Risk factors correlated with antibiotic-induced carbapenem resistant in Gram-negative bacilli in China: a retrospective cohort study

Background: Increasing resistance to carbapenem, particularly common in Gram-negative bacilli (GNB), has become a growing public health concern around the world. The objective of this study was to investigate risk factors associated with antibiotic-induced carbapenem resistant in Gram-negative bacilli (CR-GNB) among inpatients. Methods: A retrospective cohort study was conducted in one of the largest tertiary A-level hospitals including patients with GNB cultured from any of the clinical specimens who had been admitted for more than 2 calendar days from January 2017 to June 2019. Kaplan-Meier analysis and Cox proportional hazard model were used to estimate the hazard of CR-GNB induction by antibiotics. Results: 2490 patients including 7 cohorts were included. After cox proportional risk model analysis, carbapenems, β-lactamase inhibitors, and cephalosporins had significantly higher hazards than other types of antimicrobial (P<0.001). But even without using any antimicrobials, the hazard would increase with the length of hospital stay. On multivariate analysis, carbapenem was the most principal hazard factor for antibiotic-induced CR-GNB (hazard ratio HR, 2.968; 95% confidence interval CI, 1.7065.162), followed by ICU admission (HR, 1.815; 95% CI, 1.507 2.186), cephalosporin (HR, 1.605; 95% CI, 1.2881.999), tracheotomy (HR, 1.563; 95% CI, 1.251 1.952) and β-lactamase inhibitor (HR, 1.542; 95% CI, 1.2371.921). However, quinolone effects on antibiotic-induced CR-GNB were not statistically significant. Conclusions: Prior carbapenem was a strong risk factor for antibiotic-induced CR-GNB, but quinolone was not associated with that. Rational use of carbapenems should be implemented and antimicrobial stewardship policies should be adjusted according to the characteristics of each hospital. principal diagnosis, tumor, liver failure, kidney failure, heart failure, respiratory failure, diabetes, hypertension, chronic obstructive pulmonary disease, hemodialysis, venous catheterization, ventilation, urinary catheterization, tracheotomy, surgery, ICU admission during current or previous episodes.

3 past years, posing a challenge for clinical antimicrobial chemotherapy and hospital infection control [6,7].
Rational use of antibiotics can effectively treat bacterial infections and reduce the burden to patients.
But unreasonable use of antibiotics will increase the selective pressure of antimicrobials commonly used, which is one of the important factors leading to antimicrobial resistance (AMR), and some evidence showed that increased use of antibiotics could lead to the emergence of AMR [11][12][13][14][15][16][17].
However, most related studies reported the significant relationship between carbapenem consumption and carbapenem-resistant gram-negative bacilli (CR-GNB), and whether the use of other antibiotics contributes to CR-GNB is not yet certain. In addition, most of previous studies have focused on a certain type of CR-GNB, such as carbapenem-resistant P. aeruginosa (CRPA), while research on the whole CR-GNB is relatively insufficient. Accordingly, we conducted this study to assess the risk factors for isolation of antibiotic-induced CR-GNB in a large cohort.

Design and subjects
A retrospective cohort study was performed in one of the largest tertiary A-level hospitals in Sichuan Province, China from January 2017 through June 2019. The cohort study included all patients with gram negative bacilli (GNB) cultured from any of the clinical specimens who had been admitted for more than 2 calendar days. This study was approved by the Ethics Committee of Sichuan Academy of Medical Sciences and Sichuan Provincial People's Hospital. The Review Board exempted requirement for informed consent because of the retrospective study and absence negative impact on the patients.
Cohort-Carbapenems: All patients with clinical samples positive for GNB who have only used carbapenem antibiotics before GNB was isolated.
Cohort-β-lactamase inhibitor: All patients with clinical samples positive for GNB who have only used βlactamase inhibitor antibiotics before GNB was isolated.
Cohort-Antifungal: All patients with clinical samples positive for GNB who have only used antifungals before GNB was isolated.
Cohort-Quinolone: All patients with clinical samples positive for GNB who have only used quinolone antibiotics before GNB was isolated.
Cohort-Penicillin: All patients with clinical samples positive for GNB who have only used penicillins antibiotic before GNB was isolated.
Cohort-Cephalosporin: All patients with clinical samples positive for GNB who have only used cephalosporin antibiotics before GNB was isolated.
Cohort-Unused: All patients with clinical samples positive for GNB who have not used any antibiotic before GNB was isolated.

