Mortality of septic knee arthritis in Korea: risk factors analysis of a large national database

This study aimed to analyze the risk factors for mortality of septic knee arthritis in Korea through a large nationwide data research. The National Health Insurance Service-Health Screening database was used to analyze 89,120 hospitalizations for septic knee arthritis between 2005 and 2018. In-hospital, thirty-day, and ninety-day mortality, and their association with patient’s demographic factors, various comorbidities (i.e., cerebrovascular disease, congestive heart failure, and myocardial infarction) and Charlson Comorbidity Index (CCI) were assessed. Secondary outcomes of complications (osteomyelitis, knee arthroplasty, recurrence) were analyzed. The number of hospitalization with septic knee arthritis increased from 1847 cases in 2005 to 8749 cases in 2018. There was no significant difference in mortality after diagnosis of septic knee arthritis between years. The risk of mortality in patients who hospitalized with septic knee arthritis increased in comorbidities like Congestive heart failure, dementia, myocardial infarction, chronic kidney disease. Hazard ratio (HR) decreased in patients who have comorbidities such as rheumatoid arthritis, liver disease, rheumatologic disease. HR for mortality in septic knee arthritis increased in patients with CCI more than 1. The risk factors for mortality in all periods were male sex, old age, high CCI, comorbidities such as congestive heart failure, dementia, myocardial infarction, chronic kidney disease. Efforts to reduce mortality should be concentrate more on patients with these risk factors.

Factors affecting thirty-day mortality of septic knee arthritis. After diagnosis of septic knee arthritis, the total number of thirty-day mortality of septic knee arthritis were 953 cases. Table 4 shows the results of a series of univariable analyses using the Cox model. Old age over 60 years were demographic factors that increase the risk of thirty-day mortality of septic knee arthritis. Gender was not associated with thirty-day mortality of septic knee arthritis. HR for thirty-day mortality were Factors affecting ninety-day mortality of septic knee arthritis. The total numbers of ninety-day mortality of septic knee arthritis were 2403 cases. Table 5 shows the results of a series of univariable analyses using the Cox model. Among the patient's demographics, old age (over 50 years) was significantly associated with ninety-day mortality of septic knee arthritis but gender was not associated. HR for ninety-day mortality were 1.32 (95% CI = 1. 16 (Table 6).

Discussion
The National Health Insurance Service (National Health Insurance Service-HealthScreening; NHIS-HealS) based big data has great significance as a large-scale cohort study. This study is a large-scale big data study for about 15 years from 2005 to 2018, and through this, the overall prevalence and trend of septic knee arthritis in South Korea can be evaluated. It was possible to analyze the difference in the risk and mortality of each disease according to gender, age, and comorbidities. For each underlying disease, it is possible to determine the risk of complications that may occur after diagnosis of septic knee arthritis (Table 7). Although septic knee arthritis is not a common disease, it is one of the diseases that require emergency surgery or treatment 6 . The incidence of septic arthritis is about 4-10 per 100,000 patient-years per year 6,7,10,11 . Two recent nationwide studies from the United Kingdom and New Zealand reported that the incidence of septic arthritis is 6.7 and 12.2 per 100,000 14,15 . Two studies reported that the incidence of septic arthritis is increasing 14,15 . In our study, the number of cases diagnosed with septic knee arthritis increased from 1847 cases in 2005-8749 cases in 2018. Although this increase in incidence can be explained by increased reporting due to advances in diagnostic methods and advances in the medical system, it is also possible that the rate of septic arthritis has actually increased. Intra-articular infections are a rare but well-known complication of joint injections and surgery. An increase in procedures such as joint injection may be thought to be related to an increase in iatrogenic joint infection. In terms of gender, there were 39,912 male cases (44.93%) and 48, 919 female cases (55.07%), which were more common in females, and the severity and risk of the disease were also higher in female patients. In this study, HR for in-hospital mortality in female compared to male were 1.38 (95% CI = 1.25 to 1.52, p < 0.0001), respectively. Gender factor was also found to affect the severity and risk of septic knee arthritis. The other study reported that male patients had a 33% increased death risk 16 .
It is known as a disease that can lead to death if appropriate emergency treatment is not provided. In other studies, the mortality rate is reported to be 4-15%, the incidence of osteomyelitis is about 8%, and poor clinical results are reported in more than 20-30% 6,7,12,17,18 . But, the other nationwide study in Iceland reported lower mortality such as 2.7% 11 . In this study, between 2005 and 2018, in-hospital mortality was 2.01%, thirty-day mortality was 1.07%, and 90-day mortality was 2.70%. And there was no significant difference in mortality between 2005 and 2018.
In this study, the risk factors for mortality in all periods were old age, high CCI, comorbidities such as congestive heart failure, dementia, myocardial infarction, chronic kidney disease. And HR decreased in patients with www.nature.com/scientificreports/ liver disease and rheumatologic disease. The other studies also reported that, in terms of patient demographics, male sex and old age were reported as a risk factor for mortality after diagnosis of septic arthritis 16,17,[19][20][21] . As in this study, other institutions reported that coronary artery disease, cerebrovascular disease, chronic kidney disease, malignancy among comorbidities were risk factors for mortality after diagnosis of septic arthritis 16,17,20 .
In the case of elderly patients with these comorbidities, after diagnosis of septic knee arthritis, the mortality increases due to systemic infection and exacerbation of the inflammatory response. Diabetes was not statistically significant as a risk factor for mortality in this study, but in most other studies, it was reported as a risk factor for mortality 16,17,[19][20][21] .
In other studies, rheumatoid arthritis and liver disease were reported as risk factors for increasing mortality after diagnosis of septic arthritis, but in this study, liver disease, rheumatologic disease, and rheumatologic arthritis were associated with decreased HR for mortality 16,17,20,21 . Rheumatoid arthritis is a known risk factor for septic arthritis 6,22,23 . Although rheumatoid arthritis can be considered as risk factors for septic knee arthritis, it can be evaluated as better than idiopathic septic knee arthritis in terms of prognosis of septic knee arthritis. Rheumatoid arthritis is difficult to differentiate from septic arthritis. So it was treated as septic arthritis without being confirmed in culture and it is assumed that people who have those comorbidities were associated with decreased HR for mortality in this study.
There are some limitations to this study. First, the limitation of this study is that the study period (2005-2018) is relatively short and there are characteristics of the disease that require long-term follow-up, so it has the disadvantage that it does not reflect the overall natural history of the disease. Comparing the data from the 1990s and the 2000s, which achieved the development of drugs and treatments, would have been useful in analyzing whether these treatment policies and drugs were effective in preventing the aggravation of diseases. Before 2005, there was no data in the NHIS database, so there was a limit to the investigation for long term study period 13 . Second, since it is data based on the health insurance service billing data code, it is believed that there are many codes and data that are actually missing in the claim process of each hospital. In addition, there is a limitation that other factors that may affect the patient's knee joint function and condition, such as smoking, drinking alcohol, lifestyle, mental health, and the patient's condition before surgery, could not be observed or analyzed. Third, in most cases, antibiotic treatment is required for the treatment of septic arthritis. But in this report, it was investigated that some septic knee arthritis patients received no antibiotic treatment. This result is because some cases in which antibiotic treatment given to patients with septic arthritis were not covered by national health insurance service system. Due to the nature of big data research, it is impossible to determine the severity of each disease and the underlying disease can only be identified by whether or not the disease is listed in the code. Unlike other studies, diabetes mellitus was not associated with mortality. Additional study is needed in future.

