Article Text

Implementation of an adapted Sepsis Risk Calculator algorithm to reduce antibiotic usage in the management of early onset neonatal sepsis: a multicentre initiative in Wales, UK
  1. Nitin Goel1,
  2. Stephanie Cannell2,
  3. Gemma Davies2,
  4. Murali Sridhar Natti3,
  5. Vickness Kirupaalar4,
  6. Artur Abelian5,
  7. Shakir Saeed6,
  8. Rhian Smith7,
  9. Ravi Manikonda8,
  10. Prem Kumar Pitchaikani9,
  11. Dawn Davies10,
  12. Rachel May Morris1,
  13. Lynsey Edwards1,
  14. Roopashree Govindaraju3,
  15. Kate Creese4,
  16. Jane Jones5,
  17. Jalil Choudhary6,
  18. Sarah Rowley1,
  19. Chidambaram Sethuraman3,
  20. Helen Muxworthy11,
  21. Felicity Curtis11,
  22. Patricia Donnelly4,
  23. Manohar Joishy6,
  24. Ian Barnard7,
  25. Celyn Kenny1,
  26. Rajarshi Pal9,
  27. Karen Jones9,
  28. Sujoy Banerjee2
  1. 1 Department of Neonatal Medicine, University Hospital of Wales, Cardiff, UK
  2. 2 Department of Neonatal Medicine, Singleton Hospital, Swansea, UK
  3. 3 Department of Neonatal Medicine, Royal Gwent Hospital, Newport, UK
  4. 4 Department of Paediatrics, Princess of Wales Hospital, Bridgend, UK
  5. 5 Department of Paediatrics, Wrexham Maelor Hospital, Wrexham, UK
  6. 6 Department of Paediatrics, Ysbyty Gwynedd, Bangor, Gwynedd, UK
  7. 7 Department of Neonatal Medicine, Glan Clwyd Hospital, Rhyl, Denbighshire, UK
  8. 8 Department of Paediatrics, Nevill Hall Hospital, Abergavenny, UK
  9. 9 Department of Paediatrics, Glangwili General Hospital, Carmarthen, Carmarthenshire, UK
  10. 10 Department of Paediatrics, Bronglais General Hospital, Aberystwyth, Ceredigion, UK
  11. 11 Department of Women's Health, Singleton Hospital, Swansea, UK
  1. Correspondence to Dr Sujoy Banerjee, Department of Neonatal Medicine, Singleton Hospital, Swansea, SA2 8QA, UK; sujoy.banerjee{at}


Objective Assess the impact of introducing a consensus guideline incorporating an adapted Sepsis Risk Calculator (SRC) algorithm, in the management of early onset neonatal sepsis (EONS), on antibiotic usage and patient safety.

Design Multicentre prospective study

Setting Ten perinatal hospitals in Wales, UK.

Patients All live births ≥34 weeks’ gestation over a 12-month period (April 2019–March 2020) compared with infants in the preceding 15-month period (January 2018–March 2019) as a baseline.

Methods The consensus guideline was introduced in clinical practice on 1 April 2019. It incorporated a modified SRC algorithm, enhanced in-hospital surveillance, ongoing quality assurance, standardised staff training and parent education. The main outcome measure was antibiotic usage/1000 live births, balancing this with analysis of harm from delayed diagnosis and treatment, disease severity and readmissions from true sepsis. Outcome measures were analysed using statistical process control charts.

Main outcome measures Proportion of antibiotic use in infants ≥34 weeks’ gestation.

Results 4304 (14.3%) of the 30 105 live-born infants received antibiotics in the baseline period compared with 1917 (7.7%) of 24 749 infants in the intervention period (45.5% mean reduction). All 19 infants with culture-positive sepsis in the postimplementation phase were identified and treated appropriately. There were no increases in sepsis-related neonatal unit admissions, disease morbidity and late readmissions.

Conclusions This multicentre study provides evidence that a judicious adaptation of the SRC incorporating enhanced surveillance can be safely introduced in the National Health Service and is effective in reducing antibiotic use for EONS without increasing morbidity and mortality.

  • neonatology
  • health services research

Data availability statement

Data are available on reasonable request.

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.

What is already known on this topic?

  • Approximately 16% of term and near-term infants are treated with prophylactic antibiotics for the risk of early onset neonatal sepsis (EONS).

  • Perinatal use of antibiotics has been associated with microbiome modulation and long-term adverse health outcomes.

