Risk Factors and Clinical Characteristics of Catheter Line-associated Bloodstream Infection in MIMIC IV Database

Background: Central line-associated bloodstream infection (CLABSI) is a common hospital infection. The increasing use of peripherally inserted central catheters and fully implanted venous ports, data on the epidemiological and clinical characteristics of CLABSI in the overall central venous catheter (CVC) population remain limited. The Medical Information Mart for Intensive Care (MIMIC IV) database is a free, open, and public resource research database. The purpose of this study was to describe the risk factors and clinical characteristics of CLABSI in MIMIC IV. Methods: A total of 31,116 patients were included in this study. General information, CVC-related information, comorbidity information, microbiological information, and antibiotic use information were extracted to describe and analyze the clinical characteristics of patients with CLABSI. Results: According to the occurrence of CLABSI, 31,116 patients were divided into the NO CLABSI group (n=30,395) and CLABSI group (n=721). The total indwelling duration of CVC was 439,239.6 days, The incidence of CLABSI is 2.32% and 1.64/1000 catheter days. The risk factors for CLABSI were the number of CVC type, duration of CVC, number of antibiotic type, duration of antibiotics, and femoral vein and internal jugular vein intubation. The in-hospital mortality of the CLABSI group was higher than that of the NO CLABSI group, but no statistical signicance was observed (P>0.05). Gram-positive (G+) cocci and Gram-negative (G-) bacilli accounted for 80% and 16.93% of positive bacteria in catheter culture. G+ cocci and G- bacilli accounted for 59.45% and 25.62% of positive bacteria in blood culture. Drugs with the highest resistance rates in catheter culture included penicillin G benzathine, oxacillin, and erythromycin. The most commonly used antibiotics for the treatment of CLABSI included vancomycin, cefepime, piperacillin tazobactam, and cefazolin. Conclusions: This study investigated independent risk factors for


Introduction
A central venous catheter (CVC) is a catheter inserted through the great vessels and peripheral vessels with the tip located in the superior or inferior vena cava. Some of the available CVCs include double-lumen catheters, triple-lumen catheters, dialysis catheters, peripherally inserted central catheters (PICCs), and totally implantable venous infusion port (PORT). CVCs are one of the indispensable tools in modern medical practice. However, the use of CVCs may lead to catheter-related infection, which has an important impact on the prognosis of patients.
The mean incidence of central line-associated bloodstream infection (CLABSI) in acute care hospital units in the United States ranges from 0 to 2.9/1000 catheter days [1]. Studies have shown that compared with conventional CVCs, PICCs have no signi cant difference in CLABSI rate in ICU patients [2], but the rate is signi cantly reduced in onco-hematologic patients. The incidence of CLABSI in cancer patients with PORT is much lower than that of patients with PICC [3]. Between 2008 and 2014, the incidence of CLABSI decreased by 50% [4], but 30,000 new cases of CLABSI occur in the United States annually [5].
Previous studies on CLABSI in intensive care units focused on conventional double-lumen catheters, triplelumen catheters, and dialysis catheters. With the increasing application of PICCs and PORT, the overall situation of CLABSI may change. In this study, we analyzed data from the Medical Information Mart for Intensive Care (MIMIC) IV database to understand the epidemiology and microbiological distribution of CLABSI in patients with indwelling CVC during hospitalization and to explore potential association hypotheses.

Data sources
The MIMIC IV database contains medical records of patients admitted to Beth Israel Deaconess Medical Center (BIDMC) between 2008 and 2019. The database contains detailed information about in-hospital patients, including International Classi cation of Diseases, Revision 9 (ICD-9); Clinical Revision (ICD-9-CM) codes; laboratory data; vital signs; drug administration; and mortality data [6]. The database also provides disease severity scores on the rst day of each ICU admission [7].

Variable selection
In this study, we queried the derived table INVASIVA_LINE of the database system and selected all patients with indwelling catheter (n = 44,979). The exclusion criteria were as follows: 1. length of hospital stay (Los hospital) < 24 h (n = 856), 2. patients younger than 18 years old (n = 0), and 3. patients with other tubes inserted (n = 13,007), as shown in Fig. 1. CVC information was extracted on the basis of patient ID (intubation location, duration of indwelling catheter, number of tubes inserted), demographic characteristics (gender, age, weight, race, and marital status), and six different disease severity scores on the rst day of admission. The disease severity scores included Oxford Acute Disease Severity Score, Sequential Organ Failure Assessment (SOFA), Simpli ed Acute Physiology Score, Simpli ed Acute Physiology Score II, Acute Physiology Score III (APSIII), and Logical Organ Dysfunction Score. The database also queried comorbidities and hospitalizations (Los ICU, Los hospital, and discharge outcome) for each patient. The type and course of antibiotics and microbiological and drug sensitivity results of blood culture and catheter culture were determined during hospitalization.
The outcome of the event was the de nitive diagnosis of CLABSI and in-hospital mortality. Patients with CLABSI were identi ed using ICD-9 diagnostic codes 999.31 and 999.32.
The project was approved by the institutional review boards of Massachusetts Institute of Technology (MIT) and BIDMC, which was granted a waiver of informed consent.

