Impact of comorbidity burden on mortality in patients with COVID-19: a retrospective analysis of the Korean health insurance database

We aimed to investigate the impact of comorbidity burden on mortality in patients with COVID-19. We analyzed the COVID-19 data from the nationwide health insurance claims of South Korea. Data on demographic characteristics, comorbidities, and mortality records of patients with COVID-19 were extracted from the database. The odds ratios of mortality according to comorbidities in patients with COVID-19 with and without adjustment for age and sex were calculated. The predictive value of the original Charlson comorbidity index (CCI) and the age-adjusted CCI (ACCI) for mortality in patients with COVID-19 were investigated using the receiver operating characteristic (ROC) curve analysis. Among 7,590 patients with COVID-19, 227 (3.0%) had died. After age and sex adjustment, hypertension, diabetes mellitus, congestive heart failure, dementia, chronic pulmonary disease, liver, renal, and cancer were signi�cant risk factors for mortality. The ROC curve analysis showed that ACCI threshold ≥ 4 yielded the best cut-off point for predicting mortality (area under the ROC 0.92; 95% CI, 0.91–0.94). Our study revealed multiple risk factors that were associated with mortality in patients with COVID-19. The high predictive power of the ACCI for mortality in our results could support the importance of old age and comorbidities in the severity of COVID-19.


Introduction
The coronavirus disease (COVID- 19), an infectious disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was rst reported in Wuhan, Hubei, China at the end of December 2019, 1 and has rapidly spread worldwide.On March 11, 2020, the World Health Organization (WHO) declared it as a pandemic. 2As of May 30, 2020, the global death toll from COVID-19 pandemic had passed the 340,000 mark according to the WHO.Nevertheless, clinical data to guide health care professionals or policy makers in their decision-making is still scarce.Against this backdrop, the government of the Republic of Korea decided to share the COVID-19 nationwide claims data for global research. 3tection of risk factors for mortality is an important component of the strategies for managing COVID-19.This information is all the more important at a time when the demand for critical care is surging and the resources for healthcare are limited. 4,5A recent case series suggested old age and comorbidities as risk factors for severity of COVID-19. 6-8However, there is still scarce information on how the combination of these risk factors affects the severity of COVID-19.We also thought that a simple but predictable model is needed for effective health care resource allocation for this public health emergency.
The Charlson comorbidity index (CCI) has been validated for predicting mortality in patients. 9CCI quanti es the risk of mortality associated with 19 weighted comorbidities, including congestive heart failure, cerebrovascular disease, chronic pulmonary disease, and diabetes, all of which have been reported as prognostic factors of poor outcome in patients with COVID-19. 9In addition to weighting comorbidities, Charlson and colleagues also proposed the age-adjusted CCI (ACCI) by adding 1 point for every decade after 40 years of age. 10 Considering the importance of aging and comorbidities in the severity of COVID-19, we expect that ACCI could predict the mortality rate for COVID-19.
In this study, we investigated the age-and sex-adjusted odds ratio (OR) of mortality for each comorbidity and the predictive value of mortality provided by the ACCI for patients with COVID-19 from a nationwide claims data in South Korea.Our results provide valuable information for the identi cation of patients at high risk of critical illness who might need early intensive care.
The associations between the patients' demographics, clinical characteristics, and death in the all-patient cohort, Daegu-Gyeongbuk region, are shown in Table 1.In the all-patients cohort, the most common comorbidity was hypertension (n=1463, 19.3%), followed by chronic pulmonary disease (n=958, 12.6%) and diabetes mellitus (n=907, 11.9%).The non-surviving patients were signi cantly older, predominantly male, covered by medical aid, residency in the Daegu-Gyeongbuk region, and more likely to have a history of all comorbidities.These characteristics were found in patients of the Daegu-Gyeongbuk cohort.The patients' demographics, clinical characteristics, and death by sex are shown in Supplemental Table S1.
Figure 2 shows the odds ratios (ORs) of mortality by comorbidities with 95% con dence intervals (CIs), with age-and sex-adjustment in the all-patient and the Daegu-Gyeongbuk cohort.37-10.31),and cancer (OR, 1.88; 95% CI, 1.17-3.02)were signi cant risk factors for mortality in patients with COVID-19 in the all-patient cohort.A similar result was observed in the Daegu-Gyeongbuk cohort.The signi cant risk factors in both male and female patients were hypertension, diabetes, and chronic pulmonary disease (Supplemental Figure S1).