  • Introduction of the Sepsis Risk Calculator for early onset neonatal sepsis management in centres in the USA has safely reduced antibiotic treatment in term or near-term babies.

What this study adds?

  • Judicious adaptation of the Sepsis Risk Calculator (SRC) incorporating enhanced surveillance can be safely introduced in the National Health Service (NHS) for management of EONS.

  • SRC application in this multicentre NHS setting reduced antibiotic use for EONS management by nearly half, saving healthcare resources.

  • Mandated period of in-hospital observation, standardised staff training and parent awareness of sepsis are essential to the success and safety of SRC in the UK.


Early onset neonatal sepsis (EONS) in term and near-term infants is infrequent (0.5–1/1000 live births) but can be associated with high morbidity and mortality.1–4 The challenge is to identify and treat infants destined to develop true disease while avoiding invasive treatment in others. Antibiotics in early life disrupts the development of a healthy microbiome and innate immunity.5–7 Associations with lifelong health issues such as childhood obesity, allergies, type 2 diabetes mellitus and inflammatory bowel disease compel professionals to consider the optimum balance between effective treatment and avoidable interventions.8–11

The National Institute of Clinical Excellence (NICE) guidance CG149 influences management of EONS in the UK. Using a range of categorical perinatal risk factors, it identifies infants that should receive prophylactic antibiotics at birth.12 However, perinatal risk factors are common; most infants do not develop infection, and laboratory investigations have poor predictive value for true sepsis.13 14 Moreover, 50% of EONS are symptomatic at birth and 80% at the time of decision to treat.15 Following NICE guidelines is reported to treat 16% of all term and near-term newborns.16

The Kaiser Permanente Sepsis Risk Calculator (SRC) applies a Bayesian approach and multivariable modelling to predict individualised risk of EONS.17–19 In the USA, introduction of SRC is reported to reduce antibiotic use by 40%–50% without missing true EONS.17 20 21 The American Academy of Pediatrics has recently endorsed SRC for EONS management.22 Virtual comparisons of NICE and SRC for EONS management in the UK population estimate a potential 74%–88% reduction in antibiotic use, without missing additional culture-positive EONS.16 23

It is prudent to exercise caution, as the certainty of ‘no harm’ in low-incidence disease is difficult to establish in small studies. The differences in healthcare practices between USA and the UK, for example, length of hospital stay, frequency of newborn observations and staffing levels, may be significant factors affecting safety. It has been suggested that introduction of the SRC into UK clinical practice should incorporate safeguards of enhanced in-hospital observation, staff and parent education, and inbuilt quality assurance tools.15 23 24


This study aims to assess the impact of introducing a consensus clinical guideline incorporating an adapted SRC algorithm for EONS management in Wales, UK, on antibiotic usage and patient safety.


Following a review of studies undertaken in this region, the clinicians of the Wales Maternity and Neonatal network unanimously agreed to implement a new consensus guideline incorporating SRC for EONS management of term and near-term infants to facilitate prudent antibiotic stewardship.15 16 With this objective, a multidisciplinary ‘Task and Finish’ group (doctors, nurses, midwives, advanced nurse practitioners and managers) was formed. The group was tasked to undertake evidence reviews, process mapping, brainstorming and deliberation exercises to safely adapt and implement this guidance in Welsh National Health Service (NHS) Hospitals.17 20 22 25

A multifaceted programme was developed including five core elements:

  • a pragmatic evidence-based consensus SRC guideline;

  • mandated risk-stratified 24–36 hours observation period for infants in whom antibiotics were withheld by SRC;

  • standardised training for staff;

  • increased parent awareness;

  • data-driven quality assurance tools and culture of improvement.

Wales has an average annual birth rate of approximately 32 150/annum including births in hospital, midwife-led birthing units (MLUs) and home.26 The 10 participating centres included three neonatal intensive care units (NICUs), a subregional NICU and six special care units. Prior to April 2019, all centres used NICE guideline.12 MLU and home born infants were exempt from the SRC pathway (predominantly low-risk and MLU early discharge policy conflicted with mandated observation), unless referred to hospital for complications or specialist review.

The study was a service evaluation of the introduction of an evidence-based guideline, ratified as standard of care by Wales Maternity & Neonatal Network Board, specialist commissioners and the Welsh Risk Pool; its governance was mandated by a Welsh Government Health Circular, specifically for application in NHS Wales. This was backed by a rigorous quality assurance process and analysis of anonymised datasets. Based on the above information, the Health Research Council decision tool ( did not consider this study to require ethical approval. The project commenced on first April 2019 with formal evaluation planned for 12 months.

