Late-Onset Sepsis in Very Low Birth Weight Neonates: The Experience of the NICHD Neonatal Research Network
Objective. Late-onset sepsis (occurring after 3 days of age) is an important problem in very low birth weight (VLBW) infants. To determine the current incidence of late-onset sepsis, risk factors for disease, and the impact of late-onset sepsis on subsequent hospital course, we evaluated a cohort of 6956 VLBW (401–1500 g) neonates admitted to the clinical centers of the National Institute of Child Health and Human Development Neonatal Research Network over a 2-year period (1998–2000).
Methods. The National Institute of Child Health and Human Development Neonatal Research Network maintains a prospective registry of all VLBW neonates admitted to participating centers within 14 days of birth. Expanded infection surveillance was added in 1998.
Results. Of 6215 infants who survived beyond 3 days, 1313 (21%) had 1 or more episodes of blood culture-proven late-onset sepsis. The vast majority of infections (70%) were caused by Gram-positive organisms, with coagulase-negative staphylococci accounting for 48% of infections. Rate of infection was inversely related to birth weight and gestational age. Complications of prematurity associated with an increased rate of late-onset sepsis included patent ductus arteriosus, prolonged ventilation, prolonged intravascular access, bronchopulmonary dysplasia, and necrotizing enterocolitis. Infants who developed late-onset sepsis had a significantly prolonged hospital stay (mean length of stay: 79 vs 60 days). They were significantly more likely to die than those who were uninfected (18% vs 7%), especially if they were infected with Gram-negative organisms (36%) or fungi (32%).
Conclusions. Late-onset sepsis remains an important risk factor for death among VLBW preterm infants and for prolonged hospital stay among VLBW survivors. Strategies to reduce late-onset sepsis and its medical, social, and economic toll need to be addressed urgently.
Although advances in neonatal intensive care have led to improved survival of very low birth weight (VLBW) infants, late-onset sepsis continues to be an important cause of morbidity and mortality.1–10 The risk of late-onset sepsis increases with decreasing birth weight and gestational age. Ongoing infectious disease surveillance is essential because increasingly immature neonates are being provided with intensive care, require prolonged hospitalization, and are surviving. This study was undertaken to determine the current incidence of late-onset sepsis, distribution of infecting pathogens, risk factors for disease, and the impact of infection on subsequent hospital course in a cohort of VLBW infants cared for at the academic centers of the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network.
The NICHD Neonatal Research Network is a consortium of tertiary neonatal centers. The Network maintains a registry of all VLBW infants (401–1500 g) born and/or admitted to participating centers within 14 days of birth. This registry was developed to describe the populations at each participating center, to survey neonatal practice, to assess morbidity and mortality rates, and to provide information for the planning of randomized clinical trials.11 Trained research nurses collect maternal demographic, pregnancy and delivery data soon after birth and infant data until 120 days, discharge, or death.
The registry includes data on rates of late-onset sepsis, infecting organisms, and antibiotic therapy. In September 1998, infection surveillance was expanded to include detailed data on maternal intrapartum antibiotic use, results of all blood and cerebrospinal fluid cultures, C-reactive protein (CRP) results, timing and rates of infection, antibiotic use, indwelling lines, and other risk factors for infection. Blood cultures were processed by the clinical microbiology laboratories at each academic center using either the Bactec (Becton Dickinson, Sparks, MD) or BacT/Alert (Organon Teknika, Durham, NC) systems. One-half to 1 cc of blood was inoculated per blood culture.
Late-onset sepsis was defined as a positive result on 1 or more blood cultures obtained after 72 hours of life. Blood cultures positive for organisms generally considered to be contaminants, including corynebacterium, propionibacterium, penicillium, and diphtheroids were excluded from analysis. Records of patients with blood cultures positive for coagulase-negative staphylococci (CONS) were reviewed. The following definitions were used to distinguish definite or possible CONS infections from contaminants: definite infection: 2 positive blood cultures drawn within 2 days of each other or 1 positive blood culture and elevated CRP (>1) within 2 days of blood culture; possible infection: 1 positive blood culture and patient treated with vancomycin, oxacillin, or a semisynthetic antistaphylococcal agent for 5 or more days; probable contaminant: 1 positive culture without an elevated CRP or antibiotic therapy as outlined above. Patients with definite or possible CONS sepsis were included in the analysis. It is acknowledged that this definition may overestimate true rates of CONS infection. For all other pathogens, sepsis was defined by the presence of the organism in the blood culture.