Outcomes and definition
Antibiotic-induced CR-GNB was the primary outcome of this study.
Carbapenem resistant Gram-Negative Bacilli (CR-GNB): Carbapenem-resistance in GNB was defined as being non susceptible to imipenem, meropenem, ertapenem or doripenem. Species identification and in vitro susceptibility to imipenem, meropenem, ertapenem and doripenem were determined Antibiotic-induced CR-GNB: CR-GNB isolated after 2 calendar days of antimicrobial use.

Statistical analysis
Statistical analysis of the data was performed using STATA 12.0. Data were summarized using the mean and standard deviation (SD) for normally distributed variables. Categorical variables were expressed in absolute numbers and percentages.
We used both univariate and multivariate methods to analyze the data. Binary outcomes were tested using the χ2 test, and continuous data were compared using the T-test. Kaplan-Meier analysis and Cox proportional hazard model were used to estimate the hazard of CR-GNB induction by antibiotics.
P values below 0.05 were considered significant.

Patient inclusion
From January 1, 2017 to June 30, 2019, a total of 8748 inpatients with GNB isolated from any of the clinical specimens who were admitted, including 2587 patients with CRO isolated and 5417 patients with CSO isolated. After screening, 2490 patients including 7 cohorts were included in the final data set, as Fig. 1 shows.

Factors associated with antibiotic-induced CR-GNB in univariate analysis
Differences in antibiotic-induced CR-GNB effects between the 7 cohorts of antimicrobial drugs were statistically significant (, P < 0.001). In addition to antibacterial drugs, the effects of other factors on antibiotic-induced CR-GNB were also statistically significant, including gender, principal diagnosis, chronic obstructive pulmonary disease, venous catheterization, mechanical ventilation, urinary catheterization, tracheotomy, blood transfusion or use of blood products, surgery, kidney failure, respiratory failure, ICU admission now or in the past and community infections. Table 1 shows these details.

Kaplan-Meier analysis
6 Comparison of 7 cohorts showed that carbapenem, β-lactamase inhibitor, and cephalosporin had significantly higher hazards than any other type of antimicrobials. Even without the use of any antimicrobials, the hazard would increase with the length of hospital stay. Figure 2 shows these details. Log-rank test results showed that the differences were statistically significant (chi-square = 76.190, P < 0.001).