Data processing
The missing values in this study were less than 20%. The missing data were lled by multiple inference method. Ten estimated datasets were obtained using the "mice" package of R software. The classi cation variables were described as frequency and percentage values, and the differences between queues were determined by chi-square or Fisher exact test. Continuous variables were tested by Shapiro-Wilk test to verify whether they conform to normal distribution. Continuous variables were described as values in the range of mean and standard deviation values or median and quartile values, depending on whether they conform to a normal distribution.
Logistic regression was used to determine the independent risk factors for CLABSI in CVC patients. Positive LN stepwise regression was used to screen variables. The identi ed independent prognostic factors were analyzed again using logistic regression models, and the results were expressed as odds ratios (ORs) and 95% con dence intervals (CIs). The outcome of in-hospital mortality and Los hospital were analyzed by proportional hazard Cox regression analysis, and the original hazard ratio (HR) of patients with different CVC types was determined by CLABSI correlation variables. The probability of α error less than 5% (P < 0.05) was considered statistically signi cant. The calculated con dence interval was 95%. R (4.0.3) was used for statistical analysis, and P < 0.05 was considered statistically signi cant.

Baseline data
According to the occurrence of CLABSI, 31,116 patients were divided into the NO CLABSI group (n=30,395) and CLABSI group (n=721). The total catheter duration was 439,239.6 days, with an incidence of 2.32% and 1.64/1000 catheter days. Patients in the CLABSI group were younger, with a median age of 62 years, lower urine output on the rst day of admission, higher SOFA, and APS III scores, and higher infection rates in all races except whites. Moreover, they had higher congestive heart failure, peptic ulcer disease, mild liver disease, diabetes with complexcomplications, renal failure, and malignant cancer; use of more number of antibiotic types; longer duration of antibiotic use; more use of RRT; and longer duration of CVC (P<0.05). The use of indwelling CVC in the femoral vein was signi cantly different between the two groups (P<0.001). The proportion of CLABSI group increased with the number of CVC, and the difference between groups was statistically signi cant (P<0.001). The CLABSI group had higher in-hospital mortality, but the Los ICU, Los hospital, and in-hospital mortality were not signi cantly different between the two groups (P<0.05).   (Fig. 2). The top three G-bacilli were Pseudomonas aeruginosa (20.3%), Serratia (18.6%), and Enterobacter aerogenes (17.34%) (Fig. 3). Fungal infection was mainly caused by yeast (Table S1). Among the positive bacteria in blood culture, G + cocci accounted for 59.45%, while G-bacilli accounted for 25.62% (Table 4). The top three G + bacteria were Staphylococcus (49.98%), Enterococcus (18.33%), and Staphylococcus aureus (17.40%) (Fig. 4). The top three G-bacteria were Escherichia coli (31.91%), Klebsiella acidophilus (20.36%), and Pseudomonas aeruginosa (8.05%) (Fig. 5). Fungal infection was dominated by Candida albicans (Table S2). For more detailed distribution of microorganisms, see Tables S1 and S2.  Fig. 9. The most commonly used antibiotics for the treatment of indwelling CVC patients were vancomycin, cefepime, piperacillin tazobactam, and cefazolin, as shown in Fig. 10. The distribution of major antibiotic treatment was compared between the CLABSI group and the NO CLABSI group, with no statistically signi cant difference (P < 0.05), as shown in Table S3.

