Introduction

Sepsis is a leading cause of neonatal death globally and the burden of neonatal sepsis is greatest in the resource-poor countries.1, 2 Neonatal sepsis is due to a wide range of bacterial pathogens that challenges the development of effective antibiotic guidelines.3 The range and relative importance of pathogens causing neonatal sepsis may vary between geographical settings and depends upon other variables such as the timing of infection (for example, perinatal or postnatal) or whether the infection was community-acquired or hospital-acquired.3, 4, 5, 6 There has also been an increase in the prevalence of neonatal sepsis due to multidrug-resistant pathogens, particularly common as hospital-acquired infections.6, 7, 8 However, recent reviews highlight that data from the resource-limited setting remain sparse.3, 9, 10

The region of south-east Asia is one of the most populous regions in the world, and neonatal mortality is responsible for a large and increasing proportion of overall child mortality in this region.2 Yet, a recent systematic review found that there are very few publications describing the causes of neonatal morbidity and mortality, and in particular the prevalence of neonatal sepsis from the south-east Asian region, especially from the resource-limited settings within that region.11 A recent multisite study from the south-east Asia region reported the causes and outcomes for neonatal sepsis, but the sites included were neonatal intensive care units in resource-rich settings.12

We recently undertook a prospective study of all-cause morbidity and mortality among newborns admitted to a neonatal unit in central Vietnam.13 Here, we report the prevalence, causes and outcome of newborns with suspected sepsis, a summary of the pathogen isolated from blood culture and antibiotic susceptibility in vitro.

Methods

We conducted a prospective cohort study including all newborns admitted to the Da Nang Hospital for Women and Children, Vietnam, over a 1-year period (1 November 2010 to 31 October 2011). Final follow-up data for all study participants were collected up to 10 weeks after enrollment. Details of the study setting and overall methodology are presented elsewhere.13

Study setting

The neonatal unit of Da Nang Hospital is the largest neonatal unit in central Vietnam. All sick neonates born in Da Nang city are referred to the unit, as well as critically ill neonates from the surrounding provinces. Criteria for inborn admission are all newborns with birth weight <2000 g or sick neonates of any weight or gestation. The unit has a 60-cot capacity that manages ~60 to 100 inpatients each day, and is made up of a neonatal intensive care unit of 30 cots and a high dependency unit of 30 cots. The neonatal intensive care unit has a nurse:patient ratio of 1:6, has six ventilators for intermittent positive-pressure ventilation and can provide continuous positive airway pressure for an additional 15 neonates. Central catheters are used by umbilical vein or peripherally inserted central catheters, and total parenteral nutrition is provided for eligible preterm neonates. The high dependency unit has a nurse:patient ratio of 1:20. There is one basin for hand-washing per five cots, and bedside alcohol is provided. The median length of stay for all newborns during this study period was 6 (interquartile range 4 to 9) days.13

Study procedures

Data were collected on all admissions using a standardized case report form and included demographics, antenatal history, birth history, birth weight and gestational age, primary and secondary diagnoses and clinical findings. Clinical progress was closely monitored as per routine nursing care, and clinical data of progress were collected using the standardized case report until death or discharge.

A septic work-up was undertaken before commencing antibiotics for neonates with suspected sepsis on the basis of clinical investigation: such as the presence of fever, hypothermia, mottling skin, respiratory distress (which cannot be explained by respiratory conditions such as the respiratory distress syndrome or meconium aspiration), poor feeding, diminished activity and lethargy or seizures. A septic work-up included at least a venipuncture for full blood count, C-reactive protein and blood culture with additional investigations depending on the clinical presentation. The standard first-line antibiotic treatment for suspected sepsis in the neonatal unit is a combination of ampicillin and gentamicin.

For blood culture, 2 ml of venous blood were drawn and incubated in the brain–heart infusion broth at 37 °C for 7 days. If bacteria grew, Gram stains were made and the samples were subsequently cultured on blood agar and MacConkey agar to identify the bacteria. Sabouraud agar was used to isolate cases of suspected fungi. Susceptibility to antibiotics was assessed by performing a disc diffusion technique using Mueller–Hinton agar.

Confirmed sepsis was defined as a positive culture of a single pathogen from venous blood or other normally sterile sites. Early-onset sepsis (EOS) was defined as sepsis occurring before or at 72 h of age, whereas late-onset sepsis (LOS) was defined as sepsis occurring after 72 h of age. Nosocomial infection was defined as sepsis that occurred after 72 h of admission to the neonatal unit and is a subset of LOS.