An episode of sepsis was defined if a patient had a positive blood culture treated with antibiotics for 5 or more days or treated for a shorter period if the patient died. If the same organism was cultured after 10 days of appropriate antibiotic therapy or if a different organism was cultured from a subsequent culture, this was considered an additional episode. The focus of this study is on the first episode of late-onset sepsis.
Gestational age (GA) was determined by best obstetric estimates. Intraventricular hemorrhage (IVH) was graded according to the method of Papile et al12 Necrotizing enterocolitis (NEC) was classified according to the system of Bell et al13 For this analysis, respiratory distress syndrome (RDS) was defined by having each of the following: 1) need for oxygen at 6 to 24 hours of life, 2) clinical features of RDS within first 24 hours, 3) need for respiratory support (continuous positive airway pressure or ventilation) to age 24 hours, and 4) abnormal chest radiograph within first 24 hours. Chronic lung disease/bronchopulmonary dysplasia (BPD) was defined by an oxygen requirement at 36 weeks’ postmenstrual age. The clinician’s presumed cause of death was recorded for all infants who died. Coded causes of death related to infection include the following: proven sepsis/infection, suspected sepsis/infection, RDS with infection, BPD with infection, NEC with sepsis, and other (infection-related).
The participating centers and the numbers of patients each contributed to the study were: Yale University, 258; University of New Mexico, 309; Case Western Reserve University, 329; Stanford University, 336; University of Texas Southwestern Medical Center, 343; Wayne State University, 347; Emory University, 379; University of Tennessee at Memphis, 393; University of Miami, 412; Brown University, 426; University of Texas-Houston, 448; Harvard University, 462; University of Alabama, 504; Indiana University, 619; and University of Cincinnati, 650.
All analyses were based on the first episode of late-onset sepsis. Late-onset sepsis was treated as a binary variable. Associations between late-onset sepsis and maternal and neonatal variables, hospital course and death were explored. Statistical significance for unadjusted comparisons (eg, the variation in incidence of infection among clinical centers) was determined by the χ2 or Fisher exact test. When characteristics or conditions were considered antecedent or concurrent with late-onset sepsis (eg, maternal variables) adjusted relationships were evaluated using logistic regression models fit to the binary sepsis outcome. When sepsis was considered antecedent, adjusted relationships were evaluated by logistic regression if the outcome was binary (eg, the relation of sepsis to death) and by linear regression if the outcome was continuous (eg, the relation of sepsis to number of days in the hospital). The relation of variables other than GA and birth weight to late-onset sepsis were evaluated after adjusting for GA and clinical center, with other covariates included as noted in the text and footnotes to the tables. All analyses were completed using SAS software (SAS Institute, Cary, NC).14
Incidence of Infection
Between September 1, 1998, and August 31, 2000, 6956 VLBW neonates were admitted to Network centers. Because this study reviews infections occurring after 72 hours of life, neonates who died within the first 3 days of life were excluded from the analysis. The results presented are based on 6215 neonates who survived at least 3 days.
Sixty-two percent of infants (3856/6215) had 1 or more blood cultures obtained after day 3. Infants had between 1 and 39 late blood cultures obtained (24% had a single blood culture; 18% had 2; 13% had 3; and 45% had 4 or more cultures). A total of 17 712 blood cultures were obtained after day 3, of which 3584 (20%) were positive.
