Background

Antimicrobial resistance (AMR) has become a serious public health threat across the world. AMR control is a priority for the World Health Organization (WHO) [1]. According to WHO recommendations, AMR surveillance is an important part of the AMR control strategy. Furthermore, the data regarding microbiological composition and AMR profiles will guide antimicrobial prescriptions [2]. However, almost all bacterial resistance surveillances programs in China are laboratory-based surveillance strategies with potential biases resulting from optional sample collection. Therefore, it is imperative to carry out infection-defined surveillance to overcome this limitation.

Bloodstream infection (BSI) is a growing public health concern worldwide, with high mortality [3, 4]. It was estimated that the BSI incidence ranged between 113 and 204 per 100,000 in the population [5]. Inappropriate antibiotic therapy for BSI was independently associated with increased risk of mortality [6]. Microbiological epidemiology and bacterial resistance data on BSI will provide a reference for the best empirical antimicrobial therapy [7]. In the face of increasing AMR, precise surveillance has become important in defining the species distribution and resistance of pathogens causing BSI, and thus provide the basis for appropriate empirical therapy.

To comprehensively understand and accurately analyze the microbiological epidemiology and resistance profiles of BSI in China, we initiated the Blood Bacterial Resistant Investigation Collaborative System (BRICS) program in 2014. As an infection-defined surveillance initiative, information on the pathogen distribution and AMR of BSI bacteria was collected from participating hospitals covering 18 provinces in mainland China.

Material and methods

Study period and setting

A total of 52 hospitals (100,712 beds), which included 23 tertiary hospitals and 29 non-tertiary hospitals covering 18 provinces in mainland China during 2014–2019, participated in the BRICS program. All participating hospitals were equipped with a qualified microbiology laboratory and followed standardized operational programs. All laboratories participated in one external quality control program at least yearly by either the National Center for Clinical Laboratories or the local province center for clinical laboratories.

Bacterial isolate collection

Only strains isolated from blood were collected. Blood culture results of patients with only skin contaminants were considered contamination, and patients with both BSI and contamination were classified as having BSI [8]. Coagulase-negative Staphylococci, Bacillus species, viridans group Streptococci, Corynebacterium species, Propionibacterium species, Aerococcus species, and Micrococcus species from a single positive culture were excluded as contaminants, while all Brucella species were excluded due to the Biosafety Law of China. Only the first isolate of a species per patient was eligible over the full study period. All participating hospitals transferred their strains to the central laboratory quarterly. The central laboratory confirmed the identity of the isolates received using matrix-assisted laser desorption/ionization time of flight mass spectrometry (Bruker Diagnostics, Bremen, Germany) and stored the strains in Microbank® tubes at − 80 °C.

Antimicrobial susceptibility testing

The antibiotic susceptibilities (minimum inhibitory concentration, MIC) of clinical isolates were determined by agar-dilution or broth-dilution methods [9] at the central laboratory. Extended-spectrum β-lactamase (ESBL) production in Escherichia coli, Klebsiella pneumonia, and Proteus mirabilis using disk diffusion, and inducible clindamycin resistance in Staphylococcus species, Streptococcus pneumonia, and Streptococcus spp. β-hemolytic group were determined. β-hemolytic group determination using disk diffusion (D-zone test) was performed according to the Clinical and Laboratory Standards Institute (CLSI) [9]. The results of MICs were interpreted according to CLSI criteria or European Committee on Antimicrobial Susceptibility Testing (EUCAST) (https://www.eucast.org). Staphylococcus aureus ATCC 29,213, Enterococcus faecalis ATCC 29,212, S. pneumonia ATCC 700,603, E. coli ATCC 25,922, K. pneumoniae ATCC 27,853, and Pseudomonas aeruginosa ATCC 27,853 were included as quality controls.

