BACKGROUND: Group B Streptococcus (GBS) and Escherichia coli have historically dominated as causes of early-onset neonatal sepsis. Widespread use of intrapartum prophylaxis for GBS disease led to concerns about the potential adverse impact on E coli incidence.
METHODS: Active, laboratory, and population-based surveillance for culture-positive (blood or cerebrospinal fluid) bacterial infections among infants 0 to 2 days of age was conducted statewide in Minnesota and Connecticut and in selected counties of California and Georgia during 2005 to 2014. Demographic and clinical information were collected and hospital live birth denominators were used to calculate incidence rates (per 1000 live births). We used the Cochran–Amitage test to assess trends.
RESULTS: Surveillance identified 1484 cases. GBS was most common (532) followed by E coli (368) and viridans streptococci (280). Eleven percent of cases died and 6.3% of survivors had sequelae at discharge. All-cause (2005: 0.79; 2014: 0.77; P = .05) and E coli (2005: 0.21; 2014: 0.18; P = .25) sepsis incidence were stable. GBS incidence decreased (2005: 0.27; 2014: 0.22; P = .02). Among infants <1500 g, incidence was an order of magnitude higher for both pathogens and stable. The odds of death among infants <1500 g were similar for both pathogens but among infants ≥1500 g, the odds of death were greater for E coli cases (odds ratio: 7.0; 95% confidence interval: 2.7–18.2).
CONCLUSIONS: GBS prevention efforts have not led to an increasing burden of early-onset E coli infections. However, the stable burden of E coli sepsis and associated mortality underscore the need for interventions.
- CI —
- confidence interval
- CSF —
- cerebrospinal fluid
- GBS —
- group B Streptococcus
- IQR —
- interquartile range
- OR —
- odds ratio
What’s Known on This Subject:
Widespread intrapartum prophylaxis for perinatal group B streptococcal disease provoked concerns about potential increases in Escherichia coli sepsis in the first week of life, particularly among preterm infants. Approximately 30% of US deliveries are exposed to intrapartum antibiotics.
What This Study Adds:
Using active, population-based surveillance from 2005 to 2014, we documented stable rates of all-cause and E coli sepsis in the first week of life and characterized current epidemiology. Notably, rates were stable among preterm infants and for those with a birth weight <1500 g.
Infections in the first 3 days of life (early onset) remain among the leading causes of infant death in the United States and can result in lifelong sequelae among survivors.1 The widespread uptake of intrapartum antibiotic prophylaxis for the prevention of perinatal group B Streptococcus (GBS) disease, the leading cause of early-onset infections, has resulted in an 80% decline in early-onset GBS disease incidence.2 However, this decline has been accompanied by a more than doubling (30% vs 12%) in the proportion of deliveries exposed to intrapartum antibiotics compared with the preprevention era.3 Intrapartum prophylaxis has always been viewed as an interim perinatal GBS disease prevention strategy, in part because of concerns for the potential emergence of resistance among GBS to the first line, highly effective β lactam therapies, and in part because of concerns that intrapartum antibiotic exposures may increase the risk of sepsis due to non-GBS pathogens.4 Several years after the first national perinatal GBS disease prevention guidelines in 1996,5 a 1 multicenter and a series of single hospital studies reported increases in early-onset sepsis due to Escherichia coli, accompanied by high case fatality.6–10 Evidence of an increase was strongest among preterm and very low birth weight infants, populations that account for the majority of early-onset E coli infections.8,11
Although the incidence of early-onset invasive sepsis is high compared with most invasive infections in older age groups, the number of cases at individual facilities remains low because the age group (first 3 days of life) is so narrow. Single hospital studies are thus challenging for trend ascertainment, particularly if pathogen-specific trends and trends among subgroups, such as preterm or low birth weight infants, are of interest. In 2005, the Active Bacterial Core surveillance/Emerging Infections Program network established active, population-based surveillance for early-onset invasive bacterial infections. Here we describe overall, GBS, and E coli-specific early-onset incidence trends from 2005–2014 and compare the clinical and epidemiologic characteristics of GBS and E coli infections.