The details of the clinical guideline including observation schedule is available in online supplemental appendix 1, and key elements are highlighted in figure 1. Based on reported high concordance, infants in whom NICE did not recommend observation or antibiotics at birth were not put through SRC and assigned to normal care (no change from usual practice).16 All other infants were reviewed by the neonatal team. The clinicians applied SRC to calculate the infant’s prior Sepsis Risk Score (SRS) at birth and determined ‘clinical status’ through physical examination.27 The SRS and the clinical status were entered in a table with a traffic light system of graded risk used to guide management, similar to the protocol used in the study by Dhudasia et al.25 A background EONS incidence of 0.5/1000 live births was used for SRC computation, the closest population estimate from high-income countries including UK.2 3 15 Asymptomatic infants with prior SRS <0.65 received in-hospital observation for 24 hours using Newborn Early Warning Trigger and Track chart (green).28 The observation period extended to 36 hours for medium-risk asymptomatic infants (SRS 0.65–1.54) or for low-risk (SRS <0.65) infants who developed transient symptoms (amber). All clinically unwell infants or asymptomatic infants with SRS >1.54 and those where SRC recommended blood culture (Wales clinicians’ consensus) were classed as high risk and treated with antibiotics (red). Prophylactic antibiotics were recommended for any infant whose previous sibling had a confirmed group B streptococcus (GBS) sepsis, if their mother had inadequate intrapartum prophylaxis.

Figure 1

Flow chart showing the stepwise approach for management of EONS at birth using the adapted Wales SRC algorithm. EONS, early onset neonatal sepsis; SRC, Sepsis Risk Calculator.

Training was provided centrally and then cascaded locally to neonatal and midwifery staff, using a standardised training slide set. A closed social media networking group allowed quick troubleshooting and shared learning.

Parent support groups collaborated to develop information leaflets for families in English and Welsh languages. The governance, data collection and audit tools were embedded within the programme (online supplemental appendix 2) and its fidelity ensured by designated centre lead.

The primary outcome was number of infants≥34 weeks gestation on antibiotics within 72 hours of birth as a proportion of live births. Other balancing and process measures are described in table 1.

Table 1

Summary of outcome measures

Prior to this initiative, antibiotic usage data were not reliably captured by routine clinical systems. To determine a robust baseline incidence of antibiotic usage, data on all blood culture requests from postnatal wards, delivery suites and neonatal units (NNUs) within 72 hours of birth were collated from microbiology databases for a 15-month period (January 2018–March 2019) by each participating centre. As the prevailing practice in Wales was to treat all neonates investigated with blood culture to receive antibiotics and vice versa, blood culture data (after removing duplicates and infants <34 weeks’ gestation) served as a proxy for antibiotic use. Validation was undertaken using a random subset of 100 blood culture samples common to an earlier study in the region, matching this to actual antibiotic usage data collected for that study and vice versa (sensitivity and specificity 99%, 95% CI 94.5 to 99.7).16

The site teams collected predefined datasets including birth statistics, number of observations and antibiotic use on the SRC pathway, culture-positive sepsis, NNU admissions, serious illness (needing both respiratory and cardiovascular support) and sepsis-related readmissions within a week of birth. Anonymous data uploaded to a secure NHS SharePoint portal was analysed fortnightly and presented as Run Charts and Shewhart Charts using QI Chart software V.2.023 (online supplemental appendix 3).29 Categorical data are presented as proportions with mean and 95% CI and differences between groups analysed using non-parametric χ2 test; p value <0.05 was considered statistically significant. Sensitivity analysis was undertaken to assess the impact of missing data on outcomes. Monthly governance meeting assessed data completion and safety issues while sharing learning points and innovative practices. Process measures were incomplete for the last 2 months due to emerging COVID-19 pandemic.

The study was a service evaluation of the introduction of an evidence-based guideline, ratified as standard of care by professional bodies in Wales, specifically for application in NHS Wales. This was backed by a rigorous quality assurance process and analysis of anonymised datasets. Based on the above information, the Health Research Council decision tool ( did not consider this study to require ethical approval.


During the 15-month baseline period, 30 105 infants were born at gestation ≥34 weeks in the 10 participating hospitals. Of these, 4304 infants (14.3%) were identified to have undergone at least one blood culture within 72 hours of birth and served as proxy for antibiotic use. In contrast, 1917 (7.7%) of the 24 749 infants received antibiotics in the 12-month SRC implementation period. This represents a sustained mean reduction of 45.5% in antibiotic use (figure 2). For every 1000 infants on the programme, 65 avoided unnecessary exposure to antibiotics. A centre-wise breakdown of antibiotic usage before and after SRC implementation is shown in table 2 and illustrated in online supplemental figure 1a–c.