Discussion
Carbapenems are the most effective drugs for the treatment of severe infections with gram-negative bacteria due to their broad antimicrobial spectrum and high stability for hydrolysis by most βlactamases, including extended-spectrum β-lactamases (ESBLs) and AmpC cephalosporinases [22].
But unreasonable use of antibiotics leads to the emergence of CR-GNB, which is becoming more and more serious, and the treatment of some drug-resistant infections is extremely limited [8,23,24]. Therefore, understanding the hazard factors for antibiotic-induced CR-GNB is very important in the early selection of empirical antibiotic program. We demonstrated that previous carbapenem exposure was the major risk factor for antibiotic-induced CR-GNB in our study population.
During the period, cephalosporins were the most commonly used antibiotics, followed by β-lactamase inhibitors and penicillins. Unlike this, physicians in Europe preferred to prescribe more penicillins than 7 cephalosporins [25]. The difference might be explained by that most antibiotics were used for inpatient treatment in China while outpatient institutions accounted for the majority of antibiotic consumption in foreign countries [26][27][28]. The clinical application of carbapenem, as a special class antibiotic, is limited and requires pre-authorization during the using period in our country.
In this study, carbapenem, β-lactamase inhibitor, and cephalosporin had significantly higher hazards than other types of antimicrobials. Our finding that carbapenem use was the most principal hazard factor associated with antibiotic-induced CR-GNB was in accordance with some previous studies [5,[29][30][31][32][33][34][35]. The use of carbapenem may promote the production of carbapenemase, such as K pneumoniae carbapenemase and metallo-β-lactamases, which could increase carbapenem-resistant Escherichia coli (CRE) [36]. Other mechanisms of carbapenem resistance include outer membrane porin expression loss combined with extended-spectrum β-lactamase (ESBL) and AmpC enzyme, change of antimicrobial target and high expression of efflux pump [37][38][39]. This suggested that controlling the use of carbapenems could slow down the production of CR-GNB. However, unlike some studies, we did not find that previous use of quinolones was significantly correlated with antibioticinduced CR-GNB [29,34,35,[40][41][42][43][44][45]. It was speculated that quinolones might induce the high expression of efflux pump and lead to the multidrug-resistant phenotype, probably CR-GNB [46-48].
The reason for this difference might be that our design or study population was different from previous studies. Another large-scale research in China showed that fluoroquinolone consumption was not associated with CR-GNB [5]. In one way, this may be a feature of our country, and could provide basis for our choice of anti-infective therapy. Although using antibiotic was an important risk factor for drug resistance, even without the use of any antimicrobials the hazard would increase with the length of hospital stay after cox model analysis. The result might be explained by that these patients with longer hospital stay had prolonged exposure to invasive devices or use of antibiotics.
ICU stay and tracheotomy were also important hazard factors associated with antibiotic-induced CR-GNB. ICU is considered as the main source of development of multidrug-resistant bacteria and transmission due to the extremely critically ill patients, the use of invasive devices and higher intensity of selection pressure by broad-spectrum antibiotics. When patients receive tracheotomy and 8 ventilation, the normal upper respiratory barrier will be destroyed and the pipe will be gradually polluted by bacteria which will increase chance of bacterial invasion. Our results showed that the risk of antibiotic-induced CR-GNB isolation was 1.563 times higher in patients receiving tracheotomy than in patients without tracheotomy. This indicated that some preventive and control measures related to the use of tracheotomy, such as understanding the indication of tracheotomy, operation procedure and hand hygiene, could effectively reduce the CR-GNB in our hospital.
In our hospital, infection control measures against resistant bacteria are relatively comprehensive, including training for prevention and control methods for bacteria, active surveillance of the most common multidrug resistant bacteria (MDRO), computer system alerts for MDRO species, active screening for high risk inpatients, supervision of adherence to precaution measures and so on, to reduce the spread of MDROs in hospital.
There are several potential limitations in our study. Firstly, this study was a retrospective design conducted in a large tertiary A-level hospital, not a multicenter research. Secondly, we did not take into account the dose of the antibiotic and antibiotic combinations. Thirdly, the generation of CR-GNB could be varied by the practice of infection control and other risk factors, but we did not take the former into account. Despite these limitations, we believe that the main advantage of our research is its large cohort. Therefore, our research can at least provide reference value for the use of antibacterial drugs.

Conclusion
In conclusion, the use of carbapenem was strongly associated with antibiotic-induced CR-GNB in our study. Other antibiotics including cephalosporin and β-lactamase inhibitor were also related with CR-GNB. However, we did not find significant correlation between quinolones and antibiotic-induced CR-GNB. The findings will be useful for directing antimicrobial stewardship policies. Rational use of antibiotics and infection control measures are urgently needed to reduce the selective pressure of antibiotics, delay the occurrence of CR-GNB and control its cross transmission.

Ethics approval and consent to participate
This study was approved by the Ethics Committee of Sichuan Academy of Medical Sciences and Sichuan Provincial People's Hospital. The Review Board exempted requirement for informed consent because of the retrospective study and no any negative impact on the patients.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
Not applicable.

Authors' contributions
YL and JW designed the study. MC, HW, DW and CW collected the data. YL and QX performed the data analysis. JC and YL wrote the manuscript. All authors have read and critically revised the manuscript.   Results of Kaplan-Meier analysis