1 Introduction

Between 575000 to 677000 episodes of bloodstream infections (BSIs) per year have been reported in North America as the cause of 79000 to 94000 deaths annually. Reports from Europe showed that there were over 1.2 million episodes of BSIs with 157000 deaths per year (Goto and Al-Hasan, 2013). It is difficult to obtain precise estimates of nosocomial bloodstream infections (nBSIs) owing to the limited number of studies from nosocomial infection surveillance systems. Several population-based studies reported that 19.1%, 28%, and 35% of the total episodes of BSI were nosocomial in the USA (Uslan et al., 2007), Canada (Lenz et al., 2012), and Finland (Skogberg et al., 2012), respectively, since the 2000s. Patients who developed nBSIs suffered high morbidity and mortality risks and had to bear heavy financial costs (Mittmann et al., 2012; Kaye et al., 2014). Many changes have recently occurred in the epidemiology of BSIs, particularly due to the emergence of drug-resistant organisms, which has significantly increased the treatment-failure rate and the risk of adverse outcomes (Lambert et al., 2011). In China, nBSIs accounted for one-third of nosocomial infections, followed by lower respiratory tract and urinary tract infections in the intensive care unit (ICU) (Ding et al., 2009).

Traditional Chinese medicine (TCM) is very popular in China. According to the Health Statistics book of 2012, a total of 2831 TCM hospitals accounted for 12.88% of the hospitals in China (2831/21979) (National Health and Family Planning Commission of the People’s Republic of China, 2012). Patients with chronic diseases, including hypertension, diabetes, cancer, hematological malignancies, and other illnesses, often chose TCM hospitals for their treatment. Furthermore, post-operative chemotherapy of malignant tumors, diabetes, and chronic obstructive pulmonary disease patients continued to increase in our hospital for nearly a decade (Jiang et al., 2007). These patients often had a long hospital stay, were elderly, and had underlying diseases, which contributed to a higher risk of suffering from nBSIs. However, studies are limited on nBSIs in TCM hospitals and the epidemiology and microbiology of nBSIs are still unclear in TCM hospitals. Therefore, a retrospective survey was performed to evaluate the epidemiological features of nBSIs in a tertiary TCM hospital, in order to describe the characteristics of the species distribution and to identify the factors influencing mortality.

2 Materials and methods

2.1 Study design and data collection

We undertook a retrospective surveillance study of nBSIs in the First Affiliated Hospital of Zhejiang Chinese Medical University in China, from January 2009 to December 2011. The First Affiliated Hospital of Zhejiang Chinese Medical University is a 1200-bed TCM hospital. The nBSI was diagnosed if one or more cultures of the sampled blood yielded a pathogenic organism, at least 48 h after admission. If the bloodstream isolate was a potential skin contaminant (e.g., coagulase-negative staphylococcus (CoNS), Corynebacterium, Bacillus, or Propionibacterium), two or more separate blood culture sets were required for the diagnosis (Weinstein et al., 1983).

The data were obtained from the electronic medical records without any additional tests performed for the purpose of this study. The data collected for each patient included demographic information, predisposing clinical conditions, underlying diseases, microbiological data, as well as the antimicrobial treatment information and the susceptibilities of the causative pathogens.

2.2 Variables and definitions

The main clinical outcome was measured by a 28-d mortality rate. Predisposing clinical conditions included diabetes, neutropenia, septic shock, chemotherapy, parenteral nutrition, hemodialysis, and the usage of invasive devices. Based on the standard definition, septic shock is a medical condition that occurs as a result of severe infection and sepsis (Bone et al., 1992). Underlying diseases included malignancies, neurological dysfunctions, cardiovascular malfunctions, gastroenterological diseases, respiratory dysfunctions, and renal diseases. The acute severity of the illness was retrospectively assessed on the day before the diagnosis of nBSIs using the Pitt bacteremia score. This is a validated index, based on mental status, need for ventilation and vital signs, which has been used in previous studies (Hilf et al., 1989; Lim et al., 2014). Source of infection was determined from clinical and microbiological data and using CDC/NHSN criteria (Horan et al., 2008).

