Article Text

Impact of NICE clinical guidelines for prevention and treatment of neonatal infections on antibiotic use in very preterm infants in England and Wales: an interrupted time series analysis
  1. Mike Saunders1,
  2. Shalini Ojha2,3,
  3. Lisa Szatkowski2
  1. 1 University of Nottingham School of Medicine, Nottingham, UK
  2. 2 Centre for Perinatal Research, University of Nottingham School of Medicine, Nottingham, UK
  3. 3 Neonatal Unit, University Hospitals of Derby and Burton NHS Foundation Trust, Derby, UK
  1. Correspondence to Dr Lisa Szatkowski; Lisa.Szatkowski{at}


Objective To assess the impact of publication of UK National Institute for Health and Care Excellence (NICE) guidelines on the prevention and treatment of early-onset infections (EOIs) in neonates (clinical guideline 149 (CG149), published in 2012, and its 2021 update (NG195) on antibiotic use in very preterm infants.

Design Interrupted time series analysis using data from the National Neonatal Research Database.

Setting Neonatal units in England and Wales.

Participants Infants born at 22–31 weeks’ gestation from 1 January 2010 to 31 December 2022 and survived to discharge.

Interventions Publication of CG149 (August 2012) and NG195 (April 2021).

Main outcome measures Measures of antibiotic use, aggregated by month of birth: antibiotic use rate (AUR), the proportion of care days in receipt of at least one antibiotic; percentage of infants who received ≥1 day of antibiotics on days 1–3 for EOI and after day 3 for late-onset infection (LOI); percentage who received ≥1 prolonged antibiotic course ≥5 days for EOI and LOI.

Results 96% of infants received an antibiotic during inpatient stay. AUR declined at publication of CG149, without further impact at NG195 publication. There was no impact of CG149 on the underlying trend in infants receiving ≥1 day antibiotics for EOI or LOI, but post-NG195 the monthly trend began to decline for EOI (−0.20%, −0.26 to −0.14) and LOI (−0.23%, −0.33 to −0.12). Use of prolonged antibiotic courses for EOI and LOI declined at publication of CG149 and for LOI this trend accelerated post-NG195.

Conclusions Publications of NICE guidance were associated with reductions in antibiotic use; however neonatal antibiotic exposure remains extremely high.

  • Neonatology
  • Epidemiology

Data availability statement

Data may be obtained from a third party and are not publicly available. National Neonatal Research Database data may be obtained from a third party with relevant approvals.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Antibiotics are very often prescribed to preterm infants, though irrational use is associated with harm. The National Institute for Health and Care Excellence published clinical guideline 149 (CG149) in 2012 and updated guidance in 2021 (NG195), with recommendations for neonatal antibiotic prescribing. The national impact of CG149 and NG195 on antibiotic use has not been studied.


  • Interrupted time series analysis quantified the impact of publication of CG149 and NG195 on measures of antibiotic initiation and prolonged use. Publication of CG149 was associated with decreases in the proportion of care days infants received at least one antibiotic and exposure to a prolonged antibiotic course, though with smaller or no change for extremely premature infants and those with bacterial infection or recorded indication(s) for antibiotics. Publication of NG195 was associated with decreases in antibiotic use for late onset infection. Over 95% of preterm infants studied were exposed to antibiotics.


  • This study finds that publication of national guidance CG149 and NG195 was associated with reductions in neonatal antibiotic use, possibly through more conservative and consistent antibiotic prescribing. However, antibiotic use remains extremely high among preterm infants. Further research is required to find evidence-based approaches to empiric antibiotic prescribing in newborn infants to identify reliable and safe means of optimising and safely reducing avoidable antibiotic exposure.


Neonatal infection is common and burdensome.1 2 Clinical manifestations can be non-specific and difficult to distinguish from other pathology.3 Preterm infants have higher incidence and mortality of early-onset infection (EOI, within 72 hours of birth) and late-onset infection (LOI, ≥72 hours from birth) than term infants2 4 5 and have prolonged hospital stays which carry risk of healthcare-associated infection.

