Socioeconomic disparity and the risk of contracting COVID-19 in South Korea: an NHIS-COVID-19 database cohort study

This population-based observational study was conducted and reported according to the Reporting of Observational Studies in Epidemiology guidelines [12]. The study protocol was approved by the Institutional Review Board (X-2004-604-905) and the Health Insurance Review and Assessment Service (NHIS-2020-1-291). The requirement for informed consent was waived because the data analyses were performed retrospectively using anonymized data derived from the South Korean NHIS database.

The NHIS-COVID-19 cohort database was developed for medical research purposes in cooperation between the NHIS and Korea Centers for Disease Control and Prevention (KCDC). The KCDC provides data on patients diagnosed with COVID-19 from January 1, 2020 to June 4, 2020, such as COVID-19 confirmation date, treatment results, and demographic information. In addition, the data of COVID-19 patients who are receiving ongoing in-hospital treatment are not included in this database because treatment results are not yet available. Using the data on COVID-19 patients, the NHIS extracts the control population using stratification methods regarding age, sex, and place of residence as of February 2020. The NHIS-COVID-19 cohort database contains disease diagnoses according to the International Classification of Diseases (ICD)-10 codes and prescription information concerning drugs and/or procedures from 2015 to 2020. For this study, an independent medical record technician at the NHIS center unaffiliated to the study extracted the data on June 26, 2020. In the analysis, we included individuals 20 years old or older because in the NHIS-COVID-19 cohort database, the NHIS provided information on age groups considering age as a categorical variable (20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and ≥ 80 years) in conformance with the anonymized patient information in the database. In addition, individuals with an incomplete medical record were excluded; however, if an individual had missing data for only annual income levels due to the lack of information in the NHIS database, he/she was included in the analysis in the “unknown group” to avoid bias because sometimes, the annual income levels of military soldiers and individuals without health insurance coverage were not registered in the NHIS database.

All individuals in South Korea are registered in the NHIS [13] and divided into two groups: employee insured and self-employed insured. The insurance premium for employee insured individuals is determined according to income, whereas that for self-employed insured individuals is determined according to income, property, living standards, and rate of participation in economic activities. South Koreans pay a fixed rate for health insurance premiums based on their income, with approximately 67% of their medical expenses being subsidized by the government [13]. However, those who cannot afford insurance premiums or have difficulty in financially supporting themselves are included in the Medical Aid Program. In this program, the government covers almost all medical expenses to reduce the financial burden of medical costs. For this study, we obtained data on insurance premiums and used it to derive the annual income level of all participants as of February 2020. The NHIS provided information on income levels to researchers based on 20 groups created according to 5% intervals. Therefore, for analysis, we employed two methods: First, the annual income levels were classified into 20 groups according to the 5% intervals (Group 1: 0–5% [lowest] to Group 20: 95–100% [highest]) to examine the linear trends for the risk of contracting COVID-19 and risk of mortality among COVID-19 patients according to the annual income level. Participants of the Medical Aid Program (constituting approximately 2.5% of the study population) were included in the Group 1 (0–5%). Second, the annual income level was classified into four groups using quartile ratios (from Q1, the lowest, to Q4, the highest). In addition, we used the data from 2015 to 2020 to calculate the 6-year average income level and included it in the analysis. Importantly, in South Korea, all COVID-19 patients were treated in the hospital free of charge regardless of their income level.

In South Korea, all individuals with disability are registered in the NHIS database to receive various benefits. We extracted the data on registered disabilities of all participants as of February 2020. The degrees of disabilities were divided into two groups according to severity criteria: severe disability and mild to moderate disability. The types of disabilities were divided into five groups: physical disability, brain lesion disability, visual disturbance, hearing disability, and other disabilities.

The primary endpoint of this study was development of COVID-19 in the NHIS-COVID-19 database cohort. We evaluated this endpoint from January 1, 2020 to June 4, 2020. The secondary endpoint of this study was in-hospital mortality among patients diagnosed with COVID-19.

The data extracted as confounders included demographic characteristics (age and sex), place of residence (Seoul, Gyeonggi-do, Daegu, Gyeongsangbuk-do, and other areas), and the Charlson comorbidity index (CCI), which was calculated based on the registered ICD-10 diagnostic codes (Additional file 1) from January 1, 2015 to December 31, 2019. Age was divided into seven groups: 20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and ≥ 80 years. Data on the variables of age, sex, and indicators of underlying diseases, such as the CCI, were collected, and these variables were considered as confounders because they were reported to be associated with the risk of contracting COVID-19 and the risk of COVID-19 mortality [14]. The place of residence was recorded as an important confounder because there were major outbreaks of COVID-19 in Daegu and Gyeongsangbuk-do until June 4, 2020 [15], and this may have affected the results.

