Article Text

Neurodevelopmental outcomes following neonatal late-onset sepsis and blood culture-negative conditions
  1. Sagori Mukhopadhyay1,
  2. Karen M Puopolo1,
  3. Nellie I Hansen2,
  4. Scott A Lorch1,
  5. Sara B DeMauro1,
  6. Rachel G Greenberg3,4,
  7. C Michael Cotten3,
  8. Pablo J Sanchez5,
  9. Edward F Bell6,
  10. Eric C Eichenwald1,
  11. Barbara J Stoll7
  12. on behalf of the NICHD Neonatal Research Network
  1. 1 Pediatrics, Neonatology, The Children's Hospital of Philadelphia and Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA
  2. 2 Social, Statistical and Environmental Sciences Unit, RTI International, Research Triangle Park, North Carolina, USA
  3. 3 Pediatrics, Neonatology, Duke University, Durham, North Carolina, USA
  4. 4 Duke Clinical Research Institute, Duke University, Durham, North Carolina, USA
  5. 5 Pediatrics, Neonatology and Neonatology and Pediatric Infectious Diseases, Nationwide Children's Hospital, The Ohio State University College of Medicine, The Center for Perinatal Research, The Abigail Wexner Research Institute at Nationwide Children’s Hospital, Columbus, Ohio, USA
  6. 6 Pediatrics, Neonatology, University of Iowa, Iowa City, Iowa, USA
  7. 7 Pediatrics, Neonatology, McGovern Medical School at the University of Texas Health Science Center at Houston, Houston, Texas, USA
  1. Correspondence to Dr Sagori Mukhopadhyay, Newborn Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA; sagori.mukhopadhyay{at}


Objective Determine risk of death or neurodevelopmental impairment (NDI) in infants with late-onset sepsis (LOS) versus late-onset, antibiotic-treated, blood culture-negative conditions (LOCNC).

Design Retrospective cohort study.

Setting 24 neonatal centres.

Patients Infants born 1/1/2006–31/12/2014, at 22–26 weeks gestation, with birth weight 401–1000 g and surviving >7 days were included. Infants with early-onset sepsis, necrotising enterocolitis, intestinal perforation or both LOS and LOCNC were excluded.

Exposures LOS and LOCNC were defined as antibiotic administration for ≥5 days with and without a positive blood/cerebrospinal fluid culture, respectively. Infants with these diagnoses were also compared with infants with neither condition.

Outcomes Death or NDI was assessed at 18–26 months corrected age follow-up. Modified Poisson regression models were used to estimate relative risks adjusting for covariates occurring ≤7 days of age.

Results Of 7354 eligible infants, 3940 met inclusion criteria: 786 (20%) with LOS, 1601 (41%) with LOCNC and 1553 (39%) with neither. Infants with LOS had higher adjusted relative risk (95% CI) for death/NDI (1.14 (1.05 to 1.25)) and death before follow-up (1.71 (1.44 to 2.03)) than those with LOCNC. Among survivors, risk for NDI did not differ between the two groups (0.99 (0.86 to 1.13)) but was higher for LOCNC infants (1.17 (1.04 to 1.31)) compared with unaffected infants.

Conclusions Infants with LOS had higher risk of death, but not NDI, compared with infants with LOCNC. Surviving infants with LOCNC had higher risk of NDI compared with unaffected infants. Improving outcomes for infants with LOCNC requires study of the underlying conditions and the potential impact of antibiotic exposure.

  • neonatology
  • epidemiology
  • neurology

Data availability statement

Data are available on reasonable request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • Compared with unaffected infants, infants with late-onset sepsis and antibiotic-treated, blood culture-negative conditions have variably higher risks for death and neurodevelopmental impairment.

What this study adds?

  • Extremely preterm infants with late-onset sepsis had higher risk of death but similar risk of neurodevelopmental impairment, compared with infants with blood culture-negative conditions.


