Use of polymerase chain reaction as a diagnostic tool for neonatal sepsis can result in a decrease in use of antibiotics and total neonatal intensive care unit length of stay

Article metrics



To retrospectively determine if a negative 16S ribosomal RNA (rRNA) polymerase chain reaction (PCR) (PCR(−)) could lead to a decrease in the number of antibiotic doses and neonatal intensive care unit (NICU) length of stay (LOS) for infants admitted to the NICU for presumed early-onset sepsis (EOS) with negative blood culture results (BC(−)).

Study design:

Analysis included 419 infants, greater than 35 weeks gestational age, with PCR(−), BC(−) and LOS>48 h. Both the investigators and clinical care team were unaware of the PCR results. The actual number of antibiotic doses (AAD) administered was compared to an estimated number of antibiotics doses (EAD) that would have been given until PCR(−) results were available by 18 h. The number of antibiotic doses saved was calculated as (AAD−EAD). The actual NICU LOS in hours (aLOS) for a subset of infants who remained in the hospital primarily for antibiotic therapy was compared to an estimated LOS (eLOS) if infants with PCR(−) were discharged from the NICU when clinically stable. The number of hours saved was calculated as (aLOS−eLOS).


Approximately eight antibiotic doses and 85 NICU hours per infant could be saved using PCR(−) results available at 18 h.


Use of 16S rRNA PCR could decrease the number of antibiotics doses and NICU LOS for infants admitted for EOS. This may facilitate: (1) earlier NICU discharge; (2) parental satisfaction; and (3) decreased health care costs.


The timely diagnosis of early-onset sepsis (EOS) in the neonate remains a challenge to those caring for newborns. Of the term or near-term infants admitted to the neonatal intensive care unit (NICU) at Magee-Womens Hospital, 70% undergo evaluation and varying lengths of treatment for suspected or presumed infection. Yet culture-proven EOS remains a rare event. Many nonspecific clinical symptoms in the newborn can be the first presenting signs of neonatal infection. Unfortunately, neither these clinical signs1 nor currently available laboratory tests are reliable in the early detection of neonatal sepsis.2, 3, 4, 5 The blood culture has been the gold standard for the diagnosis of sepsis; however, technical pitfalls raise questions regarding blood culture reliability.6, 7 A justified concern regarding the ability to recover bacteria from the relatively small blood volumes collected for culture prompts clinicians to treat with antibiotics in spite of a negative blood culture (BC(−)). In infants with lower colony counts of bacteria, as many as 60% of culture results may be falsely negative with blood culture volumes of only 0.5 ml.8 In a post-mortem study, blood cultures were negative in 18% of infants who died with a bacterial infection.9 Furthermore, the use of intrapartum antibiotics for prevention of neonatal Group B streptococcal infection may reduce the likelihood of a positive culture in the newborn.

Whereas a positive culture confirms the diagnosis of sepsis, a negative culture does not impart the same degree of certainty and is not always utilized to determine the length of antibiotic treatment. Multiple paradigms and an array of novel laboratory tests have been studied to identify newborns with EOS. However, a recent meta-analysis has demonstrated that none of these diagnostic measures provide sufficient evidence to support their use as reliable clinical tools.5

Nucleic acid amplification tests such as polymerase chain reaction (PCR) are rapid to perform and have been used successfully to diagnose a wide range of infectious diseases, including bacteria, yeast, viral and protozoal infections.10 In a large study of 548 infants who were evaluated for infection, Jordan and Durso11 compared the results of 16S ribosomal RNA (16S rRNA) PCR with that from the automated Bactec 9240 blood culture. The rate of blood culture proven sepsis was 4.9% and there was high correlation of PCR with blood culture results with sensitivity, specificity, and positive and negative predictive values of 96, 99.4, 88.9 and 99.8%, respectively. A 16S rRNA PCR assay with high negative predictive accuracy may allow for a more timely termination of antibiotic treatment for newborn infants treated for suspected EOS.

