A Quantitative, Risk-Based Approach to the Management of Neonatal Early-Onset Sepsis

MANUSCRIPT CITATION

Kuzniewicz MW, Puopolo KM, Fischer A, Walsh EM, Li S, Newman TB, et al. A Quantitative, Risk-Based Approach to the Management of Neonatal Early-Onset Sepsis. JAMA Pediatr. 2017; 171(4):365-371. doi: 10.1001/jamapediatrics.2016.4678. PMID: 28241253

REVIEWED BY

Naomi Spotswood
Consultant Neonatologist
Department of Paediatrics, Royal Hobart Hospital, Hobart, Tasmania; Burnet Institute, Melbourne, Victoria; Department of Medicine, University of Melbourne, Melbourne, Victoria

Erin Grace
Neonatal Registrar
Department of Neonatal Medicine, Women’s and Children’s Hospital, North Adelaide, South Australia; Healthy Mothers, Babies and Children, the South Australian Health and Medical Research Institute, North Adelaide, South Australia

Amy K Keir
Consultant Neonatologist
Department of Neonatal Medicine, Women’s and Children’s Hospital, North Adelaide, South Australia; Healthy Mothers, Babies and Children, the South Australian Health and Medical Research Institute, North Adelaide, South Australia;Robinson Research Institute and the Adelaide Medical School, University of Adelaide

TYPE OF INVESTIGATION

Clinical prediction guides

QUESTION

In infants born at 35 weeks’ gestation or greater (Population), what is the effect of an early-onset sepsis (EOS) risk calculator (Intervention) compared to prior practice (Comparison) on the frequency of antibiotic use and evaluations for possible EOS (Outcome)?

METHODS

  • Design: Observational study with an interrupted time series analysis of the effect of implementation of a multivariable risk prediction model.
  • Allocation: Not applicable.
  • Blinding: Not applicable.
  • Study period: January 1 2010 to December 31 2015.
  • Setting: 14 hospitals within a North American integrated health care system.
  • Patients: 204 485 neonates ³35 weeks’ gestation born within the 14 study hospitals.
  • Intervention: A risk prediction model supplied as an online calculator interface which provided clinicians with guidance regarding the risk of EOS and clinical action required. The prediction model was developed using a Bayesian approach in a previous analysis by the study group. Data in a baseline period were collected first, then implementation of the online calculator occurred in two sequential periods:
    • The baseline period: from January 1 2010 through November 31 2012 clinical care was informed by the Centers for Disease Control and Prevention guidelines for Group B Streptococcal disease prevention.
    • The learning/familiarisation period: from December 1 2012 through June 30 2014, EOS risk was evaluated using a calculator based on gestational age plus maternal data only. Maternal data included highest maternal antepartum temperature, group B streptococcus (GBS) carriage status, duration of rupture of membranes and nature/timing of intrapartum antibiotics. This calculator provided clinicians with the probability of EOS for each case.
    • The EOS calculator period: From July 1 2014 through December 31 2015, EOS risk evaluation was based on all of the variables considered in the learning period but also incorporated patient clinical presentation. Clinical presentation was defined as either well, equivocal or clinically ill (Table 1). This calculator provided clinicians with the probability of EOS, and also gave the following clinical guidance:
      • Blood culture collection was recommended if EOS risk was 1 or greater per 1000 live births.
      • Commencement of empiric antibiotics was recommended if EOS risk was 3 or greater per 1000 live births
    • Outcomes
      • Primary outcome: Administration of antibiotics in the first 24 hours
      • Secondary outcomes: Blood culture collection in the first 24 hours, antibiotic administration between 24 and 72 hours, and number of days of antibiotic use per 100 live births.
    • Analysis and Sample Size: 204 485 neonates were included in the study: 95 343 in the baseline period, 52 881 in the learning period and 56 261 in the EOS calculator period. Infant and maternal characteristics in each period were compared using the χ2, Fisher exact, and analysis of variance tests. The effect of the intervention on the primary and secondary outcomes was estimated using an interrupted time series design with segmented regression models. A broad array of potential confounders including time-associated variations and population characteristics were evaluated and included in the model if they showed significance of <0.05 (2-sided p). Readmission for culture positive sepsis/meningitis, organism in blood culture, symptoms and infant outcomes were compared using the χ2 or Fisher exact tests.
    • Patient follow-up (% included in the analysis): 100%

Table 1: Categories of clinical presentation

Clinically Ill Equivocal Well appearing
 

1.     Persistent need for respiratory support outside of the delivery room: nasal CPAP/high flow nasal cannula/mechanical ventilation.