Discussion
Page 15/29 CVC provides effective venous access for patients in critical condition and those with renal failure and cancer.
It improves the success rate of resuscitation, alleviates the pain of patients, and prolongs their survival time.
The International Nosocomial Infection Control Consortium (INICC) reported CLABSI is 4.9/1000 catheter days [1,2]. The incidence of CLABSI in China is 5-15% and 2.9-11.3/1000 catheter days [8]. One study in Argentina found a 5.4% incidence of CLABSI and an infection rate of 8.7/1000 catheter days [9]. In this study, the incidence of CLABSI was 2.32%, and the infection rate of indwelling catheters was 1.64/1000 catheter days, which was consistent with the results of CLABSI epidemiological studies in the United States.
The patients with CLABSI in this study were younger and had severer conditions. Dialysis catheter is used for patients with renal failure requiring alternative therapy. Conventional CVC is more commonly used for monitoring treatment of cardiac failure. PICC and PORT are more commonly used for patients with malignant tumors, in which the indwelling time is long. In this study, age, severity, and comorbidities were included in the multivariate analysis, and the results were consistent with the univariate results. The number of CVC type and the duration of CVC were correlated with the incidence of CLABSI. A previous study showed that reducing the number of CVC and the duration of indwelling catheter can reduce the incidence of CLABSI [10].Parameswaran et al [11] found that the rate of femoral vein, intrajugular vein, and subclavian vein catheter-related infections was 33.3%, 22.2%, and 21.3%, respectively.Lorente et al [12] analyzed 2595 patients found that the femoral vein had a higher risk of CRBSI than the internal jugular vein and the subclavar vein.This study also con rmed a higher risk of CRBSI in the femoral vein.
Once CLABSI occurs, it requires more intensive antibiotic treatment, so it is also correlated with the number of antibiotic types and the duration of antibiotic treatment. Lee JH et al. also con rmed that antibiotic use was an independent risk factor for CLABSI [13]. Huerta et al. also found a signi cant association between duration of antibiotic use and early mortality and relapse [14].
Many studies have shown that CLABSI can signi cantly increase mortality and prolong hospital stay [15].
However, this study found an increase in in-hospital mortality in the CLABSI group, which was not statistically The study has con rmed that the incidence of CLABSI caused by PICC is signi cantly lower than that of other types of CVC [19]. The in-hospital mortality of PICC CLABSI and PORT CLABSI has not been well studied, thus affecting the overall mortality. CVC is increasingly maintained outside the hospital, and out-of-hospital CLABSI needs further study.
In this study, the positive rate of catheter culture was signi cantly higher than that of blood culture. The six most common microorganisms in catheter culture were Staphylococcus epidermidis, Staphylococcus aureus coag, Pseudomonas aeruginosa, Serratia, and Enterobacter aerogenes. G + cocci accounted for 80.48%, followed by G-bacilli and fungi. Fungi were mainly yeast. A number of studies in China showed that the most common microbe was Acinetobacter, followed by Staphylococcus epidermidis and Candida albicans [20]. A study in Spain showed that G + cocci accounted for the majority, followed by G-bacilli and yeasts [21], which was consistent with our results. According to CLABSI data from the Centers for Disease Control and Prevention and the National Healthcare Safety Network, fungal infections are high in adult ICUs, and Enterobacteriaceae and fungi are increasing in oncology wards [22]. Due to the different prevention and treatment plans of hospitals in different regions, the distribution of pathogenic microorganisms may be different. Iatrogenic factors may greatly in uence catheter microbial colonization and catheter-associated infections. Hospital managers need to strengthen environmental and human monitoring and nosocomial infection control.
The distribution of microorganisms in blood culture was not consistent with that in catheter culture. The proportion of G + cocci decreased, while those of G-bacilli and fungi increased. Staphylococcus, Enterococcus, Staphylococcus aureus coag, Escherichia coli, Candida albicans, and intestinal bacteria increased signi cantly. Enterogenic bacteria enter the blood after intestinal barrier destruction, which is an important mechanism of blood-borne infection [23]. The proportion of Candida albicans in fungi increased.
Although yeast prevalence has geographical differences, the overall level of yeast prevalence is increasing globally [24,25]. In addition, repeated exposure to broad-spectrum antibiotics, complex surgical procedures, glucocorticoids, and other factors increases the risk of fungal infections, especially Candida [26].
The three most commonly monitored antibiotics in the both cultures included gentamicin, erythromycin, vancomycin, The ve antibiotics with the highest drug resistance in the both cultures included penicillin G, oxacillin, erythromycin, levo oxacin, and ampicillin. Three kinds of sensitive antibiotics were found: daptomycin, linezolid, and carbapenems (meropenem, imipenem), which are all third-line antibiotics. Other sensitive antibiotics included enzyme inhibitors (piperacillin tazobactam), sulfa (trimethoprim sulfamethoxazole), rifampin (anti-tuberculosis), and amikacin. In the NO CLABSI group and CLABSI group, the distribution of antibiotics used for treatment was consistent, and the difference was not statistically signi cant. The overall distribution showed that the most commonly used antibiotics were vancomycin, cefepime, piperacillin tazobactam, and cefazolin. The sensitivity of blood culture was 62-82%, and that of catheter culture was 51-91%, with vancomycin and cefazolin having the highest and lowest sensitivities.
Studies have shown that on the basis of etiological predictions, early source control, when initiated with antimicrobial therapy, improves prognosis [27]. Antibiotics are often used empirically at the beginning, advocating broad spectrum and high e ciency, and replaced with narrow-spectrum and high-sensitivity drugs according to etiology and drug sensitivity results. To prevent the development of resistance, antibiotic management needs to be followed [28]. At the same time, economic factors should be taken into account.
This study showed that the use of antibiotics basically conformed to the results of drug sensitivity. Currently, research on the application of antibiotics for the treatment of CLABSI is insu cient.

Limitations
This was a single-center retrospective study. The location of catheter implantation was not recorded in some patients. The diagnosis time of CLABSI was unclear. We could not determine which catheter was the cause of CLABSI when multiple indwelling CVCs were used. All of the above factors can lead to selection bias and experimental errors.

Conclusions
In this study, the risk factors of CLABSI were con rmed by analyzing the data of CVC patients in the MIMIC IV database from 2008 to 2019. The in-hospital mortality of the CLABSI group was higher than that of the NO CLABSI group, but no statistical signi cance was observed. The etiological distribution of catheter culture and blood culture was analyzed, and the distribution of antibiotics for clinical treatment was further analyzed on the basis of drug sensitivity results.