Polymicrobial bacteremia is defined as a bacteremic episode due to at least two different organisms isolated from the same blood sample, while monomicrobial bacteremia is defined as only one organism isolated in the blood sample [10]. Contaminants are defined as a growth of bacteria in the blood culture bottle that were not present in the patient’s bloodstream but were introduced during sample collection [11]. For surveillance purpose, carbapenem-resistant Enterobacteriaeceae was defined as demonstrating resistance based upon antimicrobial susceptibility test results to at least one of the following carbapenems: ertapenem, meropenem, or imipenem according to the Centers for Disease Control and Prevention (CDC) of USA [12]. Multidrug-resistant (MDR) is defined as non-susceptibility to at least one agent in three or more antimicrobial categories and pan-drug-resistant (PDR) is defined as resistance to all antibiotic classes available for empirical treatment [13].

Data and statistical analysis

To compare resistance profiles in different economic areas, a developed or developing area was defined as having a per capita gross domestic product (GDP) ≥ 11,000 USD or < 11,000 USD in 2019. WHONET 5.6 was used to analyze the distribution and resistance rates of the isolates. To assess trends in proportions of resistant isolates over time, the Cochran–Armitage χ2 test was used. The Chi-square test was used to compare rates for different hospital levels and regions. Significance was assumed at a p-value < 0.05. Statistical analysis was performed using SPPS (version 9).

Ethics requirements

For this observational study, the need for patient consent was not required. Data were not identifiable back to the patients from whom they originated; an ethics approval was waived.

Results

Bacterial isolate collection

A total of 52 hospitals from 2014 to 2019 participated in BRICS. There were 8 tertiary hospitals and 13 non-tertiary hospitals that participated over the entire duration of the surveillance program, while other hospitals participated for 1 to 5 years. The number of participating hospitals per year were as follows: 21 in 2014, 26 in 2015, 27 in 2016, 30 in 2017, 39 in 2018, and 49 in 2019.

Over the 6-year period, 27,899 isolates were collected in total; Gram-positive organisms (GPO) accounted for 29.5% (8244/27,899) and Gram-negative organisms (GNO) for 70.5% (19,655/27,899). GNO increased from 58.4% in 2014 to 73.0% in 2019. E. coli and K. pneumoniae were the most common organisms overall, followed by S. aureus, coagulase-negative Staphylococci (CoNS), Acinetobacter baumannii, and P. aeruginosa (Fig. 1). Notably, E. coli increased 29.1% to 38.5% during the study period. A higher proportion of CoNS were detected in tertiary hospitals and developing areas (10.1% and 12.1%, respectively) than those in non-tertiary hospitals or developed areas (9.4% and 7.7%, respectively). The percentage of E. coli was higher in non-tertiary hospitals (38.8%) than in tertiary hospitals (32.5%). P. aeruginosa and A. baumannii were more often detected in tertiary hospitals (Additional file 1: Table S1).

Fig. 1
figure 1

The bacterial spectrum of bloodstream infections by year (relative proportions) from 2014 to 2019

Full size image

Bacterial resistance and trends over time

Resistance in Gram-positive organisms

The overall prevalence of methicillin-resistant Staphylococcus aureus (MRSA) declined during the study period from 39.0% to 25.9% (p < 0.05) (Table 1, Fig. 2). No vancomycin resistance, or “MIC creep” was detected over time. Of note, the prevalence of MRSA was higher in developed regions (33.9% vs 28.3% in developing regions), but no differences across different hospital levels was observed (Additional file 1: Table S2). Moreover, we observed a decline in the proportion of rifampicin-resistant S. aureus during the study period (5.3% in 2014 and 1.3% in 2019, p < 0.05).

Table 1 The susceptibility and resistance trends of Gram-positive pathogens to antimicrobial agents
Full size table
Fig. 2
figure 2

Trends in the prevalence of major drug-resistant bacteria over the study period. CR-AB, carbapenem-resistant A. baumannii; CR-EC, carbapenem-resistant E. coli; CR-KP, carbapenem-resistant K. pneumoniae; CR-PA, carbapenem-resistant P. aeruginosa; ESBL-EC, extended-spectrum β-lactamase-producing E. coli; ESBL-KP, extended-spectrum β-lactamase-producing K. pneumoniae; MRSA, methicillin-resistant S. aureus; VRE, vancomycin-resistant E. faecium

Full size image

The methicillin-resistant, coagulase-negative Staphylococci (MRCNS) prevalence was 74.3% and no glycopeptide-resistant MRCNS were detected (Additional file 1: Table S2). Compared with MRSA, MRCNS exhibited higher resistance rates to fluoroquinolones and trimethoprim/sulfamethoxazole (Additional file 1: Table S3).