Detailed information was recorded on the antimicrobial drug treatments during the period from 3 d prior to the collection of the first positive blood culture up to 7 d after it. Empirical antibiotic therapy was considered appropriate when an antibiotic, that had in vitro activity against the causative organism, was given during the first 48 h. If the isolates were resistant or if the dose or the route of administration was insufficient, treatment was considered to be inappropriate (Pedersen et al., 2003; Schonheyder and Sogaard, 2010; Lim et al., 2014).

The Ethics Committee of our hospital approved the procedures, and the data collected from this study were kept confidential.

2.3 Microbiological test

One or double blood specimens were collected from each patient at different sites within 5 min. The identification of blood isolates was carried out according to routine methods by the staff of the microbiology laboratory in our hospital. Antibiotic susceptibility testing was performed using the modified broth microdilution method. Minimum inhibitory concentration (MIC) breakpoints and quality control protocols were used according to the standards established by the Clinical and Laboratory Standards Institute (2007). Antibiotic susceptibility data was analyzed with WHONET 5.6 software.

2.4 Statistical analysis

The results were expressed as the mean±standard deviation (SD) for continuous variables, or as a proportion of the total number of patients or isolates for categorical variables. Univariate analysis was used to examine the association between predictor variables and the risk of mortality at 28 d. Variables that were statistically significant in the univariate analysis were then included in a multiple logistic regression analysis to eliminate potential confounding factors. for source of infection, central line associated blood stream infection (CLABSI) was used as the baseline category when performing regression analysis, as in the previous study of Melzer and Welch (2013). This was chosen because CLABSIs are well defined and are associated with a large number of BSIs. P values of <0.05 were considered to be statistically significant. The SPSS 16.0 for Windows software package (Version 16.0) was used for this analysis.

3 Results

3.1 Study population and patient characteristics

During the three-year study period, there were 482 significant nBSI episodes. The incidence rate of nBSIs was 5.7/1000 hospital admissions. Patient characteristics and distribution of nBSIs are shown in Table 1.

Table 1
figure Tab1

Patient characteristics of the 482 patients with nBSIs

Full size image

3.2 Microbiological features

The poly-microbial accounted for 27 out of 482 episodes (5.6%). Of all the nBSI isolates, Gram-negative organisms were the cause of proximately three-fifths (61.8%); Gram-positive organisms accounted for 33.6%; and, fungi accounted for only 4.6%. Escherichia coli (25.5%) was the principal organism responsible for nBSIs, followed by CoNS (14.1%) and Klebsiella pneumoniae (11.2%). When analyzed by the clinical department, the following patterns emerged: E. coli, K. pneumoniae, and S. aureus were the top three isolated pathogens for the general wards, except for the ICU, where CoNS (29, 22.8%) and Acinetobacter spp. (17, 13.4%) were more frequently isolated (Table 2).

Table 2
figure Tab2

Distribution of the most frequently isolated pathogens causing nBSIs and number of deaths caused by nBSIs in different clinical departments

Full size image

3.3 Antimicrobial susceptibility

Antimicrobial resistance levels for the most common organisms causing nBSIs are shown in Tables 3 and 4. Methicillin resistance was detected in 60.0% of the 30 S. aureus and 90.7% of the 54 CoNS isolates (Table 3). Resistance to the third-generation cephalosporins was found in 38.6% of the 105 E. coli and 35.5% of the 31 K. pneumoniae isolates in nBSI patients (Table 4). Extended-spectrum β-lactamase (ESBL)-positive E. coli and K. pneumoniae accounted or 79.8% and 57.1% of isolates, respectively. Imipenem resistance was only detected in 6.1% of the 114 E. coli and 6.5% of the 31 K. pneumoniae isolates. However, resistance to imipenem was observed in 50.0% of Acinetobacter spp. isolates.

