Bacterial bloodstream infections in a tertiary infectious diseases hospital in Northern Vietnam: aetiology, drug resistance, and treatment outcome

Background Bloodstream infections (BSIs) are associated with high morbidity and mortality worldwide. However their aetiology, antimicrobial susceptibilities and associated outcomes differ between developed and developing countries. Systematic data from Vietnam are scarce. Here we present aetiologic data on BSI in adults admitted to a large tertiary referral hospital for infectious diseases in Hanoi, Vietnam. Methods A retrospective study was conducted at the National Hospital for Tropical Diseases between January 2011 and December 2013. Cases of BSI were determined from records in the microbiology department. Case records were obtained where possible and clinical findings, treatment and outcome were recorded. BSI were classified as community acquired if the blood sample was drawn ≤48 h after hospitalization or hospital acquired if >48 h. Results A total of 738 patients with BSI were included for microbiological analysis. The predominant pathogens were: Klebsiella pneumoniae (17.5%), Escherichia coli (17.3%), Staphylococcus aureus (14.9%), Stenotrophomonas maltophilia (9.6%) and Streptococcus suis (7.6%). The overall proportion of extended spectrum beta-lactamase (ESBL) production among Enterobacteriaceae was 25.1% (67/267 isolates) and of methicillin-resistance in S. aureus (MRSA) 37% (40/108). Clinical data was retrieved for 477 (64.6%) patients; median age was 48 years (IQR 36–60) with 27.7% female. The overall case fatality rate was 28.9% and the highest case fatality was associated with Enterobacteriaceae BSI (34.7%) which accounted for 61.6% of all BSI fatalities. Conclusions Enterobacteriaceae (predominantly K. pneumoniae and E. coli) are the most common cause of both community and hospital acquired bloodstream infections in a tertiary referral clinic in northern Vietnam.


Background
Bloodstream infections (BSIs) are an important cause of sepsis-related morbidity and mortality worldwide [1]. A systematic review of BSI burden in population based studies estimated the annual numbers of BSI episodes ranged from 575.462-677.389 in North America and 1,213,460-1,381,590 in Europe with the annual numbers of BSI associated deaths were 79,466-93,655 and 57,750-276,318 respectively [2]. In developed setting, the BSI and BSI associated septic shock still accounted for 6% and 3% of all admission to ICUs with the inhospital mortality rates of 40% and 49% respectively [3]. Additionally, BSIs are associated with long-term risk of mortality among survivors [4]. In a population based study in Denmark, while the incidence of community and hospital acquired BSIs annually decreased 3.7% and 4.2% respectively between 1998 to 2008, the incidence rate of healthcare-associated bloodstream infection BSI remained unchanged [4]. Data from low and middle income countries (LMIC) on the causes of bloodstream infection are limited [5], but crucial for enabling clinicians to direct antimicrobial therapy appropriately and for the design of preventive measures. Useful lessons can be learned from analysing the bacterial aetiology of BSIs as most cases reflect severe illness and bacteria detected are usually directly responsible for disease [6,7]. The causative agent will be affected by a number of factors, particularly the focus of infection, comorbidities like immunodeficiency, chronic renal and liver disease, in addition to socio-economic, climatic and other geographical factors [8]. The level of economic development can also impact aetiology distribution through changing exposure to environmental organisms associated with urbanisation, changing farming and food consumption practices, vaccination and comorbidities associated with increased wealth and improved life expectancy [9]. Consequently, data from tropical emerging economies are important to understanding and updating worldwide trends of the aetiology of severe bacterial disease in the tropics.
