Risk Factors and Opportunities for Prevention of Early-onset Neonatal Sepsis: A Multicenter Case-Control Study
Background. Early-onset group B streptococcal (GBS) prevention efforts are based on targeted use of intrapartum antibiotic prophylaxis (IAP); applicability of these prevention efforts to infections caused by other organisms is not clear.
Methods. Multicenter surveillance during 1995 to 1996 for culture-confirmed, early-onset sepsis in an aggregate of 52 406 births; matched case-control study of risk factors for GBS and other sepsis.
Results. Early-onset disease occurred in 188 infants (3.5 cases per 1000 live births). GBS (1.4 cases per 1000 births) andEscherichia coli (0.6 cases per 1000 births) caused most infections. GBS sepsis less often occurred in preterm deliveries compared with other sepsis. Compared with gestation-matched controls without documented sepsis, GBS disease was associated with intrapartum fever (matched OR, 4.1; CI, 1.2–13.4) and frequent vaginal exams (matched OR, 2.9; CI, 1.1–8.0). An obstetric risk factor—preterm delivery, intrapartum fever, or membrane rupture ≥18 hours—was found in 49% of GBS cases and 79% of other sepsis. IAP had an adjusted efficacy of 68.2% against any early-onset sepsis. Ampicillin resistance was evident in 69% of E coliinfections. No deaths occurred among susceptible E coliinfections, whereas 41% of ampicillin-resistant E coliinfections were fatal. Ninety-one percent of infants who developed ampicillin-resistant E coli infections were preterm, and 59% of these infants were born to mothers who had received IAP.
Conclusions. Either prenatal GBS screening or a risk-based strategy could potentially prevent a substantial portion of GBS cases. Sepsis caused by other organisms is more often a disease of prematurity. IAP seemed efficacious against early-onset sepsis. However, the severity of ampicillin-resistant E colisepsis and its occurrence after maternal antibiotics suggest caution regarding use of ampicillin instead of penicillin for GBS prophylaxis.
- group B streptococcal
- Escherichia coli
- antibiotic resistance
- risk factors
- GBS =
- group B streptococcus •
- IAP =
- intrapartum antibiotic prophylaxis •
- ROM =
- rupture of membranes
Numerous studies identified obstetric risk factors for early-onset group B streptococcal (GBS) disease,1 and clinical trials addressed the efficacy of intrapartum antimicrobial prophylaxis.2–4 Less information is available regarding risk factors for early-onset sepsis attributable to other organisms.5 ,6 The risk-based approach to preventing GBS disease involves intrapartum antibiotic prophylaxis (IAP) when unscreened women present with threatened preterm delivery at <37 weeks, rupture of membranes (ROM) ≥18 hours, or intrapartum fever. The potential for this approach to prevent other causes of early-onset sepsis has not been fully assessed.
Recent attention to prevention of perinatal transmission of GBS through the use of intrapartum antimicrobial prophylaxis4 7–9 has generated renewed concern regarding perinatal infections caused by antimicrobial resistant organisms.5 ,10 ,11 Concerns about whether increased prophylaxis will promote infection with organisms resistant to the prophylaxis agent (eg, penicillin or ampicillin) must be addressed.5 ,10 ,12 Baseline incidence data on the pathogens that cause early-onset neonatal sepsis in diverse populations will help monitor changes throughout time.
Neonatal sepsis, including that caused by GBS, is more common among minority populations.13 ,14 Although prevention strategies are recommended for use in all obstetric programs, the feasibility of the two consensus approaches in hospitals serving primarily minority populations is not well documented, and missed opportunities for prevention have not been categorized. We conducted a multicenter study in institutions primarily serving minority populations to identify risk factors for early-onset sepsis and to assess the efficacy of intrapartum antibiotics for prevention of early-onset sepsis.
We conducted surveillance for early-onset sepsis and meningitis among infants born in the following hospitals from April 1995 through December 1996: University of Miami Medical Center-Jackson Memorial Hospital, Miami, FL; University of Mississippi Medical Center, Jackson, MS; University Hospital, Carolina Area, and Caguas Regional Hospital at Puerto Rico, San Juan, PR; University of Tennessee at Memphis Regional Medical Center, Memphis, TN; and Broward General Medical Center, Fort Lauderdale, FL. Surveillance was also conducted at Grady Memorial Hospital, Atlanta, GA (May 1995 through December 1996) and at the Medical College of Virginia, Richmond, VA (March 1995 through March 1997). There were 52 406 births at the participating hospitals during surveillance.