Data analysis

All analyses are presented treating each admission as an independent event, ignoring the fact that occasionally there were repeat admissions in the same infant. As repeat admissions were rare (<1%), this was not expected to affect the results. The prevalence of EOS and LOS is presented as the proportion of neonates with sepsis relative to the total number of admissions. Risk factors including maternal infection, gestational age and birth weight for EOS or nosocomial sepsis were explored using logistic regression, with results presented as odds ratios and 95% confidence interval. Initially, each predictor was assessed using a separate univariable model, with multivariable regression then used to identify independent predictors among those significant (P<0.05) in the univariable model. Variables included in the multivariable model for nosocomial sepsis included gestational age, use of central catheter, having had surgery or transfusion. Gestational age was highly correlated with birth weight, and so the birth weight was not included in multivariable analysis. We did not include invasive respiratory support into the multivariable model because respiratory support is also a consequence of infection. The numbers and proportions of pathogens were isolated and their antibiotic susceptibility were calculated.

Ethical approval

The study was approved by the Scientific Board of Da Nang Hospital and the Human Research Ethics Committee of the University of Melbourne.

Results

Overall, there were 2555 neonatal admissions and 220 deaths (8.6%) during the 1-year study period. Of the admissions, 1364 (53%) were inborn, 684 (27%) were born at home and presented directly to the hospital, and 507 (20%) were referred in from outlying health facilities. There were 616 neonates having 729 episodes of suspected invasive sepsis with blood cultures performed, and of the episodes, 223 (31%) presented as suspected EOS and 506 (69%) as suspected LOS. Pathogens were isolated from blood in 115 (16%) episodes of suspected sepsis, and comprises 75 bacterial pathogens and 40 fungal pathogens. There were no pathogens isolated from other sites. There were five newborns with a clinical diagnosis of meningitis, but none had a positive culture from the cerebrospinal fluid.

The 115 episodes of confirmed sepsis with positive blood cultures occurred in 106 neonates, which represented 4% of all admissions. There were 99 neonates with 1 pathogen isolated from a single episode. The other seven neonates had a positive blood isolate from two (n=5) or three (n=2) separate episodes, with it being two or more different pathogens in six of these. The median (interquartile range) time to detect a pathogen was 5 (refs. 4, 5, 6, 7) days.

Pathogens and characteristics associated with onset of sepsis

Twenty neonates (19% of those with confirmed sepsis) had EOS, equating to a prevalence of 8 (95% confidence interval: 4 to 11) per 1000 admissions, and there were 8 deaths. The commonest causes of confirmed EOS were coagulase-negative Staphylococcus (CoNS), Staphylococcus aureus and Acinetobacter (Table 1). Eighty-six neonates (81%) had LOS, equating to a prevalence of 34 (95% confidence interval: 27 to 41) per 1000 admissions, and there were 41 deaths. The commonest causes of confirmed LOS were Candida, and other fungi, followed by Acinetobacter, CoNS and Klebsiella pneumoniae. Of the 86 neonates with LOS, 69 (80%) also fulfilled the criteria for nosocomial sepsis, and so the spectrum of pathogens for nosocomial sepsis was similar to that for LOS, with fungal pathogens being the causative pathogen in around half of the nosocomial sepsis cases.

Table 1 Type and frequency of blood isolate in neonates with early-onset sepsis, late-onset sepsis and nosocomial sepsis
Full size table

Of the 75 episodes of bacterial sepsis, 19 (25%) were EOS and 56 (75%) were LOS (Table 1). Gram-positive bacteria were more commonly isolated in cases of confirmed bacterial sepsis with confirmed EOS (13/19 or 68%) than in those with confirmed LOS (15/56 or 27%) in which Gram-negative bacteria predominated. Almost all (93%) cases of fungal sepsis were among those classified as nosocomial sepsis. Fungal sepsis occurred in 35 neonates representing a prevalence of 14 per 1000 admissions. Among neonates with fungal sepsis, 69% were preterm (<37 weeks) neonates and 60% were <34 weeks’ gestational age. Of the confirmed nosocomial sepsis cases, 44% occurred in very low birth weight (<1500 g) babies.

Neutropenia (white cell count <5000 mm−3) and thrombocytopenia (platelet count <100 000 mm−3) were more commonly associated with episodes of confirmed sepsis compared with episodes with suspected but unconfirmed sepsis (13% versus 4% and 39% versus 14%, respectively; P<0.001 for both). A C-reactive protein ⩾48 mg l−1 was also significantly more common (53%) in episodes of confirmed sepsis than for suspected but unconfirmed sepsis (53% versus 31%, respectively; P<0.0001).