Neonatal antibiotic use is empirical (given prior to confirmation of infection) and based on risk and clinical assessment,4 as investigations may yield false negative results3 and delay could risk deterioration.6 Antibiotics are very commonly prescribed in neonatal units7–9 and save lives but use risks drug toxicity, side effects4 and antibiotic resistance. Prolonged antibiotic use is associated with harm, including necrotising enterocolitis (NEC), neurological injury, LOI,8 10–12 invasive fungal diseases13 and disruption of gut microbiota.3

The National Institute for Health and Care Excellence (NICE) produces and publishes evidence-based clinical guidelines in the UK. Before 2012, there was no UK national guidance on preventing and treating neonatal infections. NICE published a consensus-based clinical guideline (CG149) in August 20122 with recommendations for intrapartum antibiotic prophylaxis, risk factors for and possible clinical signs of EOI, and indications for investigation and empirical treatment of EOI.

CG149 had potential limitations. Preterm birth after spontaneous labour with prolonged preterm rupture of membranes necessitated antibiotic use, meaning many premature babies receive antibiotics as default. Additionally, in practice a low threshold is used to manage maternal perinatal fever as sepsis14 which obligated neonatal antibiotic prescribing. Clinical indicators of neonatal sepsis were non-specific,1 based on low-quality evidence2 and risked subjective interpretation. Additionally, antibiotic course length recommendations were unclear, the impact of intrapartum antibiotic prophylaxis on neonatal management was not described, and there were no recommendations for LOI.

New guidance (NG195) superseded CG149 in April 2021,15 which introduced recommendations for managing LOI. The main recommendations in this revision were again introduced without any evidence-base or evaluation of impact of antibiotic exposure on babies. Based on evidence from the USA, NG195 permits the alternative of using the Kaiser Permanente neonatal sepsis risk calculator16 (KP-SRC) to inform clinical decisions on EOI management. The KP-SRC, however, is only applicable to babies born at ≥34 weeks’ gestation and the recommendations for most preterm babies therefore remained consensus based.

It is not known whether the introduction of NICE CG149 and NG195 (summarised in online supplemental figure 1) altered antibiotic prescribing patterns. This is pertinent to ensure antibiotics are used wisely and prevent avoidable harms. This study aims to assess their impact on measures of antibiotic use in very and extremely preterm infants admitted to neonatal units in England and Wales.


Data management and analysis were conducted using R V.4.3.217 and Stata V.18 (Stata, College Station, Texas, USA).

Data source and study population

We used data from the UK National Neonatal Research Database (NNRD),18 derived from the electronic patient records of all infants admitted to NHS neonatal units in England and Wales, for very preterm infants (born at 28–31 weeks gestational age, GA) and extremely preterm infants (22–27 weeks GA) born from 1 January 2010 to 31 December 2022. Infants were excluded if missing data on sex, birth weight, final discharge destination, episodes of care; died; or were discharged for ongoing care. We also excluded infants with implausible birth weight for GA z-scores >4 SD above or below the mean, admitted >24 hours after birth or born with a lethal congenital anomaly (online supplemental table 1).

Outcome measures

NNRD data on antibiotic exposure in the first 14 days are accurate.18 Antibiotic use was identified by searching the NNRD daily drugs field for character strings matching antibiotic names with intravenous preparations (online supplemental table 2), identified from the British National Formulary for Children (version September 2021).19 Prophylactic, oral and topical preparations were excluded.

We calculated several measures of antibiotic use aggregated by month of birth. The antibiotic use rate (AUR) was the aggregate percentage of inpatient days in receipt of at least one antibiotic. We calculated the percentage of infants exposed to one or more days of antibiotics initiated day 1–3 (presumed to be for EOI) and initiated day 4 or later (presumed to be for LOI), with day of birth defined as day 1. We identified the percentage of infants who received a prolonged antibiotic course (≥5 consecutive days) for EOI or LOI. LOI antibiotic prescribing was defined using a ‘washout’ period of 2 days without antibiotics to exclude a continuing EOI antibiotic course. Finally, for the first day of antibiotic prescription for EOI and LOI, we identified which antibiotic(s) were prescribed.