We defined a case of early-onset disease if a bacterial pathogen was isolated from blood or cerebrospinal fluid from an infant <7 days old who was born in one of the study hospitals, with the following exception: coagulase-negative Staphylococcus was included only if the organism was isolated from 2 or more blood cultures and clinical signs of illness were present. Organisms considered contaminants (eg, diphtheroids) were excluded from analysis.
Cases were identified by reviewing microbiology laboratory records. No specimens were collected for this study. No systematic efforts were made to identify cases of sepsis among infants born within a study hospital but cared for at other facilities.
For the case-control study, patients with sepsis or meningitis in all facilities except for Grady Memorial Hospital were enrolled. Up to 2 controls were matched to each case-patient based on gestational age category (<34, 34–36, ≥37 weeks), closest calendar month of birth (within 6 months), and hospital of delivery.
Intrapartum antibiotic use was defined as receipt of at least 1 dose of antibiotic therapy during labor but before delivery. For assessment of missed opportunities for prevention, we considered there to be an opportunity to provide antibiotics if a woman was in the hospital after the risk condition occurred and at least 1 hour before delivery. We assumed that women in prenatal care could have had a late gestation GBS test performed, and that late screening has a sensitivity of 87% of detecting GBS among women colonized at delivery.15
Information on clinical and demographic characteristics was collected by reviewing hospital records using a structured data abstraction form. Data analysis used SAS version 6.12 (SAS Institute, Cary, NC). Comparison of categorical data used the χ2 test and continuous data were compared using the Wilcoxon sign rank test for nonparametric distributions. The case control data were analyzed using conditional logistic regression. As with other evaluations of preventive interventions such as vaccines, we estimated the efficacy of intrapartum antibiotics using the formula: (1 − matched odds ratio) × 100%.
The protocol was approved by the institutional review boards of the CDC and the participating hospitals.
One hundred eighty-eight cases of early-onset disease were identified based on 183 blood isolates, 2 cerebrospinal fluid isolates, and 3 cases for which organisms were isolated from both blood and cerebrospinal fluid specimens. The overall incidence of early-onset disease was 3.5 cases per 1000 live births, but varied significantly by center from 6.6 per 1000 (Grady Memorial Hospital) to 1.2 per 1000 (University of Miami Medical Center) (P < .0001; Fig 1). The frequency of organisms identified is shown in Table 1. The rates of invasive GBS and Escherichia coli disease were 1.4 and 0.6 cases per 1000 live births, respectively.
Early-onset GBS disease differed from that caused by other organisms (Table 2). GBS case patients were more likely to be born at term and had higher birth weights. GBS early-onset disease was less often fatal and was more likely to have onset of illness (defined as date at which positive culture was collected) occur on the day of birth. Figure 2 shows the etiologies of early-onset disease cases by day of onset. Although GBS accounted for 53% of infection presenting on the first day of life, no GBS occurred after day 3, in contrast to other organisms. E coli was the dominant organism on days 4 through 6 (accounting for 50% of these infections). Early-onset disease caused by GBS was less likely to occur in the presence of ROM ≥18 hours, compared with sepsis caused by other pathogens. One or more of the obstetric complications which are used as an indication for GBS prophylaxis, according to the risk-based strategy, was present in less than half of the early-onset GBS cases, compared with nearly 80% of other early-onset sepsis cases (Table 3).
One hundred two early-onset cases (41 GBS; 61 other pathogens) and 200 gestation- and hospital-matched controls were enrolled in the case-control study. A similar proportion of case and control infants were male, African-American, and born to mothers <20 years old. Exposure to tobacco and antenatal antibiotics did not differ between case and control mothers. Exposure to illicit drugs and to alcohol use during pregnancy was infrequently noted in the medical records of either case or control patients. No differences in previous pregnancy complications such as stillbirths or spontaneous abortions were identified.
