Objective. To determine the pathogens associated with fulminant (lethal within 48 hours) late-onset sepsis (occurring after 3 days of age) in a neonatal intensive care unit (NICU) and the frequency of fulminant late-onset sepsis for the most common pathogens.
Methods. A retrospective study was conducted of sepsis in infants in a NICU over a 10-year period (1988–1997).
Results. There were 825 episodes of late-onset sepsis occurring in 536 infants. Thirty-four of 49 (69%; 95% confidence interval [CI]: 55%–82%) cases of fulminant late-onset sepsis were caused by Gram-negative organisms, including Pseudomonassp., 20 (42%); Escherichia coli, 5 (10%);Enterobacter sp., 4 (8%); and Klebsiellasp., 4 (8%). The frequency of fulminant sepsis was highest forPseudomonas sp., 20 of 36 (56%; 95% CI: 38%–72%) and lowest for coagulase-negative staphylococci, 4 of 277 (1%; 95%CI: 0%–4%). The very low frequency of fulminant sepsis caused by coagulase-negative staphylococci did not increase during the period when oxacillin was used instead of vancomycin as the empiric antibiotic for Gram-positive organisms.
Conclusions. These data suggest that empiric antibiotics selected for treatment of suspected sepsis in infants >3 days old need to effectively treat Gram-negative pathogens, particularlyPseudomonas sp., because these organisms, although less frequent, are strongly associated with fulminant late-onset sepsis in the NICU. Avoiding empiric vancomycin therapy seemed to be a reasonable approach to late-onset sepsis, because of the very low frequency of fulminant sepsis caused by coagulase-negative staphylococci.
Late-onset sepsis (occurring after 3 days of age) is an important cause of morbidity and mortality in the neonatal intensive care unit (NICU).1 The distribution of pathogens causing sepsis in a specific hospital unit is usually considered when empiric antibiotics are selected, but this approach assumes that all pathogens causing sepsis have equal likelihood of causing severe complications and death. In reality, when selecting empiric antibiotics for late-onset sepsis in the NICU, neonatologists should be especially concerned about fulminant late-onset sepsis, in which infants die in <48 hours of onset of illness, often before pathogens and antibiotic susceptibilities have been identified.
In May 1994, the Centers for Disease Control and Prevention published a draft statement in the Federal Register, which recommended guidelines for prudent use of vancomycin in an attempt to prevent the spread of vancomycin-resistant enterococci.2 Infectious disease specialists discussed these guidelines with the neonatologists at our institution. In October 1994, the neonatologists changed empiric antibiotic therapy for late-onset sepsis in our NICU from vancomycin and cefotaxime to oxacillin and gentamicin. We believed that not starting vancomycin therapy for a day or two would not be harmful, based on the assumption that coagulase-negative staphylococcal infections, albeit common, rarely cause rapid deterioration.2
For this study, the hypothesis was that the distribution of pathogens causing fulminant late-onset sepsis in the NICU could potentially differ significantly from that causing late-onset sepsis in general. The specific aims were: 1) to compare the distribution of pathogens causing late-onset sepsis to the distribution of pathogens causing fulminant late-onset sepsis in a large referral NICU; 2) to determine and compare the frequencies of fulminant late-onset sepsis for the most common pathogens; and 3) to determine the impact of avoiding empiric vancomycin therapy for late-onset sepsis on the duration of coagulase-negative staphylococcal sepsis and the frequency of fulminant cases.
The study population included all infants admitted to the NICU at Children's Hospital of The King's Daughters over the 10-year period from 1988 through 1997. This hospital contains the regional referral nursery for southeastern Virginia and northeastern North Carolina. Infants with late-onset sepsis were identified from a computerized microbiology database. Demographic and clinical data were obtained from a computerized neonatal database and from each infant's medical records. The institutional review board of the Eastern Virginia Medical School approved the study.
Late-onset sepsis was defined as 1 or more positive blood cultures obtained after 3 days of age from infants with clinical features of sepsis.1 Cultures with the same organism within 30 days of first isolation were considered 1 episode. Diagnosis of coagulase-negative staphylococcal sepsis required at least 2 positive blood cultures. Duration of each sepsis episode was defined as the interval from the first positive to the last positive blood culture. Episodes of polymicrobial sepsis were counted as multiple individual episodes of sepsis. Sepsis was identified as fulminant when death occurred within 48 hours of the first positive blood culture (ie, based on time-case relationships). The frequency of fulminant late-onset sepsis was determined by dividing the number of fulminant cases by total cases for each pathogen.
Blood cultures were collected and processed according to standard microbiologic techniques. Between 1988 and 1991 each specimen was inoculated directly into a DuPont Pediatric Isolator tube (DuPont, Wilmington, DE) and a supplemental peptone broth bottle. Between October 1991 and 1997, Bactec Peds Plus/F (Becton Dickinson and Company, Sparks, MD) culture vials were used. The change in microbiology culture methods in 1991 was occasioned by conversion to a pediatric automated blood culture system for purposes of efficiency, cost containment, and decrease in the frequency of blood culture contamination. All positive vials were Gram-stained and subcultured for organism identification on sheep's blood, chocolate, and MacConkey agars.