E. faecium showed increased resistance rates to rifampicin over time (p < 0.05) (Table 1). Of note, the overall prevalence of vancomycin-resistant E. faecium (VRE) was low. Daptomycin- or linezolid-resistant E. faecium was rare (≤ 0.1% and ≤ 0.5% overall, respectively). The prevalence of MDR E. faecium was similar across different hospital levels and economic regions (Additional file 1: Table S4). 7.4% E. faecalis were resistant to ampicillin (Table 1).

S. pneumoniae was susceptible to penicillin G as defined by the breaking point of non-meningitis injection of penicillin G (Additional file 1: Table S5). None of the 388 Streptococcus spp. β-hemolytic group were resistant to penicillin G, while 1.4% of α-hemolytic Streptococci strains was resistant (Additional file 1: Table S2).

Resistance of Enterobacteriaceae

Carbapenem-resistant (CR) E. coli (CR-E. coli) and CR K. pneumoniae (CR-K. pneumoniae) increased from 1.2% and 7.0%, respectively, in 2014 to 1.7% and 19.6%, respectively, in 2019 (p = 0.397 and p < 0.05, respectively) (Table 2, Fig. 2). Unlike E. coli, MDR K. pneumoniae increased significantly over a 6-year period (p < 0.05). Overall, the resistance rate of CR-E. coli to ceftazidime/avibactam was 60.7% whereas 8.3% CR-K. pneumoniae were resistant to ceftazidime/avibactam (Additional file 1: Tables S6 and S7). It is interesting to note that the prevalence of CR-K. pneumonia was higher in tertiary hospitals and developed regions compared with non-tertiary hospitals and developing regions (both p < 0.05). However, there was no significant difference in the prevalence of CR-E. coli between developed and developing regions (Additional file 1: Table S4).

Table 2 The susceptibility and resistance trends of Gram-negative pathogens to antimicrobial agents
Full size table

The prevalence of ESBL-positive E. coli (ESBL-E. coli) and ESBL-positive K. pneumonia (ESBL-K. pneumonia) declined over the period of the study (both p < 0.05) (Table 2, Fig. 2). In general, higher resistance rates of pathogens were more common in tertiary hospitals compared with non-tertiary hospitals (Additional file 1: Table S2). However, the prevalence of ESBL-E. coli and ESBL-K. pneumonia was higher in developing regions compared with developed regions (58.2% and 36.4% vs 48.8% and 29.6%, respectively; both p < 0.05) (Additional file 1: Table S4).

Other Enterobacteriaceae, including E. cloacae, Klebsiella species (excluding K. pneumoniae), Proteus species, Serratia species, Salmonella species, and Citrobacter species showed lower resistance to ceftazidime/avibactam, carbapenem, and amikacin (Table 2, Additional file 1: Table S5).

Resistance in glucose non-fermenting Gram-negative bacteria

MDR rates in A. baumannii and P. aeruginosa were 70.3% and 21.1%, respectively, which did not increase significantly in 6 years (p = 0.841and p = 0.488, respectively) (Table 3). However, the PDR rate in A. baumannii was more than 50% (52.5%, 558/1062), wherein 6.7% of the strains were resistant to polymyxin B and 4.3% to tigecycline (Table 1). CR-A. baumannii (CR-AB) fluctuated between a prevalence of 55.7% and 71.2% during the surveillance time (Table 1, Fig. 2). The resistance rates in A. baumannii and P. aeruginosa were higher in tertiary hospitals than non-tertiary hospitals (Additional file 1: Table S2). It noted that the prevalence of CR-A. baumannii and CR-P. aeruginosa were higher in tertiary hospitals settings (71.3% and 25.5%, respectively, vs 52.0% and 12.6%, respectively; both p < 0.05) (Additional file 1: Table S4). However, the prevalence of CR-A. baumannii was higher in developing areas than developed areas (72.9% vs 56.7%, respectively; p < 0.05) (Additional file 1: Table S4).