In all patients with COVID-19, the receiver operating characteristic (ROC) curve analysis showed that an ACCI threshold ≥ 4 yielded the best cut-off point for predicting mortality (area under the ROC (AUROC); 0.92; 95% CI, 0.91-0.94)with a corresponding sensitivity of 86.8%, a speci city of 84.1%, a positive predictive value (PPV) of 14.4%, and an negative predictive value (NPV) of 99.5% (Table 2).The predicting performance of the ACCI was superior to that of the original CCI threshold ≥2 (P < 0.001) (Figure 3).In the Daegu-Gyeongbuk cohort, an ACCI threshold ≥ 4 also yielded the best cut-off point for predicting mortality (AUROC, 0.90; 95% CI, 0.88-0.91)with a corresponding sensitivity of 89.5%, a speci city of 78.2%, a PPV of 17.0%, and an NPV of 99.3%.The predicting performance of the ACCI in the Daegu-Gyeongbuk cohort was also superior to that of the original CCI threshold ≥2 (P < 0.001).
Predicting performances of the CCI and ACCI in mortality among patients with COVID-19, according to sex, are shown in Supplemental Table S2 and Figure S2.In male patients, an ACCI threshold ≥ 3 yielded the best cut-off point for predicting mortality.
Figure 4 shows the mortality rate according to the CCI and ACCI in the all-patients and the Daegu-Gyeongbuk cohort.The mortality rate showed an increasing trend with the ACCI.The distributions of the CCI and ACCI in the all-patients cohort and the Daegu-Gyeongbuk cohort are shown in Supplemental Table S3.In all-patient cohort, for those with an ACCI score of 4 or more, the mortality rate was 14.4% (197/1367), and for those with less than 4 points, the mortality rate was 0.5% (30/6223.In the Daegu-Gyeongbuk cohort, for those with an ACCI score of 4 or more, the mortality rate was 17.0% (180/1060), and for those with less than 4 points, the mortality rate was 0.7% (21/3174).Among the 433 nonhospitalized patients, a total of 9 patients had died and all except 1 patient had an ACCI score of 4 or more (Supplemental Table S4).

Discussion
In this study, we investigated the impact of comorbidity burden on mortality in patients with COVID-19 from a nationwide claims database.The main nding of our study was that comorbid hypertension, diabetes, congestive heart failure, chronic pulmonary disease, liver disease, renal disease, dementia, and cancer were identi ed as signi cant risk factors for mortality in patients with COVID-19 after age and sex adjustment.The predictive performance of the ACCI was superior to that of the CCI.ACCI ≥ 4 was found to be the optimal cut-off value for the prediction of death in patients with COVID-19.Our results could provide useful prognostic information to health care professionals, allowing the selection of patients in most need of medical attention and resources.
The mortality rate reported in our study was lower than that reported in studies conducted in other countries. 5,11According to the most recent reports from the WHO and the KCDC (last updated June 2, 2020), mortality rates in Europe, Americas and South-East Asia were 8.4, 5.7, and 2.8%, respectively.These differences may be explained by several factors.One possible explanation for the difference in mortality rates between different countries is the different clinical characteristics of the populations.The risk of COVID-19 mortality has been consistently reported to increase in patients with male sex, advanced age, and the presence of comorbidities, similar to the observations in our study.The patients in our study were relatively younger and had fewer comorbidities than those reported in studies from other countries. 5,11These characteristics might be associated with the COVID-19 outbreak in a relatively large number of young people in South Korea. 12,13Second, striking differences between Asian and European mortality might indicate the effect of ethnicity on disease outcome. 14However, because ethnicity is a complex entity composed of social constructs, cultural identity, genetic make-up, and behavioral patterns, 15 it might be di cult to conclude the association between ethnicity and disease outcome.In addition, differences in the organization of health care systems and strategies to contain COVID-19 among different countries may have affected the result.Korea's rapid and extensive diagnostic testing (more than 10,000 tests daily), and intensive anti-contagion policies may have contributed to the disease outcome. 16 this study, we adjusted both sex and age that could affect the prevalence of comorbidities to investigate its effects on the severity of COVID-19.Further, some studies have suggested that male sex is a risk factor for the severity of COVID-19. 17,18It has been suggested that the sex-based difference between the circulating angiotensin-converting enzyme (ACE)-2 levels, the receptor of which was associated with intracellular penetration of SARS-CoV-2, 19 or the smoking rate difference according to sex may have affected the sex difference on the severity of COVID-19. 18,20In addition, the potential association between androgen level and COVID-19 severity was suggested. 21Our results also revealed the tendency toward sex difference in the mortality of COVID-19.Further research is needed to assess the effect of sex on the severity of COVID-19.Age has consistently been reported to affect the severity of COVID-19 in several studies. 22,23In addition to the increased prevalence of comorbidities in older age, 24 physiological changes caused by aging itself may affect the severity of COVID-19.Aging leads to impaired function of various organs including immune system, resulting it greater susceptibility to in ammation or death. 25,26Therefore, we adjusted age and sex in each analysis to investigate the effects of various comorbidities on patient mortality due to COVID-19.