Nonparametric data are expressed as median (range); comparisons between groups are made with the Mann-Whitney test. Categorical data are analyzed using the χ2 or Fisher's exact test as appropriate. Significance is set at P < .05 and measures of relative risk are computed using 95% confidence intervals (CI).
During the 10-year study period, there were 6275 admissions to the NICU. During this time, 825 episodes of late-onset sepsis occurred in 536 infants, of which 49 episodes of fulminant late-onset sepsis occurred in 43 infants; 6 of the 43 infants had 2 different pathogens simultaneously.
Figure 1 shows the predominant pathogens among the 825 cases of late-onset sepsis. Coagulase-negative staphylococci were the most common, accounting for 277 cases, and were followed by Candida sp. with 143 cases. Gram-positive organisms were the next most common, including Enterococcussp. (83 cases), Staphylococcus aureus (67 cases), andStreptococcus agalactiae (36 cases). Sepsis caused byPseudomonas sp. was uncommon, with 36 cases.Klebsiella sp., Enterobacter sp., andEscherichia coli occurred even less frequently.
Fulminant Late-Onset Sepsis
Figure 2 shows the pathogens that caused the 49 cases of fulminant late-onset sepsis. In contrast to Fig 1, the most prevalent organism in this subgroup wasPseudomonas sp., causing 20 cases, followed by E coli (5 cases). Both Enterobacter sp. andKlebsiella sp. caused 4 cases and Citrobacter diversus caused one. Therefore, Gram-negative organisms caused 69% (95% CI: 55%–82%) of fulminant sepsis cases. Although Gram-positive organisms caused more than 460 cases (56%) of late-onset sepsis, S aureus, coagulase-negative staphylococci, andEnterococcus sp. together caused only 11 (22%) cases of fulminant sepsis.
Table 1 shows the frequencies of fulminant late-onset sepsis for the most common pathogens in descending order. Pseudomonas sepsis had the highest frequency of fulminant sepsis with 20 of 36 infants (56%) dying within 48 hours of obtaining blood cultures. The next highest frequency of fulminant sepsis involved 3 other Gram-negative organisms: E coli with 19%, Enterobacter sp. with 14%, and Klebsiellasp. with 13%. Gram-positive organisms had frequencies of fulminant sepsis that were considerably lower: S aureus with 6% andEnterococcus sp. with 4%. Candida sp. had a 3% rate of fulminant sepsis. Coagulase-negative staphylococci had the lowest frequency of fulminant sepsis with only 4 of 277 (1%) dying within 48 hours of initial blood cultures. The frequency of fulminant sepsis was highest for Pseudomonas sp., 20 of 36 (56%; 95% CI: 38%–72%) and lowest for coagulase-negative staphylococci, 4 of 277 (1%; 95% CI: 0%–4%), yielding a relative risk of 50 (95% CI: 16–161.)
Coexisting Illnesses in Infants With Fulminant Late-Onset Coagulase-Negative Staphylococcal Sepsis
Each of the 4 patients with fulminant coagulase-negative staphylococcal sepsis had severe coexisting conditions at the time of their sepsis. Two infants had polymicrobial sepsis. In 1 infant,Pseudomonas sp. sepsis was the primary cause of death. In the other infant, both enterococcal sepsis and culture-proven enterococcal meningitis were present, supporting the contention that disseminated enterococcal infection was the primary cause of death. The third infant had multiple major malformations including caudal regression, cloacal extrophy, omphalocele, and myelomeningocele, suggesting that coagulase-negative staphylococcal sepsis played a secondary role in the infant's death. In the fourth infant, autopsy listed the primary cause of death as a thrombus in the ascending aorta with subtotal luminal occlusion. Although coagulase-negative staphylococci may have had a role in development of the thrombus, no organisms were detected in serial sections. When these coexisting conditions are considered, it is clear that a 4 of 277 frequency of fulminant coagulase-negative staphylococcal sepsis in neonates is a worst-case estimate, and that the frequency is more likely to be in the range of 2 of 277 (0.7%).
Impact of Avoiding Empiric Vancomycin Therapy for Late-Onset Sepsis
Table 2 compares the outcomes of cases of coagulase-negative staphylococcal sepsis during the January 1988 to September 1994 interval, in which vancomycin was used for empiric antibiotic treatment of Gram-positive pathogens, to the outcomes of coagulase-negative staphylococcal sepsis in the October 1994 to December 1997 interval, in which oxacillin was used instead of vancomycin. Substitution of oxacillin for vancomycin as the empiric antibiotic for suspected Gram-positive sepsis had no impact on the frequency of fulminant sepsis for coagulase-negative staphylococci. During the interval from October 1994 to December 1997, 85% of coagulase-negative staphylococcal isolates were not susceptible to either oxacillin or gentamicin, suggesting that even in the neonate, coagulase-negative staphylococci have very limited potential to cause fulminant sepsis, even in the face of suboptimal therapy. Table 2 also shows that there was no significant increase in the duration (P = .38) of these sepsis episodes during the time interval when vancomycin was no longer used for empiric antibiotic therapy. In addition, there were no cases of infections caused by vancomycin-resistant enterococci in the NICU during the 10-year study period.
