Neonatal infections are an important cause of mortality and morbidity world wide. In their 2000–2003 report, the World Health Organization estimated that neonatal sepsis and pneumonia are responsible for about 1.6 million deaths each year, mainly in resource-poor countries.1 Antibiotic resistance is an important problem in resource-poor countries,2^–5 and a survey of neonatologists in Asian countries suggested that there is a significant problem with sepsis caused by multi-resistant Gram-negative organisms and meticillin-resistant Staphylococcus aureus (MRSA).6 Previous studies have reported rates of hospital-acquired neonatal infections that are 3–20 times higher in resource-poor than resource-rich countries.3 The most common reported organisms are Gram-negative bacilli and S aureus.3 Antibiotic resistance rates are high: in South-East Asia up to 86% of Klebsiella species are resistant to cefotaxime and gentamicin, 56% of Escherichia coli are resistant to gentamicin, and 28% of S aureus are resistant to meticillin.3

In this study, we look at the microorganisms (bacteria and fungi), the incidence and the mortality of early-onset and late-onset neonatal sepsis in mainly resource-poor countries in Asia, and at antibiotic-sensitivity patterns.

MATERIALS AND METHODS

Neonatologists in Asia identified from a database of neonatologists working in level 3 neonatal units (defined as those that can manage babies with artificial ventilation) were invited to participate in the study. Questionnaires were sent to those neonatologists who agreed.

Neonatal sepsis was defined as the pure growth of a single potentially pathogenic organism (bacterium or fungus) from the blood of a baby who was clinically septic according to defined criteria6 7 and had supportive laboratory evidence of sepsis (eg, one or more of low or high white cell count or abnormal immature: total (I:T) ratio, low platelets, or raised serum C-reactive protein, as defined previously).6 Laboratory variables were age-dependent. We did not further define clinical sepsis. We did not request two positive blood cultures because antibiotics are usually started empirically in Asia after only one set of blood cultures had been taken. Likely contaminants were excluded. The decision as to whether a baby had true sepsis or if the cultured organism was a contaminant was made by the local clinician, according to clinical judgement but using the above criteria. Early-onset sepsis was defined as the onset of sepsis within 48 h of delivery, and late-onset sepsis as the onset of sepsis more than 48 h after delivery. Outcome was defined as: died from sepsis, died possibly from sepsis, died from unrelated cause, or survived. Reported cases were excluded if they did not meet the defined criteria for sepsis or if the data were inconsistent and could not be verified. We relied on clinicians to report all cases of babies with positive blood cultures and clinical sepsis, and we did not search microbiology records for unreported episodes of neonatal septicaemia.

All babies with sepsis meeting our criteria on neonatal units in the study were included, regardless of their postnatal age. The babies were de-identified and recorded by their initials only. The following data on babies with neonatal sepsis were collected: gestational age, birth weight, gender, and details of whether the baby was born in the hospital or outside the hospital and then transferred to the nursery. Denominator data were collected on the number of babies managed at each site, categorised by birth weight. The age in days when the positive blood culture was obtained was also recorded.

The organisms isolated on blood culture and their antibiotic sensitivities were recorded. Local laboratories used recognised methods of antibiotic susceptibility testing, but these were not standardised. For Gram-negative bacilli, data were only sought on sensitivity to third-generation cephalosporins (cefotaxime or ceftazidime for Pseudomonas) and to gentamicin. The organisms were recorded as sensitive to both cephalosporin and gentamicin, sensitive to cephalosporin and resistant to gentamicin, resistant to cephalosporin and sensitive to gentamicin, or resistant to both cephalosporin and gentamicin. For S aureus, data were requested on sensitivity to meticillin.

For each episode of sepsis, it was recorded whether or not a lumbar puncture had been performed. The results of the cerebrospinal fluid (CSF) culture were recorded as well as the presence of CSF leucocytosis (defined6 as a CSF white blood cell count >100/ml). Meningitis was defined as the presence of an organism cultured from CSF or the presence of CSF leucocytosis in the presence of a positive blood culture. Co-existing conditions associated with infection, such as pneumonia or necrotising enterocolitis, were also recorded.