Preterm infants are diagnosed with late-onset sepsis (LOS), defined as positive blood or cerebrospinal fluid (CSF) culture obtained >72 hours of age, at rates varying from 10% to 30%.1 2 Infants with culture-confirmed infections are at higher risk for abnormal neuroimaging findings and neurodevelopmental impairment (NDI) compared with uninfected infants.3–10 In analyses of infants born 1993–2001 with birth weight (BW) 401–1000 g and admitted to hospitals in the Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network (NRN), infants administered antibiotics≥5 days in the absence of a positive blood culture also had a higher risk of NDI compared with uninfected infants.6

Although many studies refer to the concept of ‘culture-negative sepsis’, there is uncertainty regarding this entity, including whether in some cases, infection is truly present.11 LOS and late-onset, blood-culture negative conditions (LOCNCs) share clinical presentations and by definition, both are managed with antibiotics. While LOS is a culture-based diagnosis,1 8 9 12 LOCNC is variably defined as a combination of clinical signs suggesting sepsis with or without inflammatory markers, sterile blood cultures and the clinical team’s decision to administer variable durations of antibiotics.6 7 9 10 12 13 LOCNC is frequently considered as the equivalent of ‘missed’ LOS. Studies often describe outcomes for preterm infants with LOS or LOCNC separately and compared with uninfected infants as reference; few studies directly compare outcomes of infants with LOS and LOCNC.14 Such a comparison would reveal differences between the two conditions, measuring not whether these conditions have an effect on outcomes but whether the effects differ in magnitude or character between the two diagnoses.

Our objective was to compare risk of death or NDI among infants with LOS compared with infants with LOCNC. Infants with LOS or LOCNC were also compared with unaffected infants.



This is a retrospective cohort study of infants with gestational age (GA) 22 0/7–26 6/7 weeks, BW 401–1000 g, no major birth defect, born at NRN centres 1/1/2006–31/12/2014 and enrolled in the NRN registry of extremely preterm infants. The registry included clinical information prospectively collected during the birth hospitalisation of infants. Details of the registry are noted in the online supplemental appendix. Surviving infants were eligible for a follow-up developmental assessment. The institutional review board at each centre approved participation in the registry and the follow-up study, with waiver of consent or written parental consent as required by individual sites.

Study definitions

LOS was defined as isolation of a pathogen from blood or CSF obtained >72 hours of age and appropriate therapy for ≥5 days (≥7 days for CSF growth) or death before completed treatment. Cultures growing Bacillus, Micrococcus and Corynebacterium were considered contaminants and were excluded. Polymicrobial cultures were counted as LOS cases if at least one species was a pathogen. Cultures growing coagulase-negative staphylococci (CoNS) were counted as LOS cases unless an additional contaminant organism was also isolated. Six infants whose only positive culture grew Bacillus species were excluded. LOCNC was defined as antibiotics administered for ≥5 days or death before completed treatment, without a positive blood culture obtained >72 hours of age. Brain injury was defined as cranial imaging with ≥1 of the following: severe (≥Grade 3) intraventricular haemorrhage (IVH),15 periventricular leukomalacia, porencephalic cyst, ventriculomegaly or cerebellar haemorrhage.

Three exposure groups were identified: (1) LOS: infants with ≥1 episode of LOS and no episode of LOCNC; (2) LOCNC: infants with ≥1 episode of LOCNC and no LOS episode; (3) unaffected: infants without either LOS or LOCNC. We excluded infants with conditions whose management overlaps with LOS or LOCNC including infants with both conditions, culture-confirmed early-onset sepsis, necrotising enterocolitis16 and intestinal perforation.

Due to early mortality, many extremely preterm infants do not live long enough to suffer LOS/LOCNC.17 To decrease survival bias, we restricted analysis to infants surviving >7 days, which still allowed us to capture the majority of LOS cases (online supplemental figure 1). We also excluded survivors missing neurodevelopmental assessment.


