Data sources
In this study, we analysed data from the China Health and Retirement Longitudinal Study (CHARLS) 2011–2012 (wave 1), 2013–2014 (wave 2), and 2015–2016 (wave 3) [17, 18]. The CHARLS is a nationally representative large-scale survey targeting the population aged 45 years and above in China. The study, launched in 2011, aims to collect a set of high-quality microdata representing families and individuals aged ≥ 45 years to analyse population ageing in China and promote interdisciplinary research; this study covers 150 counties, 450 villages and approximately 10,000 households. Study participants are followed up every 2 years. Data collection in CHARLS has been described elsewhere [19]. We excluded observations with missing values, leaving respondents aged ≥ 45 years: n = 16,128 in wave 1, 16,646 in wave 2, and 17,470 in wave 3 (Table 1, Additional file 1: Fig. S1-S3).
All examinations were conducted after informed consent was obtained from participants. The Biomedical Ethics Committee of Peking University approved the survey (The approval number is IRB00001052-11015), and the conduct of the study adhered to the principles of the Declaration of Helsinki.
Measures
Health outcomes
In the CHARLS, each participant was interviewed using a standardized questionnaire and underwent a medical examination. Data included sociodemographic characteristics, biometrics, lifestyle and behavioural characteristics, cardiovascular disease risk factors, health history, and medications. Each new survey respondent was queried, “Have you been diagnosed with [chronic disease] by a doctor?” In follow-up interviews, participants were asked, “Our records from your last interview show that you have had/not had [chronic disease]; is this right?” and “Have you been diagnosed with [chronic disease] by a doctor [since your last interview in the last 2 years]?” Participants recorded yes or no responses to all questions. Those who answered affirmatively to any questions were required to provide medical or hospital records.
According to the CHARLS questionnaires, 14 chronic diseases were identified: hypertension, dyslipidaemia (i.e., elevated low-density lipoprotein, triglycerides, and total cholesterol or low high-density lipoprotein), diabetes (i.e., high blood sugar), cancer or malignant tumour (except minor skin cancers), chronic lung disease (e.g., chronic bronchitis or emphysema) excluding tumours or cancer, liver disease (except fatty liver, tumours, and cancer), heart disease (e.g., coronary heart disease, angina, congestive heart failure, other heart problems), stroke, kidney disease (except tumour or cancer), stomach and other digestive diseases (except tumour or cancer), emotional or psychiatric problems, memory-related disease, arthritis or rheumatism, and asthma.
Socioeconomic status (SES)
To measure inequalities in the prevalence of chronic disease among people with different standards of living, data on household assets and housing characteristics were used to construct a proxy index to measure living standards [20]. In this study, durable consumer goods (including an automobile, electric bicycle, motorcycle, refrigerator, washing machine, television, computer, stereo system, video camera, camera, air conditioner, mobile phone, furniture, musical instrument, valuable decorative items, jewellery, collectibles, precious metals, or artwork) and housing characteristics (including the type of structure of residence, one-story or multi-level building, toilet, electricity, running water, bathroom facilities, coal gas or natural gas, heating, source of cooking fuel, telephone, and internet connection) were combined into an index of SES to measure household living standards.
Principal component analysis (PCA) is a common approach [21] used to describe variation in a set of variables as linear combinations of the original variables, in which each continuous linear combination is derived to explain variation in the original data as much as possible while being uncorrelated with other linear combinations. Typically, the first principal component with the largest amount of information from the original variables was chosen to represent wealth status and be defined as the wealth index here [20]. In the case of PCA, the wealth index for individual is defined as follows: (see Equation 1 in the Supplemental Files)
where is the value of asset for household , is the sample mean, is the sample standard deviation, and are the weights associated with the first principal component (Additional file 1: Table S1).
Other variables
To compare the differences in chronic disease prevalence between different participant groups, demographic characteristics (including age, sex, urban or rural residence, living area) were obtained.
Concentration index
The concentration index has become one of the standard measures in the health economics literature on equality and inequality in health and health care. The concentration index is defined as twice the area between the concentration curve and the line of perfect equality (the 45-degree line), where individuals are ranked by socioeconomic level, general income, and the cumulative ranking of each individual plotted against the cumulative share of health outcomes or health care utilization [22]. Thus, in the case of no socioeconomic-related inequality, the concentration index is zero. When the outcome of interest is ill health, the convention is that the index takes a negative value when the curve lies above the line of equality, indicating that the prevalence of chronic disease is excessively concentrated among poor populations; the index takes a positive value when the curve lies below the line of equality. The concentration index can be expressed as follows: (see Equation 2 in the Supplemental Files)
where is the indicator of the health status of individual ; is the fractional rank of individual in the living standards distribution, with for the poorest and for the wealthiest individuals; and is the mean of the health outcome variables [22, 23]. The concentration index is bound between –1 and +1. When the concentration index is used to compare inequality across time, place, and subpopulations, calculating the concentration index for binary outcomes is potentially problematic because the possible range of the concentration index differs according to the mean of the outcome variable [24-27]. To resolve this issue, the concentration index is divided by (1 − mean of the outcome variable), referred to as the adjusted concentration index herein, following previous studies [25, 26].
Statistical analyses
We produced an initial estimation of socioeconomic inequalities in health by comparing the prevalence rates of 14 chronic diseases across wealth index tertiles. Rates were standardized for age and sex using logistic regression, including a random effect to account for clustering at the community level.
To measure socioeconomic inequalities in chronic disease prevalence among participants of CHARLS waves 1–3, we first calculated the concentration index for the prevalence of each chronic disease in each wave. After obtaining the total concentration index, we calculated concentration indices grouped by sex and urban or rural residence to compare the differences in chronic disease prevalence inequities among different populations. We used a SAS macro and the bootstrap method to estimate the concentration index and confidence interval [28]. We plotted values of the concentration index for each province against per capita gross domestic product (GDP) in the province. We used Kendall’s rank correlation coefficient (Kendall’s tau) to measure the strength and direction of the association between per capita gross GDP and concentration indices. All data preparation and data analysis were performed in SAS version 9 (SAS Institute Inc., Cary, NC, USA).