Bacterial etiology of bloodstream infections and antimicrobial resistance in Dhaka, Bangladesh, 2005–2014

Bloodstream infections due to bacterial pathogens are a major cause of morbidity and mortality in Bangladesh and other developing countries. In these countries, most patients are treated empirically based on their clinical symptoms. Therefore, up to date etiological data for major pathogens causing bloodstream infections may play a positive role in better healthcare management. The aim of this study was to identify the bacterial pathogens causing major bloodstream infections in Dhaka, Bangladesh and determine their antibiotic susceptibility pattern. From January 2005 to December 2014, a total of 103,679 single bottle blood samples were collected from both hospitalized and domiciliary patients attending Dhaka hospital, icddrb, Bangladesh All the blood samples were processed for culture using a BACT/Alert blood culture machine. Further identification of bacterial pathogens and their antimicrobial susceptibility test were performed using standard microbiological procedures. Overall, 13.6% of the cultured blood samples were positive and Gram-negative (72.1%) bacteria were predominant throughout the study period. Salmonella Typhi was the most frequently isolated organism (36.9% of samples) in this study and a high percentage of those strains were multidrug-resistant (MDR). However, a decreasing trend in the S. Typhi isolation rate was observed and, noticeably, the percentage of MDR S. Typhi isolated declined sharply over the study period. An overall increase in the presence of Gram-positive bacteria was observed, but most significantly we observed the percentage of MDR Gram-positive bacteria to double over the study period. Overall, Gram positive bacteria were more resistant to most of the commonly used antibiotics than Gram-negative bacteria, but the MDR level was high in both groups. This study identified the major bacterial pathogens involved with BSI in Dhaka, Bangladesh and also revealed their antibiotic susceptibility patterns. We expect our findings to help healthcare professionals to make informed decisions and provide better care for their patients. Also, we hope this study will assist researchers and policy makers to prioritize their research options to face the future challenges of infectious diseases.


Background
Bloodstream infection (BSI) due to bacterial pathogens is a global concern. It is often associated with increased length of hospital stay, a significant amount of healthcare related costs and most significantly, a high rate of morbidity and mortality [1]. Depending on the age, severity of infection and other risk factors, the mortality rate for BSI varies between 4.0 and 41.5% [2][3][4][5][6][7]. Recent studies have reported a rapid increase in the number of bloodstream infections from both community and nosocomial sources [8,9]. Staphylococcus aureus, Streptococcus pneumoniae and Escherichia coli have been found to be the most commonly isolated pathogens associated with BSI worldwide [3,6].
The epidemiology of BSI varies depending on the geographic location, age and co-morbid illnesses. As an example, Salmonella enterica is a frequently isolated pathogen from blood samples in both African and Asian regions, however their serotypes differ substantially [10]. S. Paratyphi is the predominant organism in the Salmonella group in Africa whereas S. Typhi is the most frequently isolated organism in Asia. Besides their isolation rate, their antibiotic susceptibility pattern varies substantially [10]. So, understanding of local epidemiology may play an important role in making proper empirical treatment choices before laboratory test results are available. This is especially true for Bangladesh and other developing countries where healthcare systems operate on poor hygiene system and lack proper facilities to contain infections. In these countries, early treatment is usually based on the patient's clinical symptoms rather than diagnostic results. Therefore, patient's early prognosis to final outcome might be much improved by available epidemiologic data for the most frequently isolated pathogenic organisms. However, complete data on BSI causing organisms from Dhaka, Bangladesh is scarce.
In this study, we aimed to identify the most prevalent bacterial pathogens involved in BSI in hospital and domiciliary patients from Dhaka, Bangladesh. We also determined pathogen antibiotic susceptibility patterns to evaluate the changing trend of antimicrobial susceptibility in this region.