GBS case-mothers were four times as likely as matched control-mothers to have intrapartum fever (Table 4) and nearly three times as likely to have had ≥6 vaginal exams intrapartum. Multivariate analysis did not reveal an association between these features and early-onset disease caused by other organisms.
Efficacy of Intrapartum Antibiotics
Intrapartum antibiotics were not associated with disease or its prevention in univariate analysis. Because antibiotics were often given for therapeutic purposes, rather than prophylaxis, we used bivariate analysis to assess the role of intrapartum antimicrobial use for prevention of GBS and non-GBS infections. We adjusted the efficacy estimates by accounting for the presence of maternal fever during labor and delivery. Adjusting for the presence of fever, receipt of intrapartum antibiotics was associated with an efficacy of 68.2% (95% CI, 23.4%–86.8%) against all early-onset disease. Adjusting for the presence of fever, efficacy was 85.0% against GBS early-onset disease (95% CI, −42.3%–98.4%) and 63% against early-onset disease attributable to other organisms (95% CI, 1.5%–86.1%). Adjusted efficacy was 81.8% (95% CI, 32.1%–95.1%) against any sepsis occurring on the first day of life, but was 58.9% (95% CI, −83.7%–90.8%) against sepsis occurring on the third day of life or thereafter. The sample size was insufficient to assess efficacy after 1 or 2 doses of antibiotics or after antibiotics initiated at differing time intervals before delivery.
Missed Opportunities for Prevention
Of 188 cases, 42 (23.0%) occurred despite maternal receipt of intrapartum antimicrobial agents (ie, represented antibiotic failures). Among the 42 case-mothers who had received intrapartum antibiotics, 18 (44%) received antibiotics only after fever was recognized. Sixty-two percent (26 out of 42) of early-onset case-mothers who received antibiotics got a single dose intrapartum (range, 1–26 doses).
Most mothers of GBS case-patients had an opportunity for either the risk-based or screening-based GBS prevention strategy to be applied. Of 76 GBS case-mothers, 31 (40.8%) had at least one obstetric risk factor (gestation <37 weeks; ROM ≥18 hours; intrapartum fever) present at least 1 hour before delivery—ie, an opportunity to intervene with antibiotic prophylaxis. Of the 69 GBS case-mothers with sufficient records to assess prenatal history, 67 (97%) had received prenatal care, but only 10 (14.9%) of these had a prenatal GBS test collected. From these data, lack of prenatal care utilization does not seem to be an obstacle to using a screening-based approach in these populations yet screening was not common during this study period.
Antimicrobial Resistance Among Early-onset Sepsis
Antimicrobial susceptibility of pathogens isolated from early-onset cases was measured in each of the study hospitals as part of routine clinical microbiology procedures. Resistance to ampicillin was detected in 22 of 32 available isolates of E coli. Of the 22 cases who had an ampicillin-resistant E coliinfection, 13 (59%) were born to mothers who had received intrapartum antibiotics. Eight of the 13 women received only a single dose of antibiotics, and only 5 women received intrapartum antibiotics for prophylaxis, the remaining 8 women received antibiotics for treatment of maternal infections present during labor and delivery. Case fatality of resistant and susceptible E coli infections was 40.9% (9 out of 22) and 0% (0 out of 10) respectively (P = .017). Resistant cases were more often preterm (91%) compared with susceptible cases (20 out of 22 vs 2 out of 10, P < .001). A higher proportion of mothers of those infants who had sepsis caused by resistant E coli had received intrapartum antibiotics (59.1% vs 0%, P = .004).
This multicenter study assessed the incidence of culture-confirmed sepsis in the first week of life. Early-onset disease occurred at a rate of 3.5 cases per 1000 live births and 16.0% of patients died. Consistent with earlier reports, group B streptococcus was the predominant organism identified, accounting for more than 40% of cases. The aggregate rate of early-onset GBS disease during the 1995 to 1996 study period—1.4 per 1000 births—was similar to the estimate of 1.3 per 1000 births derived from population-based surveillance in 4 geographic areas during 1995.16 Nevertheless, substantial variation in the incidence of early-onset disease caused by GBS and other organisms was identified in the 7 study sites.