The data were recorded on standard proformas by local clinicians, and these were either posted or e-mailed back. The data were then collated into a dedicated database created for the purpose. Mortality per 1000 live births was calculated using only data from neonatal units attached to maternity hospitals. Statistical calculations were performed using SPSS V15.

This study was approved by the ethics committee of the Royal Alexandra Hospital for Children. Local sites were asked if they required local ethics approval and all felt that the Royal Alexandra Hospital for Children ethics approval was sufficient.

RESULTS

Seventeen level 3 neonatal units were approached. Data were received from eight units (47%), two from China and one each from Hong Kong, India, Iran, Kuwait, Malaysia and Thailand. We present the prospective data from 1 January through 31 December 2005. Five of the neonatal units were attached to maternity units, and we were able to obtain the number of live born infants in the maternity unit over the year. The other three neonatal units only accepted babies born elsewhere. The neonatal units had similar case mixes; all were able to ventilate preterm babies and all cared for babies who underwent abdominal surgery.

There were 453 episodes of sepsis documented in the study, affecting 394 babies. The overall rate of a baby developing either early or late sepsis in our study, calculated from neonatal units attached to maternity hospitals, varied from 3.0 per 1000 live births in Hong Kong to 15.0 per 1000 live births in Kuwait. Forty-one babies had two episodes of infection, six had three episodes of infection, and one had five episodes of infection. There were 41 deaths reported as a direct result of sepsis, giving a mortality of 10.4%. If mortality in only the maternity hospitals was analysed, the overall mortality as a direct result of sepsis can be expressed as 0.69 deaths/1000 live births. If we include deaths probably occurring as a result of sepsis in the analysis, the overall mortality increases to 1 death/1000 live births.

The most common associated conditions or infections were meningitis (76 episodes), pneumonia (49 episodes), necrotising enterocolitis (38 episodes) and skin sepsis (12 episodes). Thirty-four of the 38 bacteraemic episodes of necrotising enterocolitis (89%) were associated with episodes of late infection, and 44 of the 49 episodes of pneumonia (90%) occurred in association with episodes of late infection. In 248 episodes of infection, no co-existing condition was recorded.

Early-onset sepsis

Early-onset infections, defined as infection with onset within 2 days of birth (47 episodes in total), caused 10.4% of all infections reported. The rate of early-onset sepsis from any organism was 0.72 infections/1000 live births. Six out of 47 babies with early-onset sepsis died, a mortality of 13%. Two babies with early-onset sepsis went on to have late-onset sepsis (one baby had one episode of late-onset sepsis, and one baby had two episodes of late-onset sepsis).

Table 1 shows the organisms that caused early-onset sepsis. Group B streptococcus (GBS) was the most common organism, causing 18 out of 47 episodes of early-onset sepsis (38%). There were eight episodes of early GBS sepsis reported from Kuwait, five from Malaysia, four from Hong Kong, and one from China. Seventeen episodes of early-onset GBS infection occurred in inborn babies in maternity hospitals, giving an incidence of 0.51 per 1000 live births for early-onset GBS. Four of the 18 babies with early-onset GBS died, a mortality of 22%. This was higher than the mortality from Gram-negative infection in early-onset sepsis (2 of 16, 12.5%) and the total overall mortality in early sepsis (12.8%), but the differences did not reach statistical significance.

There were 17 early-onset episodes of Gram-negative bacillary infection. The most common Gram-negative bacillus in this group was E coli (five episodes). The incidence of Gram-negative bacillary early-onset sepsis was 0.15 episodes per 1000 live births, and mortality was 12% (two deaths from 17 infections).

Sensitivities were reported for 16 episodes of early-onset Gram-negative bacillary sepsis. Seven organisms (44%) were resistant to either a third-generation cephalosporin or gentamicin, six organisms (37%) were resistant to both, and three organisms (19%) were sensitive to both antibiotics.

There were 30 episodes of early-onset sepsis in which Gram-positive organisms were isolated. Eighteen of these episodes (60%) were due to GBS. Eight episodes were reported to be due to coagulase-negative staphylococcus (CONS), but we were unable to exclude that they may have been contaminants. One of the two S aureus isolated in early sepsis was meticillin resistant.

Six episodes of early sepsis (12.8%) were associated with meningitis (see below and table 5).