Active laboratory surveillance for invasive (culture positive from blood or cerebrospinal fluid [CSF]) bacterial infections among infants 0–2 days of age was conducted by the Centers for Disease Control and Prevention Active Bacterial Core surveillance/Emerging Infections Program network from 2005 to 2014 in the following catchment areas: 20 hospitals capturing 95% of births in the San Francisco Bay area in California; 18 hospitals representing 98% of births in 8 metropolitan Atlanta counties in Georgia; 30 hospitals representing 99% of births in Connecticut; and 140 hospitals representing 97% of births in Minnesota. Surveillance officers in each area reviewed microbiology records at clinical laboratories serving the catchment populations on a regular basis and collected demographic, intrapartum, and clinical information from the labor and delivery record and case medical records by using a standardized form. Antimicrobial susceptibility interpretations were abstracted from the medical record starting in 2007; diagnosis of chorioamnionitis was abstracted starting in 2011. The surveillance catchments combined included ∼200 000 live births annually. Live birth denominators for the hospitals in the surveillance catchment were obtained from state vital records files; the 2013 live birth denominator was used for both 2013 and 2014, and in California, live births from 2007 were used to estimate those in 2008. Among the live birth denominator data, missing values for race (8%), gestational age (1%), and birth weight (<0.5%) were distributed based on the distribution of those with known values within surveillance sites.
Stillborn infants, infants <23 weeks of gestation, infants born at home, and infants with a first positive culture collected >12 hours after death were excluded. Additionally, cultures yielding ≥3 organisms were considered contaminants, as well as cultures yielding single organisms belonging to the following groups: Aerococcus, Bacillus, Burkholderia, Capnocytophago, Corynebacterium, Cupriavidus, Flavimonas, Gemella, Granulicatella, Haemophilus other than H influenzae, Lactobacillus, Micrococcus, Morganella, Mycobacteria other than M tuberculosis, Neisseria other than N meningitides, Ochrobacterum, Paenibacillus, Previotella, Propionibacterium, Roseomonas, Staphylococcus other than S aureus, Stetrophomonas, Stomatococcus, and Tatumella.
When race, gestational age, or outcome were missing from the medical record they were supplemented with values from vital records; remaining unknowns were distributed based on the distribution of those with known values within surveillance sites. Because of the limited number of cases with Asian, American Indian, or Pactific Islander race, race in multivariable analyses was categorized as black versus all others (“nonblack”). Clinical syndrome refers to the clinical presentation abstracted from the medical record and was categorized as meningitis, primary bacteremia, pneumonia, or other. Preterm was defined as <37 weeks of gestation, and was additionally categorized into early (<34 weeks of gestation) and late preterm (34–36 weeks of gestation). Similarly, standard birth weight categories were considered (very low birth weight: <1500 g; low birth weight: 1500–2499 g; 2500–3999 g; ≥4000 g). Incidence rates were calculated as cases per 1000 live births. We calculated 95% confidence intervals (CIs) for the incidence rates based on the assumption that the annual case count followed a Poisson distribution.
We tested for linear trends in incidence by using the Cochran–Armitage test. Univariate differences were assessed by χ2 for categorical variables and by the Kruskal–Wallis test for continuous variables. Multivariable analysis was conducted by using manual backward, stepwise logistic regression, starting with all univariate factors that were significant at P < .15. Final multivariable models included all factors significant at P < .05 and were assessed for interaction and collinearity.