We hypothesized that application of 16S rRNA PCR to screen blood for bacteria, with results within 18 h, could decrease the number of antibiotic doses and subsequent length of stay (LOS) for infants admitted to the NICU for presumed EOS.


This study is a single-site, non-interventional, non-randomized investigation. The Magee-Womens Hospital Institutional Review Board (IRB) approved this project to be conducted without prospective consent by the infant's parent/guardian because normally discarded clinical blood samples from infants admitted to the NICU were used for the 16S rRNA PCR analysis. Approval by the IRB also included retrospective chart review for clinical outcomes of all infants with both blood culture and PCR results. Infants born from 9 January 2000 to 4 January 2004 were eligible for inclusion. After 15 April 2003, outcome data were collected only for infants whose mother had signed the IRB-approved Magee-Womens Hospital Research Registry to access medical records for such retrospective reviews.

The bacterial DNA targeted for PCR amplification used the previously described RW01 and DG-74 primers to generate a 380 bp product from the 16S rRNA gene.12, 13 These PCR primers will detect all clinically relevant Gram-positive and Gram-negative bacteria. For this project, the retrieved blood sample was processed on day and evening shifts. Therefore, we assumed a conservative time estimate from blood draw to PCR result to be a maximum of 18 h. In the following analysis of antibiotic use and NICU LOS, 18 h is consistently used as the time interval for PCR testing.

Infants admitted to the NICU were included if they met the following criteria: born at greater than or equal to 36 completed weeks gestation; evaluated for EOS with both a complete blood count (CBC) and blood culture at less then 72 h of life; received antibiotics; remained in the NICU for more than 48 h; and had both BC(−) and negative PCR (PCR(−)) results. Infant blood samples were rejected for PCR analysis for specimen volumes of <100 μl or if samples were either grossly hemolyzed and/or clotted. Infants with radiographic evidence of congenital pneumonia or those with an NICU LOS of less than 48 h were excluded from this analysis. In our institution, infants are not transferred to the well-baby nursery with ongoing antibiotic therapy.

During the study period, a list of infants admitted to the NICU who met the gestational age inclusion criteria was transmitted three times per week to the clinical lab to allow for retrieval of discarded CBC blood samples in a timely manner. The 16S rRNA PCR was performed in the study lab by technicians masked to the infant's clinical condition. Neither the clinicians in the NICU nor the study investigators were aware of the results of the PCR testing.

After paired blood culture and PCR results were determined, the infant's information was entered into the study database. Chart review was performed by an NICU discharge nurse coordinator for the following data: maternal diagnosis of chorioamnionitis, maternal and infant antibiotic treatment, infant admission date/time, discharge date/time, admission diagnoses, date/time of resolution of admission clinical signs and symptoms, CBC, blood culture results and radiologist interpretation of chest radiograph.

Antibiotic doses were recorded as follows: the actual number of each antibiotic dose (AAD) received and the estimated number of antibiotic doses (EAD) that would have been given had PCR(−) results been available at 18 h. Using standard dosing intervals, the estimated number of doses given is dependent on the specific antimicrobial regimen. For example, infants who received ampicillin and gentamicin would have received three doses (two doses of ampicillin given every 12 h and one dose of gentamicin when given every 24 h); and infants who received ampicillin and cefotaxime would have received four doses of antibiotics (two doses each of ampicillin and cefotaxime) until PCR(−) results were available. The potential antibiotic doses saved were calculated as (AAD−EAD).

A subset of infants was evaluated to assess the effect of PCR(−) results on LOS. Infants were included in the LOS analysis only if they received more than three doses and up to 7 days of antibiotics and were discharged within 24 h of their last antibiotic dose. This cohort represents those infants whose LOS would be most affected by the length of antibiotic treatment. In the review of each infant's medical record, the discharge nurse coordinator recorded the date/time when the infant met each of the NICU-specific discharge criteria. These criteria are as follows; minimum of a 24-h period of stable metabolic and cardiovascular status, date/time of resolution of admission clinical signs and symptoms, adequate urine output (>1 ml/kg/h), full volume oral feedings (>100 ml/kg/day), breathing less than 60 breaths/min in room air (for 24 h). Using the latest of these dates, an estimated LOS (eLOS) was formulated. Actual LOS recorded as aLOS. The potential hours NICU LOS saved was calculated as (aLOS−eLOS).