Methods
In this retrospective study, blood samples were obtained from patients attending out-patient and in-patient services at Dhaka hospital, icddrb, Bangladesh which is a primary care hospital with 200 in-patient beds. A total of 103679 blood samples were processed from January 2005 to December 2014. In the consecutive ten years of the study 9600, 9364, 9189, 9523, 9523, 11578, 11179, 10822, 10161 and 12740 samples were received and processed from 2005 to 2014 respectively. All the blood samples were processed for culture using a BACT/Alert blood culture machine to identify the presence of bacterial pathogens. Antimicrobial susceptibility tests were performed on the isolated pathogens using standard microbiological procedures [11].

Bacterial isolation
Collected blood samples were directly inoculated into adult (more than 12 years of age) and pediatric (up to 12 years of age) FAN blood culture bottle. Bottles were incubated in the BACT/Alert machine for up to 5 days. Positive culture samples were directly inoculated onto MacConkey (MC) agar, chocolate agar and blood agar (5% sheep blood) plates. MC plates were then incubated at 35°C in aerobic condition. Chocolate and blood agar plates were incubated at 35°C in microaerophilic condition (containing 5% CO 2 ). Bacterial pathogens were identified using standard bacteriological procedures [11]. API identification strips (bioMérieux, France) were used as supportive tests for further identification.

Statistical analysis
Blood-borne pathogen trends for isolation and antimicrobial susceptibility over the last ten years was determined using χ 2 (chi-square) test for trend in SPSS version 16.0. P value ≤ 0.05 was considered significant. Standard error of the mean (SEM) was calculated for mean isolation rates. Odd ratios for age and sex were calculated to show their association with infection by particular organisms, and SPSS version 16.0 was used for this purpose. For age, an odds ratio greater than 1.0 indicated the association of a pathogen with age group of less than five years old, and an odds ratio of less than 1.0 indicated the association of a pathogen with age group of more than five years old. For sex, an odds ratio of less than 1.0 indicated the association of a pathogen with female patient group and an odds ratio of more than 1.0 indicated an association of that pathogen with male patient group.

Results
From January 2005 to December 2014, a total of 103,679 blood samples were received from both hospitalized and domiciliary patients and among them 14015 samples were found to be culture positive. Over these past ten years 11.9, 12.2, 13.9, 16.5, 16.0, 17.3, 12.7, 12.9, 10.7 and 11.9% of the samples were found to be culture positive (P < 0.001) respectively. The mean culture positive rate was 13.6 ± 0.7%. Table 1 shows the distribution of organisms found throughout this study period. Blood samples were received from patients with age range of 1 day to 115 years and the mean age was 18 years. S. Typhi was the most frequently isolated blood-borne bacterial pathogen in this study, accounting for 36.9% of the total isolates. Other frequently isolated pathogens included coagulase-negative Staphylococcus species (21.5%), Pseudomonas species (12.5%), S. Paratyphi A, B (8.9%) and Acinetobacter species (5.1%).
Pseudomonas species and Acinetobacter species were the two major non-fermenter bacteria isolated between 2005 and 2014. Pseudomonas species showed a sharp increase in isolation rate, from 40.6 to 79.2% (χ 2 = 105.1, P < 0.001) of the total non-fermenter bacteria, between 2005 and 2010. However, for the next four years their isolation rate decreased slightly and reached 74.3% in 2014. Acinetobacter species showed a decreasing trend in their isolation rate over this study period, from 49.8 to 24.4% (χ 2 = 95.5, P < 0.001). Salmonella species accounted for 46.7% of the total blood-borne pathogens. We observed an overall decreasing trend in their isolation rate from 2005 to 2014 (χ 2 = 306.5, P < 0.001). Besides nonfermenter and Salmonella species, other Gram-negative bacteria constituted only 7.7% (χ 2 = 72.0, P < 0.001) of the total blood-borne pathogens. E. coli, Enterobacter species and Serratia species had a steady isolation rate over the ten year period, while Klebsiella species showed an increasing trend in their isolation rate, from 22.6 to 50.9% (χ 2 = 29.1, P < 0.01) of the total Gram-negative bacterial group (excluding Salmonella species). Gram-positive bacteria accounted for 27.7% of the total blood-borne pathogens. Their isolation rate increased from 20.6 to 30.8% (χ 2 = 351.7, P < 0.001) over study period. From 2005 to 2014, Streptococcus pneumoniae showed a decreasing trend in their isolation rate, from 20.8 to 5.6% (χ 2 = 158.7, P < 0.001) of the total Gram-positive bacteria.