2.     Need at ³2 hours of age for supplemental oxygen (outside of the delivery room) to maintain oxygen saturations ³90%.

3.     Haemodynamic instability requiring vasoactive drugs.

4.     Neonatal encephalopathy /Perinatal depression as indicated by: seizures and/or 5 minute Apgar score of <5.

 

1.     Single persistently abnormal observation at ³4 hours or ³2 abnormal observations lasting ³2 hours:

•      Heart rate ³160

•      Respiratory rate ³60

•      Respiratory distress

•      Temperature ³38˚C or <36.4˚C

 

 

No persistent abnormal observations

MAIN RESULTS

While statistically significant differences between study periods were present for birthweight, gestational age and SGA infants, these were small and not clinically meaningful. Similarly, there were statistically significant but small differences in the distribution of different ethnicities between the study periods. In the EOS calculator period compared to the learning and baseline periods, positive maternal GBS status was less common (22.0%, 23.4% and 22.5% respectively) and prolonged rupture of membrane time occurred more frequently (17.1%, 16.4% and 15.8% respectively). The incidence of culture confirmed EOS showed no statistical difference between study periods: 0.03% in the baseline period, 0.03% in the learning period and 0.02% in the EOS calculator period.

Compared to the baseline period, the implementation of the EOS calculator (EOS calculator period) saw reductions in the rates of sepsis evaluations, antibiotic use in the first 24 hours, and antibiotic use in days per 100 infants. The use of antibiotics between 24 and 72 hours did not increase. Results are detailed in Table 2. Of EOS cases in the baseline and EOS calculator periods (24 and 12 infants respectively), there were no statistical differences in pathogen types, readmission rates or illness severity. Individual case analyses of infants with EOS in the calculator period show that 11 of the 12 cases declared symptoms which prompted evaluation and treatment. Two infants in the EOS calculator period who would have received additional evaluation or earlier treatment under CDC guidelines were treated once a positive blood culture was identified: both remained well and had sterile additional pre-antibiotic cultures, suggesting transient bacteraemia had occurred.

Table 2: Antibiotic administration and Blood Culture Collection between Baseline and EOS calculator periods

Baseline Period,

%

EOS calculator Period,

%

Unadjusted Absolute difference, % (95%CI) Adjusted Absolute difference, % (95% CI)
Antibiotic administration in first 24 hours (%) 5 2.6 -2.3 (-2.1 to -2.5) -1.8 (-2.4 to -1.3)
Blood culture collection in first 24 hours (%) 14.5 4.9 -9.6 (-9.3 to -9.9) -7.7 (-13.1 to -2.4)
Antibiotic administration >24 hours, <72 hours (%) 0.5 0.4 -0.1 (-0.05 to 0.2) 0.05 (-0.1 to 0.2)
Antibiotic days/100 live births (days)* 16 8.5 -7.6 (-6.7 to -8.5) -3.3 (-6.1 to -0.5)

*Antibiotic days = number of calendar days the infant received at least one dose of intravenous antibiotics

CONCLUSION

The use of the early-onset sepsis risk calculator reduced the proportion of newborns ³35 weeks’ gestation undergoing laboratory testing and receiving empirical antibiotic treatment, without missed cases or morbidity due to delayed therapy.

COMMENTARY

Antibiotics are life-saving for babies who have early-onset sepsis (EOS), and a false negative assessment for this disease can be lethal. However, the risk evaluation process is challenging: early signs are subtle, and frequently occur with other pathologies. For this reason, a careful and conservative approach is employed in clinical guidelines (1) and by clinicians charged with the decision of whether to prescribe antimicrobials to a newborn who has a small, but quantifiable, risk of infection.

As is the case for any intervention, antibiotics are not without associated risk. They are generally prescribed to neonates intravenously, and involve risks inherent to invasive vascular devices and parenteral medications (2). The impact of antibiotic exposure to the developing microbiome is becoming better understood, including its association with allergic pathologies (3) and metabolic diseases including obesity (4). Further, the contribution of excess antibiotic use to the global escalation of antibiotic resistance cannot be ignored (5). Thus, while true cases of EOS must be identified and treated, there is a strong case for reducing potentially avoidable antibiotic prescriptions to newborns.