Table 3
figure Tab3

Rates of antimicrobial resistance among Gram-positive bacteria most frequently isolated from patients with nBSIs

Full size image
Table 4
figure Tab4

Rates of antimicrobial resistance among Gram-negative bacteria most frequently isolated from patients with nBSIs

Full size image

3.4 Antimicrobial drugs

A total of 479 patients (99.4%) used antibiotics during the period from 3 d before the collection of the first positive blood culture to 7 d after it. Of those, 85.2% used antibiotics before blood specimen collection. Only about one-third of patients (157/482) received appropriate empirical therapy (Table 1). Carbapenems were the antimicrobial drugs that were the most frequently given (n=208; 43.2%), followed by glycopeptides (n=128; 26.6%) and cefoperazone-sulbactam (n=123; 25.5%), as shown in Fig. 1.

Fig. 1
figure 1

1 Most frequent use of antimicrobial drugs

The most frequent antimicrobial drug treatments during the period from 3 d prior to the collection of the first positive blood culture up to 7 d after it

Full size image

3.5 Clinical outcomes and risk factors for the 28-d mortality

The overall 28-d mortality rate in all the patients with nBSIs was 16.8% (81/482). The top three pathogens associated with the fatalities included Acinetobacter spp., Enterococcus faecium, and Pseudomonas aeruginosa, with mortality rates of 77.3% (17/22), 33.3% (8/24), and 29.2% (7/24), respectively. Although the prevalence rates of CoNS and E. coli isolates were the highest in ICU and general wards respectively, the corresponding mortality was not the highest. The pathogen that was most frequently associated with fatality in the ICU was Acinetobacter spp. (82.3%, 14/17), while in hematology and oncology it was K. pneumoniae (43.8%, 7/16; 37.5%, 3/8) (Table 2).

Univariate analyses of 28-d mortality are presented in Table 5. A number of variables were statistically significantly associated with 28-d mortality, including septic shock, central venous catheter, urinary catheter, ventilator, hemodialysis, parenteral nutrition, appropriate empirical therapy, Pitt bacteremia score, and source of infection. The results of multiple logistic regression analysis are presented in Table 6. After adjustment for potential confounding factors, we found that septic shock (odds ratio (OR)=2.77, P=0.01), hemodialysis (OR=3.29, P=0.012), Pitt bacteremia score >4 (OR=12.88, P<0.001), and urinary tract infection (OR=11.12, P<0.001) were strongly associated with 28-d mortality from nBSIs, while appropriate empirical therapy was a protective factor for 28-d mortality (OR=0.23, P=0.001).

Table 5
figure Tab5

Univariate analysis of 28-d mortality in nBSIs

Full size image
Table 6
figure Tab6

Multivariate logistic regression analysis of 28-d mortality in nBSIs

Full size image

4 Discussion

Hospital surveillance studies on nosocomial infections could be useful to discover specific issues related to species distribution and resistance patterns of the pathogens causing nBSIs. It is essential to evaluate the etiologic agents and clinical symptoms together to make sure that the surveillance data are reliable. Therefore, we performed a clinically oriented surveillance by collecting the data on the defined infections rather than solely evaluating the culture results in this study.

The incidence of nBSIs in this study was 5.7/1000 hospital admissions, which was not as high as expected and close to the incidence levels of the US with 6.0/1000 admissions (Wisplinghoff et al., 2004). Other studies reported that the incidence of nBSIs was 11.6 to 15.5 episodes per 1000 admissions in Thailand (Hortiwakul et al., 2012) and 16.0 to 31.2 episodes per 1000 admissions in Europe (Rodríguez-Créixems et al., 2008). The relatively low incidence rate in this study might be explained by following reasons. One probable reason might be the low number and incorrect timing of blood cultures, which may have a significant effect on diagnostic of BSIs (Rampini et al., 2011; Schmitz et al., 2013). Another reason might be the small proportion of surgical and critically ill patients in our hospital. Many studies showed that these patients had a significantly higher incidence of nBSIs (Ding et al., 2009; Trethon et al., 2012). The incidence of nBSIs was up to 34.25/1000 in the cardiac surgery ICUs (Trethon et al., 2012).

