Preterm and Term Infants Evaluated for Sepsis: Differences in Management and Clinical Outcomes Free

Our freestanding, academic, children’s hospital has 673 beds and is affiliated with the state university college of medicine. We have 2 associated emergency departments and 7 urgent care centers. There were >1.6 million combined inpatient and outpatient visits in 2019.

We conducted a single-center, retrospective analysis of the electronic medical records of infants aged 0 to 60 days presenting to our hospital from January 2014 to January 2019 for temperature dysregulation, as defined by International Classification of Diseases (ICD), Ninth and 10th Editions, group codes for fever and hypothermia (778.4, 780.6, 995.85, P81.9, R50.9, 778.3, 780.65, 991.6, P80.9, R68.0, T68XXXA). This list was then confirmed with our local data analytics team for completeness. We performed additional electronic medical record queries looking for urinary tract infection, bacteremia, meningitis, pneumonia, skin soft tissue infection, and viral infections. Subjects with reported fever (≥38°C) or hypothermia (≤36.5°C) at home and 1 of the above ICD codes were included. Subjects were included irrespective of temperature documented at the time of presentation to a health care setting given that term normothermic infants with history of fever at home have been shown to be at risk for invasive bacterial infection and SBI.¹⁷  Initial presenting temperature was also recorded for analysis, with fever defined as temperature ≥38.0°C and hypothermia defined as temperature ≤36.5°C. We elected to include both febrile and hypothermic infants, because hypothermia is an independent risk factor for sepsis in the preterm population in particular.^15,16  The study was approved by the institutional review board.

From the cohort that met inclusion criteria, we defined 2 groups for comparison: preterm infants (32–36 6/7 weeks) and term infants (37–42 weeks). Given that the initial group of term infants was significantly larger than the group of preterm infants, we then selected a sample of term infants approximately double the number of preterm infants using randomization software. Infants were excluded if:

1. they had received antibiotics in the 48 hours before presentation;

2. they had any preexisting medical conditions apart from prematurity; or

3. they had indwelling devices.

Initial home births, admissions to the newborn nursery, or initial NICU admissions were excluded. The primary record of interest was limited to a readmission separate from these birthing events. Records of patients meeting inclusion criteria were manually reviewed by at least 2 investigators.

We collected information on the diagnostic evaluation for the index visit for each infant, including temperature at presentation, initial laboratory, and imaging ordered. If an infant presented for >1 visit during the study period, the subsequent visits were not included in evaluation. Testing for common respiratory pathogens was performed by multiplex polymerase chain reaction (PCR) (BioFire FilmArray) with nasopharyngeal swabs. Testing for enterovirus, parechovirus, and herpes simplex virus from blood and superficial sites was performed by laboratory-developed and validated PCR assays.^18,19  Testing for common pathogens of meningitis was performed on cerebrospinal fluid (CSF) by multiplex PCR (BioFire FilmArray).^18,20 

We recorded results of laboratory tests and rates of viral and bacterial infections. Elevated inflammatory markers were defined as follows: bands ≥10% of total white blood cells, C-reactive protein (CRP) ≥1.2 mg/dL, procalcitonin ≥0.5 ng/mL (CRP and procalcitonin based on local laboratory definitions of abnormal). Urine culture was considered positive if a catheterized urine specimen yielded a potential uropathogen in quantities of either:

1. 10 000 to 49 000 colony-forming units/mL with pyuria or bacteriuria on urinalysis; or

2. ≥50 000 colony-forming units/mL regardless of urinalysis results.^21–23 

Bacteremia was defined by isolation of a pathogenic organism from the blood, and bacterial meningitis was defined by CSF culture or identification of a bacterium in CSF by multiplex PCR. Cultures were considered to be contaminants if they contained an organism frequently known to cause contamination,²⁴  and were treated as such and documented by the medical team.

We additionally collected information on the rates of hospitalization, location of admission (general floor versus ICU, including transfer to the PICU or NICU setting, Supplemental Table 4), and use and duration of IV antibiotics and antiviral therapy. We compared the use of vasopressors, radiant warmers, respiratory support, and mortality among those infants who were admitted to an ICU. We also evaluated the 7-day and 30-day unplanned sick visits to a primary care office or return visits to an acute health care facility (urgent care or the emergency department) and readmission to an inpatient unit.

Continuous variables are presented as median and interquartile range (IQR), and categorical variables are presented as frequencies and percentages. We compared the diagnostic evaluation, therapeutic management, and clinical outcomes between preterm and term infants. Additional analyses were performed between moderate (≥32, <34 weeks) and late (≥34, <36) preterm infants. Unadjusted associations between gestational age and categorical variables were assessed using Pearson’s χ² test and Wilcoxon’s rank-sum test for continuous variables. Logistic regression models were used to estimate the relationship between selected outcomes (SBI, viral infection, admission to hospital, and ICU transfer) and predictor variables gestational age, chronological age, and presenting temperature. Similarly, a linear regression model was used to estimate the association between length of stay and the interaction of predictor variables gestational age, chronological age, and presenting temperature. Length of stay was log-transformed to improve model diagnostics, and estimates were back-transformed onto the original scale. For each regression model, estimated marginal effects were plotted and contrasts between preterm and term infants were summarized with corresponding 95% confidence intervals. Except where indicated, all P values presented are unadjusted for multiple comparisons, and complete case analysis was used for instances of missing data. Statistical significance was assessed at α = .05. Analyses were performed using the R system for statistical computing (R core team, 2022), version 4.2.1.

