Antibiotic stewardship in the context of management of early onset neonatal sepsis-an evolving paradigm shift

Abstract

Early onset neonatal sepsis (EOS) is a rare but potentially life-threatening condition in term and late preterm infants. Clinicians face the challenge of promptly identifying and treating the small subset of neonates with true sepsis, while avoiding unnecessary intervention in the majority who remain unaffected. The clinical overlap between sepsis and normal physiological adaptation in the first 72 hours of life, combined with the limitations of current diagnostic modalities, has historically led to a culture of presumed safety, resulting in the overtreatment of many healthy infants with antibiotics. Such early antibiotic exposure is associated with numerous invasive procedures, disruption of mother-infant bonding, increased healthcare utilisation, and significant costs. Moreover, the global rise of antibiotic resistance, the slow pace of novel antibiotic development, and growing concerns regarding microbiome dysgenesis and its long-term health consequences have catalysed a paradigm shift. Neonatal antibiotic stewardship now stands as a cornerstone of EOS management, seeking to balance the imperative for rapid, effective therapy against the risks of antimicrobial overuse. This review traces the evolution of antibiotic stewardship within neonatal care, highlighting the development and implementation of risk-based guidelines and strategies and decision-support models designed to curtail inappropriate antibiotic use without compromising clinical safety. By synthesising contemporary evidence and practice innovations, we emphasise the necessity for ongoing evaluation and refinement of stewardship approaches to optimise outcomes for neonates at risk of EOS. The journey of neonatal antibiotic stewardship is dynamic and ongoing, demanding sustained collaboration, research, and a commitment to safe, evidence-based clinical practice.

Key Words: Infant; Newborn; Early-onset neonatal sepsis; Neonatal antibiotics; Antibiotic stewardship; Sepsis risk calculator; Serial physical examination; Antimicrobial resistance

Core Tip: Early-onset neonatal sepsis is rare yet fear of missed serious infection continues to drive widespread empirical antibiotic use, exposing large numbers of well newborns to avoidable harm. This review synthesises contemporary evidence underpinning the paradigm shift in neonatal antibiotic stewardship, critically comparing traditional categorical guidelines with multivariate risk prediction tools and serial physical examination strategies. By integrating epidemiology, emerging United Kingdom and international data, and evolving clinical practice, it provides practical insights into how antibiotic exposure can be safely reduced without compromising outcomes, supporting more precise, family-centred, and evidence-based care.

INTRODUCTION

Early-onset neonatal sepsis (EOS) is an infrequent but potentially serious complication in the newborn with high mortality and morbidity. EOS is variably defined as blood or cerebrospinal fluid culture-proven infection within the first 48-72 hours of birth[1,2], although some guidance, including United States Centers for Disease Control (CDC) recommendations, extends this to the first 7 days[3]. For this review, EOS is defined as infection occurring within the first 72 hours of life. The pathogenesis of EOS is thought to involve ascending colonisation of the maternal genital tract, followed by subsequent colonisation and infection of the fetus or newborn, often occurring in the absence of overt maternal systemic illness[4].

Although there is wide variation in reported incidence in the literature, in large population-based studies, the global incidence of EOS is quoted as 0.3-0.9 per 1000 live births[5-8]. The overall incidence in the United Kingdom is approximately 0.7 per 1000 live births with incidence in term and late preterm infants reported as 0.3-0.6 per 1000 live births[6]. Group B streptococcus (GBS) and Escherichia coli (E. coli) are the most frequent pathogens. The incidence of early onset GBS sepsis is 0.22 per 1000 live births in United States and 0.3 per 1000 live births in the United Kingdom[6,7]. The incidence of EOS has been falling in the last two decades following the widespread use of intrapartum antibiotic prophylaxis[6].

The incidence of EOS is nearly 20-fold higher in preterm infants than in term or late preterm infants. The preterm infants also have a 10-fold higher mortality than their term counterparts[9]. The question for screening and need for treatment if they develop sepsis is therefore very different for preterm and term infants due to their differences in incidence and mortality. For the preterm and very low birth weight infants, majority with risks of EOS will need to be treated and we need to establish ‘who not to evaluate or treat’ as opposed to term and late preterm infants where most will not require treatment and the focus must be on ‘who to evaluate and treat’. In the preterm population, the focus should be more on timely discontinuation of antibiotic when sepsis has been ruled out or infrequently on not initiating antibiotics where the risk is minimal (elective caesarean sections for maternal reasons).