S. Typhi was found to be consistently sensitive to ceftriaxone and cefixime over this study period; however, a strain of ESBL S. Typhi was reported [22]. At the same time, decreasing trends of resistance against ampicillin The bolded data indicates the total number and percentage of that group (χ 2 = 540.4, P < 0.001) and co-trimoxazole (χ 2 = 740.3, P < 0.001) were observed. S. Paratyphi A, B was found to be consistently susceptible to ampicillin, co-trimoxazole, ceftriaxone and cefixime over the ten year period. Noticeably, both S. Typhi and S. Paratyphi A, B showed a trend of increasingly reduced susceptibility against ciprofloxacin (χ 2 = 24.3 and 66.1 respectively, P < 0.001) throughout this study period (Table 3). E. coli and Enterobacter species showed an increasing trend of resistance against gentamicin (χ 2 = 52.44 and statistically insignificant respectively, P < 0.001) and ceftriaxone (χ 2 = 52.4 and 16.5 respectively, P < 0.001). A reducing trend of susceptibility against ciprofloxacin (χ 2 = 89.9 and 86.8 respectively, P < 0.001) was also observed (Table 4 and Additional file 5: Table S4). Additionally, E. coli showed an overall increase in resistance, from 69.0 to 90.0%, against co-trimoxazole and ampicillin between 2005 and 2014. Klebsiella species showed an increasing trend of resistance against imipenem (χ 2 = 79.6, P < 0.001), and a reducing trend of susceptibility against ciprofloxacin (χ 2 = 36.1, P < 0.001) (Additional file 6: Table S3). Over the same time period an overall increase in resistance, from 62.0 to 76.0%, was observed against gentamicin and ceftriaxone respectively for Klebsiella species. Additionally, Klebsiella species showed an overall 86.0, 33.0 and 70.0% resistance against cefixime, meropenem and azithromycin respectively from 2010 to 2014.
S. pneumoniae was found to be consistently susceptible to ampicillin and ceftriaxone between 2005 and 2014 (Table 5). Over this period, their resistance to cotrimoxazole reduced (χ 2 = 92.5, P < 0.001). However, it increased against penicillin G (χ 2 = 195.3, P < 0.001) and erythromycin (χ 2 = 293.2, P < 0.001). Also, a trend of increased susceptibility was observed against ciprofloxacin (χ 2 = 79.2, P < 0.001). Other Streptococcus species showed an increasing trend of resistance against gentamicin (χ 2 = 44.9, P < 0.001) and erythromycin (χ 2 = 70.0, P < 0.001), while there was a reducing trend in susceptibility against ciprofloxacin (χ 2 = 67.3, P < 0.001) (Additional file 7: Table S5). S. aureus was found to be highly resistant to ampicillin (up to 100%) throughout   The bolded data indicates the total number and percentage of that group this study (Table 6). For this species, resistance against erythromycin increased from 36.0% in 2005 to 75.0% in 2014 (P < 0.09) while susceptibility to ciprofloxacin reduced (χ 2 = 37.2, P < 0.001). S. aureus remained consistently susceptible to vancomycin for the last five years of the study and during the same time period they showed an overall 42.0% resistance against ceftriaxone. E. faecalis was found consistently susceptible to vancomycin between 2010 and 2014, while this species showed an overall 31.0, 45.0, 60.0, 39.0 and 81.0% resistance against ampicillin, gentamicin (120), ciprofloxacin, penicillin G and cotrimoxazole respectively (Additional file 8: Table S6).