Table 3 The susceptibility and resistance trends of non-fermentative pathogens to antimicrobial agents
Full size table

Cefoperazone/sulbactam and levofloxacin were the most effective agents against Stenotrophomonas maltophilia (with 95.0% and 89.2% susceptibility, respectively) (Additional file 1: Table S5). Overall, 18.3% of S. maltophilia were resistant to trimethoprim/sulfamethoxazole. The prevalence of ceftazidime-resistant Burkholderia species was 14.3%, and 24.4% of the strains were resistant to trimethoprim/sulfamethoxazole (Additional file 1: Table S5).

Discussion

During the surveillance period from 2014 to 2019, E. coli and K. pneumoniae were the main BSI pathogens and the proportion of MRSA and ESBL-E. coli declined, while the frequency of CR-K. pneumoniae continuously increased. We also determined that the prevalence of antimicrobial resistant pathogens, especially Gram-negative organisms (GNO), varied by hospital types and the levels of local economic development. The declining trend could be attributed to the results of a special national campaign of antimicrobial stewardship initiated in 2012 [14]. As the first national, infection-defined, surveillance program in China, the BRICS has provided more precise data to help clinicians to improve antimicrobial therapy and contain the spread of AMR.

The frequency of blood cultures is recommended for 100 to 200 blood cultures sets per 1,000 patient days [15]. The 2018 annual report of the European Antimicrobial Resistance Surveillance Network (EARS-Net) varied substantially between countries, specifying only 5.3 blood cultures per 1,000 patient days in Lithuania, compared to 206.9 in Portugal [16]. However, 100 to 200 blood cultures sets per 1,000 patient days are far from routine use in China, as the China Antimicrobial Resistance Surveillance System shows that blood culture samples account for less than 10% of total microbiological samples [17]. In China, clinical microbiology is only a branch of laboratory medicine, which is different from established clinical microbiology and infectious disease departments in other countries. As a bloodstream-infection-definition surveillance, BRICS carried out AMR studies to raise attention to blood culture and promote antimicrobial stewardship programs by providing precisive surveillance data.

The surveillance revealed that GNO constituted a major proportion of all BSI causes, which increased during the monitoring period. Furthermore, the proportion of E. coli increased rapidly, which is consistent with previous reports [18]. The SENTRY Antimicrobial Surveillance Program also revealed that detection of GNO had increased and the proportion of GNO, such as E. coli and K. pneumoniae, also increased [19]. The emergence of MDR GNO, such as carbapenem-resistant strains, has increased rapidly [20], which makes treatment options extremely limited in clinical practice [21] as antibiotic development is lagging behind resistance for GNO. The spillover of the resistant bacteria, which is a consequence of the fact that antibiotic-resistant bacteria can be transmitted from person to person, could be the reasons for the increasing proportion of GNO [22].

During this surveillance, it is interesting to note that MRSA decreased over time, which is in line with other surveillance results [23]. Lawes et al. suggested that antibiotic stewardship and infection control strategies might have played an important part in the reduced prevalence of MRSA [24]. However, these strategies do not fully explain why pathogens other than MRSA are becoming more prevalent rather than declining over the same period. Although specific factors responsible for changes in the rates of MRSA infection remain uncertain, molecular epidemiology might provide a better understanding of MRSA population dynamics. The phenomenon of clonal replacement, whereby clones that were once widely disseminated during a certain period may become less dominant and are replaced by other epidemic clones, has been observed worldwide [25, 26]. In a Portuguese tertiary hospital, the Brazilian (ST239-IIIA) clone was replaced by the arrival of epidemic EMRSA-15 (ST22-IV) [27]. In China, the structural change in the population of MRSA was also observed when ST5-t437 replaced ST239-t030 as the predominant genotype [28].