Assessing term and late preterm infants for risk of EOS is one of the most common clinical tasks undertaken by neonatologists. As the use of preventive intrapartum antibiotic therapies has increased and the incidence of EOS has decreased; physicians are challenged to identify those newborn infants who are at the highest risk of infection. Most physicians are particularly concerned about initially well appearing infants with identified risk factors for EOS for fear of missing the opportunity to intervene before infants become critically ill. This had led to widespread use of antibiotics by medical professionals as a defensive ‘safe rather than sorry’ approach. However, the global rise of antibiotic resistance, the slow pace of novel antibiotic development, and growing concerns regarding microbiome dysgenesis in the early neonatal period and its long-term health consequences has resulted in a call for rethink and paradigm shift. Antibiotic stewardship is now an integral strategy for managing EOS, seeking to balance the imperative for rapid, effective therapy against the risks of antimicrobial overuse.

In the pursuit of both optimal neonatal EOS management and responsible antibiotic stewardship, clinicians have adopted structured risk management tools. Three main approaches have emerged: Categorical univariate risk assessment guidelines from the United Kingdom National Institute for Health and Care Excellence (NICE) and CDC; the Kaiser Permanente Sepsis Risk Calculator (SRC); and the serial physical examination (SPE) model. Each strategy has distinct advantages and limitations, influenced by the clinical context in which it is applied, with implications for antibiotic use, clinical outcomes, and resource utilisation.

This narrative review synthesises contemporary evidence on EOS in term and late preterm infants, focusing on the case for antibiotic stewardship, potential harms of avoidable antibiotic exposure, and comparison of current management strategies to balance safety with prudent reduction in antibiotic use. References to perinatal context or preterm cohorts are included only where they clarify the disruptive effects of antibiotics on the developing microbiome and help explain biologically plausible, clinically relevant long-term consequences.

METHODOLOGY

The review integrates epidemiological data, international guidelines, cluster randomised and observational studies evaluating three principal approaches to EOS risk assessment and management: Categorical risk-based guidelines, multivariate risk prediction tools (SRC), and SPE strategies.

A targeted literature search was conducted in December 2025 using electronic databases including CINAHL Plus with Full Text (via EBSCOhost), Ovid MEDLINE, EMBASE, and the Maternity and Infant Care Database. Search strategies combined subject headings and free-text terms relating to population (“newborn”, “neonat*”, “infant*”), condition (“early onset sepsis”, “EOS”, “neonatal sepsis”), and management approaches (“serial physical examination”, “SPE”, “sepsis risk calculator”, “SRC”, “Kaiser Permanente”). Boolean operators (AND/OR) were used to optimise retrieval. Searches were limited to English-language studies published between January 2015 and December 2025.

To contextualise clinical practice, key national and international guidelines were included, particularly those from the NICE (CG149 and NG195) and the CDC, alongside relevant clinical reports from the American Academy of Pediatrics (AAP). These were selected for their relevance to contemporary EOS management and their influence on clinical decision-making frameworks.

Given the narrative nature of this review and its focus on evolving clinical paradigms, a formal systematic screening process was not undertaken. Instead, studies and guidelines were selected based on relevance, recency, and contribution to understanding the balance between antibiotic stewardship and patient safety. This approach allowed integration of diverse evidence sources to provide a comprehensive and practice-oriented overview of current strategies in EOS management. Grading of evidence for included references used in the study are summarised in Table 1.

Table 1 Grading of evidence for included references using the Oxford Centre for Evidence-Based Medicine 2011 guidance.

Ref.	Design	Level
[1]	Guideline	5
[2]	Guideline	5
[3]	Guideline	5
[4]	Cohort (model derivation)	2b
[5]	Narrative review	5
[6]	Cohort (surveillance)	2b
[7]	Cohort	2b
[8]	Implementation study	2b
[9]	Cohort	2b
[10]	Retrospective comparison	2b
[11]	Observational comparison	2b
[12]	Retrospective cohort	2b
[13]	Implementation study	2b
[14]	Multinational cohort	2b
[15]	Narrative article	5
[16]	Cohort	2b
[17]	Cohort	2b
[18]	Cohort	2b
[19]	Cohort	2b
[20]	Cross-sectional	4
[21]	Cohort	2b
[22]	Case control	3b
[23]	Review	5
[24]	Cohort (metagenomics)	2b
[25]	Cohort	2b
[26]	Cohort	2b
[27]	Cross-sectional	4
[28]	Economic model	-
[29]	Guideline	5
[30]	Cohort	2b
[31]	Guideline	5
[32]	Review	5
[33]	Cohort	2b
[34]	Multicentre cohort	2b
[35]	Observational study	2b
[36]	Systematic review + meta-analysis	1a
[37]	Cluster randomised controlled trial	1b
[38]	Systematic review + meta-analysis	1a
[39]	Review	5
[40]	Cohort	2b
[41]	Prospective observational implementation study	2b
[42]	Population cohort	2b
[43]	Population cohort	2b
[44]	Observational study	2b