Table 2 explores patient characteristics as predictors of confirmed EOS and confirmed nosocomial sepsis in all admissions, that is, including those with suspected but unconfirmed sepsis plus those without suspected sepsis. There was evidence that maternal infection was associated with an increased odds of confirmed EOS, but not with nosocomial sepsis. There was little evidence that gestational age or being small for gestational age were associated with confirmed EOS. Confirmed nosocomial sepsis was strongly associated with prematurity <32 weeks and with the use of transfusions and central catheter insertion. Eighty (75%) of the neonates with confirmed sepsis had invasive respiratory support at some stage during their stay in the unit. This was more common for LOS cases than EOS cases, 83% (71/86) and 45% (9/20), respectively.

Table 2 Characteristics associated with the occurrence of confirmed early-onset sepsis and confirmed nosocomial sepsis
Full size table

Neonates with suspected but unconfirmed sepsis had a higher gestational age compared with neonates with confirmed sepsis (mean of 37 (s.d. 3) weeks versus mean 35 (s.d. 4) weeks, respectively, P<0.0001) and a lower case fatality rate (14.1% versus 46.2%, P<0.0001). Comparison of clinical features between LOS cases with suspected sepsis that were confirmed and those that were not confirmed found that lethargy, respiratory distress, apnea and need for increased respiratory support were all significantly associated with confirmed sepsis (data not shown).

It is uncertain how many of the CoNS isolates were contaminants. For example, only 2 of the 23 CoNS had a central catheter at the time of the blood culture. On the other hand, nine CoNS cases were EOS and CoNS, and would be considered an unusual cause of community-acquired EOS. Further, the use of respiratory support or transfusion was less common in those with CoNS than in those with other bacterial isolates: 35% (8/23) and 17% (4/23) versus 81% (42/52) and 60% (31/52), respectively.

Of the 40 neonates with fungal sepsis, 22 had a birth weight of <2000 g and all 22 received total parenteral nutrition. Those with fungal sepsis were more likely to have required respiratory support or had a transfusion than those with bacterial sepsis: 93% (37/40) and 70% (28/40) versus 67% (50/75) and 47% (35/75), respectively, Fisher’s exact test P=0.003 and 0.02. However, when CoNS were excluded from this analysis, there were no significant differences for these variable between fungal and bacterial sepsis cases.

Antibiotic susceptibility and usage

Results of antibiotic susceptibility testing for all 75 bacterial isolates are listed in Table 3. In vitro resistance to multiple antibiotics was common among Gram-negative bacteria, especially to cephalosporins and aminoglycosides. Klebsiella species were commonly resistant to four different antibiotic groups, and imipenem was the only antibiotic to which most Gram-negative pathogens showed in vitro susceptibility. Multidrug-resistant Gram negatives such as Klebsiella species and Pseudomonas were common in LOS or nosocomial sepsis cases (Table 3). All isolates of Staphylococcus showed in vitro susceptibility to vancomycin.

Table 3 Prevalence of in vitro susceptibility to antibiotics among bacterial isolatesa
Full size table

Antibiotics were prescribed and given to 1732 (68%) of all admissions; 14% of all admissions received four or more antibiotics, with a maximum number of nine different antibiotics being used in one patient. Among these 1732 newborns, 350 (20%) received antibiotics to treat noninvasive umbilical or skin sepsis, 616 (36%) for suspected invasive sepsis with blood culture performed, and a further 766 (44%) cases received antibiotics for suspected sepsis but did not have blood taken for culture.

Outcomes

Overall, the case fatality rate for those with confirmed sepsis was 46% (49/106). This compares with a case fatality rate of 17% (87/510) for the newborns that had at least one episode of clinically suspected sepsis that was not confirmed from blood culture, and of 10% (74/766) for the other newborns that had at least one episode of clinically suspected sepsis but did not have blood culture taken. The case fatality rate was similar for confirmed EOS and LOS, 40% and 48%, respectively. The case fatality rate was also similar between those with confirmed bacterial sepsis, 45% (32/71), and for those with confirmed fungal sepsis, 49% (17/35). Of neonates with suspected sepsis who had blood culture performed, the case fatality rate was higher for those with confirmed sepsis compared with those with unconfirmed sepsis: 40% versus 18% for EOS (P=0.02) and 48% versus 14% for LOS (P<0.001).

Confirmed sepsis was associated with 22% (49/220) of all deaths that occurred in the neonatal unit over the 1-year study period, and there were a further 40% (87/220) of deaths where sepsis was suspected. Of deaths due to confirmed sepsis, 84% (41/49) were in neonates with LOS.

Discussion

This study provides original data from the largest neonatal unit in central Vietnam, and shows that neonatal sepsis is common and is associated with a high mortality in this setting. The incidence of confirmed EOS and LOS per 1000 admissions was higher than that reported recently from a large multisite study of neonatal sepsis that included neonatal intensive care units in China, Malaysia, Thailand and Hong Kong.12 Further, the overall case fatality rate of 46% found in our study was three times higher than that reported from the multisite study.12 These differences are likely to reflect differences in socioeconomic indicators, available resources and practices for perinatal and neonatal care, as well as differences in the spectrum of pathogens causing sepsis and blood culture practices affecting microbiological yield.