Figure 2

P chart: proportion of live births ≥34 weeks who received antibiotics in Wales during baseline and SRC implementation phase. SRC, Sepsis Risk Calculator.

Table 2

Distribution of live births, infants on antibiotics and culture positive sepsis by individual centres

Direct antibiotic usage data with recorded blood cultures were retrospectively validated in a consecutive sample of 565 infants in the implementation period (29.5% of all infants on antibiotics) from two different centres (1 and 5). All infants on antibiotics had blood culture recorded, while six infants with blood cultures did not receive antibiotics (sensitivity 99%, 95% CI 97.7 to 99.6; specificity 100%, 95% CI 99.9 to 100).

Excluding contaminants, 19 infants grew pathogenic organism in their blood cultures in the implementation period, compared with 21 in the baseline period. This corresponds to an EONS incidence rate of 0.69/1000 livebirths ≥34 weeks’ GA in the baseline period and 0.76/1000 in the implementation period for hospital-based births. For each confirmed EONS, table 3 provides demographic details, clinical presentation and management recommendations by NICE and Wales SRC pathway. There were no deaths and all confirmed cases were discharged home in good health. None of the confirmed cases were born in MLU.

Table 3

Individual characteristics of culture positive cases

Of these 19, 10 received antibiotics at birth as per Wales SRC pathway (seven symptomatic at birth and three with high SRS score). If compared, NICE would have recommended antibiotics to eight and observation for the other two infants. Other nine infants were treated when symptomatic: two were without risk factors and another infant would have completed recommended observation period on both systems. The remaining six infants each underwent 24-hour observation on the pathway and would have undergone 12-hour observation on NICE. However, three of these six infants would have been symptomatic only after completing their 12-hourobservation on NICE (symptomatic at 17, 22 and 24 hours). Overall, 17 of the 19 confirmed EONS infants (89.4%) became symptomatic during their hospital stay, 14 (73.7%) with respiratory symptoms. Two infants required multiorgan support (ventilation and inotropes) from sepsis: one was symptomatic from birth and the other had no identifiable risk factors on either SRC or NICE. No special cause variation was seen in ‘days between culture-positive sepsis’ in the baseline or implementation period (figure 3).

Figure 3

T and C charts showing stable balancing measures during the implementation period.

There was no change in NNU admissions for infants 34 weeks’ gestation before and during the intervention period (figure 3). There were two readmissions with culture-positive sepsis to children’s ward within 7 days of birth, but neither infant had any risk factors of sepsis (NICE or SRC). Process measures remained stable during the intervention period (online supplemental figure 2).

The only missing data was a consecutive 9-month period of baseline from centre 5. Sensitivity analysis excluding centre 5 showed no statistically significant difference in the baseline incidence (p=0.65).


This is the first report of a prospective multicentre application of the SRC outside USA. It demonstrates a large-scale reduction in antibiotic use across all centres in Wales, consistent with findings from other prospective studies.17 30–33 The effect size was consistent in large tertiary centres, while variations were more evident in smaller centres. It is unclear if the variations noted in these centres were secondary to the risk profile of the centre population, smaller birth cohort or variability in the interpretation and application of the new guidance.

We used uniform EONS incidence of 0.5/1000 on SRC for computation of all cases. Suggestions of using disease-specific incidence rates, for example, chorioamnionitis, into the calculator to reflect the higher risk is flawed and cannot be recommended as extrapolation to subpopulations in regression models significantly overestimates risk, particularly if the model contains covariates of the same condition,for example, fever.34–36

All confirmed EONS cases were appropriately managed by the study pathway with no additional risks of missed cases or delayed diagnosis. The ‘days between culture-positive sepsis’ remained stable throughout, refuting concerns that avoiding prophylactic antibiotics may increase the number of culture-positive sepsis.

