Impact of coronavirus disease-2019 on chronic respiratory disease in South Korea: an NHIS COVID-19 database cohort study

This population-based cohort study was conducted according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines [13]. The study protocol was approved by the institutional review board of Seoul National University Bundang Hospital (X-2004–604-905) and the Health Insurance Review and Assessment Service (NHIS-2020–1-291). Informed consent was waived because data analyses were performed retrospectively using anonymized data derived from the National Health Insurance Service (NHIS) in South Korea.

The NHIS-COVID-19 cohort database was developed for medical research in cooperation between the NHIS and Korea Centers for Disease Control and Prevention (KCDC). The KCDC provides information on patients who were diagnosed with COVID-19 from 1 January 2020 to 4 June 2020, such as the confirmation date of COVID-19, the results of treatment, and the demographic information. However, NHIS and KCDC did not provide the data regarding patients with COVID-19 who were undergoing treatment in the hospital as of June 26th 2020, as their treatment results have not yet been determined. By using this information of patients with COVID-19, the NHIS extracted the control population using stratification methods with regard to age, sex, and residence in February 2020. In the NHIS-COVID-19 cohort database, all disease diagnoses by the International Classification of Diseases (ICD)-10 codes, and prescription information concerning drugs and/or procedures from 2015 to 2020 were included. An independent medical record technician at the NHIS center who was unaffiliated with this study performed data extraction on 26 June 2020. In this NHIS-COVID-19 cohort database, individuals who were ≥ 20 years old were included in the study. In South Korea, patients who were diagnosed with COVID-19 were admitted to the hospital if they had severe symptoms such as pneumonia. However, if they had mild or no symptoms, they were isolated and closely monitored in certain government-managed centers. In addition, the KCDC tested all individuals for COVID-19 in South Korea, who had either direct or indirect contact with COVID-19 patients in the community or hospital.

The following diseases were considered as CRDs and the patient data extracted for this study included: COPD (J44*), asthma (J45*), interstitial lung disease (ILD, J84.9), lung cancer (C34*), lung disease due to external agent (J60-J70), obstructive sleep apnea (OSA) (G47.33), and tuberculosis of the lungs (TB, A15). The ICD-10 codes from 2015 to 2019 were used to evaluate CRDs in the study population. The study groups included the CRD group with individuals who were diagnosed with any type of CRD and the control group included the other individuals without any respiratory illnesses. In South Korea, as sole public insurance coverage, the CRDs should be registered in the NHIS database after diagnosis by physicians to receive financial coverage for treatment.

The primary endpoint of this study was the diagnosis of COVID-19, and it was evaluated from 1 January 2020 to 26 June 2020. The secondary endpoint of this study was hospital mortality among patients who were diagnosed with COVID-19.

Additional data collected included (1) demographic information (age and sex), (2) place of residence (Seoul, Gyeonggi-do, Daegu, Gyeongsangbuk-do, and Other areas), (3) underlying disability, (4) income level in 2020, and (5) the Charlson comorbidity index (CCI), which was calculated based on registered ICD-10 diagnostic codes (Additional File 1) from 1 January 2019 to 31 December 2019. The income level was divided into four groups using quartile ratio, and the patients were divided into seven age groups (20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and ≥ 80 years). The annual income level of all individuals in South Korea is registered to determine yearly NHIS premiums. For this study, income levels were classified into quartile groups. In addition, in South Korea, the total disability of all individuals should be registered in the NHIS database to receive various benefits; it includes physical and brain lesion disabilities; visual disturbance; hearing and speech disabilities; autism; intellectual, mental, renal, heart, and respiratory disorders; hepatopathy; facial disfigurement; intestinal and urinary fistulae; and epilepsy.

The baseline characteristics of all individuals in this study were presented as percentages for categorical variables and mean values with standard deviations for continuous variables. Comparison of characteristics between the CRD and control groups was performed using the t-test for continuous variables and the chi-squared test for categorical variables. We constructed a multivariable logistic regression model to investigate whether CRD was associated with the progression of COVID-19 in South Korea, and it was defined as multivariable model 1. All covariates were included in the model for multivariable adjustment, however, the CCI was included in the other model to avoid multi-collinearity with other underlying diseases that were used for the CCI calculation. In addition, as a sensitivity analysis, we divided the CRD group into seven disease groups and included them in the multivariable logistic regression model for the analysis of the progression of COVID-19 in South Korea, and it was defined as multivariable model 2. The CCI was included in the multivariable model 2 for adjustment, while other underlying diseases that were used for the CCI calculation were not included in the multivariable model 2.