Late-onset sepsis

There were 406 episodes of late-onset sepsis, affecting 347 babies. Gram-negative bacilli caused 189 (46.6%) of the episodes. Table 2 shows the organisms causing late-onset sepsis. The most common Gram-negative bacillus was Klebsiella pneumoniae (69 episodes), followed by Enterobacter species (29) and E coli (26).

Sensitivities were recorded for 180 of the 189 Gram-negative bacilli (95.2%): 44% were sensitive to both gentamicin and to a third-generation cephalosporin, 27% were resistant to either gentamicin (6%) or a third-generation cephalosporin (21%), and 30% were resistant to both (table 3). Resistance was most common in Bangalore, India, where 21 (75%) of 28 Gram-negative bacilli were resistant to both third-generation cephalosporins and gentamicin, and only three (11%) were sensitive to both.

There were 187 episodes of late-onset Gram-positive sepsis. CONS was the single most common isolate, causing 136 episodes of Gram-positive sepsis (33.5% of all late sepsis episodes). The next most common Gram-positive organism was S aureus, accounting for 22 late-onset infections (5.4%). Two of the 12 S aureus infections (17%) for which sensitivities were reported were meticillin resistant.

Regarding the 406 episodes of late-onset sepsis, 36 (8.9%) babies died as a direct result of sepsis. If the 18 babies who probably died from sepsis are included in the analysis, mortality from late-onset sepsis increases to 13.3%. If the 136 episodes of CONS infection are excluded from the total (because they are potential contaminants), mortality as a direct result of sepsis increases to 13.3% and as a direct result of definite plus probable sepsis to 20%. Gram-negative bacillary septicaemia had a higher mortality as a direct result of infection (14.8%) than CONS (1.5%) and S aureus (4.5%), but this did not reach statistical significance.

The incidence of late-onset infection was inversely proportional to birth weight (table 4). The proportion of babies developing late-onset sepsis varied from 2.0 per 1000 live births in Hong Kong to 22.0 per 1000 in Thailand. The overall figure was 11.6 per 1000 live births (table 4).

Meningitis

Meningitis was reported in 76 episodes of sepsis (16.8% of all episodes) affecting 75 babies (table 5). Six of 47 babies with early-onset sepsis (13%) were reported with meningitis compared with 70 of 406 episodes of late-onset sepsis (17.2%) (p>0.05). It was not possible to calculate the incidence of meningitis in early sepsis per 1000 live births, as none of the episodes of meningitis occurred in a maternity hospital. One of six babies with early-onset meningitis died.

Seventy episodes of late sepsis (17.2%) were associated with meningitis (table 5). Gram-negative bacilli caused 53 episodes of late-onset meningitis (75.7%). Meningitis occurred in 28.0% of the 189 episodes of late-onset Gram-negative sepsis. The predominant species causing late-onset Gram-negative meningitis were Klebsiella (25, of which 23 were reported as Klebsiella pneumoniae) and E coli (nine episodes). Meningitis was reported in 18 (9.6%) of 187 episodes of late-onset sepsis due to Gram-positive cocci: 10 were coagulase-negative staphylococci, two S aureus, and only one GBS (table 5). The reports of meningitis due to CONS may represent contaminants, but we were unable to verify this. One episode of meningitis was caused by an anaerobe. We received no reports of fungal meningitis.

Meningitis was significantly more common with late-onset Gram-negative bacillary sepsis than with sepsis due to Gram-positive cocci (χ² = 11.3, p = 0.001).

The episodes of meningitis associated with late sepsis were spread across all gestations. One baby had two episodes of late-onset meningitis. Fourteen of 70 babies with late-onset meningitis died (20%).

DISCUSSION

The rates of sepsis in our study varied from 2 per 1000 live births in Hong Kong to 22 per 1000 live births in Thailand, with an overall figure of 11.6 per 1000 live births, a number dominated by the data from Kuwait. Our study dealt with neonatal units and is not representative of neonatal sepsis in the community. The rates of sepsis we calculated were from units with attached maternity wards, and the figures we obtained were comparable to the reported overall incidence of neonatal sepsis in Asia, of 7.1–38.0 per 1000 live births,2 8^–12 and higher than the rates usually reported from resource-rich countries in North America,7 Europe7 and Australia6 of 1.0–8.1 per 1000 live births. The mortality figures in this study, of 13% for early sepsis and 8.9% for late sepsis, are similar to recent mortality figures from the USA,7 and are very similar to the Australian mortality figures for 1991–2 of 15% for early-onset sepsis and 9% for late-onset sepsis.6 The mortality from late-onset Gram-negative bacillary sepsis in the present study was also comparable to recent North American13 and Australian14 data.

