Neonatal Early-Onset Sepsis: Comparing Three Risk-Assessment Strategies
Weighing the sepsis risk calculator, categorical culture-based workup, and structured observation against the twin harms of antibiotic overtreatment and missed infection in well-appearing newborns.
One-Minute Takeaway
- Neonatal early-onset sepsis affects roughly 0.5 per 1,000 term births, yet 5–15% of newborns receive empirical antibiotics under traditional categorical workup — the overtreatment these three strategies target (highest-tier evidence: one cluster-randomised trial plus meta-analyses).
- In a 2025 cluster-randomised trial, the sepsis risk calculator reduced empirical antibiotics to 7.2% versus 26.6% with categorical guidelines (absolute reduction 19.0%, 95% CI 11.3–26.7).
- A meta-analysis of 13 studies (175,752 newborns) found antibiotic use fell to 56% of the conventional rate with the calculator (RR 0.56, 95% CI 0.53–0.59).
- Serial physical examination is a low-technology alternative: in one cohort, only 1.9% of observed at-risk newborns received antibiotics, with normal outcomes throughout.
- Safety for rare missed infection remains uncertain: a sensitivity meta-analysis estimated 14–22 of 75 culture-positive cases could be missed or delayed by the calculator versus stricter categorical rules.
Why Neonatal Early-Onset Sepsis Matters
Neonatal early-onset sepsis is bacterial infection presenting within the first 72 hours of life. Group B Streptococcus and Escherichia coli remain the leading pathogens in term and late-preterm infants.
Widespread intrapartum antibiotic prophylaxis has lowered incidence. The AAP places it near 0.5 per 1,000 term live births, rising toward 1 per 1,000 among late-preterm infants. The condition is now uncommon but still potentially fatal.
The clinical problem is asymmetric. Infection is rare, yet its early signs are subtle, so clinicians have long treated many well-appearing at-risk newborns empirically. Under categorical guidelines, an estimated 5–15% of newborns receive antibiotics, while true infection touches a fraction of one percent.
That imbalance carries costs. Empirical antibiotics disrupt the developing microbiome, prompt NICU admission, separate mother and infant, and drive resistance. Laboratory tests such as complete blood count and C-reactive protein are neither sensitive nor specific enough to resolve the question on their own.
The core tension: every strategy must minimise two competing harms at once — overtreating thousands of uninfected newborns, and missing the rare infant whose infection is genuinely developing. No approach eliminates both.
The AAP endorses three approaches for infants born at 35 weeks’ gestation or later: categorical risk assessment, multivariate risk assessment via the sepsis risk calculator, and risk assessment based on the newborn’s evolving clinical condition through enhanced observation. This review compares how those three perform.
The Options Compared
The three strategies differ in what triggers action and how much they rely on maternal risk factors versus the infant in front of you.
Categorical, culture-based workup
The traditional approach treats defined risk factors — chorioamnionitis exposure, prolonged rupture of membranes, inadequate prophylaxis — as thresholds. Crossing one prompts a blood culture, laboratory tests, and often empirical antibiotics, regardless of how well the infant appears.
The sepsis risk calculator
The multivariate calculator, derived from a Kaiser Permanente cohort of more than 600,000 births, converts five objective inputs into an individualised risk estimate. Gestational age, highest maternal intrapartum temperature, GBS status, duration of membrane rupture, and intrapartum antibiotic type combine with the newborn’s clinical examination to yield a recommendation.
Serial physical examination (observation)
Enhanced observation replaces reflexive testing with structured, repeated clinical assessment at standardised intervals. Antibiotics are withheld from well-appearing infants unless clinical signs emerge, using the newborn’s own condition as the most sensitive early indicator.
The comparative evidence below draws only on studies whose figures were confirmed against the primary source, spanning one randomised trial, three meta-analyses, and large implementation cohorts.
