Data sources
This study utilizes longitudinal data from the China Health and Retirement Longitudinal Study (CHARLS), which is coordinated by the National Development Research Institute at Peking University. CHARLS aims to collect comprehensive data across multiple waves to assess various aspects of the aging population in China. This includes demographic information, family composition, employment and retirement status, income and expenditure levels, as well as detailed health metrics such as general health status, insurance coverage, and the utilization of health services.
For the purpose of this analysis, the study examined data from four survey waves—2013, 2015, 2018, and 2020. The initial 2013 survey serves as the baseline for this study, with data from 2015, 2018, and 2020 used to construct a balanced panel dataset. The baseline survey was conducted between June 2011 and March 2012, targeting household members aged 45 and older. Spouses of the primary respondents were automatically included, leading to a comprehensive initial sample of 18,605 participants.
From this initial cohort, the study applied specific inclusion and exclusion criteria to refine the sample for longitudinal tracking. Households were excluded if they did not include any members aged 60 years or older or lacked crucial data points.To further focus on the financial impact of health issues, the analysis only included households that did not report catastrophic health expenditures in 2013. These selected households were then followed biennially, culminating in a final sample of 6,050 participants by the 2020 wave for the current analysis. Figure 1 illustrates the sample selection process across the different survey waves, detailing both the attrition and inclusion at each stage.
Dependent variable
Catastrophic Health Expenditure (CHE) is a critical metric used to evaluate the financial impact of out-of-pocket health care expenditures on household finances. CHE occurs when these expenditures exceed a predefined threshold, representing a significant financial burden on households. Generally, CHE is operationalized using two widely recognized thresholds to determine whether health expenditures can be considered catastrophic:
(1)10% of Total Household Expenditure: This threshold considers out-of-pocket healthcare payments that exceed 10% of a household's total annual expenditure[15, 16]. This includes all household spending, encompassing daily living costs, utilities, transportation, and other non-discretionary spending.
(2)40% of Non-Food Expenditure: This threshold evaluates whether out-of-pocket healthcare payments constitute more than 40% of the household's total non-food expenditure[17]. Non-food expenditure is often used as a proxy for discretionary spending and is considered a more accurate reflection of a household’s financial capacity to absorb health-related costs
Data Quality and Socioeconomic Status Measurement: In many low- and middle-income countries, income data are often unreliable or incomplete. Therefore, consumption expenditure is frequently utilized as an alternative measure of socioeconomic status due to its generally more accurate reporting. The broader scope of consumption data includes but is not limited to household expenses on utilities, communication, transportation, domestic services, entertainment, personal care, and education.
Operational Definition in Current Study: For the purposes of this analysis, only the health expenditures incurred by the respondent and their spouse are considered. These expenses encompass costs associated with outpatient visits, hospital admissions, and medications incurred over the past year.
Calculation of Non-Food Expenditure (CTP): In this study, Total Non-Food Expenditure (CTP) is calculated by deducting essential subsistence needs (i.e., food) from total household expenditure. This measure includes expenditures on communication, utilities, fuel, local transportation, domestic help, entertainment, daily necessities, dining out, education, out-of-pocket health expenses (excluding portions covered by insurance such as Medicare), personal fitness and beauty, rent, maintenance, taxes, donations, and household durable goods.
If a surveyed household's annual health spending crosses these thresholds, the dependent variable is coded as 1, indicating the presence of catastrophic health expenditures. If it does not, it is coded as 0, indicating absence.
Independent variable
The primary independent variable in this study is chronic disease status, defined using a binary metric based on the presence or absence of chronic diseases among participants. This operationalization draws on data collected through the China Health and Retirement Longitudinal Study (CHARLS), which systematically captures health information across a nationally representative aging population.
Chronic Conditions Assessed: In CHARLS, participants were asked to report if they had been diagnosed with any of the following 14 chronic conditions by a healthcare professional: including hypertension, dyslipidemia, diabetes, cancer or malignant tumor, chronic lung diseases, liver diseases, heart attack, stroke, kidney diseases, stomach and other digestive diseases, emotional, nervous and psychiatric problems, memory-related diseases, arthritis and rheumatism, and asthma.
Coding of Variable: For analytical purposes, the chronic disease variable is coded dichotomously. A value of 1 is assigned to respondents who report having one or more of the listed chronic conditions, indicating the presence of chronic disease. Conversely, a value of 0 is assigned to those with no reported chronic conditions, indicating the absence of chronic disease.
Justification for Use as an Independent Variable: The inclusion of chronic disease status as an independent variable is predicated on the substantial body of evidence linking chronic health conditions with both increased healthcare utilization and heightened financial vulnerability. Chronic diseases often require long-term management and treatment, which can lead to significant out-of-pocket healthcare expenditures and potentially catastrophic financial impacts on households.
Analytical Considerations: The binary nature of this variable allows for clear distinction in the analysis of health expenditure patterns between individuals with and without chronic conditions, thus facilitating a more nuanced understanding of the financial burden of health care among aging populations in China.
Household and respondent characteristic
This study systematically categorizes household and respondent characteristics to elucidate their potential impact on health expenditures.
Household Characteristics: Place of Residence: Categorized as urban or rural, this variable reflects the geographical diversity of the sample. Marital Status: Defined as living with a spouse for those who are married or cohabiting, and not living with a spouse for those who are unmarried, separated, divorced, or widowed. This variable reflects the family support structure of the sample.
Respondent Characteristics: Demographic Information: Includes gender (male, female) and age (in years), providing a demographic profile of the cohort.Educational Attainment: Classified into three categories—elementary school and below, junior high school, and high school and above. This categorization helps assess the correlation between education level and health outcomes or expenditures.
Health-related Variables: Consist of the prevalence of chronic diseases (yes, no) and health insurance coverage (yes, no). Insurance coverage is further subdivided into three major types of basic medical insurance programs in China: basic medical insurance for urban employees, basic medical insurance for urban residents, and the new rural cooperative medical scheme. These variables are crucial for evaluating the burden of healthcare costs.
Statistical analysis
Initial analysis employed descriptive statistics to outline the baseline characteristics of the study population. This included computing means and standard deviations for continuous variables, and frequencies and percentages for categorical variables.
The core of our analytical framework is a binary logistic regression model, which was utilized to explore the potential predictors of catastrophic health expenditures (CHE). CHE was defined using the threshold of non-food expenditures exceeding 40% of total household expenditures. This threshold is particularly relevant for low-income households as it mitigates the bias introduced by essential food expenditures.
We conducted stratified analyses to assess how chronic diseases influence catastrophic health expenditures (CHE) across different demographic segments, including varying genders and chronic disease statuses. This approach allows for a deeper understanding of the factors influencing CHE in different subpopulations and helps analyze the impact of single chronic diseases and multimorbidity on household economic conditions.
Sensitivity analyses were performed using alternative definitions of CHE to validate the robustness of our findings. This step is crucial to ensure that our results are not overly dependent on the chosen threshold.
Data from the CHARLS questionnaire were processed and cleaned using R programming language, ensuring accurate and efficient handling of the dataset. The statistical analyses were performed using Stata17. A two-sided p-value of less than 0.05 was considered statistically significant for all tests.