Vietnam is a low middle income country (LMIC), an emerging economy and a hotspot for emerging infectious diseases [10]. Levels of antibiotic resistance, an important cause of treatment failure and subsequent mortality are also high in Vietnam [11,12]. In the Asian Network for Surveillance of Resistant Pathogens with 11 participating countries in 2001, Vietnam had a highest prevalence of pneumococcal multidrug resistance (defined as resistance to at least three classes of antibiotics) (71.4%) which much higher than the second and third highest countries and territories of Korea (54.8%) and Hong Kong (43.2%) [12]. In another study in 20 hospitals in 5 Asia-Pacific countries, the prevalence of carbapenem non-susceptible Gram-negative pathogens was highest at 35.0% in Vietnam, in which the prevalence of carbapenem non-susceptiblity among Pseudomonas aeruginosa, Enterobacteriaceae and Acinetobacter baumannii were 46.7%, 5.6% and 89.5% respectively [11].
In-hospital mortality among patients with sepsis is reduced by early initiation of adequate empiric antibiotic therapy [13,14] and strategies to promote appropriate antibiotic use may help to prevent antimicrobial resistance and optimise use of hospital resources. However, the epidemiology of BSI in Vietnamese adults and associated treatment outcomes are not well studied. An understanding of BSI in Vietnam, its resistance patterns and their impact on patient's outcomes is important to guide clinical management and appropriate antibiotic use. This article reports on a retrospective analysis of three years of BSI data at the National Hospital for Tropical Diseases (NHTD), a large tertiary referral hospital for infectious diseases in Hanoi, Vietnam.

Study design and data collection
This is a retrospective, cohort study of patients with bacterial bloodstream infection (BSI) admitted to the National Hospital for Tropical Diseases (NHTD) between January 1st, 2011 and December 31st, 2013. NHTD is a 300-bed tertiary teaching hospital that provides care to individuals suspected of an infectious disease living within its catchment area. NHTD is also a referral hospital assigned to all provincial hospitals in the north of Vietnam for infectious diseases that cannot be managed locally. As a result the hospital receives patients both directly from the community and transferred from hospitals in the region. The study included all positive blood cultures with recognized bacterial pathogens among patients who were hospitalized during the study period and additional clinical information from a subgroup of patients whose patient records were available to the investigators. A standard data collection form was created to extract demographic & clinical data, microbiology and outcome information from routine patient and microbiology records. Where patients had been transferred from another hospital, the transfer letter was examined for any record of antibiotic therapy having been provided at the referring hospital.

Study definitions
A recognised bacterial pathogen identified as an organism which is not on the CDC's National Healthcare Safety Network (NHSN) common commensal list unless there were 2 bottles from the same set were positive for the same organism or a single positive bottle plus an evident focus of infection (abscess or valvular vegetation). An episode of BSI was defined as the isolation of the organism(s) from one or more positive blood bottles which were explainable by a common source of infection and the time between two isolations was not interrupted by an asymptomatic period. When the clinical information was not available for judgment, the same organism(s) isolated within 48 h of another positive culture was considered as the same episode [15].
A case of community-acquired BSI (CABSI) was defined as a patient with a positive blood culture from a blood sample drawn ≤48 h after hospitalization at any health care facility (to ensure correct categorisation of patients referred from another hospital) for the current illness. Hospital-acquired BSI (HABSI) cases were defined as patients with >48 h between admission at any health care facility and the blood being drawn using the referring letter when applicable [16,17]. These data were only available from patients whose clinical records were available. Any BSI caused by Streptococcus suis or Burkholderia pseudomallei, was classified as CABSI regardless of when samples were taken with the assumption that the acquisition of those infection due to occupational exposure to pigs and environmental exposure to contaminated environment are unlikely in the hospital setting respectively [18,19].
Blood for culture, approximately 5-10 mL of blood, was collected in a single commercially sourced aerobic bottle and incubated in an automated system (Bactec, Becton Dickinson, USA) for five days (5-10 mL blood). Bottles that flagged positive were Gram-stained, subcultured and identified using standard microbiological techniques, including VITEK and API testing (bioMérieux, France). Antibiotic susceptibilities were performed using CLSI guidelines with breakpoints from 2013 [20]. For S. suis, the susceptibility was evaluated using CLSI breakpoints for Streptococcus viridans group [21].