The key findings in this study are that Gram-negative organisms caused 69% of fulminant late-onset sepsis in neonates and thatPseudomonas sp. were the most prominent pathogens causing 42% of fulminant sepsis cases. The frequency of fulminant late-onset sepsis for Pseudomonas sp. was 56%, in striking contrast to a frequency of <1% for coagulase-negative staphylococci. Our study confirms the 50% case fatality rate reported by Leigh et al3 in their case control study of 22 cases ofPseudomonas aeruginosa sepsis and meningitis in neonates. These findings are especially disturbing considering a recent report of increasing incidence of Gram-negative rod bacteremia in a NICU.4
In 1996, Stoll et al1 suggested that it would be appropriate to include vancomycin in empiric antibiotic therapy for suspected late-onset sepsis if coagulase-negative staphylococci were responsible for >50% of cases in very low birth weight (VLBW) infants. Nevertheless, they were concerned about the dangers of overuse of vancomycin, especially if many coagulase-negative staphylococcal isolates represented contaminants rather than true infections. In a study of the epidemiology of vancomycin usage at a children's hospital, 1993 to 1995, Sinkowitz et al5 reported that the highest rate of vancomycin usage on the inpatient medical services was on the neonatology service with a rate of 28 per 100 admissions. Use of vancomycin is a risk factor for infection and colonization with vancomycin-resistant enterococci.6,7 In 1995, the Hospital Infection Control Practices Advisory Committee reaffirmed its 1994 draft publication recommending avoidance of empiric vancomycin therapy, especially in VLBW infants, in an effort to prevent the emergence and spread of vancomycin-resistant enterococci.8 Fortunately, in their study of vancomycin usage in a children's hospital, Sinkowitz et al5 reported a significant decrease in neonatologists' usage of vancomycin from 34 per 100 admissions in 1993 to 19 per 100 admissions in 1995 (P < .001), perhaps in response to the Centers for Disease Control and Prevention recommendations. Matrai-Kovalskis et al9 reported a study of 239 episodes of coagulase-negative staphylococcus-positive blood cultures and highly selective vancomycin usage in neonates. Only 22 (9%) episodes were treated with vancomycin and only after coagulase-negative staphylococci were identified. The other 217 cases were managed either without antibiotics or with empiric antibiotics, which did not include vancomycin (usually ceftazidime ± ampicillin). Morbidity and mortality were very low (<2%) regardless of treatment given. Our results are consistent with this report, showing that avoidance of vancomycin as an empiric antibiotic had no impact on the very low rate of fulminant sepsis for coagulase-negative staphylococcal sepsis in neonates. Furthermore, we observed that the practice of starting vancomycin therapy only after coagulase-negative staphylococci were identified did not prolong the duration of sepsis caused by these organisms.
The National Institute of Child Health and Human Development (NICHD) Neonatal Research Network reported that late-onset sepsis occurred in at least 25% of VLBW infants and that sepsis caused 45% of deaths in VLBW infants after 2 weeks of age.1 Although Gram-positive organisms caused 73% of late-onset sepsis in VLBW infants, Gram-negative organisms had the highest mortality rate, 40% (odds ratio = 4.53 for death from Gram-negatives versus other organisms;P < .001).1 Among Gram-negative organisms, P aeruginosa causes a particularly fulminant form of sepsis in VLBW infants with a 50% case fatality rate and death occurring rapidly within 48 to 72 hours of onset.3 Both the high mortality from Gram-negative sepsis and the rapidly fatal course of Pseudomonas sepsis suggest that empiric treatment of late-onset sepsis in neonates may not be focused on the most common causes of fulminant late-onset sepsis.
Our data suggest that empiric antibiotics selected for treatment of suspected sepsis in infants greater than 3 days of age need to effectively treat Gram-negative pathogens, particularlyPseudomonas sp., as these organisms, although less frequent, are strongly associated with fulminant late-onset sepsis in the NICU. Although coagulase-negative staphylococci were the most prevalent pathogens in late-onset sepsis, the very low frequency of fulminant sepsis suggests that avoiding empiric vancomycin therapy until culture results and susceptibilities are known is a reasonable approach to management of late-onset sepsis. However, before these recommendations can be generalized to other NICUs, cooperative research groups, such as the NICHD Neonatal Research Network, should determine the distribution of pathogens causing fulminant late-onset sepsis, as there may be regional differences that might alter specific recommendations.
- Received December 10, 1999.
- Accepted February 18, 2000.
Reprint requests to (M.G.K.) 601 Children's Ln, Norfolk, VA 23507.
- NICU =
- neonatal intensive care unit, CI, confidence interval •
- VLBW =
- very low birth weight •
- NICHD =
- National Institute of Child Health and Human Development
- Centers for Disease Control and Prevention
- Edmond MB,
- Ober JF,
- Weinbaum DL
- Murray BE
- Copyright © 2000 American Academy of Pediatrics