The characteristics of COVID-19 patients and those in the control group were compared using the Student’s t-test for continuous variables (CCI [because the CCI followed a normal distribution]) and the chi-square test for categorical variables (all other variables). First, we investigated the relationship between the risk of contracting COVID-19 and income level in 2020 using restricted cubic splines (RCS). Second, we constructed a multivariable logistic regression model to analyze the development of COVID-19 among the NHIS-COVID-19 database cohort. All confounders were included in the model for multivariable adjustment. Income level in 2020 was included as two types of independent variables to avoid multicollinearity in the model: 1) categorical variables using quartile ratios and 2) continuous variables using 5%-increase intervals to examine if there were linear trends for the risk of contracting COVID-19 and risk of mortality among COVID-19 patients according to the annual income level. In the sensitivity analysis, the 6-year average income level was included in a separate multivariable model to investigate whether the average income level for 6 years (2015–2020), was associated with risk of COVID-19 and mortality among COVID-19 patients. The grade and type of disability were also included in the separate multivariable model to avoid multicollinearity in the model. Additionally, the CCI and diseases used to calculate the index were included in the separate model to avoid multicollinearity. Third, we performed subgroup analyses according to age. All participants were classified into three subgroups according to age (20–39, 40–59, and ≥ 60 years groups). Therefore, multivariable model 1 included the income level in 2020 as a categorical variable using quartile ratios; model 2 included the income level in 2020 as a continuous variable (per 5% decrease); model 3 included the average income level for 6 years (2015–2020) for sensitivity analysis. In addition, the grade of disability, and CCI were included in multivariable model 1, whereas the type of disability and specific underlying diseases, which were used to calculate the CCI, were included in multivariable model 2. The subgroup analyses were performed using the same methods as those described for the main analyses. Finally, we performed multivariable logistic regression analysis for in-hospital mortality among patients diagnosed with COVID-19 to investigate whether income level in 2020 affected their mortality after the South Korean government subsidized all hospital charges for COVID-19 patients during the period of the study.

As of June 4, 2020, the NHIS-COVID-19 database cohort comprised 8070 COVID-19 patients and 121,050 individuals in the control population. Among them, 4790 individuals younger than 20 years and 2290 with incomplete medical records were excluded from the analysis. Thus, 122,040 individuals were included in the final analysis, and 7669 individuals were diagnosed with COVID-19 during the study period. Of these, 251 (3.2%) died due to COVID-19 during hospitalization (Fig. 1). Table 1 presents the results of comparison of characteristic between COVID-19 patients and control population in South Korea.

The RCS in Additional File 2 shows that the log odds of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection risk increased in those with an income level above the median value in 2020. In the 20–39-year-old subgroup, the RCS in Additional File 3 shows a U-shape between income level in 2020 and the log odds of SARS-CoV-2 infection risk, suggesting that both the highest and lowest income levels were associated with higher risks of COVID-19. In the 40–59-year-old subgroup, the RCS in Additional File 4 shows a similar pattern for all individuals as shown in Additional File 2. However, in the ≥60-year-old subgroup, the RCS in Additional File 5 shows that the log odds of SARS-CoV-2 infection risk gradually increased as the income level in 2020 decreased, suggesting that income level and COVID-19 risk were inversely related.

Table 2 presents the multivariable logistic regression models for the diagnosis of COVID-19 in South Korea. Regarding income levels in 2020, the Q1 group had a 1.19-fold higher COVID-19 risk than the Q4 group (OR: 1.19, 95% CI; 1.12–1.27; P < 0.001; model 1), whereas the Q3 (P = 0.06) and Q2 (P = 0.26) groups showed no significant differences. An income level decrease of 5% was associated with an increase of 1% in COVID-19 risk (OR: 1.01, 95% CI: 1.01–1.02; P < 0.001; model 2). In the sensitivity analysis, the Q1 group had a 1.10-fold higher COVID-19 risk than the Q4 group (OR: 1.10, 95% CI: 1.01–1.19; P = 0.028 model 3). In addition, individuals with severe disabilities had a 1.45-fold higher COVID-19 risk than individuals without disabilities (OR: 1.45, 95% CI: 1.26–1.67; P < 0.001; model 1), whereas mild to moderate disabilities were not significantly associated with a higher risk for contracting COVID-19 (P = 0.088). Among the types of disabilities, brain lesion disability was specifically associated with a 1.32-fold higher COVID-19 risk (OR: 1.32, 95% CI: 1.02–1.70; P = 0.033; model 2). Hosmer–Lemeshow statistics showed appropriate goodness-of-fit in the three models (all P > 0.05). The area under the curve of the multivariable models in the ROC analyses was 0.81 (95% CI: 0.80–0.81).

Table 3 presents the results of subgroup analyses according to age. Regarding income levels in 2020, in the 20–39-year-old group, the Q3 and Q2 groups had 11% (OR: 0.89, 95% CI: 0.81–0.97; P = 0.001) and 22% (OR: 0.78, 95% CI: 0.90–1.11; P < 0.001) lower risks of COVID-19, respectively, than the Q4 group, whereas the Q1 group showed no significant difference (P = 0.99). In the 40–59-year-old group, the Q1 group had a 1.23-fold (OR: 1.23, 95% CI: 1.10–1.37; P < 0.001) higher COVID-19 risk than the Q4 group, whereas the Q2 (P = 0.49) and Q3 (P = 0.11) groups showed no significant differences. In the ≥60-year-old group, the Q1, Q2, and Q3 groups showed 1.39-fold (OR: 1.39, 95% CI: 1.24–1.56; P < 0.001), 1.29-fold (OR: 1.29, 95% CI: 1.12–1.48; P < 0.001), and 1.14-fold (OR: 1.14, 95% CI: 1.01–1.30; P = 0.05) higher risks of COVID-19 than the Q4 group.

Table 4 presents the results of multivariable logistic regression models for in-hospital mortality in COVID-19 patients in South Korea. Regarding income levels in 2020, the Q1 (P = 0.30), Q2 (P = 0.53), and Q3 (P = 0.32) groups were not associated with in-hospital mortality among COVID-19 patients compared with that of the Q4 group. In addition, individuals with severe disabilities had a 2.82-fold higher hospital mortality risk than individuals without disabilities (OR: 2.82, 95% CI: 1.68–4.73; P < 0.001), whereas individuals with mild to moderate disabilities did not have a significantly higher mortality risk (P = 0.557).