Surveillance identified 1484 invasive early-onset sepsis cases from 2005 to 2014. GBS was the leading cause (n = 532) followed by E coli (n = 368). Remaining organisms with a frequency of at least 10 cases included: viridans streptococci (n = 280); H influenzae (n = 67), S aureus (n = 52), Enterococcus (n = 46), group D Streptococcus (n = 21), Listeria monocytogenes (n = 19), Klebsiella pneumoniae (n = 14), and S pneumoniae (n = 14). Overall, primary bacteremia was the leading syndrome documented (82.9%), followed by pneumonia (5.0%) and meningitis (4.2%); 97.8% of cases were diagnosed by blood culture with the remaining from CSF. Preterm birth was common (42.2% of cases). Polymicrobial infections were rare (1.6%, n = 23). The median length of hospitalization was 10 days (interquartile range [IQR]: 6–22 days). Eleven percent of newborns died; among survivors at hospital discharge, 6.3% (n = 94) had documented sequelae (Table 1); the most common sequelae included oxygen requirement on discharge (51%), hearing loss (35%), and seizures (21%).
During the 10-year period, the overall incidence of invasive early-onset sepsis was stable (2005: 0.79 cases per 1000 live births; 2014: 0.77 cases per 1000 live births; P = .052); similarly, the incidence trend for E coli was stable (2005: 0.21 cases per 1000 live births; 2014: 0.18 cases per 1000 live births; P = .25) whereas the incidence of GBS decreased (2005: 0.27 cases per 1000 live births; 2014: 0.22 cases per 1000 live births; P = .02; Fig 1). Additionally, although the observed incidence of E coli early-onset sepsis remained lower than that of GBS early-onset sepsis for each of the years under surveillance, the 95% CIs around GBS and E coli incidence estimates overlapped with the degree of overlap more marked in the second half of the surveillance period. The relative incidence of E coli to GBS invasive early-onset infections also varied by state. In California, E coli cases were more common than GBS cases in each of the 10 surveillance years; the other 3 surveillance areas had at least 1 year with more E coli than GBS cases. Incidence trends for viridans streptococci, the most common group among the non-GBS and E coli sepsis cases, were stable overall with a low of 0.12 cases per 1000 live births in 2007, 2012, and 2013 and a high of 0.16 cases per 1000 live births in 2008 and 2014.
For both GBS and E coli sepsis, incidence was at least an order of magnitude higher among infants <34 weeks of gestational age compared with infants born at term (Fig 2 A and B). Similarly, early-onset GBS and E coli sepsis incidence was notably higher among very low birth weight infants compared with infants with birth weights of 2500 grams or more (Fig 3 A and B). However, there was no evidence of a trend toward increasing incidence among either the low gestational age (GBS: P = .11; E coli: P = .49) or very low birth weight subpopulations (GBS: P = .09; E coli: P = .15) for either pathogen. In fact, the point estimates for both GBS and E coli incidence among very low birth weight infants declined (Fig 2 A and B).
When all-cause early-onset sepsis was stratified by both race and gestational age, the incidence among black infants was significantly higher than among nonblack infants for term infants and for some calendar years among infants <34 weeks of gestation. Black infants <34 weeks of gestation had the highest all-cause sepsis incidence (10.2 cases per 1000 live births in 2012; Supplemental Information). Notably, however, none of the subpopulations showed evidence of an increasing trend during the study period (Supplemental Information).
Black race (34.9% vs 20.3%; P < .0001), birth at <34 weeks of gestation (32.2% vs 3.1%; P < .0001), and very low birth weight (23.3% vs 1.5%; P < .0001) were overrepresented among early-onset sepsis cases compared with the surveillance catchment population. Compared with population-level estimates based on a representative sample of live births from the same Active Bacterial Core surveillance areas in 2003 to 2004,3 early-onset cases also differed importantly from the population of live births in the following characteristics: cesarean delivery (cases: 46.1%; 95% CI: 43.5%–48.7% versus population: 26.3%; 95% CI: 24.5%–28.2%), maternal intrapartum fever (cases: 20.7%; 95% CI: 18.6%–22.7% versus population: 3.9%; 95% CI: 3.1%–4.9%), exposure to intrapartum antibiotics (cases: 50.0%; 95% CI: 47.5%–52.5% versus population: 32.7%; 95% CI: 30.7%–34.7%), prolonged membrane rupture before delivery (cases 42.0%; 95% CI: 39.5%–44.5% vs population: 7.7%; 95% CI: 6.5%–9.0%), and suspected maternal chorioamnionitis (cases: 29.9%; 95% CI: 25.4%–34.5% versus population: 3.4%; 95% CI: 2.5%–4.5%). Maternal chorioamnionitis (Table 1) was more common among infants with E coli (51.6%) than among those with GBS (23.8%) or viridans streptococci (17.4%).