Towers et al12 recently reported that when early-onset GBS rates declined at one institution after increased use of intrapartum antimicrobial prophylaxis, the rate of E coli sepsis increased significantly. Although our surveillance includes 50 000 births, the relatively small numbers of cases in each hospital limited our ability to correlate GBS and non-GBS rates of disease. Longitudinal surveillance is needed to determine if prevention of GBS disease by increased use of intrapartum antibiotics will significantly affect rates of early-onset disease attributable to other organisms.
Comparison of early-onset disease caused by GBS with that attributable to other organisms identified several differences. GBS was significantly more likely to cause disease in full-term infants and to present on the day of birth. Sepsis attributable to other organisms was significantly more likely to result in death (22.3% vs 6.7%), perhaps as a result of the predominance of preterm infants among those infants who developed sepsis caused by other agents.
A subset of early-onset cases were compared with controls to identify risk factors for sepsis caused by GBS or other organisms. As in previous studies, intrapartum fever was significantly more likely to occur in mothers whose infants developed GBS disease. A nearly 3-fold risk of early-onset GBS disease occurred when 6 or more vaginal exams were performed before delivery. Previous studies have also associated frequent vaginal exams with amnionitis, early-onset sepsis, and neonatal infections.6 ,17 ,18 Unlike previous studies, early onset GBS disease was not significantly associated with maternal age <20 years.14 ,19 The limited racial diversity within each hospital's study patients may explain the lack of an association between GBS disease and race. Although population-based surveillance continues to identify higher rates of early-onset GBS disease among African-Americans, rates of disease have declined more rapidly among this group, resulting in a reduced racial disparity.16
We did not find that risk factors for early-onset GBS sepsis were associated with increased risk of sepsis attributable to other organisms. Intrapartum fever, ROM ≥ 18 hours before delivery, and frequent vaginal exams were not associated with sepsis attributable to other organisms. This may be partially explained by differences in the route of transmission between GBS and other organisms. For example, GBS is clearly acquired from the mother in nearly all cases of early onset disease; only 7% of GBS cases presented after the first day of life. By contrast, 39% of non-GBS cases presented after the first day of life, and thus, may represent infections acquired during hospitalization or through contact with others. Prematurity may be the overwhelming risk factor for non-GBS infections presenting on the day of birth.
Prevention of early-onset group B streptococcal disease by use of a risk-based strategy to select women for IAP is increasing.7 Whether this approach will also reduce infections caused by other organisms is not yet known. The applicability of the risk-based approach to prevention of other infections will depend on the prevalence of risk conditions among these cases and on the effectiveness of intrapartum antibiotics against other causes of early-onset disease. In this study, at least one of the obstetric risk factors was present in 79% of early-onset cases caused by other organisms. Although non-GBS cases were more likely to have an obstetric risk factor, IAP seemed to have lower efficacy against other pathogens compared with GBS, although the small number of study participants enrolled resulted in wide confidence intervals.
We estimated the efficacy of intrapartum antibiotics against GBS early-onset disease to be 85%, however, this estimate was not statistically significant. Because antibiotics were frequently given for maternal fever, it was difficult for us to distinguish prophylactic from therapeutic use. We therefore reported efficacy based on bivariate adjustment for the presence of fever. However, this may overestimate the true efficacy of intrapartum prophylaxis. Our study was also too small to assess the efficacy of a single dose of antibiotics or of antibiotics initiated 4 hours before delivery. Both the number of prophylactic doses and the duration of intrapartum prophylaxis before delivery were proposed as criteria to consider in determining the need for diagnostic testing of newborns whose mothers received antibiotics for GBS prevention.4 ,9 Because limited data exist to support these criteria,20 future studies are needed to characterize adequacy of prophylaxis.
We found higher efficacy of intrapartum antibiotics against infections presenting on the first day of life (81.8%) compared with those presenting 3 or more days after birth (58.9%). This observation may account for the difference between efficacy of prophylaxis against GBS versus other pathogens, because most GBS disease presents on the first day of life. Intrauterine acquisition of the organism is more likely for cases with the earliest onset, and those episodes, whether caused by GBS or other organisms, may be the most amenable to maternal antibiotic administration.
