Antimicrobial resistance among pathogens that infect the bloodstream: a multicenter surveillance report for 1998–2017

Background Bloodstream infections (BSIs) are a common consequence of infectious diseases and cause high morbidity and mortality. Appropriate antibiotic use is critical for patients’ treatment and prognosis. Long-term monitoring and analyzing of bacterial resistance are important for understanding the changes in bacterial resistance and infection control. Here, we report a retrospective study on antimicrobial resistance in BSI-associated pathogens. Methods Data from the Hubei Province Antimicrobial Resistance Surveillance System (HBARSS) from 1998–2017 were retrospectively analyzed using WHONET 5.6 software. Results Data from HBARSS (1998–2017) revealed that 40,518 Gram-positive bacteria and 26,568 Gram-negative bacteria caused BSIs, the most common of which were Staphylococcus aureus and Escherichia coli. Salmonella typhi was a predominant BSI-associated pathogen in 1998–2003. Drug susceptibility data showed that the resistance rates of E. coli and Klebsiella pneumoniae to cefotaxime were signicantly higher than those to ceftazidime. Carbapenem-resistant (CR) E. coli and K. pneumoniae have also emerged. In 2013–2017, K. pneumoniae showed resistance levels reaching 15.8% and 17.5% to imipenem and meropenem, respectively, and Acinetobacter baumannii showed high resistance rates ranging from 60–80% to common antibiotics. The detection rate of Salmonella typhi resistance to third-generation cephalosporins and uoroquinolones was less than 5%. Control of methicillin-resistant Staphylococcus aureus (MRSA) remains a major challenge, and in 2009–2017, the MRSA detection rate was 40–50%. The number of extensively drug-resistant A. baumannii and P. aeruginosa has been increasing since 2008. From 1998 to 2017, the total detection rates of extensively drug-resistant A. baumannii and P. aeruginosa were 34.38% (493/1434) and 7.45% (140/1879), respectively.

monitoring network has been extended to secondary and tertiary hospitals across the entire province, and more than 50 hospitals have joined the monitoring network to date. The proportion of network hospitals from the registered hospitals for all of Hubei Province reached 14.45% (50/346) in 2018. All hospital administrators log into the national bacterial resistance monitoring network of the Ministry of Health (http://www.carss.cn) using their own user name and password, then enter the provincial login to submit data.
Bloodstream infections (BSIs) are a major cause of morbidity and mortality in adults and children [3][4].
Appropriate use of antibiotics is critical for their treatment and prognosis. At present, China is one of the largest users of antibiotics worldwide [5], and antibiotic overuse remains a serious problem worldwide [6].
Here, we report a 20-year analysis of HBARSS for 1998-2017. Our ndings provide a reference for monitoring changes of bacterial resistance and management of antibiotics.

Study design and procedures
To effectively analyze the accumulated susceptibility data and determine the trend in drug resistance for the major pathogens, only data from the initial 15 hospitals in 1998-2002, 16 hospitals in 2003-2004 and 17 hospitals in 2005-2017 were analyzed. Each network hospital independently cultured, identi ed and conducted susceptibility testing of the strains, and the data were submitted to HBARSS annually.
Blood culturing was performed on patients who satis ed the clinical standards [ 7]. Automated blood culture instruments, including the BD 9120, 9240 and FX 400 (BD Co., NJ, USA) or the 3D 120, 240 and 720 (Bio Mérieux, Lyon, France), were used in each hospital in the monitoring network. Strains were identi ed following each laboratory's protocol, which combined various automated instruments or an IVD-MALDI Biotyper (Bruker, Karlsruhe, Germany) with manual biochemical experiments. Either the diskdiffusion method or an automated instrument was used for the antimicrobial susceptibility tests. From 1998-2010, all hospitals used the disk-diffusion method for drug susceptibility testing. From 2011-2017, six hospitals used automated instruments, and 11 used the disk-diffusion method. Automated instruments for drug-sensitivity testing included the Vitek-2 Compact system (Bio Mérieux, Lyon, France) and the domestic drug-sensitivity testing system (Dier, Zhuhai, China). Antimicrobial susceptibility tests were performed strictly in accordance with Clinical Laboratory Standards Institute (CLSI) standards. Each hospital routinely carried out indoor quality control and participated in the External Quality Assessment of the Ministry of Health of China. Laboratory quality control experiments strictly followed the CLSI guidelines of the corresponding year, and standard strains were tested once weekly.
Because CNS, Corynebacterium, Bacillus, Propionibacterium and other potential skin contaminants frequently contaminate blood cultures, whether these organisms were colonizing, pathogenic or contaminating bacteria was determined from the available clinical data [8].