Compared with infants with GBS disease, infants with invasive E coli were more likely to be of younger gestational age (60.6% vs 20.9%; P < .0001) and very low birth weight (43.6% vs 15.6%; P < .001) (Table 1). Infants with E coli infections were also more likely than infants with GBS infections to be delivered by cesarean section, to have older mothers, to have mothers with a range of intrapartum risk factors, to be exposed to intrapartum antibiotics, and to have longer hospitalizations and poorer outcomes (Table 1). When limited to infants <34 weeks of gestation (n = 334), the only characteristics differing significantly between infants with E coli (n = 223) and GBS (n = 111) infections included black race (E coli: 39.4%; GBS: 52.7%; P = .02), membrane rupture of ≥18 hours (E coli: 73.1%; GBS: 35.1%; P < .0001), and exposure to intrapartum antibiotics (E coli: 90.1%; GBS: 64.0%; P < .001).
The vast majority of infants exposed to intrapartum antibiotics were exposed to a β-lactam or cefazolin (Table 1). Among GBS cases, 63.3% (88/139) of cases with an indication for prophylaxis according to national guidelines received intrapartum prophylaxis,4 although median durations were shorter than the recommended 4 hours (Table 1); a majority of cases (73% or 377/516 with complete information) did not have a prophylaxis indication. Among E coli cases with susceptibility interpretations documented in the medical record (255/293 or 87% of E coli cases from 2007 to 2014), ampicillin resistance was documented for 168 cases (66%) and gentamicin resistance was documented for 26 cases (10%; 25 of these 26 also had resistance to ampicillin). Among infants with E coli infections with susceptibility results, 84% (141/168) of those exposed to intrapartum antibiotics had ampicillin-resistant infections compared with 65% (50/77) of unexposed infants (P < .001). The proportion of infants with E coli infections who died was not significantly higher among those with ampicillin-resistant infections (20.8% vs 18.2%).
Factors associated with mortality in univariate analysis are shown in Table 2. In multivariable analysis, birth weight and pathogen (GBS, E coli, or other) were the only significant factors; however, they had a significant interaction. When we stratified by birth weight, among infants weighing <1500 g (n = 345; deaths = 123), E coli was not significantly more likely than GBS to result in death (odds ratio [OR]: 1.3; 95% CI: 0.7–2.2). Among preterm infants with E coli infection, 27% of deaths (21/77) occurred on day 0 and 31% (24/77) on day 1; 86% of these deaths (66/77) were exposed to intrapartum antibiotics. Among infants weighing ≥1500 g (n = 1139), deaths were much more rare (n = 40); for this population, the odds of death among infants with E coli infection were significantly greater than among those with GBS infection (OR: 7.0; 95% CI: 2.7–18.2), as were the odds of death among infants with infections due to other pathogens compared with infants with GBS infection (OR: 2.7; 95% CI: 1.0–6.9). Among survivors, factors associated with sequelae at discharge on univariate analysis included: pathogen class, clinical syndrome as denoted in the medical record, race, birth weight category, gestational age category, prolonged membrane rupture, and exposure to intrapartum antibiotics. On multivariable analysis, only meningitis syndrome (adjusted OR: 3.53; 95% CI: 1.74–7.16) and very low birth weight (adjusted OR: 3.88; 95% CI: 1.47–10.22) had significant associations.