The median number of antibiotic doses and NICU LOS saved per infant was analyzed using Wilcoxin rank sum test.


Antibiotic analysis

The demographic and clinical data for the 419 infants who met inclusion criteria with both PCR(−) and BC(−) results are shown in Table 1. Admission signs and symptoms for these infants are shown in Table 2. The majority of these infants exhibited some form of respiratory symptoms, including respiratory distress, cyanosis or apnea. Ampicillin and gentamicin were the antibiotic regimen given to 96% of the infants; all others received ampicillin and cefotaxime. Infants received an average of 11 antibiotic doses ((mean) 10.9±(s.e.) 0.3). We estimated that patients would receive only three doses before PCR results were available. Therefore, an estimated eight doses per infant ((mean) 7.9±(s.e.) 0.3, P=0.000) could have been avoided (a total of 3310 antibiotic doses) if PCR(−) results were available and utilized within 18 h of testing.

Table 1 Antibiotic analysis – maternal and infant characteristics (n=419 infants)
Table 2 Antibiotic analysis – admission signs and symptoms (n=419)

LOS analysis

A total of 163 infants met criteria for inclusion in the LOS analysis. The demographic and clinical data for these infants are shown in Table 3. Admission signs and symptoms for this subset of infants are shown in Table 4. Actual NICU LOS for infants in this study was approximately 85 h ((mean) 84.5±(s.e.) 3.4). Based on NICU-specific discharge criteria, we estimated a LOS of approximately 42 h ((mean) 41.7±(s.e.) 2), a savings of an average of 43 h ((mean) 42.8±(s.e.) 2.6, P=0.000) per baby. Therefore, for these 163 infants, a total of 6976 h of NICU stay was the estimated savings if PCR(−) results were utilized to discontinue antibiotics allowing for discharge from the NICU when the infant was clinically stable.

Table 3 LOS Analysis – maternal and infant characteristics (n=163)
Table 4 LOS Analysis – admission signs and symptoms (n=163)


Use of 16S rRNA PCR to screen blood for bacteria has the potential for rapid and accurate diagnosis of EOS. Furthermore, the high negative predictive value will allow the clinician to avoid the unnecessary use of antimicrobials.

At the present time, no test, used either alone or in combination, provides an ideal predictive value for diagnosis and treatment of newborn infection. In the absence of accurate laboratory evaluation to predict EOS, the decision of which infant to treat and for how long has relied solely on the judgment of the clinician. In the face of a BC(−) in a term or near-term infant, antibiotic treatment varies from as few as 2 days to greater than 10 days.14 In fact, the blood culture as the ‘gold standard’ may not be optimal in the newborn because of antibiotic exposure in utero and the small blood volume obtained for culture. It has been estimated that 20 to 40% of all infants with sepsis will have a negative blood culture.8, 9, 15 An inherent limitation of our study is that the PCR(−) results, for lack of a better measure, can only be compared to the BC(−) results, a standard that we acknowledge is flawed. Routine laboratory evaluation including neutrophil indices and C-reactive protein as potential markers of infection in newborn infants has shown limited value in the diagnosis of EOS sepsis.3, 16 Recent interest in the contribution of proinflammatory cytokines and acute phase reactants to diagnose newborn sepsis has also been disappointing.4, 5, 17 Available studies to date are inconsistent with regard to sample size, methodology and test availability, which limits the clinician's confidence in applying these tests to practice. Furthermore, these investigations have shown the difficulty in standardization of measures of inflammation across postnatal age and gestational age groups. Several reports have discussed the use of sepsis screening pathways with the use of C-reactive protein, CBC parameters, risk factors and clinical symptoms to rule out sepsis.14, 18, 19 To improve negative predictive accuracy in infants at-risk for infection, the sepsis screen was not applied until 12 to 24 h after birth.14, 19 These nonspecific screens may reflect a physiologic response that is not related to infection or the infant's inability to mount an inflammatory response. In contrast, the PCR assay will detect the etiologic infectious agent.