E. coli (25.5%), CoNS (14.1%), and K. pneumoniae (11.2%) were the top three bacteria isolated in this study. The patterns of bacterial distribution in this study were consistent with those reported in a teaching hospital in Beijing (Lu et al., 2013) and the nationwide surveillance research of China (Wei et al., 2012). In the present study, antibiotic resistance within the group of Gram-negative bacteria was worrisome, especially in the case of Acinetobacter spp. with a rate as high as 50% for resistance to imipenem. Imipenemresistant Acinetobacter spp. mainly occurred in ICU. Since resistance to imipenem was usually associated with multidrug resistance, very few antibacterials remain potent (Ye et al., 2010). This could explain why patients with Acinetobacter spp. infection had the highest mortality rate (82.3%, 14/17).

As for the use of antibiotics, almost every patient was treated with broad-spectrum antibiotics. About 85.2% of them used the antibiotics before blood specimen collection, which was much higher than the 20% reported by a multi-center study and included the 1100 ICU patients with nBSIs (Tabah et al., 2012). However, compared with that study, the most frequent use of antibiotics for treating BSIs was the same; they were carbapenems and glycopeptides. Moreover, only one-third of patients (157/482) received appropriate empirical therapy in our hospital, which is much lower than those in other hospitals (Townell et al., 2014).

In the present study, septic shock, hemodialysis, Pitt bacteremia score >4, and source from urinary tract infection were shown to be the most common risk factors for the 28-d mortality of nBSI. The Pitt bacteremia score is a marker of severity of illness in patients with nBSIs and patients with scores >4 are considered to be critically ill. In line with the results of previous studies (Neuner et al., 2011; Retamar et al., 2011; Melzer and Welch, 2013), in our hospital the Pitt bacteremia score was strongly correlated with 28-d mortality. Compared with CLABSI, we demonstrated a significant association between urinary tract source of BSI and death at 28 d. This result has also been validated in a multi-center study in the UK (Melzer and Welch, 2013). However, appropriate empirical therapy can protect patients from mortality at 28 d. The mortality rate was reduced by approximately 4-fold in patients receiving appropriate empirical therapy compared to those without (OR=0.23, P=0.001). As shown in other studies, appropriate empirical antibiotic treatment has been known to be the most effective treatment for BSIs (Sogaard et al., 2011) and inappropriate empirical antibiotic therapy has been one of the most important risk factors for mortality in nBSIs (Tabah et al., 2012). Therefore, how to improve the empirical use of antibiotics in the early period of BSIs in order to reduce the 28-d mortality in our hospital should be elaborated in future studies.

In conclusion, the incidence of nBSIs was low in the TCM hospital but the proportion of nBSIs due to antibiotic-resistant organisms requires more attention. Efficient control methods are needed to decrease antibiotic drug resistance and to ensure patients receive effective treatment. A high Pitt bacteremia score was identified as one of the most important risk factors for mortality in nBSIs. Adequate empirical therapy is independently associated with decreased mortality in patients with nBSIs. Therefore, programs to improve the quality of empirical therapy in patients with suspected nBSIs and optimization of definitive therapy should be implemented.

Compliance with ethics guidelines

Jian-nong WU, Tie-er GAN, Yue-xian ZHU, Jun-min CAO, Cong-hua JI, Yi-hua WU, and Bin LV declare that they have no conflict of interest.

This article does not contain any studies with human or animal subjects performed by any of the authors.