| Study / Source | Design | N | Strategies compared | Key verified result | Conf. |
|---|---|---|---|---|---|
| van der Weijden 2025 | Open-label cluster-RCT (10 hospitals) | 1,830 (915/arm) | Calculator vs categorical guideline | Empirical antibiotics 7.2% vs 26.6%; absolute reduction 19.0% (95% CI 11.3–26.7). Readmission 3 vs 2; adverse-event rates similar between arms. | V |
| Achten 2019 | Systematic review / meta-analysis (13 before-after studies) | 175,752 | Calculator vs conventional | Antibiotic use RR 0.56 (95% CI 0.53–0.59). Missed EOS OR 0.96 (95% CI 0.26–3.52). | V |
| Pettinger 2020 | Systematic review / meta-analysis (11 studies) | 75 EOS cases | Calculator vs NICE categorical | 14 (best case) to 22 (worst case) of 75 culture-positive cases missed or delayed by the calculator. | V |
| Kuzniewicz 2017 | Before-after cohort (Kaiser N. California) | 204,485 | Calculator vs prior practice | Blood-culture use 14.5%→4.9% (adjusted difference −7.7%, 95% CI −13.1 to −2.4); empirical antibiotics also fell. | V |
| Cavigioli 2022 | Single-centre prospective, three periods | 3,002 | Categorical vs SPE vs calculator + universal SPE | Laboratory tests 33%→6.6%; antibiotics 8.5%→1.4% (both p<0.01); EOS and mortality unchanged. | V |
| Berardi 2016 | Retrospective multicentre cohort | 2,092 (216 via SPE) | Serial physical examination (observation) | Of infants managed by observation, 5.6% needed later workup and 1.9% received antibiotics; all had normal outcomes. | V |
Confidence key: V Verified against the primary source this review was prepared from. RR = relative risk; OR = odds ratio; SPE = serial physical examination; CI = confidence interval.
Empirical Antibiotic Exposure — How the Approaches Differ
Reducing unnecessary antibiotics is where the evidence is strongest and most consistent. Every comparative study favoured the calculator or observation over categorical workup on this outcome.
The single cluster-randomised trial provides the firmest evidence. Randomising 10 hospitals to the calculator or categorical guidelines, it found empirical antibiotics started in 7.2% of at-risk newborns under the calculator versus 26.6% under categorical care, an absolute reduction of 19.0% (95% CI 11.3–26.7).
The high categorical rate reflects a deliberately enriched population — every enrolled newborn carried at least one risk factor — so these figures describe the at-risk subgroup, not all births. The direction and magnitude nonetheless favour the calculator clearly.
Pooled observational data agree. In the Achten 2019 meta-analysis, antibiotic use with the calculator fell to 56% of the conventional rate (RR 0.56, 95% CI 0.53–0.59) across 175,752 newborns. Because the underlying studies were before-after comparisons, this is an association rather than a randomised effect, but its consistency is notable.
Observation performs comparably. In the Cavigioli 2022 program, combining the calculator with universal serial examination was associated with antibiotic treatment falling from 8.5% to 1.4%. Observation used alone, in the Berardi 2016 cohort, left just 1.9% of monitored infants on antibiotics. Both strategies converge on a similar destination from different starting points.
Detecting Neonatal Early-Onset Sepsis Safely — How the Approaches Differ
Fewer antibiotics only help if genuine infections are still caught. Here the evidence is thinner, and the comparison depends heavily on which categorical standard the calculator is measured against.
Against typical United States categorical practice, the calculator shows no clear safety penalty. The Achten 2019 meta-analysis found comparable proportions of missed cases: 28% under the calculator versus 29% under conventional management, with a pooled odds ratio of 0.96 (95% CI 0.26–3.52). The confidence interval is wide and rests on small case counts, so it neither confirms nor excludes a meaningful difference.
Against the stricter NICE-style categorical approach, a different picture emerges. The Pettinger 2020 sensitivity meta-analysis pooled 11 studies containing 75 culture-positive cases and estimated that 14 to 22 of them would have been missed or delayed had the calculator governed treatment instead of NICE guidance.
Interpretation: the calculator is less aggressive than the most cautious categorical rules. Whether that trades acceptable safety for fewer antibiotics depends on your baseline infection rate and how much residual observation backstops a low-risk score.
Serial observation carries its own logic on safety. Because it acts on the infant’s evolving examination rather than a birth-time estimate, it can detect infection that a static score assigns to a low-risk band. In the Berardi 2016 cohort, all infants managed by observation had normal outcomes, though the number of true infections in that low-risk group was small.
Every strategy assumes a functioning safety net: reliable serial monitoring, clear escalation criteria, and caregiver access if signs appear after discharge. A low-risk calculator output is a recommendation to observe, not a clearance to stop watching.
Laboratory Testing and Mother–Infant Separation — How the Approaches Differ
Beyond antibiotics, the strategies differ in how much testing they generate and how often they separate mother and infant — outcomes that shape families’ first days and unit workload.
The calculator reduces blood draws. In the Kuzniewicz 2017 Kaiser cohort of 204,485 infants, blood-culture use fell from 14.5% to 4.9% after implementation, an adjusted difference of −7.7% (95% CI −13.1 to −2.4). Fewer cultures mean fewer painful procedures and fewer incidental abnormal results to chase.