Statistical methods

The study period is divided into three periods: (1) before CG149 publication (January 2010–July 2012); (2) between publication of CG149 and NG195 (August 2012–March 2021); (3) after NG195 publication (April 2021–December 2022). Characteristics of the study population by period were described.

We used segmented regression to investigate the impact of publications on the outcomes.20 The regression models estimate (1) the monthly trend in each outcome in period 1; (2) any immediate absolute change in magnitude or trend in period 2 relative to period 1 and (3) any immediate absolute change in magnitude or trend in period 3 relative to period 2.

We built a parsimonious model for each outcome through backwards elimination of non-statistically significant variables, based on a significance level of 0.05. The autocorrelation function, partial autocorrelation function and Ljung-Box test demonstrated no significant autocorrelation of model residuals.

Subgroup analysis

We conducted prespecified subgroup analyses to investigate variations in impact defined by characteristics considered likely to influence antibiotic use: (1) very versus extremely preterm infants; (2) infants with versus without evidence of bacterial infection; (3) infants with versus without a recorded antibiotic indication(s) (based on presence of diagnosis codes for bacterial infection, isolation of an antibiotic-resistant organism or NEC). We identified characteristics for subgroup analyses 2 and 3 by searching for character strings in daily and episodic diagnoses records matching a list of diagnoses (online supplemental table 3). We excluded diagnoses indicating risk of infection, suspected but unconfirmed infection and infections not specific to a bacterial cause.

Type(s) of antibiotic prescribed

We plotted the most frequently prescribed antibiotic monotherapy and combinations for EOI and LOI by birth year to illustrate changes over time. We present the top seven most frequently prescribed antibiotic(s) with the remainder categorised as ‘other’.


Data were available for 97 387 infants born at 22–31 weeks GA during the study period and admitted to neonatal units in England and Wales. After exclusions (online supplemental table 4), 84 626 infants were included. Table 1 describes the population by time period.

Table 1

Description of characteristics of study population

On average, 546 infants were born each month, with relatively fewer per month in period 3. The proportion of infants first admitted to a neonatal intensive care unit increased over time but there were no clinically relevant differences in median GA, birth weight, birth weight for GA z-score and length of stay.

Table 2 describes overall use of antibiotics and indications for use. Almost all infants (96.0%, n=81 278) received at least one antibiotic during their stay. The percentage of infants who had a recorded antibiotic indication, evidence of bacterial infection or NEC declined over time, but there was little difference in the percentage with an antibiotic-resistant organism.

Table 2

Antibiotic use and indications for use by study period

Table 3 shows the results of the interrupted time series analysis, and figures 1 and 2 show the fitted regression lines from the parsimonious models.

Figure 1

Changes in the overall antibiotic use rate before and after the publication of CG149 and NG195. CG149, clinical guideline 149; NG195, NICE guideline 195.

Figure 2

Changes in the percentage of infants who received ≥1 day, and ≥1 prolonged course, of antibiotics for EOI (A and C) and LOI (B and D). CG149, clinical guideline 149; NG195, NICE guideline 195; EOI, early-onset infection; LOI, late-onset infection.

Table 3

Month-on-month absolute percentage changes (with 95% CIs) in antibiotic use before and after the publication of CG149 and NG195

The median (IQR) AUR across the study period was 19.5% (18.5–20.2) of total care days per month. In period 1, AUR was increasing by 0.03% per month. This immediately declined by 1.02% in period 2 and the trend reversed, resulting in an absolute decrease of 0.02% per month, which did not change during period 3. Over the study period, the AUR declined from 20.5% to 17.5% of total care days, which equates to approximately 1000 fewer days of antibiotic use per month (approximately 2 fewer days of antibiotics per infant) on average.

Almost all (93.4%, n=79 006) infants received antibiotics for EOI. The prevalence was stable in periods 1 and 2, though declined in period 3 by 0.20% per month. Just over half (54.0%, n=45 684) received antibiotics for LOI. Prevalence was increasing by 0.02% per month in period 1 and did not change until period 3, where it declined by 0.23% per month.