We identified 363 preterm infants who presented with temperature dysregulation; 336 infants met inclusion criteria. Of 2331 term infants who presented with temperature dysregulation, we randomly selected a sample of 600, of whom 554 met criteria for inclusion (Fig 1). Forty-nine percent of the preterm group were 0 to 28 days of age versus 46% of the term group. The median chronological age of preterm infants was similar to term infants (30 days versus 31 days). Of the 336 preterm infants, 31% presented with hypothermia, compared with 6.5% of the 554 term infants (P < .001) (Table 1). There were no statistically significant differences observed between the diagnostic approaches, rates of infection, and management between moderate (≥32, <34 weeks) and late (≥34, <36) preterm infants on the basis of results of Pearson’s χ² test and Wilcoxon’s rank-sum test, and thus the groups were examined in aggregate.

Diagnostic evaluations differed between preterm and term infants in several tests performed (Table 1). Compared with term infants, preterm infants had a higher frequency of CRP testing (31% vs 25%, P = .034), but a similar frequency of procalcitonin testing (26% vs 23%, P = .2). Preterm infants had chest x-rays performed more frequently than term infants (50% vs 32%, P < .001). There were no differences between preterm and term infants in rates of lumbar punctures (55% vs 50%, P = .2) or blood cultures (67% vs 72%, P = .2).

We also examined rates of diagnostic testing of preterm infants compared with term infants when stratified by chronological age, 0 to 28 days and 29 to 60 days. Term infants in both the 0 to 28 days and 29 to 60 days groups were less likely to receive x-rays compared with the preterm counterparts. No further statistical differences in invasive testing, such as blood cultures and lumbar punctures, were detected between these 2 age groups (Supplemental Table 5).

When comparing/evaluating laboratory values and clinical outcomes between preterm and term infants, we found that overall preterm infants had lower white blood cell counts (9.0 X10³/μL versus 9.9 X10³/μL, P = .002) and absolute neutrophil counts (2464 mm³ vs 3038 mm³, P < .001) (Table 1).

Although preterm infants had similar rates of any bacterial infection compared with term infants (7.1% vs 6.3%, P = .6), they had more than double the rate of bacteremia (5.9% vs 2.5%, P = .035). Causative organisms were primarily group B Streptococcus (n = 4) and E. coli (n = 3). Preterm infants had lower rates of viral infections compared with term infants (33% vs 42%, P = .015), particularly respiratory viral infections (26% vs 33%, P = .001).

Preterm infants were hospitalized more frequently than term infants (72% vs 63%, P = .006) (Table 2). Preterm infants had a longer duration of treatment with intravenous antibiotics compared with term infants (38 hours versus 33 hours, P < .001). Preterm infants also received acyclovir more frequently (51% vs 43%, P = .006.)

Preterm infants required ICU level of care more frequently than term infants (32% vs 5%, P <.001). For preterm infants, 40% of ICU stays were in the NICU, whereas no term infants were admitted to a NICU. Preterm infants in the ICU needed radiant warmers more often than term infants (33% vs 12%, P = .032) (Table 3). Despite the need for a higher level of care, preterm infants did not have increased rates of intubation (P = .6), vasopressor support (P = .5), or mortality (P > .9). Preterm infants admitted to the ICU required noninvasive positive pressure about half as often as term infants (21% vs 44%, P = .015).

Preterm infants were also hospitalized longer than their term counterparts (49 hours versus 39 hours, P < .001), though there was increased variability in length of stay (IQR 6–111 hours versus IQR 4–51 hours) (Table 2). Preterm infants did not have statistically significant increased return visits >or readmission rates.

We repeated the analyses for the subgroup of 231 preterm and 518 term infants with fever, excluding those with hypothermia. The results of this analysis were similar to the full cohort analysis with a few exceptions. Febrile preterm infants were nearly twice as likely to have bacteremia as term infants, though this difference was not statistically significant (2.7% vs 5%, P = .2) as compared with when both febrile and hypothermic infants were examined in aggregate. (Supplemental Table 6). Preterm infants continued to require ICU admissions more frequently than term infants (13% vs 3%, P < .001), although overall hospital admission rates for both groups decreased (Supplemental Table 7).

Next, we examined the subgroup-specific effects for gestational age, chronological age, and presenting temperature with selected outcomes. Point estimates and 95% confidence intervals are provided in Fig 2. Infants presenting with fever had the highest probability of serious bacterial infection and viral infection; however, preterm infants aged 29 to 60 days presenting with hypothermia (n = 23) had similar probabilities of bacterial and viral infections as the fever group. Furthermore, preterm infants with hypothermia were more likely than term infants to require ICU level care and had longer lengths of stay. Expected length of stay was 2.3 times greater for 0 to 28-day-old preterm infants presenting with hypothermia compared with the corresponding term infants (P = .069).

