The data used in this study are from the Chinese Longitudinal Health Longevity Survey (CLHLS), which is a large population-based study conducted by the Centre for Healthy Aging and Family Studies at Peking University and the Chinese Center for Disease Control and Prevention. The CLHLS sampled older adults aged 65 and over from 22 out of the 31 provinces of mainland China, the population in these provinces constitutes approximately 85% of the total population . The CLHLS has established the sampling frame with all centenarians from the sampled counties/cities. Based on the prepared code, one octogenarian and one nonagenarian were randomly selected that matched each sampled centenarian. And for every three sampled centenarians, four older adults aged 65–79 were randomly selected from a nearby geographical unit. All respondents in the CLHLS were surveyed by face-to-face interviews using internationally compatible questionnaires. Proxy respondents (a spouse or other family member) were instead interviewed when the participants were unable to answer questions, while questions about cognitive function were answered by the participants themselves. Data of the study were obtained from the CLHLS in 2018, which surveyed 15874 older adults. To ensure the analysis effectively, the samples with missing values or answers of “don't know” in any variables of interest would be excluded. Finally, a total of 10665 respondents were included in this study.
The CLHLS study was approved by the Ethics Committee of Peking University (IRB00001052-13074), and all respondents or their proxies provided written informed consent.
Explained variable: Cognitive impairment
Cognitive impairment of the respondents was assessed by the Chinese version of the Mini-Mental State Examination (MMSE) adapted from the scale developed by Folstein and colleagues , which has been proven to be reliable and valid for elderly Chinese [29,30]. MMSE tests 24 items from the five aspects of cognitive function: orientation, reaction, attention & calculation, recall, and language. The total score ranged from 0 to 30, and the higher score indicated better cognitive ability. Education-based MMSE cutoff points are widely used to screen for cognitive impairment in the elderly population with low education . As quite a number of respondents (48.29%) in this study had no formal education, we used education-based MMSE cutoff points to define those who with cognitive impairment: < 18, respondents with no formal schooling; < 21, respondents with 1 to 6 years of schooling; and < 25, respondents with more than 6 years of schooling [32,33].
Explanatory variables of the study consisted of socioeconomic factors, social support, and social participation.
Socioeconomic factors included education level, occupation, and economic status. Respondents were asked “How many years did you attend school?”, according to the answers of 0 years, 1~6 years, and 7 years or more, education level was classified into three categories: illiterate (1), primary school (2), and middle school or more (3). The categories of occupation in the questionnaire included professional and technical personnel, governmental, institutional or managerial personnel, commercial, service or industrial worker, self-employed, agriculture, forestry, animal husbandry or fishery worker, house worker, and others. In this study, the occupation was recoded into non-white-collar (1) and white-collar (2). Economic status was assessed by asking “the per capita of household income in the last year?” and then divided into quintiles with quintile 1 (1) indicating the poorest and quintile 5 (5) indicating the richest.
Social support was measured by three elements: emotional support, financial support, and living arrangement. Emotional support was obtained by the question “The first person to whom you usually talk frequently in daily life (including spouse, son/daughter, son-/daughter-in-law, grandchildren and their spouse, other relatives, friends/neighbors, social workers/housekeeper, network friends, and nobody).” The answers were grouped into four categories: nobody (1), relatives/friends/neighbors and others (2), children (3), and spouse (4). Financial support was calculated by the answers to 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?”. If the answers to the three questions add up to 0, we assumed the respondent had no financial support (1), otherwise, he/she had financial support (2). Living arrangement was categorized into lived alone (1), lived in an institution (2), and lived with household members (3).
Social participation included two aspects of organized activities and informal activities. The questions for organized activities and informal activities were “Do you take part in some social activities (organized) at present?”, “Do you take part in the following activities (e.g. square dance, series, interact with friends, play cards/mah-jongg) at present?”, respectively. And the options of two questions both included “almost every day”, “not daily, but once for a week”, “not weekly, but at least once for a month, “not monthly, but sometimes”, and “never”. If the respondents select “never”, we assumed the respondent had not participated in organized activities or informal activities (1), otherwise, he/she had participated in organized activities or informal activities (2).
Covariates included demographic factors such as sex, age, marital status, and residential area. Sex was defined as male (1) and female (2). Age was obtained by self-reported and then divided into three groups: aged 65~74 years old (1), aged 75~84 years old (2), and aged 85 years or above (3). Marital status was categorized into two groups: separated/divorced/widowed/never married (1), married and living with spouse (2). The residential area was defined as urban/town (1) and rural (2).
The Stata 15.1 software was used for data analysis. Difference analysis was performed with the Chi-square test. As the dependent variable is dichotomous, binary logistic regression was used to estimate the effects of demographic factors, social support, and social participation on cognitive impairment among older adults. In addition, the concentration curve and concentration index (C) were used to reflect the income-related inequality in cognitive impairment. In this study, the distribution of cognitive impairment was examined by economic status quintiles. The C is defined as twice the area between the concentration curve and the line of equality, and the concentration curve is obtained by plotting the cumulative percentage of cognitive impairment (Y-axis) against the cumulative percentage of the population ranked by economic status (X-axis). The C can be calculated using the following formula :
where h is the health outcome (cognitive impairment in this study), µ is the mean of h, and r denotes the fractional rank of individuals in the distribution used (economic status quintiles). The C ranges between − 1 and + 1, a value of zero represents absolute fairness and there exists no income-related inequality. When C is positive, suggesting that cognitive impairment is more concentrated among rich people (pro-poor). Conversely, if the C takes a negative value, indicating cognitive impairment is more concentrated among poor people (pro-rich). As the outcome variable in the study is binary, the bounds of C do not vary between − 1 and + 1. To correct this issue, we followed Wagstaff’s suggestion , normalizing the C by dividing estimated C by 1 minus the mean (1−µ).
Decomposition analysis was further performed to determine the contribution of each factor to the inequality, in which the contribution of each factor is the product of the sensitivity or elasticity of cognitive impairment with respect to that factor and its degree of inequality.