Statistical analysis
Data were analyzed using WHONET 5.6 software. To avoid the effects of repeated subculturing on bacterial resistance, only the rst strain was used in the analysis. Interpretation criteria for the antimicrobial susceptibility results were based on CLSI 2018 Guidelines [9].
Antimicrobial susceptibility of Gram-negative bacteria Both E. coli and K. pneumoniae showed higher resistance to the third-generation cephalosporin, cefotaxime, than to ceftazidime. The resistance rates of E. coli to ceftazidime and cefotaxime were 10.5-30.1% and 31.75-67.3%, respectively, whereas those of K. pneumoniae were 24-31.6% and 41.7-49.7%, respectively (Fig 4 and Fig 5). The resistance rate of E. coli to uoroquinolones was signi cantly higher than that of K. pneumoniae. The resistance rates of E. coli to cipro oxacin and levo oxacin were 47.3-55.6% and 45.2-52.8%, respectively, and those of K. pneumoniae were 18.1-27.7% and 11.9-25.5%, respectively (Fig 4 and Fig 5). The resistance rate of K. pneumoniae to carbapenems was signi cantly higher than that of E. coli. The resistance rates of K. pneumoniae to imipenem and meropenem were 2.4-15.8% and 1.8-17.5%, respectively, whereas those of E. coli were 0.8-2.3% and 0.8-1.3%, respectively (Tables S1 and S2). S. typhi showed resistance to third-generation cephalosporins and uoroquinolones, but the resistance rate was less than 6% ( Fig 6). The resistance rate of S. typhi to ampicillin increased signi cantly from 6.9% in 1998-2002 to 38.5% in 2013-2017 (Table S3).

Epidemiology of methicillin-resistant Staphylococcus aureus (MRSA)
MRSA strains included S. aureus that expressed mecA or another methicillin resistance mechanism such as changes in the a nity of penicillin-binding proteins (PBPs) for oxacillin (modi ed S. aureus strains) [9]. Cefoxitin was tested as a surrogate for oxacillin. Isolates that tested resistant to cefoxitin on the minimum inhibitory concentration (MIC), disk, or oxacillin MIC tests were considered to be MRSA [9,10]. The MRSA detection rate was 10-30% in 1998-2003, which (Tables S7 and S8).