In the current study, we found that preterm infants aged 0 to 60 days presenting with temperature dysregulation underwent more extensive diagnostic workup than term infants of the same age. Preterm infants had higher rates of bacteremia and more often presented with hypothermic illness. In concordance with these findings, preterm infants were hospitalized and required ICU level of care more frequently and had longer hospitalizations, even when adjusting for chronological age and presenting temperature. Importantly, these infants did not have increased rates of mortality or readmission. We found that hypothermia was an important risk factor for severity of illness in both preterm and term infants.

The recent American Academy of Pediatrics febrile infant guidelines discuss the value and use of inflammatory markers in guiding evaluation of infants at risk for sepsis.⁷  In our study, we found that preterm infants underwent inflammatory marker testing more frequently than term infants. Given the lack of guidance and uncertainty with managing preterm infants at risk for sepsis, clinicians may use these tests to risk stratify. Similarly, despite evidence showing minimal diagnostic utility of chest x-rays for the evaluation of infants at risk for sepsis,^25–27  we found that preterm infants were more likely to receive imaging studies. It is possible that preterm infants were more likely to undergo radiographic testing given increased severity of illness at time of presentation, with more NICU stays/ICU level of care required despite lower rates of viral infections and less frequent need for noninvasive positive pressure ventilation.

Importantly, we found that rates of blood culture testing were similar between preterm and term infants. It may be prudent to evaluate preterm infants with blood cultures more frequently given that preterm infants had higher rates of serious bacterial infections than term infants, and with more than twice the rate of bacteremia. Almost half of preterm infants with bacteremia were hypothermic at time of presentation, reflecting that preterm infants with serious bacterial infection may present differently than term infants, and caution should be taken when approaching this population. Additionally, we found that preterm infants had lower absolute neutrophil counts, making it harder to extrapolate these results to the newest febrile infant guidelines that use absolute neutrophil count cutoffs of >4000 mm³.

Preterm infants were not only admitted/hospitalized more frequently, but also required ICU level of care more frequently than their term counterparts, particularly those with hypothermia. This finding may have several potential explanations. First, given increased concerns for sepsis in prematurity,²⁸  it is possible that providers were more likely to admit for concerns of potential deterioration. Second, given the frequency of hypothermia in this population, it may be that premature infants required ICU level of care for rewarming alone, which may or may not have been related to an infectious etiology.¹⁶  Finally, we found that a greater number of these premature infants were admitted to a NICU rather than a PICU, which may be related to the medical equipment available in the NICU, such as an incubator, or potentially related institutional staffing and admitting policies in times of high patient census.

These findings are consistent with previous studies that have established that preterm infants have significantly more medical comorbidities and costly hospitalizations than term infants in the immediate postnatal period outside of the NICU.^28,29  Our study suggests that these health outcome discrepancies may continue outside the initial birthing hospitalization and well into the second month of life, because preterm infants continue to be hospitalized more frequently and have longer lengths of stay. However, despite these differences, discharge outcomes for these infants did not seem to be affected, because the rates of return visits and readmissions for preterm infants were similar to those of term infants in our study.

Our study had several limitations. First, it is a single-center review and it may lack generalizability. Although our findings may not capture the potential variability at other institutions, to our knowledge, our study is 1 of the largest manual reviews comparing the evaluation and management of preterm and term infants aged <60 days with temperature dysregulation who are evaluated for sepsis. Second, some of the microbiology assays available at our institution may not be universally available, such as blood and mucosal viral PCR testing (including herpes simplex virus, enterovirus, and parechovirus) allowing clinicians to readily incorporate results into risk-stratifying models. Additionally, other facilities may not have the ability to care for infants in a level IV NICU, and therefore disposition and management of this population may be different. Furthermore, although we did additional queries of the chart for meningitis, bacteremia, urinary tract infection, skin soft tissue infections, and pneumonia, and these were verified by our local data analytics team, we may have potentially missed some infants with SBI and invasive bacterial infection by using ICD codes for patient identification.

However, despite these limitations, our study cohort was large and may help inform the design of future prospective studies, because national guidelines for the evaluation and management of sepsis in preterm infants outside the NICU do not exist.

The results of our study suggest that previously healthy preterm infants are a population clinically distinct from term infants. At our institution, preterm infants underwent more extensive diagnostic evaluations, and showed higher rates of bacteremia and longer hospitalizations compared with term infants. Further, those presenting with hypothermia, which is likely multifactorial and associated with concomitant issues of prematurity, required higher acuity care more often. Larger prospective studies are urgently needed to determine the best strategies to evaluate and manage young, healthy preterm infants at risk for sepsis.

We thank The Ohio State University Center for Clinical and Translational Science for their collaboration and Beth Burkhart from Nationwide Children’s Research Institute for her contributions.