Conclusion
We studied the trend and specific comorbidities associated with mortality of septic knee arthritis. Demographic factors for mortality of septic knee arthritis were old age. Patients with comorbidities such as congestive heart failure, dementia, myocardial infarction, and chronic kidney disease showed increased HR after diagnosis of septic knee arthritis. In the case of patients with these comorbidities, treatment should always be carried out with the possibility of exacerbation of other related complications, and management and monitoring of the patient's general condition should be essential.

Materials and methods
This study is retrospective cohort study using customized data provided by the National Health Insurance Service (National Health Insurance Service-HealthScreening; NHIS-HealS) 13 . This study used NHIS-NSC data made by National Health Insurance Service (NHIS). The Institutional Review Board of National Health Insurance Service Ilsan Hospital (NHIMC 2021-11-012) approved this retrospective Health Insurance Portability and Accountability Act-compliant cohort study and waived the informed consent from the participants, because this study was expected to present no or minimal risk of harm to the participants, and all the data used were anonymized 13 . www.nature.com/scientificreports/ All methods were performed in accordance with relevant guidelines and regulations 13 . The authors alone are responsible for the content and writing of the paper. In Korea there is an obligatory National Health Insurance system with universal coverage 13 . NHIS-HealS database has reimbursement records from all medical institutions in Korea 13 . Septic knee arthritis was investigated by their diagnostic codes in the NHIS-HealS database. Among patients diagnosed with septic knee arthritis (diagnostic codes : M0096), patients above 20 years of age who had been charged for hospitalization between January 1, 2005 and December 31, 2018 were included in the study 13 . In Korea, the diagnosis code M00.96 is often selected after confirming the knee joint fluid analysis before the bacterial culture result is confirmed, and other codes (M00.06, M00.16, M00.26, M00.86) are excluded. Patients who underwent total knee arthroplasty were excluded. Patients hospitalized for septic knee arthritis between 2002 and 2004 were also excluded. A total of 89,120 hospitalizations were included. In this case, we investigated all-cause mortality and cases of mortality were divided into groups according to the period from the day of hospitalization to the day of death (in-hospital mortality, thirty-day mortality, ninety-day mortality). Definition of 'in-hospital mortality' is the mortality cases within in-hospital period and definition of 'thirty-day mortality' is the mortality cases within 30 days after discharge, and definition of 'ninetyyear mortality' means death between the day of discharge to 90 days. This was to confirm the difference between the period of death after hospitalization and to determine whether the effect of comorbidity differs by the period from hospitalization to mortality 13 . Patient demographic factors including age and sex was investigated and comorbidities was collected using International Classification of Diseases 10 diagnostic codes reported in NHIS 24 . We calculated the Charlson Comorbidity Index (CCI) for each patient and divided CCI into four categories : 0, 1, 2 and ≥ 3, as originally proposed by Charlson et al 25 . The comorbidities included for the analysis were gout, rheumatoid arthritis, osteoarthritis, cerebrovascular disease, congestive heart failure, chronic pulmonary disease, hypertension, dementia, diabetes, liver disease, malignancy, myocardial infarction, peripheral vascular disease, rheumatologic disease, chronic kidney disease.
For all analyses, SAS Enterprise 7.1 (SAS Inc., Cary, NC, USA) was used. We performed conditional Cox proportional hazards model analyses of the comorbidities. Hazard ratios (HRs) and 95% confidence intervals (CIs) are presented. The level of significance was maintained at a P value < 0.05.