Discussion
Surveillance data from 1998-2017 in Hubei Province showed that the most common BSI-associated Gram-negative and Gram-positive bacteria were E. coli and S. aureus, respectively. This nding was consistent with that of the European Antimicrobial Resistance Surveillance Network (EARS-Net, formerly EARSS) report for 2002-2009 [11] but differed from reports from Malawi, Africa, which showed that nontyphoid Salmonella, S. typhi and Streptococcus pneumoniae were the main BSI-associated pathogens [4].
Our study showed that S. typhi was also a main BSI-associated pathogen in Hubei Province from 1998-2003. Typhoid fever is a poverty-related disease, mainly occurring in Africa and Asia, with a low incidence in economically developed regions such as Europe and the United States [12][13][14][15][16]. Typhoid fever is transmitted mainly through contaminated food and drinking water [17]. The incidence of S. typhi-related BSIs in rural children was reported to be 2-3 times higher than that in urban children [18]. The different incidences in different areas may be related to local medical and health conditions and vaccination rates. These factors may also have contributed to the high incidence in Hubei Province during 1998-2003.
Reports from Africa suggested that S. typhi and non-S. typhi were consistently the most common pathogens of BSIs [4]. Salmonella infections are frequently associated with human immunode ciency virus infections, very young or elderly patients, clinical malaria and malnutrition, and can be fatal in up to 20-25% of patients [19][20]. Reports from Africa showed that Salmonella was often resistant to rst-line antibiotics such as chloramphenicol, sulfonamide and ampicillin [21][22]. In our study, the resistance rate of S. typhi to ampicillin increased from 6.9% in 1998-2002 to 38.5% in 2013-2017, and resistance rates to other antibiotics were lower than 10% in 2013-2017. Resistance to uoroquinolones and thirdgeneration cephalosporins has also been reported in several African countries [23][24]. Our data showed that S. typhi resistance to third-generation cephalosporins and uoroquinolones has emerged, but in 1998-2017, the detection rate was less than 5%.
Antibiotic susceptibility tests showed that the resistance rates of E. coli and K. pneumoniae to thirdgeneration cefotaxime were signi cantly higher than those to ceftazidime, which is consistent with the 30-year data reported from CHINET in China [1]. Wang et al. showed that CTX-M was the most important ESBL type in China and that cefotaxime resistance might be a sign of ESBL-producing bacterial strains [25]. E. coli and K. pneumoniae showed low resistance to amikacin, cefoperazone/sulbactam and imipenem; thus, these antibiotics might be used as empirical treatment options. Notably, in 2013-2017, the rates of K. pneumoniae resistance to imipenem and meropenem reached 15.8% and 17.5%, respectively. Studies have con rmed that mortality rates of patients infected with carbapenem-resistant (CR) K. pneumoniae strains are signi cantly higher than those of patients infected with carbapenemsensitive strains [26][27]. CR K. pneumoniae strains often exhibit combined resistance to cephalosporins, uoroquinolones, aminoglycosides, beta-lactamase inhibitors and other antimicrobial agents [28]. Few antimicrobial agents, including tigecycline and polymyxin, can be used to treat CR K. pneumoniae [29].
This study revealed that P. aeruginosa and A. baumannii were the most common non-fermentative Gramnegative bacteria that cause BSIs. Susceptibility tests showed that resistance rates of P. aeruginosa to most antibiotics were less than 30%. However, these results differed from those reported in a multicenter epidemiological study on the risk factors and clinical outcomes of nosocomial intra-abdominal infections in China (the Chinese antimicrobial resistance surveillance of nosocomial infections [CARES] 2007-2016), which indicated that P. aeruginosa showed high resistance to a variety of antimicrobial agents, except amikacin, whose susceptibility rate was 83.4% [30]. The antimicrobial susceptibility pro les of A. baumannii isolates from BSIs were similar to those of A. baumannii isolates from abdominal infections. A. baumannii was alarmingly resistant to diverse antibiotics, including third-generation cephalosporins, aminoglycosides, uoroquinolones and carbapenems [30]. In this study, resistance rates of A. baumannii to common antibiotics increased signi cantly in 1998-2017. In 2003-2007, the antimicrobial resistance rate of A. baumannii was less than 50%, but by 2013-2017, the resistance rate reached 60-80%. The emergence of multidrug-resistant A. baumannii, especially extensively drug-resistant and fully drugresistant strains, has made clinical treatment di cult. According to CLSI guidelines, S. maltophilia showed standard resistance levels to minocycline, levo oxacin and trimethoprim/sulfamethoxazole as determined by disk-diffusion tests, but MIC testing showed break points for ticarcillin/clavulanic acid, ceftazidime and chloramphenicol, minocycline, levo oxacin and trimethoprim/sulfamethoxazole [9]. Therefore, some hospitals could increase the drug sensitivity test results of some drugs after changing disk-diffusion tests to MIC tests. For example, for S. maltophilia, disk diffusion method had only three drug break points, while MIC method had six drug break points. As a result, clinicians had more choices in the empirical treatment. However, the disadvantage of the change of drug sensitivity test methodology was that the cumulative drug sensitivity data were inevitably biased when comparing data for many years. In this study, the resistance rate of S. maltophilia to ceftazidime increased to 58.1% in 2013-2017, whereas the resistance rates of S. maltophilia to other antimicrobial agents were less than 25%. Whether the increase in ceftazidime resistance was related to its wide clinical application requires further investigation and analysis.
Surveillance data on BSIs from 1998-2017 showed that the resistance rate of A. baumannii to common antibiotics has reached a high level, and the prevalence of CR K. pneumoniae has increased signi cantly, resulting in signi cant di culties in clinical treatment. Our data show that vancomycin, teicoplanin, linezolid and trimethoprim/sulfamethoxazole can be used to treat MRSA. The resistance rate of MRSA to trimethoprim/sulfamethoxazole has decreased signi cantly, possibly related to the decreased of use of this drug in recent years. Studies from China, South Korea and France have shown that the antimicrobial resistance rates of S. aureus, K. pneumoniae, E. coli, P. aeruginosa and Candida albicans also decreased with the decreased clinical use of these antimicrobial agents [31][32][33][34]. Tigecycline and polymyxin can be used to empirically treat CR K. pneumoniae, E. coli and A. baumannii.
This study had several limitations. The BSI incidence in Hubei Province was often reported from single research centers. We failed to nd accurate data on the BSI incidence for all of Hubei Province from 1998-2017. Previous reports lacked demographic data. One shortcoming of this study was that the accurate BSI incidence was not calculated for Hubei Province. Another limitation was that different hospitals used different strain identi cation methods, including manual biochemical experiments and an IVD-MALDI Biotyper, and these results were undistinguishable once combined. Different hospitals adopted different drug sensitivity test methods, and the same hospital may change the drug sensitivity test method used between 1998 and 2017. Although each hospital strictly followed the CLSI guidelines, the inconsistency of test methods and the difference of drug sensitive consumables may lead to deviation in the analysis of drug resistance. The weakness of the analysis of the resistance mechanism involved in Gram-negative resistance to beta-lactams and more particularly to carbapenems was also a limitation of this study. We will increase the content of drug resistance mechanism research in the future. Conclusion CR K. pneumoniae, extensively drug-resistant A. baumannii and MRSA present major challenges to controlling BSIs. S. typhi resistant to the third generation cephalosporins and quinolones has emerged, but the drug resistance rates were all less than 5%. The study protocol was approved by the Tongji Hospital ethics committee for research in health. The Tongji Hospital ethics committee also approved the waiver of informed consent to participate in this study due to its retrospective design. All patient data were anonymous prior to the analysis.