ANTIBIOTIC USE FOR SCREENING AND MANAGEMENT OF EOS IN NEWBORN INFANTS

There is widespread variation in the volume of antibiotics prescription in different regions of the world to manage the risk of EOS in neonates, ranging from less than 2% to as high as 16% of all term and late preterm live births[4,8,10,11]. Despite the very low incidence of confirmed EOS in the United Kingdom, when following the NICE CG149 guideline, approximately 12%-16% of all newborns were reported to have undergone evaluation and antibiotic prescription for managing EOS[10,11]. Up to 40% of EOS cases lack identifiable risk factors, making detection through risk-based methods impossible without ongoing clinical vigilance and heightened parental awareness[12]. With an EOS incidence of 0.5/1000 live births, an approximate 60% detection rate using categorical risk factors, a 10%-15% prevalence of antibiotic use would expose 330-500 newborns to empirical antibiotics for each culture-confirmed EOS case. Subsequent modification of the NICE guideline, NG195, is reported to have reduced the use of antibiotics significantly (26%-35%) thereby reducing the treatment: Detection ratio[13].

Although not as widespread, similarly concerning pattern is also evident internationally. A large retrospective cross-sectional analysis of neonatal antibiotic prescribing in 11 high-income countries in Europe, North America and Australia reported that for each confirmed case of EOS, 58 neonates were commenced on antibiotics, amounting to a cumulative exposure of 273 antibiotic-days per case[14].

This high treatment-to-disease ratio reflected the inherent challenges clinicians face to distinguish early signs of EOS from normal transitional physiology. As evidence of harm from antibiotic overuse emerges in all age groups including newborn infants, there is call for more prudent approaches to management of neonatal EOS that balances clinical safety with responsible antibiotic use.

THE CASE FOR RESPONSIBLE STEWARDSHIP: EVIDENCE OF HARM FROM UNNECESSARY ANTIBIOTIC EXPOSURE

Although, safety is paramount in EOS management, the adverse impact of unnecessary antibiotic exposure, both immediate and long-term are now well established. High antibiotic use in term and late preterm infants on postnatal wards invariably increases the number of invasive procedures. Routine blood sampling for monitoring inflammatory markers, drug concentrations or organ function during antibiotic treatment leads to repeated phlebotomies, causing pain, discomfort, and family anxiety. Frequent cannulations risk extravasation, thrombophlebitis, and procedural pain. Administering multiple medications raises the risk of adverse events and prescribing errors, especially due to weight-based dosing and narrow safety margins[15].

Prolonged antibiotic therapy in the postnatal ward can disrupt maternal–infant contact, impeding breastfeeding, bonding, and consequent neurodevelopment. Early neonatal antibiotic use has been linked to a four-fold increase in delayed initiation of breastfeeding, a two-fold increase in non-medically indicated formula supplementation, higher levels of maternal anxiety and stress, and increased incidence of postpartum anxiety and depression[16].

Although not directly linked to EOS management, there are lessons to be learned from well conducted studies on associations of adverse health issues with early life exposure to antibiotics. Multiple cohorts consistently link antibiotic exposure in the first months of life with later childhood overweight/obesity, including the Danish Birth Cohort; ALSPAC, Canadian and ISAAC global cohort[17-20]. Maternal treatment during pregnancy also appears relevant: In a mother–infant cohort of 436 pairs, antibiotic use in the second or third trimester was associated with an 84% rise in obesity at 7 years[21]. Taken together, these findings suggest that disturbances to early microbial development may alter host–microbial metabolic interactions, with downstream effects on energy regulation and long-term adiposity risk.

Similarly, a large Swedish case–control study reported association of coeliac disease and related pathology with prior antibiotic exposure, and the association persisted after excluding prescriptions in the year before diagnosis, supporting a plausible microbiome-immune link[22].

Much of the current understanding of neonatal microbiome disruption and antibiotic-related harm is derived from preterm cohorts, where higher disease burden and intensive monitoring have facilitated detailed studies. In preterm infants, necrotising enterocolitis is preceded by reproducible dysbiosis (↑Proteobacteria; ↓Firmicutes/Bacteroidetes), plausibly amplifying gut inflammation (e.g., via TLR4 signalling)[23]. Strain-level data also implicate uropathogenic E. coli[24]. Clinical correlation comes from studies that show that prolonged empirical antibiotics in early days without proven infection are consistently associated with higher risk of necrotising enterocolitis and late-onset sepsis in extremely preterm cohorts[25,26]. While this is a different cohort, the physiology helps us understand potential pathogenetic mechanisms, reinforcing the need to minimise avoidable exposure to antibiotics.

