Second, differences in the proportion of individuals reporting excellent, good, or poor self-rated health were examined for each socioeconomic and religious factor for each country using χ2 tests. Third, multinomial logistic regression analysis, given the three categories of self-rated health, was used to assess the associations of socioeconomic and religious factors with self-rated health both without and with a full set of studied potential confounders within and between China and Korea.
While the ordinal nature of self-rated health would suggest using ordinal regression, multinomial logistic regression was appropriate in view of the violation of the proportional odds assumptions of ordinal regression. In this study, multinomial logistic regression was used to predict the risks of reporting either poor health or good health versus excellent health (reference). The adjusted odds ratios with 95% confidence intervals (OR, 95% CI) indicated the associations of the independent variables with good health versus excellent health (Table 3) or with poor health versus excellent health (Table 4).
The regression models were run for each country separately to explore possible national differences in the effects of socioeconomic and religious factors on health. Variance inflation factors (VIFs) for the independent variables were also within acceptable limits (VIF < 3.4), indicating no serious problems of collinearity. All data analyses were performed using SAS, version 9.2 (SAS Institute Inc., Cary, NC, USA).