The primary outcome was survival with NDI or death at >7 days age and before follow-up. Secondary outcomes were death and NDI assessed separately. Neurodevelopmental assessment: Surviving infants were assessed at 18–22 months (births before 7/1/2012) or 22–26 months (births on or after 7/1/2012) corrected age (CA). Neurodevelopmental outcomes assessed at <14 months or >30 months CA (2% of those assessed) were considered missing data. Assessment included a physical examination of the child and an interview with the primary caretaker to review clinical history. A neurological examination and a developmental evaluation using the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-3) were administered by certified examiners.18 Motor function was classified using the gross motor function classification system (GMFCS) and cerebral palsy (CP) was classified using the GMFCS scores: mild (level 1), moderate (level 2–3) or severe (level 4–5).19 Bilateral blindness was defined as corrected vision <20/200 in both eyes. Hearing impairment was defined as permanent hearing loss with or without amplification. NDI was defined as ≥1 of the following: bilateral blindness, hearing impairment, GMFCS level ≥2 with or without CP or a Bayley-3 cognitive composite score <85.

Statistical analysis

Statistical significance for unadjusted comparisons between infants was determined by χ2 test for categorical variables and student’s t test for continuous variables. Poisson regression models with robust variance estimators20 were used to assess risk of outcomes in the exposure groups while adjusting for the following pre-exposure covariates: maternal education, insurance, race/ethnicity, antenatal antibiotics, antenatal steroids, antepartum haemorrhage, infant GA, BW, sex, temperature at ≤60 min of birth, intubation at birth, maximum respiratory support ≤24 hours of age, enteral feeds started ≤3 days of birth, receipt of antibiotics for ≥5 days starting ≤72 hours of age, severe IVH diagnosed ≤7 days of birth and centre. Maternal hypertension, chorioamnionitis, delivery mode and membrane rupture were not associated with death/NDI in univariate comparisons and not included in the models. Categorical variables with missing values for ≥1% of infants were entered in models with a level indicating missing. Risk of NDI associated with multiple episodes was assessed in a separate model that categorised exposure group as 0, 1, 2 or 3+ LOS or LOCNC episodes. Adjusted relative risks, 95% CI and p values by the Wald χ2 test from these models were reported. Analyses were performed using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA).


Study population

After exclusions, 3940 infants were included (figure 1): 786 (20%) infants were diagnosed with LOS, 1601 (41%) with LOCNC and 1553 (39%) were unaffected. Across study sites, the prevalence of infants diagnosed with LOS ranged from 8% to 37% and with LOCNC ranged from 23% to 63% (online supplemental figure 2). All centres, except two, had a higher frequency of LOCNC diagnosis than LOS, and there was no relation between centre-specific rates of LOS and LOCNC. Overall, 579 (15%) infants in the study cohort died before follow-up and 3881 (87%) surviving infants were evaluated for NDI. Follow-up visits were completed between October 2007 and August 2017.

Figure 1

Derivation of the study cohort. BW, birth weight; GA, gestational age; LOCNC, late-onset blood culture-negative condition; LOS, late-onset sepsis; NDI, neurodevelopmental impairment; NEC, necrotising enterocolitis; SIP, spontaneous intestinal perforation.


LOS pathogens were isolated only from blood in 742 cases, only from CSF in 13 cases and from both in 31 infants (online supplemental table 1). The most common organisms isolated from blood were CoNS (52%), Staphylococcus aureus (13%) and Escherichia coli (7%). Among 44 cases of meningitis, CoNS (17/44, 39%) was the most common organism. The median age of LOS diagnosis was 14 days (IQR 9–22) for the first episode and 29 days (IQR 18–41) for the second episode (online supplemental figure 1). The age of onset was not recorded for LOCNC.

Clinical characteristics

Most maternal characteristics did not differ between the three groups, although several infant characteristics differed, mainly between infants with LOS or LOCNC and unaffected infants (table 1). Infant characteristics were not different between infants with LOS and LOCNC except infants with LOCNC were born at higher GA (p=0.03) and a greater proportion of them had received early (≤72 hours) antibiotics for ≥5 days without culture-confirmed infection (44% LOS vs 57% LOCNC, p<0.001).

Table 1

Maternal, delivery and neonatal characteristics

In-hospital morbidities were also higher among LOS and LOCNC infants vs unaffected infants (table 2). Infants with LOS had higher frequency of brain injury than infants with LOCNC (29% vs 24%, p=0.005) and lower survival at 36 weeks CA (77% LOS vs 91% LOCNC, p<0.001). Among infants surviving to 36 weeks CA, bronchopulmonary dysplasia was more prevalent in infants with LOCNC (66%) than infants with LOS (60%), p=0.01.