Beyond clinical harm, empirical antibiotic use for suspected EOS carries substantial resource and cost implications. In a large United States cohort of neonates ≥ 35 weeks, an estimated $76.7 million was spent on birth admissions for infants started on antibiotics yet discharged within ≤ 3 days, suggesting a sizeable proportion of potentially avoidable expenditure[27]. Consistent with this, Guan et al[28] estimate that broad application of empiric antibiotics in infants evaluated for EOS increases healthcare utilisation through longer length of stay, additional laboratory testing, and intensified clinical monitoring, even when infection is ultimately excluded.

APPROACHES TO SCREENING AND MANAGEMENT OF NEONATAL SEPSIS

In the pursuit of both optimal neonatal EOS management and responsible antibiotic stewardship, clinicians have adopted structured risk management tools. Three main approaches have emerged: Categorical univariate risk assessment guidelines from the NICE and CDC; the Kaiser Permanente SRC; and the SPE model. Each strategy offers distinct advantages and challenges in different contexts, impacting antibiotic exposure, clinical outcomes, and resource allocation.

The early traditional risk based categorical methods for screening EOS

Management and prevention of EOS in the United Kingdom have traditionally been guided by NICE and Royal College of Obstetricians and Gynaecologists guidelines[1,2,29]. NICE first issued neonatal infection management guideline (CG149) in August 2012, which was replaced by NG195 in April 2021 and subsequently updated in March 2024, December 2024 and December 2025. The guidance set out a framework for identifying which infants required septic screening at birth based on specific risk factors and used both biochemical and clinical parameters to inform subsequent screening and management. This framework incorporates threshold-based criteria and dichotomous clinical outcomes, alongside added weighting for specific high-risk indicators (‘red flags’), to support decision-making around empiric antibiotic initiation in infants at risk of EOS.

A major limitation of this approach is that while it does include escalation with evolving clinical indicators of sepsis after birth, it does not incorporate the vast majority of infant’s stable postnatal clinical condition (‘wellness’) into account to downgrade the risk assessment. In practice, an infant’s early clinical appearance is a powerful predictor of true sepsis risk. Late preterm and term infants who are born in good clinical condition e.g. vigorous, well-perfused, and with no respiratory compromise, experience a significant reduction in their actual likelihood of EOS, even when intrapartum risk factors are present. Large cohort studies demonstrate that a well-appearing clinical status at birth reduces the risk of EOS by approximately 60%-70%, reflecting the strong negative predictive value of a normal early examination[4,30-32]. This means that for many infants, the initial risk estimated from maternal factors alone may considerably overstate their true postnatal risk. By not accounting for this immediate postnatal wellness assessment, the model may over-recommend empiric antibiotic initiation, potentially exposing clinically well infants to avoidable treatment, separation from parents, and prolonged hospitalisation.

Similarly, in the United States, following the initial 1996 CDC guidelines (updated 2002) recommendations, a multidisciplinary review in 2009 led to the development of the 2010 CDC guidelines[3], which were subsequently endorsed by major professional organizations. The updated guidance includes expanded recommendations for laboratory identification of GBS, clearer criteria for reporting GBS bacteriuria in pregnancy, revised screening and intrapartum antibiotic prophylaxis algorithms for women presenting with preterm labour or preterm premature rupture of membranes, an updated penicillin G dosing regimen, alternate prophylaxis options for women with penicillin allergy, and a revised neonatal management algorithm for assessing early-onset GBS disease.

The other issue of some of these algorithms is the reliance on laboratory markers in addition to blood cultures in directing the need for antibiotics. Although frequently used in the evaluation of suspected EOS, isolated laboratory markers such as C-reactive protein (CRP), procalcitonin (PCT) and white cell count (WCC) have limited diagnostic utility when interpreted on their own[32]. Their predictive performance is constrained by both biological and timing-related factors. CRP, for example, is an acute-phase reactant that rises in response to inflammation but does not distinguish between infectious and non-infectious causes. In the first hours after birth, CRP values are often still within the normal range even in truly septic infants, because CRP typically takes 6-12 hours to rise after the onset of infection. Conversely, elevated CRP concentrations may reflect physiological processes, such as labour, meconium aspiration, fetal distress, or even prolonged rupture of membranes-leading to high false-positive rates. Similarly, the diagnostic utility of PCT in EOS remains limited. Physiological postnatal elevation of PCT in the first 48-72 hours reduce its specificity, and studies in term neonates indicate that while PCT may rise earlier than CRP, it cannot reliably distinguish infected from non-infected infants when used in isolation[33]. WCC indices (total WCC, absolute neutrophil count, immature-to-total neutrophil ratio) are inherently variable in newborns in the peripartum period, influenced by gestational age, mode of delivery, perinatal stress, and sampling timing. Abnormal values are common among well infants, whereas normal counts do not reliably exclude EOS, contributing to poor sensitivity. As a result, the positive and negative predictive values of these tests, when used in isolation, are insufficient to confidently diagnose or exclude EOS. Their clinical usefulness increases only when interpreted serially over time, or in combination with the infant’s clinical condition and other risk factors, rather than as stand-alone indicators, by which time the infants are already prescribed antibiotics. This difficulty is compounded by the large number of infants, who fall into this category. However, these guidelines have been widely used for many years and have served an important role in providing structure and standardisation in the early evolution of EOS management across the United States, the United Kingdom, and Europe. Their relative simplicity and ease of application have supported reproducibility across different clinical settings, while aligning with institutional risk tolerance by prioritising safety and minimising the risk of missed sepsis. Owing to this perceived safety profile in managing infants considered at risk of EOS, clinicians have traditionally been comfortable adopting and continuing to use these approaches.