Fatal cases were defined as patients who died in hospital or were discharged with diagnosis of brain death and the expectation that they will die imminently within hours or days (palliative discharge to die at home, a common practice in Vietnam).

Data entry and statistical analysis
Data were entered in Epidata (EpiData Association, Odense, Denmark) and analysed using SPSS 22.0.0 (IBM Co., Armonk, NY, USA). Standard descriptive statistics were calculated for categorical (in percentage) and continuous variables (median and interquartile, IQR). Bivariate analyses were performed using Pearson's chi-squared test or Fisher's Exact Test for categorical variables as appropriate. The possible selection bias was assessed by chi-square and t-test analyses to compare gender and age differences between those with and without clinical outcomes, respectively. Cox proportional-hazards regression was used to identify variables that predict clinical outcomes. Cox's proportional hazards regression was performed to analyse factors associated with all-cause in-hospital mortality. BSI episodes of polymicrobial infection and S. maltophilia infections were excluded from regression analysis due to a small number and an uncertain pathogenicity respectively. In cases where there were multiple blood cultures positive, we only included the first positive blood culture. All tests were two-tailed and differences were considered statistically significant at p values ≤0.05.

Study population's demographics
There were 894 positive blood cultures corresponding to 835 BSI episodes in 825 patients recorded between January 2011 and December 2013. Notes from 543/ 825 (65.8%) patients could be retrieved to collect clinical details of individual patients while 282 (34.2%) were not available ( Fig. 1). Sensitivity analysis of age and gender showed no differences between cases with and without clinical outcome (chi-square test, p = 0.664 and t-test, p = 0.053 respectively). Among collected records, sixty-six cases (12.2%, 66/543) were subsequently classified by cultured organism as having a pseudo-bacteremia due to contamination, including 16 cases with Burkholderia cepacia, 19 cases with Alcaligenes species, 5 cases with Ralstonia pickettii and 4 cases with Pseudomonas species (2 P. putida and 2 P. stutzeri). There were 3 cases with Serratia marcescens, 3 S. maltophilia, 1 K. pneumoniae and 1 E. coli which were classified as contaminants by treating doctors at the time of results return, based on patient's survival and recovery without antibiotic treatment for bacterial disease and with an alternative possible viral diagnosis. Other contaminated cultures were 3 cases with Ochrobactrum anthropi, 1 case with Chryseobacterium indologenes, 1 case with Chryseomonas luteola, 1 case with (non-diphtheroid) Corynebacterium spp., 1 case with Staphylococcus spp. and 7 cases with single culture of viridans group streptococci without an endovascular or other focus. As a result, 477 cases were included for the clinical analysis.
In patients with CABSIs, 12.5% (50/400) had multiple positive blood cultures with either S. aureus or K. pneumoniae. Whilst most patients with CABSI were admitted directly from the community, 153/400 patients (38.3%) were transferred from other hospitals within 48 h, with 62/153 patients (40.5%) recorded as having received antibiotics before admission to the study hospital.
The overall proportion of extended spectrum betalactamases (ESBL) production among Enterobacteriaceae was 25.1% (67/267 isolates, 78 missing result). Tables 3 and 4 display the antibiotic resistance of K. pneumoniae and E. coli. The prevalence of ESBL production among   Table 5).
K. pneumoniae isolates associated with HABSI more frequently showed antibiotic resistance compared to CABSI isolates, though statistical significance was only reached for quinolones, while the pattern of resistance among E. coli isolates did not differ between the hospital and community (Tables 3 and 4). MRSA rates differed between community (38.7%, 12/31) and hospital settings (0/7), though this did not reach statistical significance (p = 0.074). Case notes were not retrievable in 69/110 (62.7%) cases of S. aureus bacteremia.

Discussion
This retrospective study includes 3-years of complete microbiological data on aetiology of BSI for 738 patients and almost two thirds (64.6%, 477/738) of these patients for clinical outcome analysis in a tertiary hospital in the north of Vietnam.