Table 2

In-hospital morbidities

Death and NDI

LOS versus LOCNC: The adjusted relative risk for the primary composite outcome of death or NDI was significantly higher for infants with LOS compared with those with LOCNC, as was the risk of death alone before follow-up (table 3). Among survivors assessed at follow-up, the proportion of infants with NDI was similar in the LOS (32%) and LOCNC (33%) groups and the infants had a comparable adjusted risk for NDI. Surviving infants with LOCNC had a higher risk for GMFCS level ≥2 than infants with LOS.

Table 3

Mortality and neurodevelopmental outcomes

LOS/LOCNC versus unaffected infants: Infants with either LOS or LOCNC were at significantly higher risk for death/NDI than unaffected infants (table 3). Infants with LOS had a greater risk of death than unaffected infants but infants with LOCNC did not. While the adjusted risk estimates for NDI among survivors with LOS and LOCNC were higher than that for unaffected infants, the risk was statistically significant only for LOCNC infants. Differences in risk for components of NDI were found between those with LOCNC and unaffected infants but not between infants with LOS and unaffected infants.

Multiple episodes

Among infants assessed at follow-up, more infants had repeated episodes of LOCNC (36%) than LOS (23%) (table 4). Risk of NDI was increased for infants who had ≥2 episodes of LOCNC compared with those with one episode. While the proportion of infants with NDI increased from 31% to 35% to 42% among infants who had 1, 2 or ≥3 LOS episodes the adjusted risk of multiple episodes was not significantly different compared with one episode.

Table 4

Neurodevelopmental impairment (NDI) by number of infection episodes


While infants with LOS and LOCNC did not differ in most clinical characteristics, those with LOS had higher risk of the combined outcome of death or NDI and of death alone before follow-up, when compared with either infants with LOCNC or to unaffected infants. Risk of death did not differ for infants with LOCNC and unaffected infants. The relationships between LOS, LOCNC and NDI among surviving infants were less straightforward. While brain injury was more common among infants with LOS than with LOCNC and in both groups approximately one-third of surviving infants had NDI compared with one-fourth of unaffected infants, the adjusted relative risk of NDI for infants with LOS compared with unaffected infants did not reach statistical significance. In contrast, the adjusted risk of NDI was greater for infants with LOCNC compared with those with no identified infection. The risk of NDI also increased significantly with repeated episodes of LOCNC, but not with LOS episodes. The smaller number of surviving infants with LOS (compared with LOCNC) may have contributed to these findings.

Few studies have specifically compared infants with LOS to infants with LOCNC.14 In doing so, we found a significantly increased risk of mortality among infants with LOS (table 3). We eliminated from the analysis 875 infants (11.9% of eligible) with both LOS and LOCNC, to ensure a clear comparison. We defined LOCNC by administration of ≥5 days of antibiotics in the absence of blood or CSF-confirmed infection. This diagnosis can be due to non-systemic bacterial infection, missed bacterial infection or to non-bacterial or non-infective conditions.21 22 Aetiologies such as urinary tract infection have a lower mortality risk than bacteraemia and may have contributed to our findings. Clinically, however, LOCNC diagnoses frequently reflect a concern for ‘false negative’ blood cultures.21 22 Inadequate inoculant volume is reported as one driver of false negative cultures.23–25 The lower mortality observed among LOCNC infants (compared with infants with LOS) would not support the concern that LOCNC cases were predominantly due to false negatives from inadequate inoculant. A second concern is that of ‘low-level’ bacteraemia.26 The vast majority of organisms between 1 and 10 CFU/mL can be detected reliably with 1 mL of blood.27 28 Whether the majority of LOCNC cases occur due to low-level bacteraemia and whether standard antibiotic regimens are effective or necessary in such cases requires further study.11 Finally, LOCNC may also simply reflect non-infectious causes of infant decompensation difficult to distinguish from bacterial sepsis using clinical judgement and laboratory markers—causes that would not respond to antibiotic therapy and may have a differential risk of mortality and morbidity.21 22 29