Comparative retrospective analyses of confirmed EONS cases have recently reported lower sensitivity of SRC to NICE in recommending prophylactic antibiotics during asymptomatic phase.15 24 While the reported difference is a legitimate concern, the very low EONS incidence translates into a handful of cases spanning several years. Prophylactic antibiotic treatment may not be most appropriate strategy to capture these infants as it results in several thousand unnecessary treatments of others.15 16 We have argued that such cases are not necessarily ‘missed’ and can be safely identified and treated (similar to 44% of EONS without risk factors) through a system of high-quality observation. Prospective studies in healthcare settings where similar in-hospital surveillance is in place have not shown any evidence of harm.17 30–33 Prudent healthcare programmes balance benefits with risk and cost, and therefore sensitivity alone (without measures of specificity) cannot determine the effectiveness of an intervention. Our results add to the pool of prospective data confirming that SRC can be safely implemented when backed by such surveillance programme.

Inappropriate antibiotic use in this critical phase of life is not without consequences,5–11 and recent papers suggest a significant cost–benefit of SRC over traditional guidelines.37 38 A 24–36 hours observation programme instead of presumptive antibiotic treatment (36–48 hours stay) means avoidance of antibiotics exposure annually for more than 2000 babies in Wales without prolonging hospital stay. In our study, nearly 90% of EONS cases became symptomatic within 24 hours. Extending 24 hours of observation to all babies (to capture EONS without risk factors) may not be practical, cost-effective or acceptable to families. As in our programme, the prudent way to capture EONS in infants without risk factors would be to raise family and staff awareness of sepsis symptoms through education, parent information leaflets and standardised staff training.

Our cautious approach followed Dudhasia et al 25 protocol in reducing threshold for presumptive antibiotics at birth to a prior SRS of 1.5/1000 (three times the population risk) compared with 3/1000 in the original SRC. If original SRC recommendations were followed, only four infants with sepsis might experience a short delay in introducing antibiotics, while their symptoms were being evaluated for possible sepsis. Other regions contemplating SRC implementation programme may wish to consider these findings in choosing their SRS threshold and observation protocol.

Our findings have limitations. Blood culture data served as proxy for antibiotic use in the baseline period. Prevailing clinical practice in the UK to treat infants with antibiotics if they meet the threshold for blood culture and our validation data both in baseline and implementation period provides assurance that the proxy representation is accurate. Similar incidence is also reported from other studies.16 39 The limitation of routine data infrastructure made it difficult to collect demographics, length of stay and feeding data on large number of infants not on the SRC pathway. We could not exclude the rare possibility of EONS readmissions outside Wales.

Our collaborative project was undertaken without dedicated funding and still demonstrated a high degree of fidelity in its application. The guidance is now ingrained in routine clinical practice in Wales. Other regions considering similar initiative may want to seek dedicated funding and resources for implementation and sustainability.


Through network-wide multidisciplinary collaboration, a judicious adaptation of the SRC reduced antibiotic use for management of EONS across Wales by nearly half, without adverse consequences and potentially saving healthcare resources. A mandated period of in-hospital observation, standardised staff training and parent awareness of sepsis were key to the success and safety of the programme.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The study was a service evaluation of the introduction of an evidence-based guideline, ratified as standard of care by Wales Maternity & Neonatal Network Board, specialist commissioners and the Welsh Risk Pool; its governance was mandated by a Welsh Government Health Circular, specifically for application in NHS Wales. This was backed by a rigorous quality assurance process and analysis of anonymised datasets. Based on the above information, the Health Research Council decision tool ( did not consider this study to require ethical approval.


We would like to thank the Wales Maternity and Neonatal Network especially Alan Hather, Amy Stabler, Elizabeth Gallagher and Claire Richards for logistic support. We would like to thank all the junior doctors, advanced neonatal nurse practitioners, nursery nurses, midwives and neonatal nurses in Wales for enthusiastically adapting their clinical practice and collecting data for the Quality Improvement (QI) and safety monitoring. We would like to thank Dr Jean Matthes for her support and wisdom, Professor Karen Puopolo for being an external advisor during the design phase of the project and for sharing the clinical guideline used in CHOPS hospitals and Dr Alan Willson for his guidance with the QI methodology. Finally, we would like to acknowledge the input from parent support groups in Cardiff and Swansea in offering their insight and help in designing the parent information leaflet.


Supplementary materials


  • Contributors SB and NG conceived this project, authored the guideline, devised training package and data collection tools, analysed data and produced the first draft of the manuscript for publication. SC, GD, MN, VK, AA, SS, RS, RM, PKP, DD, RM, LE, RG, KC, JJ, JC, SR, CS, HM, FC, PC, MJ, IB, CK, RP and KJ contributed to the development of guideline, information leaflets, led local training and audit, and collected and submitted data for analysis. All authors critically appraised and approved the final submitted version of the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • 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.

Linked Articles

  • Fantoms
    Ben J Stenson