Between 2005 and 2014, a period of widespread intrapartum prophylaxis for prevention of early-onset GBS disease, we documented stable incidence rates of all-cause and E coli invasive early-onset sepsis and a modest decline in GBS invasive early-onset sepsis in a population representing ∼5% of US live births. Additionally, the observed incidence of E coli infections among early-preterm and very low birth weight infants declined during the 8-year period, although the declines were not statistically significant. These trends suggest that exposure of approximately one-third of live births to ampicillin or penicillin for GBS prophylaxis has not resulted in an increase in Gram-negative sepsis. Moreover, reports from the early years of GBS prevention raising concern about increasing E coli incidence among very low birth weight or preterm infants are not borne out by our observations. More recent observations from single institutions and hospital networks are consistent with our results.12,13
GBS remained the most common invasive early-onset pathogen in each surveillance year, followed by E coli, with other pathogens notably less frequent. An assessment of implementation of perinatal GBS disease prevention guidelines in these surveillance areas among a representative sample of live births in 2003 to 2004 already showed strong implementation of universal antenatal screening and administration of intrapartum prophylaxis to colonized women.3 The case-only data presented in this study do not reflect population-level implementation because it is enriched for implementation failures; our data do suggest that there may be potential for small additional decreases in GBS incidence based on the observation that 37% of cases with an indication for prophylaxis did not receive it.
Although overall GBS remained the leading pathogen across surveillance years, the CIs around the incidence rates for GBS and E coli overlapped. In the most recently reported years of multisite surveillance from the National Institute of Child Health and Development’s Neonatal Research Network (2009),14 and the large Pediatrix network (2010),13 GBS early-onset incidence also remained higher than that of E coli. Nationwide surveillance in the Netherlands15 and population-based surveillance in Italy16 also continue to show GBS as more common than E coli in their most recent reporting years (2011 and 2009–2012, respectively). Notably, some single institutions now report E coli as the most common cause of invasive early-onset sepsis12; moreover, in one of the surveillance areas (California), E coli was more common than GBS for all surveillance years. Additionally, a study from 2005 to 2012 of bacteremia among febrile young infants admitted to general care units (rather than the ICU) found E coli as the lead cause.17 Thus, regionally and globally, the relative pathogen prevalence likely varies and should be considered in the context of management and prevention strategies.
Early-onset sepsis incidence was significantly higher among black term infants with less evident differences for infants 34 to 36 weeks of gestation and <34 weeks of gestation. This may in part reflect the preponderance of GBS cases among term infants, given that GBS disease risk is higher among black infants.3 Case fatality rates were highest among preterm and very low birth weight infants. Consistent with other recent surveillance,9,13,14 E coli was associated with most early-onset sepsis deaths, primarily due to its predominance among very low birth weight infants. For this subpopulation, E coli was not significantly more likely to result in death than GBS. It is likely in this vulnerable population that pathogen virulence may not be strongly associated with risk of death. However, among infants ≥1500 g at birth, where death was less frequent, E coli infections were associated more often with severe outcomes. The large catchment in our surveillance may have given us the power to detect this trend, which was not noted in other, smaller studies.14,16 Clonal changes among E coli associated with early-onset sepsis and, in particular, emerging ampicillin resistance, which was documented in two-thirds of our cases, may contribute to the severity of E coli outcomes,18 although on univariate analysis, death among infants with E coli was not associated with ampicillin resistance. Aminoglycoside resistance remained rare but notably, gentamicin resistance was strongly associated with ampicillin resistance, highlighting the importance of continued evaluation of regimens for first-line early-onset sepsis treatment.19 Our observation that more than half of preterm E coli cases died very close to birth, despite exposure to intrapartum prophylaxis, further supports this need.