Approximately one-third (35.7%, n=30 201) of infants received a prolonged EOI antibiotic course. The prevalence dropped by 1.87% in period 2 and continued to decline by 0.08% per month thereafter. Similarly, one-third (32.9%, n=27 796) were exposed to a prolonged LOI antibiotic course. Prevalence initially increased by 0.12% per month but immediately declined by 0.02% per month in period 2, accelerating to a decline of 0.14% per month in period 3.

Subgroup analysis

Online supplemental table 5 and online supplemental figure 2 show the subgroup analyses. The number of infants born per month in some subgroups was small, resulting in greater monthly outcome variability and reduced power to detect small changes.

Extreme prematurity, bacterial infection and a recorded antibiotic indication were associated with a higher AUR and exposure to antibiotic(s) for ≥1 day, or ≥1 prolonged course, than the respective comparison subgroup.

Very premature infants observed a greater reduction in AUR, EOI antibiotic exposure and prolonged LOI exposure than extremely premature infants in period 2. During period 3, extremely premature babies observed greater reductions in AUR and EOI and LOI antibiotic exposure than very premature babies.

In period 2, the trend in AUR reduced for infants without bacterial infection, and EOI and LOI antibiotic use immediately reduced for infants with evidence of bacterial infection, but without a change in trend. Over the same period, infants without infection saw a temporary increase in EOI antibiotic use with a reduction in trend and no changes in LOI antibiotic use.

Type(s) of antibiotic prescribed

Figure 3 shows changes over time in the type of antibiotics prescribed on the first day of prescribing for EOI and LOI (full data in online supplemental table 6). For EOI, the percentage of infants receiving benzylpenicillin plus gentamicin increased from 72.3% in 2010 to 83.0% in 2014, remaining relatively stable since. Prescribing for LOI was more variable. Flucloxacillin plus gentamicin was the most frequent combination, received by 28.2% of infants, followed by cefotaxime and vancomycin (7.6%).

Figure 3

Type and combination of antibiotics prescribed on the first day of prescribing for EOI (A) and LOI (B), by year of birth. EOI, early-onset infection; LOI, late-onset infection.


We found that publication of NICE guidance in 2012 (CG149) and its update in 2021 (NG195) were associated with some reductions in antibiotic use for extremely and very preterm infants in England and Wales, resulting in more consistent antibiotic use, with defined course lengths. This may explain the temporary changes in antibiotic prescribing for LOI seen in 2012, with sustained decreasing trends only seen following the update of the guidance in 2021 when recommendations on antibiotic use for LOI were first introduced.

Choice of antibiotics for EOI has become increasingly consistent with NICE guidance. However, choice for LOI continues to vary substantially, perhaps due to omission of recommendations for LOI in 2012 and only a broad recommendation to use ‘a combination of narrow-spectrum antibiotics’ in 2021.15

This is the first national evaluation of the impact of NICE guidelines on antibiotic use in preterms, though other studies have investigated the impact on term and near-term infants.21 22 Almost all extremely and very preterm infants received antibiotics, reiterating that neonatal units are a priority setting to reduce avoidable antibiotic use.

For EOI antibiotic use in term and near-term infants, NICE now recommends an alternative approach, the KP-SRC.16 While NICE standard recommendations are largely opinion based, the KP-SRC is based on a multivariable risk prediction model using data from large cohorts of infants born at ≥35 weeks’ gestation in California, USA.16 The KP-SRC and other strategies recommended by the US American Academy of Paediatics,23 similar to NICE, identify risk factors for EOI and clinical indicators of illness and need for repeat observations. The KP-SRC has been widely adopted in the UK and observational studies show large reductions in antibiotic use and laboratory testing compared with NICE 2012 (Goel24) but there is a concern that more conservative use of antibiotics and fewer babies screened for EOI incurs a risk of missing a significant number of infants with infection.25 Use of the KP-SRC has been subjected to numerous observational studies in the UK, but there is no randomised comparison with NICE 2021, although an ongoing trial is comparing a similar approach to NICE 2021 with the KP-SRC in the Netherlands.26 It remains a strong priority to compare the approaches, to reduce antibiotic use without risk of delayed treatment leading to harm. Furthermore, the KP-SRC is designed for ≥35 weeks’ gestation and most premature infants, who are both at a higher risk of infection and disproportionately affected by antibiotic associated harm, are excluded. The only guidance in the UK for this population is the NICE recommendations, highlighting the need for developing an evidence-based approach to rationalise antibiotic use in this group.