Observation-based programs show similar restraint. In the Cavigioli 2022 experience, laboratory testing fell from 33% to 6.6% once serial examination replaced reflexive categorical workup, while measured early-onset sepsis and mortality were unchanged across the study periods.
Both the calculator and observation keep more well-appearing newborns beside their mothers rather than in the NICU for rule-out antibiotics. Categorical assessment, by contrast, admits and separates the largest share, because a single maternal risk factor can trigger the full pathway irrespective of the infant’s condition.
These are consistent observational signals rather than randomised outcomes, and testing rates depend on local thresholds. The pattern across independent centres, however, points the same way: less reliance on maternal risk factors alone means less testing and less separation.
Subgroups That Change the Choice
The best strategy is not universal. Several factors shift the balance between approaches.
Baseline antibiotic exposure
The calculator’s advantage shrinks where overtreatment is already low. An updated Swiss national guideline declines to recommend the calculator in settings where antibiotic exposure is already below roughly 3%, favouring serial examination instead — because a tool built to cut excess has little excess to remove.
Chorioamnionitis exposure
Infants exposed to intraamniotic infection are where categorical rules are most reflexively aggressive, and where the calculator produces its largest relative reductions. This subgroup drove some of the strongest antibiotic-sparing signals in the pooled data.
Gestational age
All three strategies apply to infants born at 34 to 35 weeks or later. Genuinely preterm infants fall outside the calculator’s derivation and are managed under separate, more cautious guidance, since their infection risk and clinical fragility differ substantially.
Local infrastructure
Observation demands dependable staffing to perform and document serial examinations. Where that capacity is limited, a structured calculator recommendation may be more reliably delivered than an examination schedule that is easy to let slip.
Quality and Consistency of the Evidence
Certainty varies sharply by outcome. It is highest for antibiotic reduction and lowest for the rare, safety-critical event of a missed infection.
- Antibiotic reduction (calculator vs categorical) — moderate to high. Supported by one cluster-randomised trial and a consistent 13-study meta-analysis pointing the same direction.
- Reduced testing and separation — moderate. Multiple large cohorts agree, but the evidence is observational and before-after in design.
- Antibiotic reduction with observation — moderate. Consistent cohort data, no randomised comparison against the calculator head-to-head.
- Safety for missed infection — low. Small case counts, wide confidence intervals, and results that flip depending on the categorical comparator used.
A structural caveat runs through the literature: most implementation studies are before-after comparisons, so secular trends in practice may contribute to the reductions attributed to a given tool. The randomised trial mitigates this for antibiotic use but was not powered for rare infection outcomes.
What the Evidence Does Not Show
No adequately powered trial has compared all three strategies for the outcome that matters most — missed or delayed early-onset sepsis. Infection is too rare for existing studies to resolve small differences in that risk.
A direct randomised comparison of the calculator against serial observation is absent. Available head-to-head data pit the calculator against categorical guidelines, leaving observation evaluated mainly in single-arm cohorts.
Long-term outcomes are largely unstudied. Whether reduced neonatal antibiotic exposure translates into measurable differences in later microbiome-related or immune outcomes has not been established in this population.
Generalisability is uncertain. Much of the evidence originates from a small number of well-resourced systems, and performance in settings with different infection epidemiology, staffing, or follow-up capacity is not well characterised. Available evidence is insufficient to determine a single best strategy for every unit.
Practical Implications
The evidence supports several practical positions without dictating a single choice.
All three AAP-endorsed strategies reduce overtreatment relative to reflexive categorical antibiotics, and both the calculator and structured observation do so while maintaining the safety outcomes that current studies can measure.
Units with high baseline antibiotic exposure stand to gain the most from adopting either the calculator or serial observation, because they have the most excess treatment to remove. Units already treating few infants may find little to improve and reasonable grounds to prefer low-technology observation.
Whichever pathway a unit selects, the evidence favours pairing any birth-time risk estimate with ongoing clinical observation, and building explicit safeguards: documented serial examinations, defined escalation triggers, and clear discharge guidance for families. The AAP recommends that units choose a single institutional approach suited to local resources and audit it for infrequent adverse outcomes.
Evidence Grade and Bottom Line
Overall grade: Moderate.
- The sepsis risk calculator and serial observation both reduce empirical antibiotics and testing versus categorical culture-based workup; this reduction is the best-supported finding, anchored by one randomised trial.
- Safety for rare missed infection is not firmly established. The calculator is less sensitive than the strictest categorical rules, so a low-risk score should trigger continued observation, not disengagement.
- No powered trial separates the three strategies on missed-infection risk, and no randomised study compares the calculator with observation directly.