Our study shows the important role of pathogenic Gram-negative bacteria with a high proportion of cephalosporin-resistant E. coli and K. pneumoniae isolates and high prevalence of MRSA in CABSI. Our study shows the high contribution of Enterobacteriaceae BSI to mortality among adults admitted to our hospital in Vietnam, with over 60% of deaths associated with this group of bacteria. This study supports the observed shift within Enterobacteriaceae organisms from typhoidal and non-typhoidal Salmonella to K. pneumoniae and E. coli, the current leading causes of community acquired blood stream infections.
In Vietnam, Salmonella enterica serovar Typhi was reported to be the major cause of BSI, accounting for up to 74% of isolates from blood in the 1990s but gradually declined to 6% by 2008 [22]. With the recent emergence of K. pneumoniae and E.coli, we observed the high prevalence of extended spectrum Beta-lactamase (ESBL) production in these pathogens, especially in the hospital setting. In surveillance data from five Asia-Pacific countries in 2010, Vietnam reported the highest rate of ESBL positive Enterobacteriaceae isolates with an overall prevalence of 55.1% and 80.1% in ICU patients [11]. The high prevalence of antibiotic resistance in Vietnam may be attributed to multiple factors, such as dispensing practices in the community, lack of enforcement of regulation, limited resources for microbiological diagnostics, injudicious use of antibiotics in hospital, community and agriculture and insufficient antibiotic resistance surveillance and infection control [23].
In our study, ESBL production was detected among 6.1% K. pneumoniae from CABSI, which is consistent with the low ESBL prevalence of 4.1% among K. pneumoniae oropharyngeal carriers reported previously from northern Vietnam [24]. The rate of ESBL-production in E. coli(45%) was higher than in Thailand (CABSI, 11.8%) and Korea (31.3% in HABSI and 8.8% in CABSI) and comparable to the prevalence of ESBL among E. coli isolates in BSI in Cambodia (47.7%), [15,17,25]. The high rate of ESBL carriage in E. coli from CABSI is deeply worrying and likely reflects extensive overuse of oral cephalosporins in the community, their use in agriculture or other environmental sources [26,27]. S. suis is a wellrecognized zoonotic pathogen associated with bacterial meningitis and BSI among adults in Vietnam [10] and was noted as an important Gram-positive pathogen cultured from blood with high levels of susceptibility to penicillin.
Our study showed a higher case-fatality rate (110/400 or 27.5%) among community-acquired bacterial bloodstream infections than the rate of 9% reported in a review from 1990 to 2010 in south and southeast Asia and the rate of 18.1% among patients with CABSI in a meta-analysis in Africa from 1984 to 2006 [28,29]. This difference is likely related to the changes in spectrum of aetiology with higher prevalence of Salmonella spp. (53.2% of Enterobacteriaceae isolates) in the past [28], the emergence of high levels of antibiotic resistance in Vietnam and the tertiary referral nature of the study hospital. The overall mortality in our study was slightly higher than the overall mortality of 22.5% in Cambodia  where the epidemiology and pattern of resistance are similar [15]. We reported a high mortality of 35.6% among E. coli BSI, which is higher than mortality of 18.2% and 8% and in a large surveillance in England and in a population-based study in Finland [30,31]. It is not clear about the reasons of the differences, but it may relate to pneumonia focus, severity of infection, antibiotic susceptibilities and seasonal variation [31]. As NHTD is a tertiary referral hospital, potentially more severe cases or cases that have failed therapy elsewhere may have been referred as compared to the hospitals included in other studies, leading to higher resistance and case fatality rates. In North America and Europe, the nosocomial BSI case fatality rate ranges from 12 to 32% [2]. Changes in the aetiology of HABSI began in the 1960s-1970s in the United States of America (US) with replacement of Gram-negative by Gram-positive organisms, such that 65% of BSIs were S. aureus, the most common nosocomial pathogen [32]. However, a Vietnamese population based study indicated S. aureus nasopharyngeal carriage proportion (33.8%, CI 29.4-38.8) was higher than in European regions and the States [33], and S. aureus BSI is still mainly community-acquired in Vietnam [34]. In Vietnam the proportion of BSI attributable to S. aureus did not change during the 15-years from 1994 to 2008 [22]. Gram-negative bacilli are still recognized as the most frequent nosocomial isolates in developing countries [35]. In a multicenter surveillance study in Thailand, Gram-negative bacteria accounted for 67.6% of HABSI, of which Acinetobacter spp. (16.2%) was detected most frequently, followed by K. pneumoniae (13.9%) and S. aureus (13.9%) and the in-hospital mortality was 26.6% [36]. The overall case fatality rate of 36.4% among HABSI in our study population was higher than seen in North America and Europe. This may be related to higher levels of antibiotic resistance, differences in the availability of resources or the nature of admissions to NHTD, as described earlier.