Bacteraemia/meningitis in premature infants is known to be associated with white matter abnormalities5 30 and NDI in early childhood6 12 31 that persists at school age.8–10 32 However, the association of LOCNC and NDI is less clear. A prior NRN study found higher odds of NDI among infants with culture-confirmed infection and with LOCNC compared with uninfected infants.6 Culture-confirmed infection in that study, however, included early-onset sepsis and did not address infants with exclusive LOS versus LOCNC. Two other studies of infants born at <28 weeks gestation, in contrast, reported no difference in developmental outcomes at 10 years of age for infants with presumed infection compared with uninfected infants.10 33 A similar lack of association has been reported by other investigators.9 12 Some of the differences between studies may be from different definitions of presumed infection and inclusion of localised infection in the definition of culture-confirmed infection.9 10 We found significantly increased risk for NDI with LOCNC that increased with multiple episodes (tables 3 and 4). The mechanisms for neurological injury attributed to LOCNC are likely related to its aetiology. When attributable to bacteraemia/meningitis not isolated in culture, or to localised bacterial infection, LOCNC-associated NDI may share pathogenesis with LOS. However, when viral infection or non-infectious causes of clinical decompensation lead to LOCNC diagnosis, injury may be attributed to a failure to provide appropriate therapies, as well as to dysbiosis from ineffective antibiotics.21 22 34–36

Our study is limited by the fact that dates of LOCNC episodes were not collected, and therefore we could not do an age-based comparison between the three groups. Specific information about the clinical conditions recorded as LOCNC, and clinical parameters associated with LOS and LOCNC that could define severity of illness were not collected. Thus, we cannot say with certainty whether LOCNC infants actually were infected with a bacterial/fungal pathogen the team failed to isolate, or if they were evaluated and treated with antibiotics for clinical instability that was non-infectious in origin, or due to a viral pathogen—and the reason for that instability led to neurodevelopmental consequences. Future studies that include data on type and severity of the clinical instability, for both LOS and LOCNC episodes, may be better able to distinguish specific outcome patterns to inform prevention and intervention strategies.


LOCNC is a commonly diagnosed condition, occurring twice as often in our study compared with LOS and with more frequent recurrences. Conservatively, this translated to ~2.8 times more antibiotic courses for LOCNC than for LOS. While less life-threatening than LOS, LOCNC was associated with worse neurological outcomes compared with unaffected infants. It remains unclear whether this injury is due to the aetiology of decompensation or due to management decisions, including use of antibiotics. We use the term ‘late-onset culture-negative condition’ rather than ‘late-onset culture-negative sepsis’ to underscore the uncertainty in this diagnosis and highlight the need for better diagnostic tools to evaluate sick newborns. Quality improvement efforts to reduce the incidence of LOS have successfully relied on interventions targeting LOS pathophysiology.1 37 Poor understanding of both the aetiology of LOCNC and of the impact of antibiotic exposures on brain development presents a barrier to improving outcomes.11 13 34 35 In the era of antibiotic stewardship, knowing the suspected adverse consequences of antibiotic misuse, a more accurate diagnosis would allow us to limit antibiotic usage to infants with clear need and devise targeted interventions for other aetiologies of LOCNC.

Data availability statement

Data are available on reasonable request.

Ethics statements


Supplementary materials


  • SM and KMP contributed equally.

  • Contributors SM conceptualised and designed the study, contributed to interpretation of results, drafted the initial manuscript and reviewed and revised the manuscript. KP conceptualised and designed the study, contributed to interpretation of results and reviewed and revised the manuscript. NIH carried out the statistical analyses, critically reviewed the manuscript and approved the final manuscript as submitted. SL, SBDM, RG, BJS, CMC, PJS, EFB and EE contributed to the study concept, reviewed and revised the manuscript, and approved the final manuscript as submitted.

  • Funding This work was conducted and supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) that provides grant support for the generic database and follow-up studies of the Neonatal Research Network. Please see online supplemental appendix of acknowledgments for full funding information. SM is supported by a grant from the NICHD (K23HD088753).

  • Competing interests RG has received support from industry for research services ( The other authors have no conflicts of interest relevant to this article to disclose.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.