Twenty-five percent of infants (1536/6215) had at least 1 positive blood culture after day 3. We excluded the culture results of 168 infants with a single blood culture positive for an organism considered to be a contaminant (see definitions above) and 55 infants with a single positive blood culture that grew >1 organism (dual infections). The analyses we report are based on 1313 infants with late-onset sepsis—ie, 34% of all infants evaluated with a blood culture (1313/3856) or 21% of all infants studied (1313/6215). The majority of infants had a single episode of late-onset sepsis (943/1313, 72%); 20% (262/1313) had 2 episodes; 6% (76/1313) 3 episodes, and 2% (32/1313) 4 or more episodes. Infants with the lowest birth weights were more likely to have multiple episodes (401–500 g, 38%; 501–750 g, 39%; 751-1000 g, 25%; 1001–1250 g, 21%; 1251–1500 g, 13%; P < .001). The focus of this study was on the first episode of late-onset sepsis. There was considerable center-to-center variability in the incidence of late-onset sepsis, with rates ranging from 10.6% to 31.7% of VLBW infants at study sites, (P < .001). This difference remains significant after adjusting for birth weight, GA, race, and sex.
The vast majority (70%) of first episode late-onset infections were caused by Gram-positive organisms (Table 1). CONS were the most common late-onset pathogens (48% of all infections, 68% of Gram-positive infections). Using the definitions outlined above, 276 (44%) CONS were definite infections and 353 (56%) were possible infections. If possible infections are excluded, CONS still remains the single most common agent isolated (276/1313, 21%). Other Gram-positive organisms included Staphylococcus aureus, Enterococcus spp, and group B Streptococcus. Gram-negative pathogens accounted for 18% of first episode infections. Escherichia coli, Klebsiella, Pseudomonas, Enterobacter, and Serratia were the most frequent Gram-negative pathogens. Fungal organisms were responsible for 12% of first episode late-onset sepsis. Of all pathogens, Candida albicans was the third most frequent organism isolated (76/1313, 6%).
The average age for the first episode of late-onset sepsis was 22 ± 0.5 days (median: 17 days; 75th percentile: 28 days; 95th percentile: 57 days). The age for first infection did not differ significantly by infecting pathogen (data not shown).
Maternal demographic data and clinical histories were reviewed. After adjustment for GA and clinical center, there were no statistically significant associations between infection and maternal age, marital status, prenatal care, multiple gestation, hypertension or preeclampsia, duration of rupture of membranes, presence of labor, mode of delivery, or antenatal antibiotic or steroid therapy (data not shown).
Birth weight and GA were strongly associated with risk of late-onset sepsis. Forty-three percent of neonates with birth weights of 401 to 750 g had at least 1 episode of blood culture-proven late-onset sepsis. The rate of infection decreased with increasing birth weight to 28% for infants 751 to 1000 g, 15% for infants 1001 to 1250 g, and 7% for those 1251 to 1500 g (P <.001). Similarly, the infection rate was inversely related to GA. Almost half (46%) of the neonates born at <25 weeks’ gestation developed late-onset sepsis. This rate of infection declined to 29% at 25 to 28 weeks, 10% at 29 to 32 weeks, and only 2% for infants born after 32 weeks (P < .001). Rates of first episode infection per 1000 hospital days also decreased with increasing birth weight and GA (Table 2). After adjusting for GA, birth weight, and clinical center, there was no relationship between either race or sex and the risk of late-onset sepsis (percentage of neonates with infection by race and sex: black, 23%; Hispanic, 23%; white, 19%; other, 15%; P = not significant [NS]; male, 22%; female, 21%; P = NS).
Most VLBW neonates have complex medical problems and prolonged hospitalization. Infants with late-onset sepsis had a longer length of hospitalization (79 versus 60 days; P < .001). This difference remained significant after additional adjustment for IVH, BPD, and NEC in infants hospitalized >28 days (99 versus 89 days; P <.001). There were no associations between late-onset sepsis and RDS, surfactant use, or severe IVH/periventricular leukomalacia (data not shown). However, patients with PDA (34% vs 16%, odds ratio [OR]: 1.2 [1.0–1.4]; P = .04), BPD (41% vs 16%, OR: 1.3 [1.0–1.5]; P =.02), and NEC (50% vs 19%, OR: 2.7 [2.1–3.4]: P <.001) had a higher incidence of late sepsis.