As a single test with high negative predictive accuracy, PCR could allow the clinician to discontinue antibiotics as soon as a PCR(−) result was available. In our study, the calculation of antibiotic doses saved was based on an estimation of antibiotic doses received until the PCR(−) results could be obtained. We excluded all infants with a radiographic diagnosis of pneumonia. However, approximately 50 of the 419 infants were treated with empiric antibiotics for a clinical diagnosis of pneumonia, despite a BC(−) and radiologist confirmation of normal chest radiograph. These infants were included in our analysis. However, if documentation of PCR(−) and radiology confirmation of the absence of pneumonia does not provide confirmatory evidence of the lack of infection, then our calculation of antibiotic doses saved is an overestimation by this amount.

Some clinicians may feel that a few additional doses of antibiotics pose minimal risk to the baby. However, there is a risk of acquiring yeast- or drug-resistant bacteria when unnecessary doses of antibiotics are used. In addition, administration of any drugs, particularly those given via intravenous route, poses a risk of medication error. Any attempt to decrease the number and frequency of medications may result in a decrease in errors.

Use of PCR(−) results to shorten the course of antibiotics and decrease of NICU LOS could lead to a reduction in health-care costs. Related economic costs include the pharmacist preparation time and nurse administration of intravenous antibiotics. In addition, there are both physical and social costs including pain and suffering, parental anxiety and separation of the mother/baby dyad.

We do not, at this juncture, suggest that the use of PCR will replace the clinician's judgment in treatment decisions regarding a critically ill term infant at 18 h of age, but rather add additional data on which to base those decisions. Given that small blood sample volumes and maternal pretreatment render culture results suspect, the additive value of having both a PCR(−) and a BC(−) would be the ideal. In that situation, a PCR(−) should prompt further investigation into an alternate cause of the infant's unstable clinical condition. During this study, PCR results were not made available to the clinical team caring for the infant.

Future studies will incorporate real-time PCR to determine the influence of timely PCR results on clinical decision making to either discontinue therapy or select the most appropriate antibiotic.


The benefit of PCR is its rapid availability of results with a high negative predictive value. As a tool to ‘rule out sepsis’, PCR can be easily incorporated into the hospital setting for term or near-term population of infants admitted to the NICU for sepsis evaluation.

In this observational study of 419 term and near-term infants, the use of PCR could result in a decrease in the use of antibiotics and a subsequent decrease in NICU LOS. We speculate that this decrease in NICU LOS could facilitate infant–parental attachments and successful establishment of breast-feeding as well as reduce the hospital costs associated with prolonged treatment.


  1. 1

    Escobar GJ . The neonatal ‘sepsis work-up’: personal reflections on the development of an evidence-based approach toward newborn infections in a managed care organization. Pediatrics 1999; 103 (Suppl E): 360–373.

  2. 2

    DaSilva O, Ohlsson A, Kenyon C . Accuracy of leukocyte indices and C-reactive protein for diagnosis of neonatal sepsis: a critical review. Pediatr Infect Dis J 1995; 14: 362.

  3. 3

    Fowlie PW, Schmidt B . Diagnostic tests for bacterial infection from birth to 90 days: a systematic review. Arch Dis Child Fetal Neonatal Ed 1998; 78: F92–F98.

  4. 4

    Mehr S, Doyle LW . Cytokines as markers of bacterial sepsis in newborn infants: a review. Pediatr Infect Dis J 2000; 19: 879–887.