Kuzniewicz et al.’s demonstration that an EOS calculator can effect safe reduction in investigations for infection and antimicrobial use is a positive step forward. The study’s large cohort and careful analysis conducted for missed cases or delayed therapy lends confidence to the EOS calculator approach. Their data confirm that contemporary rates of EOS are lower than those observed prior to the Group B Streptococcus screening era (6), underscoring the rationale of reappraising currently high antimicrobial use for suspected EOS.

There are several aspects to this study which should be interpreted in context. First, population prevalence of antibiotic use varies between neonatal health services. In this study, pre-intervention antimicrobial use was 5% and post intervention 2.6%. Recently published experience in an Australian neonatal unit showed much higher baseline antimicrobial use: 12%, reduced to 7.6% with implementation of the EOS calculator (7). By comparison, Fjalstad et al. estimated antimicrobial exposure amongst term Norwegian term infants antibiotics at 2.3%, lower than what has been published with EOS calculator use, despite a higher baseline incidence of EOS in the Norwegian study in comparison to the Australian study and Kuzniewiez et al.’s cohort (0.54/1000 versus 0.44/1000 and 0.3/1000 live births respectively) (7,8). Assessment of the EOS calculator’s efficacy in demographically similar regions where antimicrobial use is already <2.5% has not been performed, and it remains to be seen whether it might play a role where lower rates are already achieved. Further, while the EOS calculator can account for differing baseline disease incidence, its use will need careful appraisal for safety in regions with different pathogen distributions to the study’s population. This could be particularly pertinent for health services in low and middle income regions where gram negative infections are a more frequent cause of EOS (9).

In summary, this study presents a practical, feasible and highly relevant intervention for neonatal health services to safely reduce currently high rates of neonatal antimicrobial exposure.

REFERENCES

  1. Verani JR, McGee L, Schrag SJ. Prevention of Perinatal Group B Streptococcal Disease: Revised Guidelines from CDC, 2010. MMWR Recomm Rep. 2010;59(RR-10):1-36.
  2. Ben Abdelaziz R, Hafsi H, Hajji H, Boudabous H, Ben Chehida A, Mrabet A, et al. Peripheral venous catheter complications in children: predisposing factors in a multicenter prospective cohort study. BMC Pediatr. 2017;17(1):208.
  3. Hirsch AG, Pollak J, Glass TA, Poulsen MN, Bailey-Davis L, Mowery J, et al. Early-life antibiotic use and subsequent diagnosis of food allergy and allergic diseases. Clin Exp Allergy. 2017;47(2):236-44.
  4. Saari A, Virta LJ, Sankilampi U, Dunkel L, Saxen H. Antibiotic exposure in infancy and risk of being overweight in the first 24 months of life. Pediatrics. 2015;135(4):617-26.
  5. Holmes AH, Moore LSP, Sundsfjord A, Steinbakk M, Regmi S, Karkey A, et al. Understanding the mechanisms and drivers of antimicrobial resistance. Lancet. 2016;387(10014):176-87.
  6. Schrag SJ, Verani JR. Intrapartum antibiotic prophylaxis for the prevention of perinatal group B streptococcal disease: experience in the United States and implications for a potential group B streptococcal vaccine. Vaccine. 2013;31 Suppl 4:D20-6.
  7. Strunk T, Buchiboyina A, Sharp M, Nathan E, Doherty D, Patole S. Implementation of the Neonatal Sepsis Calculator in an Australian Tertiary Perinatal Centre. Neonatology. 2018;113(4):379-82.
  8. Fjalstad JW, Stensvold HJ, Bergseng H, Simonsen GS, Salvesen B, Ronnestad AE, et al. Early-onset Sepsis and Antibiotic Exposure in Term Infants: A Nationwide Population-based Study in Norway. Pediatric Infect Dis J. 2016;35(1):1-6.
  9. Investigators of the Delhi Neonatal Infection Study (DeNIS) collaboration. Characterisation and antimicrobial resistance of sepsis pathogens in neonates born in tertiary care centres in Delhi, India: a cohort study. Lancet Glob Health. 2016;4(10):e752-e60.

Leave a Reply

Your email address will not be published. Required fields are marked *