Introduction and widespread acceptance to use of Kaiser Permanente early-onset SRC

In their quest to better select those infants that would benefit from treatment and reassure those that did not need treatment in the context of EOS, The Kaiser Permanente group in Northern California applied a Bayesian analytical framework to develop a multivariate, infant-specific model for estimating the risk of EOS. This model, known popularly as the SRC, was derived and validated using a large case-control dataset of blood-culture-confirmed EOS[4]. The SRC generates a sepsis risk score (SRS) at birth by adjusting the estimated population baseline risk according to a combination of intrapartum maternal factors specific to the mother-infant dyad, such as gestational age, duration of rupture of membranes, highest intrapartum temperature, maternal GBS status, and the type and timing of intrapartum antibiotic prophylaxis. After the initial calculation at birth, the SRS is further refined by incorporating the newborn’s clinical presentation, categorised as well-appearing, equivocal, or clinically ill, yielding a final risk estimate that guides bedside management.

Although the SRS algorithm itself is grounded in evidence derived from the Kaiser Permanente dataset, the clinical thresholds that determine recommended management actions represent expert consensus rather than outcomes-validated cut-points. In practice, infants with an SRS below 1 are assigned routine newborn care without enhanced monitoring; those with an SRS between 1 and 3 undergo increased clinical observation with consideration of obtaining a blood culture; and infants with an SRS exceeding 3, or those exhibiting signs of clinical illness, are recommended to receive empirical antibiotic therapy. This structured approach aims to optimise antibiotic stewardship while maintaining patient safety.

SRC was recognised as an acceptable alternative approach to traditional categorical guidelines in the CDC guidance in 2019 and later considered in the United Kingdom during the NICE guideline update in April 2021. In its review, NICE concluded that current evidence is insufficient to determine whether the SRC or the existing NICE categorical risk-factor framework more accurately identifies infants at greatest risk of EOS. As a result, NICE advised that the SRC may be used as an alternative strategy for infants born at or beyond 34 + 0 weeks’ gestation, provided its adoption is accompanied by robust, ongoing prospective audit to ensure continued safety and appropriate clinical oversight. Apart from a Dutch Cluster randomised trial, most studies internationally and in the United Kingdom, comparing these two approaches have been before and after implementation observation studies. They have consistently demonstrated substantial reductions (40%-50% in most studies) in empirical antibiotic prescribing when the SRC is implemented[4,34-37]. Importantly, these reductions have not been accompanied by evidence of delayed treatment, missed EOS cases, or increased morbidity, although ongoing vigilance is required considering the low incidence of culture-confirmed disease. The accumulating observational data therefore suggest that use of the SRC may improve antimicrobial stewardship and reduce unnecessary exposure to antibiotics without compromising neonatal safety, particularly in populations where EOS prevalence is low and prevailing antibiotic use is high. However, comparative retrospective analyses of confirmed EOS cases have reported lower sensitivity of SRC to NICE guidelines in recommending prophylactic antibiotics during asymptomatic phase[12,38]. While the reported difference is a legitimate concern, the very low EOS incidence translates into a handful of cases of sepsis spanning several years. Prophylactic antibiotic treatment may not be the most appropriate strategy to capture these infants as it results in several thousand unnecessary treatments of others. Prudent healthcare programmes balance benefits with risk and cost, and therefore sensitivity alone (without measures of specificity) cannot determine the effectiveness of an intervention.

It can be argued that such cases are not necessarily ‘missed’ and can be safely identified and treated (similar to EOS in infants without risk factors) through a system of high-quality observation. Safety of implementation incorporating enhanced postnatal observations has been supported by United Kingdom prospective observational studies, including regional evaluations from Wales and the West Midlands, United Kingdom[34,35]. These studies reported no increase in missed EOS, sepsis-related morbidity, or mortality following adoption of the calculator.