Discussion
In this retrospective study, we aimed to identify the most prevalent pathogenic organisms involved in bloodstream infections (BSI) over a ten year period  in patients in Dhaka, Bangladesh. We also aimed to determine the antimicrobial susceptibility of the isolated pathogens against multiple antibiotics to achieve a clear outlook on the changing trend of their antibiotic susceptibility. S. Typhi, the causative agent of typhoid fever, is a major public health concern in Bangladesh and other developing Asian countries. Several studies from Bangladesh have already identified S. Typhi as a common cause of bloodstream infection in this region [23][24][25]. We found Salmonella species to be responsible for almost half of the disease burden associated with BSI in Dhaka, Bangladesh and about 80% of these infections were due to S. Typhi. However, we have observed an overall decrease in Salmonella species isolation rate over this study period. This decrease may be attributed to the improved urban water management system and sanitation practices in Dhaka city over the past years.
A significant change in the epidemiology of S. Typhi was observed in early 1990's as those years experienced a dramatic rise of MDR strains in Dhaka, Bangladesh [26]. By mid-1990s, about half of the S. Typhi strains were MDR; these were resistant against three first line antibiotics -ampicillin, cotrimoxazole and chloramphenicol [26]. Noticeably, after a few years of an initial epidemic period, a decreasing trend in the isolation rate of MDR S. Typhi strains was observed [23]. Our study confirms this trend to continue; from 2005 to 2010, the percentage of MDR S. Typhi strains isolated declined from 61.7 to 23.7% (Fig. 2). As indicated by recent studies from Bangladesh, S. Typhi still shows a high level of resistance against first line antibiotics [23]. However, we have observed a steady decrease in resistance against ampicillin and cotrimoxazole over the ten year study period. Hopefully, if this trend continues, then using the cheaper, first line antibiotics again to treat S. Typhi infections might be a possibility in near future. Ciprofloxacin or ceftriaxone is   usually the choice of treatment against MDR Salmonella strains [27]. We found all Salmonella isolates were consistently susceptible to ceftriaxone, but a very high level of reduced susceptibility exists against ciprofloxacin. Studies suggest that the presence of a mutation in the quinolone resistance-determining region (QRDR) of the gyrA gene is responsible for the emergence of this reduced susceptibility [28]. Non-fermenter Gram-negative bacilli are well known for their ability to cause nosocomial infections [29,30]. We identified Pseudomonas species and Acinetobacter species to be the third (12.6%) and fifth (5.2%) most prevalent organisms associated with BSI in Dhaka, Bangladesh. A gradual decrease in the presence of Acinetobacter species and a significant increase in the presence of Pseudomonas species were observed over the study period. This increase in BSI from Pseudomonas species was possibly from the patients increased access to the health care facilities over these years. Also noticeably, Pseudomonas species was found to be more associated with patients above five years of age; almost 70% of the infections occurring among adults (≥18 years of age). A recent study from Bangladesh also suggested that there is a very low amount of bacteremia cases due to Pseudomonas species among diarrheal children less than five years of age [31].
Acinetobacter species and Pseudomonas species are well known for their high degree of resistance against all classes of antibiotics, and the emergence of MDR strains makes it very difficult to treat them. An aminoglycosides combined with a β-lactamase-stable β-lactam is the typical choice of treatment for Acinetobacter species infection [32], and carbapenems are known to be most effective against their MDR strains [33]. We observed Acinetobacter species to develop increasingly high levels of antibiotic resistance against all aminoglycosides, and noticeably, very high levels of resistance against imipenem (64%). Usually, in the case of resistance against carbapenems, the expert's choice of treatment for Acinetobacter species infections entails colistin and tigecycline [34]. We found only 3.73% of the MDR Acinetobacter species strains to be resistant against polymyxin B (300 u), therefore it might be considered as a good treatment choice against MDR Acinetobacter species. Colistin and polymyxin B (300 u) are usually considered as the last resort for treatment of infections due to MDR Pseudomonas species [35], but we observed high levels of resistance against them. In contrast, we found imipenem to be consistently effective against all Pseudomonas species so it might be considered as a good choice of treatment.