Our study was completed before consensus guidelines were implemented in most of the centers participating in the study, and therefore provides insight into the proportion of early onset GBS disease that is potentially preventable by the two strategies recommended in mid-1996. We found that nearly all GBS case-mothers had received prenatal care, during which a screening test could have been collected, and 84.4% initiated care before the 7th month, and might have been able to receive a GBS vaccine scheduled for the late second trimester. Based on the assumption that a GBS screening culture with 87% sensitivity15 could be collected from all women with prenatal care delivering at term, that those delivering preterm were eligible for prophylaxis, and that women admitted for labor at least an hour before delivery had an opportunity for antibiotics to be offered, we estimate that 74% of GBS case-mothers could have been offered intrapartum prophylaxis according to the screening-based strategy. If efficacy of intrapartum antibiotics is 85%, 66% of early-onset GBS cases could have been prevented using the screening-based method. We found that 41% of GBS case-mothers had at least one obstetric risk factor and at least an hour in the hospital after the risk factor was evident before delivery, providing an opportunity to offer antibiotics against the disease; with 85% efficacy, we estimate that 35% of GBS cases could have been prevented with implementation of the risk-based strategy in these hospitals. Our findings suggest that the screening-based approach might have more impact on GBS cases than implementation of the risk-based approach. However, prenatal culture results must be reliably transferred to the labor and delivery setting, and failure to do so would reduce the effectiveness of this strategy.
Assessment of the impact of prevention strategies on early-onset sepsis attributable to other organisms is needed. The screening-based approach offers antibiotic prophylaxis to all women with threatened preterm delivery if a GBS culture result is unavailable. Nearly half (46%) of early onset disease caused by organisms other than GBS occurred in infants delivered before 34 weeks' gestation, and 60% of non-GBS cases delivered before 37 weeks; thus, many mothers of non-GBS cases would also have been offered prophylaxis even under the screening-based approach. Full implementation of the risk-based strategy would have resulted in antibiotic administration to 67% of non-GBS case-mothers; assuming an efficacy of 63% against non-GBS early-onset disease, 42.2% of the non-GBS cases could have been prevented by the risk-based strategy, versus 33.5% of non-GBS cases prevented by the screening-based strategy. It is worth noting that penicillin is less appropriate than ampicillin for prevention of Gram-negative sepsis such as attributable to E coli. The lower efficacy against these agents and high rates of antimicrobial resistance raise questions about the potential risks and benefits for use of broader spectrum agents for prophylaxis.12
We confirmed the severity of early-onset sepsis attributable to resistant E coli reported by Joseph et al.5 In hospitals participating in our study during 1995 to 1996, 69% of early-onset E coli disease was attributable to ampicillin-resistant organisms, and fatality in these infants was 40.9%, compared with no deaths among sepsis caused by susceptibleE coli. However, infants with resistant E coliwere born after a median of 30 weeks' gestation, compared with 39 weeks' gestation for susceptible E coli sepsis cases, and thus, prematurity confounds the association between resistance and death.
Our data suggest that control of resistant E coli infections will be important in future efforts to address neonatal mortality. More generally, we found that the problem of fatal neonatal sepsis was very closely linked to the larger problem of prematurity. Effective approaches to reduce preterm deliveries are needed.
We gratefully acknowledge financial support for this study from the Office of Minority and Women's Health, National Center for Infectious Diseases, Centers for Disease Control and Prevention.
The PENS Study Group Members include Sonyia Elder, Dennis Crouse, Joyce Fricke, Ana Gonzalez, Ariana Fajardo, Gene Burkett, JoNell Potter, Elizabeth Bowman, Karen Priester, and Marta Valcarel.
- Received April 20, 1999.
- Accepted June 8, 1999.
Reprint requests to (O.S.L.) Respiratory Diseases Branch, Division of Bacterial and Mycotic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, 1600 Clifton Rd, MS C-23, Atlanta, GA 30333. E-mail:
- American College of Obstetricians and Gynecologists
- American Academy of Pediatrics, COID/COFN
- ↵Stoll BJ, Holman RC, Schuchat A. Decline in sepsis-associated neonatal and infant deaths in the United States, 1979 through 1994. Pediatrics. 1998;102(2). URL: http://www.pediatrics.org/cgi/content/full/102/2/e18
- Copyright © 2000 American Academy of Pediatrics