Nevertheless, important limitations undermine the accuracy and generalisability of the calculator’s outputs. Methodological concerns include errors in regression intercept adjustment for oversampling, statistically imperfect incorporation of clinical examination findings, and use of upper confidence limits rather than point estimates for likelihood ratios, all of which may distort absolute risk estimates[39]. Some of these issues have been addressed in the most recent published version of the SRC. Furthermore, the SRC was derived exclusively from cohorts within the United States, where EOS incidence, obstetric practices, microbial epidemiology, and healthcare delivery differ from those in other regions. Although the calculator now allows adjustment for local EOS incidence, such recalibration does not address the fundamental limitation that the underlying regression model was developed using a United States population and may not be applicable to settings with higher EOS incidence or differing population risk profiles.

Despite these constraints, SRC retains clinical utility and has been widely adopted in more than 50% of perinatal units in the United Kingdom as a preferred tool for individualised risk stratification of EOS. Its principal value lies in supporting more selective investigation and treatment when used alongside careful clinical assessment and ongoing postnatal surveillance, an approach emphasised by both CDC and NICE guidance. However, this makes it even more challenging to compare the safety of SRC with NICE guidance in randomised controlled trial.

Overall, the SRC has facilitated a meaningful shift in EOS management practice, but its outputs must be interpreted cautiously, with explicit recognition of its technical limitations, population specificity, and the continued primacy of clinical judgement.

SPE

In contrast to traditional risk-factor driven strategies, the SPE approach represents a deliberate shift towards dynamic, patient-centred assessment. Rather than relying on static perinatal criteria to trigger investigation or treatment, SPE is grounded in structured, recurrent evaluations of the newborn over the first 24-48 hours, typically every one to four hours, allowing clinicians to respond to evolving signs rather than theoretical risk. Investigations and antibiotics are reserved for infants who develop persistent or clearly pathological features, irrespective of their initial risk profile. Although adoption remains selective, early experience from centres in Europe and North America demonstrates meaningful reductions in antibiotic exposure[40-44]. A summary of studies evaluating EOS management strategies and outcome are included in Table 2. Crucially, the effectiveness of SPE depends on the availability of clinicians who can exercise nuanced judgement and on systems with adequate staffing and electronic patient records to support consistent documentation. Unlike reactive early-warning tools such as NEWTT2, SPE is intentionally proactive and a structured surveillance strategy designed specifically to detect early signs of EOS in infants who initially appear well (Table 3).

Table 2 Summary of studies evaluating early onset neonatal sepsis management strategies.

Ref.	Design	Population (gestation)	Sample size (n)	Intervention/comparison	Key outcomes	Antibiotic reduction	Safety data	Level of evidence (OCEBM)	Main findings
Berardi et al[40], 2016	Cohort	Term infants	7628	SPE	Antibiotic use	50%	No increase in adverse outcomes	2b	SPE safely reduced unnecessary antibiotic use
Kuzniewicz et al[4], 2017	Cohort	≥ 34 weeks	608014	Development of SRC	EOS prediction, antibiotic use	48%	No increase in missed EOS reported	2b	SRC reduced empirical antibiotic use significantly
Dhudasia et al[8], 2018	Implementation study	≥ 35 weeks	2059	SRC implementation	Antibiotic exposure	42%	No increase in adverse outcomes	2b	Significant reduction in antibiotic use with maintained safety
Achten et al[36], 2019	Systematic review + meta-analysis	≥ 34 weeks	172385	SRC vs standard care	Antibiotic use, safety	44% (pooled)	No increase in mortality or severe morbidity	1a	Strong evidence supporting safety and reduced antibiotic use
Goel et al[10], 2020	Retrospective comparison	≥ 34 weeks	1187	NICE vs projected SRC	Antibiotic use	50% (projected)	Safety not directly assessed	2b	SRC projected to reduce antibiotic use vs NICE
Morris et al[12], 2020	Retrospective cohort	≥ 34 weeks	75 EOS cases	NICE vs SRC	Sensitivity		SRC less sensitive in identifying EOS cases	2b	Potential risk of delayed/missed treatment
Vatne et al[42], 2020	Population cohort	Term infants	20584	SPE	Antibiotic exposure	50%	No increase in infection-related morbidity/mortality	2b	Reduced antibiotic exposure safely
Pettinger et al[38], 2020	Systematic review + meta-analysis	≥ 34 weeks	NR	SRC sensitivity	Sensitivity		Lower sensitivity vs categorical approaches	1a	Highlights potential safety limitation
van Hasselt et al[35], 2021	Observational study	≥ 34 weeks	4000	SRC implementation	Antibiotic use	30%-40%	No safety concerns reported	2b	Reduction in antibiotic prescribing
Kimpton et al[11], 2021	Observational comparison	≥ 34 weeks	502	NICE vs SRC	Antibiotic use, sensitivity	40%	Lower sensitivity vs NICE noted	2b	Reduced antibiotic use but potential sensitivity trade-off
Goel et al[34], 2022	Multicentre implementation	≥ 34 weeks	66362	Adapted SRC	Antibiotic use, safety	40%	No increase in morbidity/mortality	2b	Safe reduction in antibiotic exposure
Cavigioli et al[41], 2022	Prospective implementation	≥ 34 weeks	4363	SRC + SPE	Antibiotic use	45%	No increase in adverse outcomes	2b	Combined approach reduced interventions safely
Bain et al[44], 2022	Observational study	≥ 34 weeks	1209	Clinical monitoring (SPE-like)	Antibiotic use	35%-50%	No increase in adverse outcomes	2b	Monitoring approach reduced antibiotic exposure safely
Yew et al[13], 2024	Implementation study	≥ 34 weeks	8856	SRC implementation	Antibiotic use	45%	No increase in adverse outcomes	2b	Reduced antibiotic exposure in real-world setting
Vatne et al[43], 2025	Population-based study	≥ 34 weeks	54671	SPE	Antibiotic use, safety	50%	No increase in mortality, NICU admission, or readmission	2b	Strong large-scale safety and effectiveness evidence
van der Weijden et al[37], 2025	Cluster RCT	≥ 34 weeks	18479	SRC vs standard care	Antibiotic use, safety	30%-35%	No increase in adverse outcomes	1b	RCT evidence supporting reduced antibiotic use with maintained safety