Enteral feeding history and the need for prolonged intravenous access reflect severity of illness. Our data document the widespread use of central catheters among VLBW infants; 46% of all infants had a percutaneously inserted central catheter (PICC) at some time during their hospital course, 15% had a peripheral arterial line (PAL), and 8% had a surgically placed central line (CVL). Umbilical lines were also used frequently: 51% had an umbilical arterial (UAC) and 40% had an umbilical venous catheter (UVC). The risk of late-onset sepsis increased as the duration of central lines and parenteral hyperalimentation increased (Tables 3 and 4). The earlier an infant was started on enteral feedings, the earlier he or she reached full enteral feedings (ie, supplemental intravenous fluids <20 μL/kg per day), and the earlier birth weight was regained, the less likely he or she was to develop infection.
The risk of infection increased in patients with increasing duration of ventilator support (Table 3). Half of the infants who were ventilated for >28 days developed late-onset sepsis versus only 9% of those who were ventilated for 1 week or less. Moreover, infants who developed late-onset sepsis had a significantly longer duration of mechanical ventilation than those without late-onset sepsis (24 vs 12 days; P < .001, Table 4). This difference remained significant after additional adjustment for IVH, BPD, and NEC in infants hospitalized >28 days (32 vs 25 days; P < .001). In a multivariate model that evaluated various risk factors for infection (GA, clinical center, central line days-PICC, UVC, UAC, surgical CVL, PAL-hyperalimentation days and ventilator days) the following remained significant: GA, center, surgical CVL, PAL, hyperalimentation, and ventilator days (each P <.01).
Fifty-six percent of all infants (3459/6215) received at least 1 course of antibiotics started after day 3. The rate of antibiotic use was inversely related to birth weight and GA. Infants who developed infection were more likely to receive antibiotics after day 3 (100% vs 44%; P <.001). The most frequently used antibiotics were gentamicin, vancomycin, cefotaxime, ampicillin, and tobramycin. Forty-four percent of all infants (2712/6215) were treated with vancomycin after day 3. Vancomycin use was inversely related to birth weight (401–500 g, 78%; 501–750 g, 75%; 751-1000 g, 60%; 1001–1250 g, 36%; 1251–1500 g, 18%). Although significantly more infants with late-onset sepsis were treated with vancomycin (93%), 30% of patients without proven infection also received this drug (P < .001).
Nine percent of all patients (578/6215) received antifungal therapy. Use of antifungal drugs was inversely related to birth weight (401–500 g, 34%; 501–750 g, 28%; 751-1000 g, 10%; 1001–1250 g, 4%; 1251–1500 g, 2%). Amphotericin B was the most frequently used drug (85% of all doses), followed by fluconazole (10% of all doses).
Among this cohort of 6215 VLBW infants who survived >3 days, 597 infants (10%) died. Infants who developed late-onset sepsis were significantly more likely to die than those who did not (18% vs 7%; P <.001). GA, study center, and male sex were all highly significant predictors of death (each, P <.001). The increased risk of death for infants with late-onset sepsis remained after adjustment for GA, study center, and sex.
Overall 242 (18%) of the 1313 infants with late-onset sepsis died. We evaluated the last positive blood culture before death. Forty-five infants (19%) died on the same day the blood culture was obtained; 54 (22%) died on the following day; 19 (8%) died on day 3; 29 (12%) died between days 4 and 7; and 95 (39%) infants died >7 days after the last positive culture was obtained. Infants with Gram-negative organisms were more likely to die acutely—ie, within 3 days of the positive blood culture (Table 5). The clinician’s presumed cause of death was evaluated for each patient with late-onset sepsis who died. Infants who died within 3 days of a positive blood culture were more likely to have a death attributed to infection (87%) than were infants who died later (4–7 days, 71%; >7 days, 46%). Infants with Gram-negative sepsis were more likely to have deaths attributed to infection (81%) than were those infected with other agents (fungi, 73%; non-CONS Gram-positives, 61%; and CONS, 49%).