Besides non-fermenter and Salmonella species, other clinically important Gram-negative bacteria were E. coli, Enterobacter species, Klebsiella species and Serratia species. E. coli has been reported as the leading cause of BSI in developed countries [1]; however, data is relatively unavailable for Bangladesh. We found E. coli to be responsible for only 3.05% of the total BSI cases. It is likely that the presence of E. coli is overshadowed by the overwhelming presence of Salmonella species in Dhaka, Bangladesh, so it would be unwise to ignore them as clinically insignificant.
A recent study from Bangladesh has shown a very high level of antibiotic resistance and MDR level among E. coli isolates from environmental samples [36]. In this study we identified a much higher level of antibiotic resistance and MDR level (36% MDR in environment samples vs. 62.38% MDR in our clinical samples) among BSI causing E. coli strains. We found carbapenems to be the  Gram-positive bacteria also pose serious threats as their incidence in BSI is increasing steadily worldwide. The problem is particularly acute in nosocomial settings where methicillin-resistant S. aureus (MRSA) and vancomycin resistant Enterococcus (VRE) are considered serious health threats [37,38]. In this study, we observed an increasing trend in the isolation rate of Grampositive bacteria. This trend may be attributed to the patient's increased access to the health care facilities where BSI from Gram-positive bacteria is common. We identified S. aureus and S. pneumoniae as the two major Gram-positive pathogenic bacteria associated with BSI in Dhaka, Bangladesh. S. pneumoniae reportedly kills about 0.7 to 1 million children under five years of age every year worldwide and 90% of these deaths occur in developing countries [39]. Pneumonia is the primary cause of childhood death in Bangladesh [40] and studies suggest that the incidence of pneumonia among preschool children in Bangladesh is surprisingly higher than in developed countries [41,42]. We found almost 75% of the BSI cases from S. pneumoniae to occur among children less than five years of age. However, a decreasing trend in their isolation rate was observed throughout this study period.
Increasing levels of antibiotic resistance are a big concern for S. pneumoniae, especially their resistance against macrolides and β-lactams, which are the first choice of treatment for most pneumonia cases. One recent review indicates a very high level of penicillin resistance among S. pneumoniae strains in different Asian countries [43]. However, we observed only a small increase in penicillin resistance in Dhaka, Bangladesh. Another study from Dhaka indicates the presence of high level of resistance against ciprofloxacin. However our study found a much lower level of resistance against ciprofloxacin and also we observed a decreasing trend in their susceptibility pattern [44]. Ampicillin and ceftriaxone were found to be consistently active against S. pneumoniae, so they might be considered as good treatment choices for them. Emergence of MRSA strain marks another very important epidemiological change for the past few decades [45]. Naturally, S. aureus are susceptible to commonly used antibiotics. However methicillin resistance is associated with resistance against a broad range of antibiotics [46]. Over the study period, we found overall 50.2% of S. aureus strains to be MDR. In the first five years of our study, we identified 34.4% of the S. aureus strains to be oxacillin (a synthetic form of methicillin) resistant; however, the percentage was much higher (58.7%) amongst the MDR strains. Vancomycin has been widely used as the preferred choice of treatment for MRSA strains [47], and we also found vancomycin to be highly effective against all S. aureus strains; only 2.1% strains were resistant. Linezolid was also found to be very effective against S. aureus; only one resistant strain was detected over the study period. Therefore, both vancomycin and linezolid might be considered as a good treatment choices against S. aureus which are resistant to commonly used antibiotics.

Limitations of the study
Due to the lack of resources, we were not able to differentiate the samples received from our hospital patients and from domiciliary patients. As a result we could not show the difference in epidemiology between nosocomial and community acquired BSI. Also due to lack of resources, we were not able to collect patient data on the clinical manifestations or any other patient characteristics, other than age and sex, which could be considered as risk factors for BSI. However, we confirm that each patient was diagnosed with signs of bacteremia by their respective physicians. Also, we were not able to perform any molecular tests on received samples due to lack of required resources.