Further, we developed a multivariable logistic model for hospital mortality among patients who were diagnosed with COVID-19 to investigate whether underlying CRD affected hospital mortality, compared to the control group. The two multivariable models (model 1 and model 2) were also constructed to investigate the association of a CRD and seven disease type of CRD with hospital mortality, separately. Hosmer–Lemeshow statistics was used to confirm the goodness of fit of multivariable models as P > 0.05, and it was confirmed that there was no multicollinearity in all multivariable models of the entire cohort with a variance inflation factor of < 2.0. The results of the logistic regression models were presented as odds ratios (ORs) with 95% confidence intervals (CIs). A receiver operator characteristic (ROC) analysis was performed to validate the use of logistic regression analysis for this study. R software (version 3.6.3; R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses, and P < 0.05 was considered statistically significant.

The cohort constituted 129,120 individuals, 4790 of whom were excluded (aged < 20 years). The final analysis included 122,040 individuals; 7669 (6.3%) were confirmed as COVID-19 cases until June 4th 2020. Among the 7669 patients with COVID-19, 251 (3.2%) passed away during hospitalization (Fig. 1). The baseline characteristics of all individuals in the NHIS-COVID-19 cohort are presented in Table 1. There was no missing data in the NHIS-COVID-19 database, except for annual income level in 2121 individuals (1.7%). However, they are not excluded in the analysis, and their income level was included as unknown to avoid bias from excluding them. A total of 36,365 individuals were diagnosed with CRD between 2015 and 2019: COPD (4488, 3.6%), asthma (33,858, 27.2%), ILD (421, 0.3%), lung cancer (769, 0.6%), lung disease due to external agents (437, 0.4%), OSA (550, 0.4%), and TB (608, 0.5%). The results of the comparison of characteristics between the CRD and control groups are presented in Table 2. The CCI in the CRD group was higher than control group (mean value: 2.7 ± 3.3 in CRD group vs 1.5 ± 2.6 in control group; P < 0.001).

Table 3 shows the results of the multivariable logistic regression model for the progression of COVID-19 in South Korea. In multivariable model 1, the CRD group was not associated with the incidence of COVID-19 compared with the control group (OR 1.04, 95% CI 0.99–1.09; P = 0.156). However, in multivariable model 2, the patients with ILD or OSA were associated with 1.63-fold (OR 1.63, 95% CI 1.17–2.26; P = 0.004) and 1.65-fold (OR 1.65, 95% CI 1.23–2.16; P < 0.001) higher incidence of COVID-19 than the control group. Patients with other CRD, such as asthma (P = 0.464), lung cancer (P = 0.533), lung disease due to external agents (P = 0.061), and TB (P = 0.372) were not associated with the incidence of COVID-19. Hosmer–Lemeshow statistics showed that the goodness of fit was appropriate in the models (P > 0.05), and the area under curve (AUC) of the multivariable models in ROC analyses was 0.81 (95% CI 0.80–0.81).

Table 4 shows the results of the multivariable logistic regression model for hospital mortality of COVID-19 patients. In multivariable model 1, the CRD group was not associated with hospital mortality in COVID-19 patients compared with the control group (OR, 1.19; 95% CI, 0.86–1.64; P = 0.299). However, in multivariable model 2, the patients with COPD and lung disease due to external agents showed 1.56-fold (OR 1.56, 95% CI 1.06–2.2; P = 0.024) and 3.54-fold (OR 3.54, 95% CI 1.70–7.38; P < 0.001) higher risk of hospital mortality of COVID-19 patients compared with the control group. Hosmer–Lemeshow statistics showed that the goodness of fit was appropriate in the models (P > 0.05), and the AUC of the multivariable models in ROC analyses was 0.83 (95% CI 0.82–0.83).