EOS: Early-onset sepsis; SRC: Sepsis Risk Calculator; SPE: Serial Physical Examination; NICE: National Institute for Health and Care Excellence; RCT: Randomised controlled trial; NICU: Neonatal intensive care unit; OCEBM: Oxford Centre for Evidence-Based Medicine; NR: Not reported.

Table 3 Comparison of early-onset sepsis management strategies.

Feature	Categorical risk-based approach (e.g., NICE/CDC)	Sepsis risk calculator	Serial physical examination
Population	Term and late preterm infants (≥ 34 weeks)	Term and late preterm infants (≥ 34 weeks)	Term and late preterm infants (≥ 34 weeks)
Management logic	Static risk factor–based (maternal and perinatal factors)	Multivariate, individualised risk estimation using Bayesian model (maternal + infant clinical status)	Dynamic, repeated clinical assessments over time (evolving clinical signs)
Approach to treatment	Low threshold for investigation and empirical antibiotics based on risk factors	Risk-stratified approach guiding observation, investigation, or antibiotics	Antibiotics reserved for infants with evolving or persistent clinical signs
Antibiotic use	Higher antibiotic exposure; risk of overtreatment	Reduced antibiotic use (commonly 30%-50% reduction)	Reduced antibiotic use (50% reduction demonstrated in observational studies)
Safety data	High sensitivity; widely accepted safety profile	Observational studies show no increase in missed EOS, morbidity, or mortality; lower sensitivity vs categorical approaches reported	Large observational studies show reduced antibiotic use without increase in adverse outcomes
Strengths	Simple, reproducible, widely implemented; aligns with institutional risk tolerance	Individualised risk assessment; supports antibiotic stewardship; widely adopted	Patient-centred; avoids unnecessary treatment; detects evolving illness
Limitations	Poor specificity; high rates of unnecessary antibiotic use	Lower sensitivity in some studies; potential delay in treatment; derived from United States populations; relies on accurate input data	Resource-intensive; requires frequent assessments, experienced clinicians, and robust systems; inter-observer variability
Implementation considerations	Easy to apply across settings; minimal training required	Requires calculator access, local incidence calibration, and audit	Requires adequate staffing, workflow support, escalation systems, and consistent documentation
Generalisability	Widely applicable across different healthcare settings	May require adaptation for different populations and EOS incidence	More feasible in well-resourced settings with adequate staffing and monitoring capacity

EOS: Early onset neonatal sepsis; NICE: National Institute for Health and Care Excellence; CDC: Centers for Disease Control and Prevention.

A recent Norwegian multicentre study by Vatne et al[43] offers one of the most robust evaluations to date of the SPE approach, examining its application across six neonatal units and more than 54000 live births. Infants born at or beyond 34 weeks’ gestation with recognised EOS risk factors or mild, transient symptoms underwent structured hourly assessments and continuous monitoring for 24-48 hours, with antibiotics reserved for those who developed persistent or clearly pathological signs. The findings were striking-the antibiotic exposure was halved, falling from 1.8% to 0.9%, with no accompanying increase in infection-related mortality, NICU admissions or readmissions. Equally noteworthy was the substantial improvement in the ratio of culture-negative to culture-positive diagnoses, indicating greater diagnostic confidence and a reduction in unnecessary treatment. Distinguished by its scale, pragmatic design and external validity across varied hospital settings, the study demonstrates that even in healthcare systems with already low baseline antibiotic use, SPE can reduce exposure further without any apparent risk of missed cases of sepsis. It promises as a scalable strategy in specific settings for improving antimicrobial stewardship in the newborn period.