Although the prevalence of ESBL-E. coli declined during the surveillance period, it was still at a high level, with a prevalence of more than 50%. This scenario has led to subsequent increased use of carbapenems, which was associated with the emergence and spread of carbapenem-resistant bacteria, especially among K. pneumoniae [29, 30]. This phenomenon was reflected in our monitoring data in which the CR-K. pneumoniae prevalence increased quickly. Other studies in China also indicated that CR-K. pneumoniae have reached higher epidemic levels in China [31, 32]. Furthermore, evidence of a particularly higher percentage of ST11 CR-K. pneumoniae [33] indicated there was clonal spread of CR-K. pneumoniae in hospitals. Our previous study also confirmed that ST11 KPC-2-producing K. pneumoniae was the common sequence types (STs) among carbapenemase-producing Enterobacteriaceae (CPE) [34]. However, sequencing was not carried out on all isolated carbapenem-resistant Enterobacteriaceae, which needs to be further studied. Accordingly, it is necessary to implement an antimicrobial stewardship program and effective infection control to contain and mitigate the risks of nosocomial transmission and outbreaks in hospitals, such as hand-hygiene education programs, contact precautions, and use of alert codes to promptly identify patients with CR-K. pneumoniae infections [35].

The CR-A. baumannii remained around 60% during the surveillance period, which poses a great challenge to patient treatment, and the prevalence was similar with other surveillance programs [36]. CR-A. baumannii is an emerging concern due to the associated high mortality rates [32]. The optimal antibiotic choice for CR-A. baumannii bacteremia is controversial. Some studies have suggested the use of polymyxin as the backbone of treatment [37, 38], Only colistin or tigecycline still presented higher susceptibility rates in this surveillance; however, the clinical efficacy of the two agents still needs to be confirmed. Hospitals should consider antimicrobial stewardship and infection control when tackling CR-A. baumannii. Indeed, Latibeaudiere et al. demonstrated that previous CR-A. baumannii colonization increased the risk of developing a CR-A. baumannii infection by eightfold [39]. Targeted antibacterial therapy against infection; rather than colonization and promoting hand hygiene, environmental cleaning, and contact precautions; could be valuable strategies in the control of CR-A. baumannii [40].

It is conceivable that resistance rates are higher in tertiary hospitals where patients with more critical conditions are admitted; larger scale operations, more frequent organ transplantations, and longer hospitalizations increase a patient’s risk of infection, which leads to more antibiotic use and easier induction of AMR [41]. Nonetheless, the higher AMR observed in developing regions may result from a poorer healthcare infrastructure, weaker infection control implementation, and less common antimicrobial stewardship activity. All hospitals should take individual actions to contain the spread of AMR.

There are some limitations to note in this study. First, participating hospitals covered 18 of the 31 provinces in mainland China. However, it has not yet covered all provinces and the participating hospitals are only a small proportion of the total hospitals in China. Therefore, it must be noted that by increasing the number of participants, the surveillance data might become more precise. Second, the failure to distinguish the pathogens isolated from community-acquired or hospital-acquired BSI was a major limitation. In China, blood culture is mainly requested for patients with symptoms of infection during hospitalization, which means most BSI is nosocomial BSI. Third, some relevant denominators, like patient-days, blood culture rates and patient characteristics, were not recorded for all patients. Last, this surveillance is not a population-based surveillance, and the incidence of BSI is not available.

Conclusions

In this report of the first national BSI surveillance program in China, E. coli and K. pneumoniae were the main BSI pathogens. The proportion of MRSA and ESBL-E. coli declined, while the frequency of CR- K. pneumoniae continuously increased. The prevalence of antimicrobial-resistant pathogens, especially ESBL-E. coli, CR-E. coli, ESBL-K. pneumonia, CR-K. pneumonia and CR-A. baumannii, varied by hospital types and the level of local economic development.