The results from our study revealed multiple risk factors that were associated with mortality in patients with COVID-19 after age-and sex-adjustment.SARS-CoV-2 binds to the target cells through the angiotensin-converting enzyme (ACE) 2 receptor expressed in epithelial cells of several organs. 19Because the expression of ACE2 is increased in patients with hypertension, diabetes, and chronic obstructive pulmonary disease (COPD), these comorbidities can increase both the risk and the severity of COVID-19 infection. 17,18In addition, evidence of myocardial or liver damage has been observed in patients with COVID-19, and pre-existing cardiovascular and liver diseases could be associated with the severity of COVID-19 infection. 29,30Recent meta-analyses have identi ed that cardiovascular diseases and COPD can greatly affect the severity of COVID-19. 31,324][35] The effects of each comorbidity on the COVID-19 mortality have been observed in our results as well, and if they are combined, the effect will be stronger on the severity of COVID-19.
Our study showed that the ACCI could be useful for predicting mortality in COVID-19 patients.A study of 52 critically ill patients with COVID-19 revealed that the median duration between the onset of symptoms and intensive care unit admission was 9-10 days, suggesting a gradual progression of the disease. 36herefore, the early detection of risk factors that can predict the severity of disease can improve the patient's prognosis and enable an e cient allocation of medical resources.To this end, we have created a simple but powerful prediction model for the mortality of COVID-19, combining age and comorbidities, known to be important risk factors for the severity of COVID-19.In addition, the high predictive power of the ACCI for mortality in our results could support the importance of old age and comorbidities in the severity of COVID-19.To date, several prognostic models for the severity of COVID-19 have been suggested. 37,38However, because most previous models require laboratory tests, they may be di cult to apply in areas where medical resources are restrictive.Further, the predictive value of ACCI in our study showed similarity to that of the recently reported clinical risk score that predicts the occurrence of critical illness in hospitalized patients with COVID-19 (development cohort: AUC 0.88 [95% CI 0.85-0.91],validation cohort: AUC 0.88 [95% CI 0.84-0.93]). 37The COVID-19 pandemic has created a shortage of medical resources, 39,40 and assessments that rely on laboratory tests may be limited in the COVID-19 crisis situation.On the other hand, we expect our results to be more widely used, as they can easily be applied by medical practitioners, health professionals, or policy-makers.
The results of our study should be interpreted cautiously for several reasons.Firstly, the data from insurance claims did not contain detailed clinical information such as vital signs or laboratory values.Secondly, the CCI does not consider the use of drugs and relies on diagnosis codes only; thus, over-or underestimation of the risk is likely to have happened.Thirdly, due to the different medical situations and resources for the COVID-19 crisis in each country, the generalizability of the results may be limited.
However, the contribution of this study is that it uses nationwide data to provide predictions on the risk of mortality, which is the most serious outcome of COVID-19 infection.Lastly, the original ACCI did not include hypertension, which is the most common comorbidity in patients with COVID-19.The development of a COVID-19-speci c comorbidity scoring system will be necessary.
In conclusion, our study identi ed that the ACCI, combined with age and various comorbidities, was associated with mortality in patients with COVID-19 in South Korea.If an increasing number of patients with COVID-19 develop severe illness, plans should be made at the national level to better manage the surge and ensure the need for critical care resources.Furthermore, because the availability of medical resources for critical care is likely to be restrictive, resource allocation policies based on risk factors should be implemented by medical professionals and policy makers.We hope our study ndings will provide important information to guide health care professionals facing the global health threat of COVID-19 in decision-making.