  5. 5

    Malik A, Hui CPS, Pennie RA, Kirpalani H . Beyond the complete blood cell count & c-reactive protein: a systematic review of modern diagnostic tests for neonatal sepsis. Arch Pediatr Adolesc Med 2003; 157: 511–516.

  6. 6

    Aronson MD, Bor DH . Blood cultures. Ann Intern Med 1987; 106: 246–253.

  7. 7

    Kellogg JA, Ferrentino FL, Goodstein MH, Liss J, Shapiro SL, Bankert DA . Frequency of low level bacteremia in infants from birth to two months of age. Pediatr Infect Dis J 1997; 16: 381–385.

  8. 8

    Schelonka RL, Chai MK, Yoder BA, Hensley D, Brockett RM, Ascher DP . Volume of blood required to detect common neonatal pathogens. J Pediatr 1996; 129: 275–278.

  9. 9

    Squire E, Favara B, Todd J . Diagnosis of neonatal bacterial infection: hematologic and pathologic findings in fatal and nonfatal cases. Pediatrics 1979; 64: 60–64.

  10. 10

    Ehrlich G . PCR-based protocols for the detection of specific organisms. In: Ehrlich GD, Greenberg SJ (eds). PCR-based Diagnostics in Infectious Disease. Blackwell Scientific Publications Inc: Cambridge, MA, 1994, pp 335–688.

  11. 11

    Jordan JA, Durso MB . Comparison of a 16S rRNA PCR assay and an automated blood culture system in the detection of neonatal bacteremia. JCM 2000; 38: 2574–2578.

  12. 12

    Greisen K, Loeffelholz M, Purohit A, Leong D . PCR primers and probes for the 16S rRNA gene of most species of pathogenic bacteria, including bacteria found in cerebrospinal fluid. J Clin Microbiol 1994; 32: 335–351.

  13. 13

    Jordan JA, Durso MB . Comparison of 16S rRNA gene PCR and BACTEC 9240 for detection of neonatal bacteremia. J Clin Microbiol 2000; 38: 2574–2578.

  14. 14

    Gerdes JS . Diagnosis and management of bacterial infections in the neonate. Pediatr Clin N Am 2004; 51: 939–959.

  15. 15

    Pierce JR, Merenstein GB, Stocker JD . Immediate postmortem cultures in an intensive care nursery. Pediatr Infect Dis J 1984; 3: 510–513.

  16. 16

    Schelonka RL, Yoder BA, Bradley A . The WBC count and differential: its uses and misuses. Contemp Obstet Gynecol 1996; 13: 124–166.

  17. 17

    Santana Reyes C, Garcia-Munoz F, Reyes D, Gonzalez G, Dominguez C, Domenech E . Role of cytokines (interleukin-1β, 6, 8, tumour necrosis factor-α, and soluble receptor on interleukin-2) and C-reactive protein in the diagnosis of neonatal sepsis. Acta Pediatr 2003; 92: 221–227.

  18. 18

    Philip AGS, Mills PC . Use of C-reactive protein in minimizing antibiotic exposure experience with infants initially admitted to a well-baby nursery. Pediatrics 2000; 106 (1): 4–8.

  19. 19

    Madan A, Adams MM, Philip AGS . Frequency and timing of symptoms in infants screened for sepsis: effectiveness of a sepsis-screening pathway. Clin Pediatr 2003; 42 (1): 11–18.

Download references


This work was supported by the NIH Grant # HD38559. The authors gratefully acknowledge assistance in data collection by Mary Kish, RNC and database management by AI Johnson, MHA.

Author information

Correspondence to B S Brozanski.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Brozanski, B., Jones, J., Krohn, M. et al. Use of polymerase chain reaction as a diagnostic tool for neonatal sepsis can result in a decrease in use of antibiotics and total neonatal intensive care unit length of stay. J Perinatol 26, 688–692 (2006) doi:10.1038/

Download citation


  • PCR
  • neonatal sepsis
  • length of stay

Further reading