Implementation of SPE at scale presents several important challenges that must be considered by health systems contemplating adoption. The Norwegian protocol required NICU admission and continuous monitoring for all infants with EOS risk factors-a level of resource availability, staffing stability and monitoring capacity that many units may struggle to provide. Such NICU-based observation also raises concerns about parent–infant separation, with potential negative effects on bonding, breastfeeding and early emotional wellbeing during a critical developmental window. Moreover, generalisability remains an issue; in settings with higher baseline antibiotic use, fewer experienced clinicians, or limited transitional care capacity, SPE may require significant adaptation to ensure both feasibility and safety. Notably, following completion of the study, Norwegian practice evolved so that only symptomatic infants now undergo NICU-based SPE, while well-appearing infants remain with their mothers. Although promising from a family-centred care perspective, this modified approach will need further evaluation. Taken together, these considerations highlight that while SPE offers a potentially safe, feasible and effective strategy to reduce unnecessary antibiotic use for EOS in well-resourced environments, careful attention to local infrastructure, workforce capability and parental experience is essential. The Norwegian data strengthens the case for dynamic, clinician-led observation as a core component of neonatal antibiotic stewardship. As health systems strive to balance safety with responsible prescribing, digitally supported SPE may represent a particularly promising path forward.

FUTURE DIRECTIONS

The Norwegian study by Vatne et al[43] adds important momentum to global antibiotic stewardship efforts and positions SPE as a pragmatic and clinically grounded next step, particularly for units that already operate with restrictive antibiotic policies and low baseline exposure. Their findings suggest that structured, clinician-led observation can meaningfully reduce overtreatment even in settings where prescribing has already been optimised. Further refinement and external calibration of EOS risk models across diverse populations will be essential to support safe implementation and wider adoption across healthcare settings with differing resources. In the future, machine-learning decision support embedded within electronic health records could enable more precise, real-time EOS risk stratification and earlier detection of subtle clinical deterioration; this should be a priority for prospective research and evaluation and must include evaluation of unintended consequences.

As these technologies mature, the combination of clinician expertise, structured observation, and intelligent monitoring systems have the potential to transform EOS management-moving neonatal care towards a model that is simultaneously more personalised, evidence-driven and judicious in its use of antibiotics.

LIMITATIONS

This paper is not intended to be a systematic review on antibiotic stewardship programme, but an expert narrative opinion derived from contemporary evidence. Selection bias is minimised using three authors to independently select relevant evidence and reaching a consensus on inclusion of available evidence.

CONCLUSION

Despite its low incidence, EOS continues to drive high rates of empirical antibiotic use in term and late-preterm infants. Traditional categorical, risk-factor-based guidelines such as NICE NG195 and CDC GBS guidance prioritise safety but identify large numbers of well infants for screening and treatment, contributing to often avoidable exposure without substantially improving detection. Multivariate tools such as the SRC offer a more individualised approach and have consistently reduced antibiotic use without evidence of harm in prospective implementation studies when combined with careful clinical observation. Dynamic strategies such as SPE further demonstrate that structured observation have potential to safely reduce antibiotic exposure. These findings support a shift from static risk-factor models toward responsive, clinically driven assessment in well-resourced settings. However, safety issues must be monitored through continuous audit of outcomes and learning embedded in any implementation. For very preterm infants, however, EOS risk is tightly linked to the circumstances of delivery, and neither SRC nor SPE is appropriate. AAP guidance emphasises early cultures, targeted empiric therapy, and prompt discontinuation when cultures remain negative-an approach supported by multiple studies linking prolonged empirical therapy with increased risks of NEC, late-onset sepsis, and mortality. Finally, no strategy captures all EOS cases, and a significant proportion occur without identifiable antenatal risk factors, reinforcing that clinical vigilance remains central. Combining multivariate risk prediction with structured examination pathways offers the most balanced approach: Reducing unnecessary treatment while maintaining safety. Continued evaluation, local audit, and integration of digital early-warning tools will be essential to advancing safe, evidence-based stewardship.

ACKNOWLEDGEMENTS

The authors would like to thank Lizzy Evans, Library Services Manager at Swansea Bay University Health Board, for her valuable support and assistance with the literature search for this review.