There were significant differences in death rates with different organisms isolated from the last positive culture before death (Table 6). VLBW infants with late-onset Gram-negative sepsis were at greatest risk of death (36% died; OR for death, Gram-negative vs other organisms, 3.5 [2.5–4.9]; P <.001). Within this high-risk group, infants with Pseudomonas infections were most likely to die (74% of infants with Pseudomonas infections died). Infants with late-onset fungal sepsis were also at increased risk of death compared with infants with other late-onset infections (32% died; OR, fungi vs other organisms, 2.0 [1.3–3.0]; P <.01). Overall, VLBW infants with late-onset Gram-positive sepsis were not more likely to die than infants who were uninfected (11% died; OR for death, Gram-positive vs uninfected 0.9 [0.7–1.2], P = NS). This is primarily attributable to the large number of infections with CONS (9% died; OR for death, CONS vs uninfected 0.8 [0.4–1.4]; P = NS). Although numbers were small and differences not statistically significant, patients infected with S aureus or with group B streptococci were more likely to die than patients who were uninfected (17% and 22%, respectively, vs 7%).
Late-onset sepsis remains an important and potentially lethal complication among VLBW infants.1–10 In this study, 21% of VLBW neonates who survived beyond 3 days of age had at least 1 episode of late-onset sepsis. The median age of onset for the first episode of late-onset sepsis was over 2 weeks of age. These infections are particularly poignant for parents and physicians because this complication affects VLBW infants who have survived early causes of mortality but remain at ongoing risk for infection. With increasing survival of ELBW preterm infants, late-onset sepsis will continue to be a challenging complication that affects other morbidities, length of hospitalization, cost of care, and mortality rates.
There was considerable variability in the incidence of late-onset sepsis between the 15 Network centers (range: 11%–32%). Studying centers with the lowest and highest infection rates might identify medical, nursing, or respiratory care practices and potential interventions that could affect infection risk. Recently, a number of health care organizations have adopted collaborative improvement models to change care and reduce adverse outcomes. The Vermont Oxford Neonatal Network has successfully used collaborative quality improvement to identify and implement “potentially better practices” related to nosocomial infection (ie, practices regarding handwashing, nutrition, skin and respiratory care, vascular access, diagnostic methods, and neonatal intensive care [NICU]“culture”) and has documented a reduction in late-onset CONS sepsis in selected intervention NICUs.15
Many studies report that CONS are the most common organisms associated with late-onset sepsis.1–5,7,8,16 It remains difficult to determine which blood culture isolates of CONS reflect true infections and which are contaminants. The finding that the 629 patients in this study classified as having definite or possible CONS were not more likely to die than patients who were uninfected, casts doubt on whether all were true infections. Alternatively, this organism—even if it invades the bloodstream—may be less virulent. The majority of CONS organisms are resistant to the routine antibiotics used to treat newborn infants, and vancomycin is often required for adequate therapy. It is alarming that 44% of infants in this cohort (whether or not they had documented CONS infection) were treated with vancomycin. The Centers for Disease Control and Prevention and others17,18 have recommended avoiding empiric vancomycin therapy in patients with suspected sepsis to prevent the emergence and spread of vancomycin resistant strains. Our data suggest that this is a reasonable approach that warrants formal study.
Over half of the patients in this cohort received antibiotic therapy; however, only 21% had blood culture proven infection. Despite numerous published guidelines on the appropriate use of antibiotics and repeated warnings about the dangers of antibiotic resistance, physicians continue to use antibiotics both inappropriately and excessively.19 Neonatologists must promote the rational use of antibiotics including optimal choice of drug (narrow spectrum when possible; broad spectrum in selected patients) and optimal duration of therapy (2–3 days in patients whose blood or other systemic cultures are negative; longer in patients with proven infection).20
Nutrition history and central catheter use were related to infection. Infants with prolonged duration of central catheters and parenteral hyperalimentation, those with delayed initiation of enteral feedings, and those with a prolonged period to reach full enteral feedings or to regain their birth weight were all at substantially increased risk of late-onset sepsis. These data suggest that efforts to initiate enteral feedings as early as possible, to minimize the use of central venous catheters, and to reduce the number of catheter days and/or days on parenteral hyperalimentation/ intralipids would not only improve nutritional status but also decrease the risk of infection.