The most common pathogen reported causing early-onset sepsis in our study was GBS, responsible for 38.3% of early sepsis. CONS was next most common, either due to rapid early postnatal acquisition of the organism or as blood culture contaminants. E coli and other Gram-negative bacilli caused most of the other early-onset infections. The pattern of organisms identified in early sepsis in Asia in this study is similar to that described in resource-rich countries.

The incidence of early-onset sepsis due to GBS in this study was 0.51 per 1000 live births, which is lower than the incidence in the USA15 and Australia6 before the widespread use of intrapartum antibiotics. GBS is usually reported to be an uncommon cause of early-onset sepsis in resource-poor countries.2 8 In previous Asian studies, the incidence of early-onset sepsis due to GBS was 0.1–0.27 cases per 1000 live births in Thailand,9 0.11–1.39 in Taiwan,10 0.27 in Singapore11 and 0.4 in Malaysia.12 In contrast, the incidence of early-onset GBS disease in the USA before the use of intrapartum antibiotics was 0.7–3.7 per 1000 live births.15

The organisms causing late-onset infection in Asian neonatal units in the present study were similar to those reported from resource-rich countries since the 1980s, with CONS being the most common cause of infection, followed by Gram-negative bacilli.1 2 6

The proportion of babies with early-onset sepsis who had meningitis in our study (12.8%) was comparable to figures from North America and Europe.12 The rate of meningitis in late sepsis (17.2%), however, was slightly higher than the rate of ∼10% generally reported from North America, Europe and Australia.6 7 This was despite our use of stringent criteria (>100 white cells/ml) to define meningitis, although lowering these criteria to >20 cells did not significantly alter our figures, as most babies with meningitis grew organisms from CSF. Mortality from late-onset meningitis was 20% in the present study, a rate that is perhaps surprisingly low and comparable to mortality in resource-rich countries.7 This may be due to selection bias and/or to a high quality of care in the nurseries that participated in the study.

Of the Gram-negative bacillary infections for which sensitivities were reported, only 30% of Gram-negative organisms responsible for late-onset sepsis and 19% of Gram-negative organisms responsible for early-onset sepsis were reported as being sensitive to both a third-generation cephalosporin and gentamicin. The rest were resistant to either antibiotic or both. This is in line with data from Asia suggesting very high rates of antibiotic resistance among Gram-negative bacilli.3 8 16 In contrast, only 17% of S aureus species responsible for late infection that were tested were meticillin resistant, whereas many Asian units report higher rates of MRSA.16

The data we obtained were surprisingly similar to those from resource-rich countries, in terms of the organisms isolated from babies with early and late sepsis, the incidence of meningitis, and mortality from meningitis and overall sepsis. The major differences are that the rate of late-onset sepsis in Asia is higher than in resource-rich countries and that a very high percentage of Gram-negative organisms are resistant to either gentamicin or a third-generation cephalosporin or both. However, this does not apparently result in increased mortality, at least in the units studied here, although, because the incidence of sepsis is higher than in resource-rich countries, more Asian babies are dying from sepsis. Furthermore, if rates of antibiotic resistance continue to rise, we expect that mortality will also rise in Asia as clinicians run out of antibiotic options.

The present study is not a systematic survey of Asian neonatal units. Selection bias is one of the major criticisms of the study, and it is likely that the neonatal units reporting data are better resourced than many other neonatal units in the same country and many others in the region. However, the data are recent, prospective and from various centres and add valuable information on neonatal infections in Asia, particularly with regard to antibiotic resistance. One of the strengths of this study is that it is an ongoing project, and hence there is the capacity to monitor changes in pathogens and their changing antibiotic-resistance patterns with time. We intend to continue collecting data for several years and to recruit more neonatal units.