- Strategy choice should follow local baseline antibiotic rates, staffing, and follow-up capacity rather than a universal ranking.
Article Information and References
For Educational Purposes Only. This is an original evidence synthesis informed by the studies listed below. It does not replace clinical judgement. Drug dosages should be verified against current prescribing information.
- Achten NB, Klingenberg C, Benitz WE, et al. Association of use of the neonatal early-onset sepsis calculator with reduction in antibiotic therapy and safety: a systematic review and meta-analysis. JAMA Pediatr. 2019;173(11):1032–1040. DOI: https://doi.org/10.1001/jamapediatrics.2019.2825
- Pettinger KJ, Mayers K, McKechnie L, Phillips B. Sensitivity of the Kaiser Permanente early-onset sepsis calculator: a systematic review and meta-analysis. EClinicalMedicine. 2020;19:100227. DOI: https://doi.org/10.1016/j.eclinm.2019.11.020
- Achten NB, Plötz FB, Klingenberg C, et al. Stratification of culture-proven early-onset sepsis cases by the neonatal early-onset sepsis calculator: an individual patient data meta-analysis. J Pediatr. 2021;234:77–84.e8. DOI: https://doi.org/10.1016/j.jpeds.2021.01.065
- van der Weijden BM, Janssen SWCM, van der Weide MC, et al. Safety and effectiveness of the early-onset sepsis calculator to reduce antibiotic exposure in at-risk newborns: a cluster-randomised controlled trial. EClinicalMedicine. 2025;87:103419. DOI: https://doi.org/10.1016/j.eclinm.2025.103419
- Kuzniewicz MW, Puopolo KM, Fischer A, et al. A quantitative, risk-based approach to the management of neonatal early-onset sepsis. JAMA Pediatr. 2017;171(4):365–371. DOI: https://doi.org/10.1001/jamapediatrics.2016.4678
- Cavigioli F, Viaroli F, Daniele I, et al. Neonatal early onset sepsis (EOS) calculator plus universal serial physical examination (SPE): a prospective two-step implementation of a neonatal EOS prevention protocol for reduction of sepsis workup and antibiotic treatment. Antibiotics (Basel). 2022;11(8):1089. DOI: https://doi.org/10.3390/antibiotics11081089
- Berardi A, Buffagni AM, Rossi C, et al. Serial physical examinations, a simple and reliable tool for managing neonates at risk for early-onset sepsis. World J Clin Pediatr. 2016;5(4):358–364. DOI: https://doi.org/10.5409/wjcp.v5.i4.358
- Vatne A, Klingenberg C, Øymar K, et al. Reduced antibiotic exposure by serial physical examinations in term neonates at risk of early-onset sepsis. Pediatr Infect Dis J. 2020;39(5):438–443. DOI: https://doi.org/10.1097/INF.0000000000002590
- Puopolo KM, Benitz WE, Zaoutis TE; Committee on Fetus and Newborn; Committee on Infectious Diseases. Management of neonates born at ≥35 0/7 weeks’ gestation with suspected or proven early-onset bacterial sepsis. Pediatrics. 2018;142(6):e20182894. DOI: https://doi.org/10.1542/peds.2018-2894
- Escobar GJ, Puopolo KM, Wi S, et al. Stratification of risk of early-onset sepsis in newborns ≥34 weeks’ gestation. Pediatrics. 2014;133(1):30–36. DOI: https://doi.org/10.1542/peds.2013-1689
- Puopolo KM, Draper D, Wi S, et al. Estimating the probability of neonatal early-onset infection on the basis of maternal risk factors. Pediatrics. 2011;128(5):e1155–e1163. DOI: https://doi.org/10.1542/peds.2010-3464
- Snoek L, van Kassel MN, Krommenhoek JF, et al. Neonatal early-onset infections: comparing the sensitivity of the neonatal early-onset sepsis calculator to the Dutch and the updated NICE guidelines in an observational cohort of culture-positive cases. EClinicalMedicine. 2022;44:101270. DOI: https://doi.org/10.1016/j.eclinm.2021.101270
- Schrag SJ, Farley MM, Petit S, et al. Epidemiology of invasive early-onset neonatal sepsis, 2005 to 2014. Pediatrics. 2016;138(6):e20162013. DOI: https://doi.org/10.1542/peds.2016-2013
- Polin RA; Committee on Fetus and Newborn. Management of neonates with suspected or proven early-onset bacterial sepsis. Pediatrics. 2012;129(5):1006–1015. DOI: https://doi.org/10.1542/peds.2012-0541