Infants who developed late-onset sepsis were on mechanical ventilation for significantly longer than those who were uninfected and were significantly more likely to develop BPD. These data suggest that decreasing the number of days on the ventilator and attention to ventilator-associated infection control issues (eg, reducing interruption of ventilator-endotracheal tube circuits) might reduce the rate of infection. Moreover, the role of infection in the cause of BPD needs additional study.
Duration of hospitalization was significantly longer in infants who developed late-onset sepsis than in those who did not. Prolonged hospitalization of VLBW neonates is a major factor in the high cost of neonatal intensive care. Therefore, late-onset sepsis has substantial financial implications for both the individual patient and VLBW care in general. Infants who developed late-onset sepsis were significantly more likely to die than those without infection. Those with late-onset Gram-negative or fungal sepsis were at greatest risk of death (36% and 32% died, respectively). As reported previously, mortality risk was influenced by infecting pathogen.3,18 Death rates were highest for infants infected with Pseudomonas, C albicans, Serratia, and E coli. Moreover, infants with Gram-negative infections were more likely to suffer a fulminant illness with acute death.18
This review of late-onset sepsis among 6215 VLBW neonates underscores the importance of this serious complication and suggests that strategies to reduce late infections in VLBW neonates and their medical, social, and economic toll are needed urgently. The use of collaborative quality improvement strategies to reduce nosocomial infections among VLBW NICU patients warrants additional study. Successful interventions should improve survival, shorten mechanical ventilation and hospital stay, decrease antibiotic usage, and reduce the high cost of caring for VLBW infants.
This study was supported by National Institutes of Health grant numbers: U10 HD27851 (Dr Stoll), U01 HD36790 (Ms Hansen), U10 HD21364 (Dr Fanaroff), U10 HD34216 (Dr Carlo), U10 HD27871 (Dr Ehrenkranz), U10 HD27856 (Dr Lemons), U10 HD27853 (Dr Donovan), U10 HD34167 (Dr Stark), U10 HD21373 (Dr Tyson), U10 HD27904 (Dr Oh), U10 HD21397 (Dr Bauer), U10 HD21415 (Dr Kornoes), U10 HD21385 (Dr Shankaran), U10 HD40689 (Dr Laptook), U10 HD27880 (Dr Stevenson), and U10 HD27881 (Dr Papile).
We thank Mazie Tinsley for preparation of the manuscript.
- Received September 25, 2001.
- Accepted February 8, 2002.
- Address correspondence to Barbara J. Stoll, MD, Department of Pediatrics, Emory University School of Medicine, 2040 Ridgewood Dr, NE, Atlanta, GA 30322
- Gaynes RP, Edwards JR, Jarvis WR, Culver DH, Tolson JS, Martone WJ. Nosocomial infections among neonates in high-risk nurseries in the United States. Pediatrics.1996;98 :357– 361
- ↵Hack M, Horbar JD, Malloy MH, Tyson JE, Wright E, Wright L. Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Network. Pediatrics.1991;87 :587– 597
- ↵SAS Institute. SAS/STAT User’s Guide. Version 8. Cary, NC: SAS Institute Inc; 1999
- ↵Horbar JD, Rogowski J, Plsek PE, et al. Collaborative Quality Improvement for Neonatal Intensive Care. Pediatrics.2001;107 :14– 22
- ↵Gray JE, Richardson DK, McCormick MC, Goldmann DA. Coagulase-negative staphylococcal bacteremia among very low birth weight infants: relation to admission illness severity, resource use, and outcome. Pediatrics.1995;95 :225– 230
- ↵Karlowicz MG, Buescher ES, Surka AE. Fulminant late-onset sepsis in a neonatal intensive care unit, 1988–1997, and the impact of avoiding empiric vancomycin therapy. Pediatrics.2000;106 :1387– 1390
- ↵Isaacs D. Rationing antibiotic use in neonatal units. Arch Dis Child.2000;82 :F1– F2
- Copyright © 2002 by the American Academy of Pediatrics