The results of our study can be compared with findings from a 1-year audit of neonatal admissions to the largest pediatric hospital in Ho Chi Minh City to which sick newborns are referred from health facilities based both within the city and outside in the provinces of southern Vietnam.14, 15 Around two-thirds of the newborns in the Ho Chi Minh City audit were referred from outlying provinces after birth, and so represent a different population to our study where the majority were babies born in the same hospital or at home in Da Nang.14 Of 5763 neonatal admissions in 1 year in the Ho Chi Minh City study, 2202 had blood cultures performed, and there were 399 episodes of culture confirmed sepsis from 385 neonates and 64 (17%) died.15 The majority had bacterial sepsis with fungal sepsis accounting for only 3% of confirmed sepsis, compared with 48% in our study. They also reported a higher proportion (84%) of LOS cases, presumably reflecting the referral pattern to that hospital. Nonetheless, their spectrum of bacterial isolates was similar to our study, with the commonest isolates being CoNS and Gram negatives such as Klebsiella and Acinetobacter species that were commonly resistant in vitro to first- and second-line antibiotics.

Our study has a number of limitations. There was a lack of consistency among clinicians in their approach to blood culture practice for neonates with suspected sepsis. It was observed that commencement of antibiotics without blood culture being taken was most common during the night or on weekends when fewer staff were available. However, it was noted that those who had blood culture performed had worse outcomes than those with suspected sepsis that did not, suggesting that blood culture practices may also have been influenced by clinical markers of severity. In addition, the blood culture technique that was used in our study is likely to provide a lower yield than an automated culture method such as the one that was used in the study in Ho Chi Minh City15, which reported a higher yield of bacterial isolates from blood culture of 18% as compared with 10% in the current study. Blood culture practice and methodology may also influence yield by biasing against the isolation of more fastidious bacteria such as group B Streptococcus. While a major focus of prevention and management of neonatal sepsis in resource-rich settings, Group B Streptococcus is often not found to cause neonatal sepsis in resource-limited settings, as was the case in our study.3, 4, 5, 16 Group B Streptococcus was also not found in the recent study from Ho Chi Minh City, but that study mainly reported causes of LOS.15 Group B Streptococcus has been reported from resource-limited settings as a cause of EOS, including in studies from the Asian region.12, 17

Nosocomial sepsis was commonly found in the neonatal unit in Da Nang. As would be expected, nosocomial sepsis was associated with very low birth weight, invasive procedures, prolonged hospital stay and a high mortality. The causes were fungi or Gram-negative bacteria with a high prevalence of in vitro resistance to multiple antibiotics. These findings are consistent with many previous studies including that in the Asian region.6, 7, 8, 12, 15, 18, 19 Lubell et al.20 reviewed susceptibility of community-acquired pathogens in neonates in low- and middle-income countries in Asia and Sub-Saharan Africa from 2002 to 2009, and identified that ~70% of Gram-negative bacteria were resistant in vitro to ampicillin, 55% were resistant to gentamicin and 30% were resistant to ceftriaxone. Fungal sepsis is also a major nosocomial infection in low birth weight neonates in many studies.21, 22, 23, 24 The frequent usage of a large number of antibiotics such as reported in our study is recognized as a risk factor for fungal infections.25, 26

The high burden of LOS, mainly nosocomial, emphasizes the need to improve infection control in the unit. Since the current study, a number of measures have been introduced that aim to reduce the risk of nosocomial sepsis in the neonatal unit in Da Nang. There has been training and implementation of practices that improve infection control and reinforce hygienic practices, such as hand washing, aseptic techniques for invasive procedures, improved skin care and the introduction of kangaroo mother care for preterm newborns.27 Given the findings on antibiotic usage in our study, there has also been efforts to improve adherence to antibiotic guidelines and reduce the overuse of broad-spectrum antibiotics in the unit. A recent study from Indonesia demonstrated marked reductions in nosocomial infections and inappropriate antibiotic usage in a pediatric setting by introducing infection control procedures and antibiotic stewardship.28 An audit of recent data (November 2013 to March 2015) from the neonatal unit in Da Nang shows a marked fall in antibiotic usage from 64 to 16% of admissions. Further, the overall case fatality rate in the unit over the same period has been <4% overall compared with 8.6% at the time of this study13 (unpublished data).

In summary, neonatal sepsis was common and associated with a high inpatient mortality in the largest neonatal unit in central Vietnam. The finding that nosocomial sepsis due to fungi or Gram-negative bacteria is resistant to multiple antibiotics highlights the need for the implementation of improved infection control measures in the unit.