Article Text

Original research
Validation of a paediatric sepsis screening tool to identify children with sepsis in the emergency department: a statewide prospective cohort study in Queensland, Australia
  1. Patricia Gilholm1,
  2. Kristen Gibbons1,
  3. Paula Lister1,2,
  4. Amanda Harley1,3,
  5. Adam Irwin3,4,
  6. Sainath Raman1,3,4,
  7. Michael Rice5,
  8. Luregn J Schlapbach1,6
  9. On behalf of the Queensland Statewide Sepsis Collaborative
  1. 1Child Health Research Centre, The University of Queensland, South Brisbane, Queensland, Australia
  2. 2Paediatric Critical Care Unit, Sunshine Coast University Hospital, Sunshine Coast, Queensland, Australia
  3. 3Queensland Paediatric Sepsis Program, Brisbane, Queensland, Australia
  4. 4Queensland Children's Hospital, South Brisbane, Queensland, Australia
  5. 5Clinical Excellence Queensland, Queensland Health, Brisbane, Queensland, Australia
  6. 6Department of Intensive Care and Neonatology, and Children's Research Center, University Children's Hospital Zurich, Zurich, Switzerland
  1. Correspondence to Dr Luregn J Schlapbach; l.schlapbach{at}


Objective The Surviving Sepsis Campaign guidelines recommend the implementation of systematic screening for sepsis. We aimed to validate a paediatric sepsis screening tool and derive a simplified screening tool.

Design Prospective multicentre study conducted between August 2018 and December 2019. We assessed the performance of the paediatric sepsis screening tool using stepwise multiple logistic regression analyses with 10-fold cross-validation and evaluated the final model at defined risk thresholds.

Setting Twelve emergency departments (EDs) in Queensland, Australia.

Participants 3473 children screened for sepsis, of which 523 (15.1%) were diagnosed with sepsis.

Interventions A 32-item paediatric sepsis screening tool including rapidly available information from triage, risk factors and targeted physical examination.

Primary outcome measure Senior medical officer-diagnosed sepsis combined with the administration of intravenous antibiotics in the ED.

Results The 32-item paediatric sepsis screening tool had good predictive performance (area under the receiver operating characteristic curve (AUC) 0.80, 95% CI 0.78 to 0.82). A simplified tool containing 16 of 32 criteria had comparable performance and retained an AUC of 0.80 (95% CI 0.78 to 0.82). To reach a sensitivity of 90% (95% CI 87% to 92%), the final model achieved a specificity of 51% (95% CI 49% to 53%). Sensitivity analyses using the outcomes of sepsis-associated organ dysfunction (AUC 0.84, 95% CI 0.81 to 0.87) and septic shock (AUC 0.84, 95% CI 0.81 to 0.88) confirmed the main results.

Conclusions A simplified paediatric sepsis screening tool performed well to identify children with sepsis in the ED. Implementation of sepsis screening tools may improve the timely recognition and treatment of sepsis.

  • Quality in health care

Data availability statement

No data are available. The code used to implement the models is available online ( Data inquiries should be addressed to the corresponding author. Approval of data release will be governed by the University of Queensland, and the Queensland Sepsis Collaborative.

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.


  • This study identified the screening criteria which were most predictive of paediatric sepsis and derived a simplified sepsis screening tool.

  • This study validated a paediatric sepsis screening tool used in a statewide quality improvement initiative, thereby addressing a key gap highlighted by the 2020 Surviving Sepsis Campaign.

  • Although the screening tool performed well, the average acuity of the cohort was low, consistent with contemporary paediatric cohorts in the emergency department.

  • Limitations include risk of selection bias, and the study was not designed to demonstrate patient-centred outcome benefits of applying the sepsis tool.

  • These finding require external validation, particularly in low-income and middle-income settings.