We observed a predominance of males among patients with BSI (72.3%). This phenomenon was similar as in the cause of death reports in another tertiary hospital in the north of Vietnam where the top three of causes of death were septic shock, intracerebral haemorrhage and pneumonia. The proportion of males in this report was higher than females (1577/2340 or 67.4%) [37].
Among common isolates causing BSI, we reported the high prevalence of S. maltophilia (9.6%. 71/738) which is only rarely considered a pathogen, particularly from the community [32]. Whilst S. maltophilia was described as a recognised cause of CABSI among patients with  comorbidities in a systematic review [38] and a recent publication from Taiwan [39], another publication describes contamination of blood culture bottles with S. maltophilia due to growth of the bacteria in disinfectant [40]. Furthermore, we have observed pseudoepidemics of S. maltophilia in this hospital (multiple patients from geographically disparate sites with different infection syndromes, newly admitted on the same ward with S. maltophilia cultured from blood at the same time). The retrospective nature of this study makes it difficult to define whether these isolates were true pathogens or represented contamination. Given how frequently this organism was isolated, further prospective investigation is required. Our study has some important limitations. Firstly, NHTD is a tertiary referral hospital and specializes in infectious diseases. Therefore, the pattern of aetiology, resistance and spectrum of clinical disease is different from that seen in district and provincial level hospitals due to referral of certain syndromes, more severe and non-responsive cases.
Secondly, during the study period, only a single blood culture was routine collected and as a cost saving technique, routine laboratory practice was that cultures showing obvious contamination (multiple organisms, skin flora etc.) were discarded before full work up without recording this finding. This will have lowered estimations of blood culture sensitivity, which can be improved with more blood cultures and higher volumes of blood [41,42]. Additionally, although the isolates of Enterobacteriaceae members are almost always interpreted as true bloodstream infections, single blood culture practice may underestimate the roles of fastidious pathogens in the era of increased catheters and prosthetic devices-related bloodstream infections [43].
In our study hospital acquired BSI was defined as cases with positive blood culture after any hospitalization for more than 48 h. Thus, patients who in reality had CABSI but with persistent bacteraemia, who were transferred after an inpatient stay at another hospital may have been misclassified as HABSI. The significance and role of certain bacteria, such as S. maltophilia or Aeromonas species, which have been associated with both pseudobacteraemia and real infection was not easy to establish because of a lack of repeat sampling in most of patients and the retrospective nature of the study. Lastly, clinical outcome analysis was only applied for 64.6% (477/738) of patients during the study period because the remaining patient's notes could not be retrieved. This has resulted in significant gaps in the data and may have introduced bias if the sample was not random.

Conclusions
Gram-negative bacteria are the most common cause of both community and hospital acquired blood stream infection in adults presenting to our tertiary referral hospital, with high associated mortality. The extended spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae are prevalent in both community and health care setting. Our findings further support the need, as described in the GARP situation analysis [44] and the 2013 Vietnamese National Action Plan for AMR [45], for establishment of a comprehensive microbiological and antibiotic susceptibility surveillance system in Vietnam, with representation of the different levels of healthcare and to inform treatment guidelines and other public health measures.