A number of maternal, intrapartum, and demographic features differed between invasive GBS and E coli cases in univariate analysis. Black race (more common among GBS cases) and prolonged membrane rupture and intrapartum antibiotic exposure (more common among E coli cases) were the only factors that remained when controlling for gestational age. The overrepresentation of intrapartum antibiotic exposure among infants with E coli infection compared with those with GBS may reflect, in part, that intrapartum regimens used for GBS prevention (most typically penicillin or ampicillin) are not effective in preventing early-onset E coli infections. The high proportion of chorioamnionitis in this group suggests that intrapartum intervention may be too late for prevention but may still hold value for initiation of early newborn treatment.
Although our surveillance benefitted from a large, population-based catchment population and detailed labor and delivery record review to capture intrapartum histories, it captured only limited clinical information on disease management and course. Maternal chorioamnionitis was also only collected from 2011 to 2014. Additionally, a predetermined contaminant definition was applied to all cases, which in some instances may have resulted in inclusion as cases of some instances of contamination (eg, a portion of the cases attributed to viridans streptococci) and may in other instances have excluded true cases (eg, all coagulase negative staphylococci were excluded, and a portion may have represented true sepsis). The case-only data allowed for evaluation of sepsis risk factors only in instances where population-level data were available. Finally, during this surveillance period, information on antimicrobial susceptibility was limited to the drug susceptibility interpretation for GBS and E coli, when recorded in the medical chart. More detail on multidrug resistance for all pathogens would be of interest in the future, ideally using a standard panel of drugs and comparable laboratory methods.
Observations from our multisite surveillance allay persistent concerns that GBS prevention efforts might have resulted in an increased burden of early-onset E coli infections. However, the stable burden of E coli early-onset sepsis we observed underscores the need for a prevention strategy. Although our surveillance identifies that many of the risk factors identified for GBS are similar for E coli, intrapartum prophylaxis has not resulted in declines, consistent with previous observations.20 Efforts to identify interventions targeting early-onset E coli infections specifically, and very preterm delivery more broadly, should remain a priority, as well as ongoing efforts to pursue maternal GBS vaccine development to protect newborns in the first days of life.
We thank Mia Apostol, Wendy Baughman, Pam Daily, Corinne Holtzman, Brenda Jewell, Gayle Langley, Melissa Lewis, Carmen Marquez, Patricia Martell-Cleary, Londell McGlone, Craig Morin, Stephanie Thomas, Amy Tunali, Michelle Wilson, and Carolyn Wright for their contributions to data collection and management.
- Accepted September 12, 2016.
- Address correspondence to Stephanie Schrag, DPhil, MS C25, Centers for Disease Control and Prevention, 1600 Clifton Rd, Atlanta, GA 30329. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This work was supported by the Centers for Disease Control and Prevention’s Emerging Infections Program Network/Active Bacterial Core surveillance.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2016-3038.
- Heron M
- Verani JR,
- McGee L,
- Schrag S
- Bizzarro MJ,
- Dembry LM,
- Baltimore RS,
- Gallagher PG
- Hyde TB,
- Hilger TM,
- Reingold A,
- Farley MM,
- O’Brien KL,
- Schuchat A
- Stoll BJ,
- Hansen NI,
- Sanchez PJ, et al
- Bekker V,
- Bijlsma MW,
- van de Beek D,
- Kuijpers TW,
- van der Ende A
- Berardi A,
- Baroni L,
- Bacchi Reggiani ML, et al
- Biondi E,
- Evans R,
- Mischler M, et al
- Weissman SJ,
- Hansen NI,
- Zaterka-Baxter K,
- Higgins RD,
- Stoll BJ
- Kent A,
- Kortsalioudaki C,
- Monahan IM, et al
- Schrag SJ,
- Hadler JL,
- Arnold KE,
- Martell-Cleary P,
- Reingold A,
- Schuchat A
- Copyright © 2016 by the American Academy of Pediatrics