Data collection and study participants
This study analyzed data from the 9th to 11th wave (2014–2016) of the Korea Health Panel Survey (KHPS), which has been conducted annually since 2008 [18] by the Korea Institute for Health and Social Affairs and the National Health Insurance Corporation. The surveyor visits the target household and conducts the interview using Computer Assisted Personal Interviewing methods. They surveyed monthly household consumption expenditures and annual household income based on the previous year according to household ledgers, receipts, or self-reports. Additionally, annual household income was converted to an equalized household income value considering the number of household members. Information on healthcare spending was collected retroactively on a yearly basis using household ledgers, receipts, and insurance usage information to prevent omissions or errors in recall due to the length of time between healthcare service use and the date of the investigation.
In addition, the state of economic activity is surveyed on the last day of the year. The dining out expenses needed to accurately calculate the CHE have been investigated since 2014, and the most recent year when the information was released is 2016. Therefore, the 2014–2016 KHPS data was used to analyze the experience of CHE in 2015 in accordance with changed economic activities from 2014–2015.
Of the 18,130 participants in 2015, 14,530 remained, with those under 19 years removed. Of these, one person was removed from the study population as the CHE could not be calculated, and other participants without information on the control variables were removed, bringing the total number to 13,709. Finally, a total of 12,454 participants were analyzed as after eliminating people whose economic activity was unknown in 2014 or 2015.
Catastrophic health expenditure experience
The dependent variable is the experience of CHE in 2015. The World Health Organization proposed methods for calculating CHE and defines CHE as a situation when medical expenses are more than 40% of one’s capacity to pay [10]. The method considered a household’s capacity to pay by subtracting essential food expenditures or household subsistence spending from the sum of household spending, and households were defined as experiencing CHE when household out-of-pocket payment on medical expenses accounted for over 40% of the household capacity. If healthcare expenses accounted for more than 40% of the amount that could be spent, then the household was classified as the Yes group who experienced CHE. If household expenses were less than the criteria, then the household was classified as the No group that did not experience CHE. The items on household consumption expenditure in the survey questioned the costs spent over the past year, so the 2016 survey used 2015 expenditure values. Healthcare expenses were defined as expenses for hospitalizations, outpatient services, emergency services, and prescription drugs throughout 2015.
Changes in economic activity
The independent variable is the change in economic activity in 2014–2015. The status of economic activity was classified as yes if they worked for income purpose and; or no if they did not. As survey was conducted at the end of last year, asking whether the individual participated in economic activity or not. The economic activity on December 31, 2014, surveyed in the 2015 survey, and the economic activity on December 31, 2015, surveyed in the 2016 survey, were categorized into the following four categories: active no change, inactive then active, active then inactive, or inactive no change.
Control variables
The control variables included individual and household level variables. The variables at the individual level are sex, age, marital status, education level, medical aid, disability status, Charlson comorbidity index (CCI), unmet medical need, current smoking, and alcohol consumption statuses. The CCI was calculated according to the method defined in a previous study and sorted into three categories: 0, 1, 2 + [19]. The variables at the household level are the region of residence, head of household, household income level, lagged household income level (i.e., household income level in 2014), and lagged dependent variable (i.e., CHE experience in 2014). KHPS was surveyed by defining the person who represents the household as a householder, regardless of whether the residence was owned or their income. Based on the survey responses, the head of household was categorized into householder, otherwise non-householder. The household income level and lagged household income level were categorized using equalized income.
Supplementary analysis
A subgroup analysis of the association between dependent and independent variables was performed according to sex, age, health-related variables (disability status, CCI, unmet medical needs), and household level variables (region, head of household, household income level, lagged household income level, lagged dependent variable) among the covariates. Additionally, to analyze the association between the discontinuation of economic activity by subgroup analysis and the CHE, the interview questionnaire includes the question “Why are you economically inactive?”. The answers for reasons for not engaging in economic activities were categorized as follows and considered as economically inactive factors: family-related factors (housekeeping, upbringing, and care), health-related factors (inability to work, illness or health deterioration, and accidental damage), retirement, other (academic work, job preparation, military service, etc.), and voluntary selection (no will to work or no job search). A sensitivity analysis was also performed by changing the CHE criteria to not only 40%, but also 30%, 20%, and 10% to confirm the results [8, 12].
Statistical analysis
Using the Chi square test, this study examined the differences between groups according to the independent variables. A p value of < 0.05 was considered significant. A logistic regression analysis was performed to investigate the association between changes in economic activity and CHE with calculations expressed as odds ratio (OR) and 95% CI (confidence interval). We checked for multicollinearity in the statistical model through tolerance, variance inflation factors (VIFs), and collinearity diagnostics. In this model, the tolerance values were all under the 0.1 and VIFs were all less than 2.5, and the results of collinearity diagnostics were acceptable; therefore, the results indicate no problem with the correlations between the dependent and independent variables [20]. All statistical analyses were performed using SAS statistical software package version 9.4. (SAS Institute Inc., Cary, NC, USA).