Methods
This study was reviewed and approved by the institutional review board (IRB) of Seoul National University Hospital (IRB No. E-2004-165-1119), and the requirement of informed consent was waived because the data did not contain any identi able information.All methods were performed in accordance with the approved guidelines and regulations. 41is retrospective cohort study analyzed data from the health insurance claims relevant to COVID-19 submitted to the Health Insurance Review and Assessment (HIRA) of South Korea by May 15, 2020.The HIRA currently provides the data on its website under the project OpenData4Covid19 (https://hira-covid19.net/). 3The current dataset includes a total of 234,427 patients who visited the hospital for the diagnosis of COVID-19 and their health insurance claims from January 2017 to May 15, 2020.The dataset was merged with the COVID-19 con rmation and mortality data from the Korea Centers for Disease Control and Prevention (KCDC).Given the obligatory nature of the National Health Insurance system (NHIS), our data covers virtually all Koreans (about 50 million people) and captures the clinical data from all healthcare institutions, i.e., clinics, pharmacies, and hospitals. 42Each healthcare institution must submit all patients' information regarding diagnosis, treatment, medical services rendered, and drug prescriptions to the NHIS, to receive reimbursement. 43 South Korea, COVID-19 diagnostic tests, mostly SARS-CoV-2 real-time polymerase chain reaction (RT-PCR), are applied to persons who have been in contact with COVID-19 positive patients or to those with symptoms suspicious of COVID-19. 44,45We de ned the con rmation of COVID-19 using the con rmation code provided by the HIRA based on the KCDC data. 3Information on the demographics of participants, type of insurance (health insurance and medical aid), and comorbidities was based on claims codes.Patients with comorbidities were de ned as those who had more than three claim records between January 2019 and before COVID-19 testing.This was done to avoid over-estimating the comorbidities and to include only those comorbidities that could more probably affect the patient's recent medical condition.The ACCI values were drawn from the claims data using the International Statistical Classi cation of Diseases and Related Health Problems, 10 th revision (ICD-10) coding algorithm proposed by Quan and colleagues. 46The algorithm was applied and validated for the national health insurance claims data in South Korea. 47,48The ACCI is a weighted measure that incorporates age into the original CCI. 10 In the ACCI, an additional point is added for each decade after 40 years of age (from 1 point for the age group 50-59 years to 4 points for the age group greater than 80 years old).Primary hypertension (ICD-10 code I10.x), which is not included in the CCI, was also identi ed by the abovementioned method.
The main outcome of our study was patient death due to COVID-19.We did not investigate the effects of the comorbidities on the infection rate or the hospitalization rate of COVID-19.We thought that not only medical conditions but also non-medical conditions such as the capacity of the medical resources or social factors in the area could have affected the infection rate or hospitalization rate of COVID-19.For example, the mass outbreaks in some speci c groups in South Korea occurred mainly in young age groups with active social activities. 12,49These occurrences may result in bias regarding the medical conditions of the patients with COVID-19.
We performed subgroup analysis for patients with COVID-19 in the Daegu and Gyeongbuk regions with a population of about ve million, where most con rmed cases of COVID-19 were reported in South Korea.In South Korea, there has been a surge in the number of con rmed COVID-19 cases in a religious sect gathering called the Shincheonji. 50This rapid spread has led to a shortage of hospital beds and healthcare professionals in this area. 51We thought the situation in this region better re ected the current crisis in other countries with high COVID-19 prevalence.In addition, we performed a comparative analysis between male and female to investigate the effect of sex on COVID-19 mortality.

Statistical analysis
The baseline characteristics are presented as mean ± standard deviation (SD) for continuous variables and as frequencies with percentages for categorical variables.Continuous variables were compared using unpaired Student's t-test, and categorical variables were compared using either the chi-square test or Fisher's exact test, as appropriate.We calculated the age-and sex-adjusted ORs of mortality by comorbidities included in the CCI, using the logistic regression model.For age adjustment, age was classi ed into ten-yearly age groups (0-9, 10-19, 20-29, 30-39, ...years) as a categorical variable.The ability of the CCI and ACCI to predict in-hospital mortality among all patients with COVID-19 or hospitalized patients with COVID-19 was determined using the ROC) curves and their area under the curve (AUC).To evaluate the predicting performance, we calculated the AUC, sensitivity, speci city, PPV, and NPV.DeLong's test was used to compare the ROC curves.All statistical tests were performed with SAS, version 9.4 (SAS Institute; Cary, NC), or R, version 3.6.

Table 2 .
Sensitivity and speci city of the Charlson comorbidity index (CCI) in predicting mortality among patients with coronavirus disease (COVID-19).PPV, positive predictive value; NPV, negative predictive value