Preterm infants often have clinical indicators of possible EOI as listed in the 2021 NICE guidance (eg, apnoea, feeding difficulty, temperature instability) due to prematurity and related comorbidities. Both iterations of NICE guidance combine recommendations for mature and preterm infants, overlooking the impact of prematurity on clinical presentation. Consequently, almost all very and extremely preterm infants receive antibiotics which may cause harm in the short term, such as increased risk of NEC, and via alterations in gut microbiome, impact long-term health of survivors of prematurity.

Interrupted time series analysis is the strongest quasi-experimental study design to evaluate the impact of national guidance over time.20 There were few substantial changes in the study population case-mix over time, reducing the likelihood of confounding. This analysis assumed an instantaneous impact of guidance on antibiotic use, though this may not reflect reality. Anticipatory and lag effects were not assessed given the absence of literature to inform a defined time period of potential effect. It is also unclear to what extent NICE guidance is consistently implemented. An attempt to assess guidance compliance in did not specifically assess antibiotic prescribing.27 Adherence to a similar guideline in the Netherlands was low,28 especially non-prescribing. Similarly, German neonatal units also found gaps between national guidance and practice.29

While our findings demonstrate a temporal association between guidance publication and antibiotic use, we cannot infer causality. This segmented regression approach assumes preintervention trends would have continued without intervention, which cannot be verified. Changes in trends may be alternatively explained by other contemporaneous changes in practice or interventions, such as local quality improvement projects.

From NNRD data, it is challenging to distinguish suspected from confirmed infection, and indications for prescribing are not recorded. Excluding infants who died, who may have had very high AURs which would have skewed the overall measure, likely under-reports the full extent of neonatal antibiotic prescribing.

Reducing avoidable antibiotic prescribing can prevent antibiotic-associated harm, antibiotic resistance and health system costs.30 Future guidance must consider the impact of prematurity on presentation and treatment to avoid antibiotic prescribing in preterm infants in situations where risk exceeds benefit. We corroborate international findings of overall reductions in antibiotic use though neonatal antibiotic use remains extremely high. It will be essential to monitor the impact of NG195 on antibiotic prescribing and apply findings to guidance development.

Data availability statement

Data may be obtained from a third party and are not publicly available. National Neonatal Research Database data may be obtained from a third party with relevant approvals.

Ethics statements

Patient consent for publication

Ethics approval

The NNRD has NHS Research Ethics Committee (REC) approval for compilation of electronic patient record data from individual neonatal units (ref: 16/LO/1930). Parents are able to withdraw consent for inclusion of their child’s data in the NNRD. Ethics approval for this study was granted by the Yorkshire & The Humber Leeds East REC (18/YH/0209), the South East Scotland REC01 (IRAS323099) and the Health Research Authority and Health and Care Research Wales (IRAS323099).


Electronic patient data recorded at participating neonatal units that collectively form the UK Neonatal Collaborative are transmitted to the Neonatal Data Analysis Unit to form the National Neonatal Research Database (NNRD). We are grateful to all the families that agreed to the inclusion of their baby’s data in the NNRD, the health professionals who recorded data and the Neonatal Data Analysis Unit team. This project is funded by the National Institute for Health and Care Research (NIHR) under its Research for Patient Benefit (RfPB) Programme (Grant Reference Number NIHR203590). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • X @DrMikePSaunders, @shaliniojha7

  • Correction notice This article has been updated since it was first published. An Orcid ID has been added to the first author's name.

  • Contributors MS designed and conceptualised the study, analysed and interpreted the data and drafted and revised the manuscript. LS designed and conceptualised the study, participated in analysis and interpretation of data, drafted and revised the manuscript, and is guarantor. SO designed and conceptualised the study, participated in interpretation of data and revised the manuscript.

  • Funding National Institute for Health and Care Research, Research for Patient Benefit Programme, NIHR203590.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.