Survival of patients with sepsis strongly depends on timely recognition and initiation of treatment.1 2 In neonates and children, sepsis remains accountable for an estimated three million annual deaths, and mortality rates of those requiring admission to intensive care continue to be high.3 4 The WHO and the Global Sepsis Alliance highlight the importance of improving the early recognition of sepsis.5 The 2020 paediatric Surviving Sepsis Campaign guidelines specifically recommend that healthcare institutions implement systematic screening of children who present acutely unwell to enhance the time-critical recognition of septic shock and other sepsis-associated organ dysfunction, based on very low quality evidence.6

Although observational evidence indicates the benefit of protocolised care for paediatric sepsis,1 7–14 most quality improvement programmes do not report on the accuracy of screening for sepsis. By consequence, there is a lack of evidence on features to include in sepsis screening tools. Only a small number of—mostly single-centre studies—have reported specifically on screening for sepsis, including two-tiered approaches with electronic health record-based triggers followed by clinician assessment.9 15–17 While screening tools favour sensitivity at the expense of specificity, the implementation of poorly performing screening may fail to improve care for children with sepsis. In addition, oversensitive screening may jeopardise resource allocation to broader emergency department (ED) patient groups, and potentially favour unnecessary antibiotic usage.18 19

In the state of Queensland, Australia, a statewide Sepsis Collaborative was launched in 2018 to address the priorities stated in the National Action Plan on sepsis.20 This quality improvement programme incorporated a systematic sepsis screening tool embedded in a sepsis pathway designed specifically for children. We hypothesised that this screening tool would perform well to identify children with clinician-diagnosed sepsis. In addition, we aimed to derive a simplified screening tool for future application.


Multicentre, prospective observational cohort study including children aged less than 18 years who were screened for sepsis on the paediatric sepsis pathway in EDs participating in the Queensland Sepsis Collaborative (online supplemental material 1). Details on this sepsis quality improvement programme have been published elsewhere.21

Briefly, the Queensland Sepsis Collaborative is an statewide quality improvement programme, with the aim to improve the recognition, treatment and outcomes of patients with sepsis in Queensland.22 A dedicated paediatric sepsis pathway was designed in iterative multidisciplinary workshops incorporating evidence and information from published or online available paediatric sepsis tools.1 23 24 The pathway consisted of a screening tool, a management tool and antimicrobial stewardship guidance. Paper printed pathways were distributed to the participating EDs, and clinicians were trained to manually enter data on the pathway sheets, which were collected with the patient charts. Data were then entered by trained ED nurses at the participating sites into the electronic case report form on REDCap hosted by Clinical Excellence Queensland, with regular auditing for data quality assurance. The sepsis pathway including the screening tool was implemented across 16 EDs in the state of Queensland, Australia, of which 12 contributed data for this project.

We included children assessed between 4 August 2018 and 31 December 2019 who were screened for sepsis on the pathway. We excluded children with an interhospital transfer who were directly admitted to intensive care and children where no pathway had been used.

Paediatric sepsis screening tool overview

The screening portion of the paediatric sepsis pathway contained a total of 32 criteria, which were grouped into four sepsis screening blocks: (1) nine Sepsis Indicators (ie, triggers to screen), (2) six Sepsis Risk Factors, (3) eight Severe Illness Features and (4) nine Moderate Illness Features (online supplemental material 2). These 32 criteria were used as predictors for the analyses.

Screening was initiated if the ED clinician (doctor or nurse, at triage or during initial assessment) suspected that the child may have sepsis, and/or the child met one or more of the Sepsis Indicator criteria. During screening, the treating clinician marked on the form if each of the specified criteria were present. Children who met any of the Severe or Moderate Illness Features were escalated to senior medical officer (SMO) review.


The primary outcome was sepsis, defined as treatment for sepsis, which was operationalised by two criteria: (1) decision of the SMO that ‘sepsis is likely’ and (2) treatment with intravenous antibiotics in the ED. These criteria were chosen as they represent a pragmatic intention-to-treat cohort.25 26 Children who were screened but who did not meet both of these criteria were classified as no sepsis. If a child was missing either of the two criteria for the outcome, namely, SMO review and antibiotics, it was assumed that these procedures were not performed, and these children were coded as no sepsis.

Patient and public involvement

During this project, the Queensland Sepsis Collaborative established a paediatric sepsis family group, which was represented as well at the Steering Group, and which contributed to workshops on the screening tool design and implementation. This led to the development of the Queensland Paediatric Sepsis Programme Family Support Structure.27

Statistical analysis methods

Demographics and clinical characteristics were compared between those classified as sepsis and no sepsis. Continuous variables were not normally distributed, confirmed through visual inspection of the Q-Q plot, and as such the Wilcoxon rank-sum test was used. The χ2 test or Fisher’s exact test (for low expected cell counts) were used for categorical variables. A p<0.05 was used to denote statistical significance for these comparisons. The number and percentage of children with each criterion were reported to ascertain the prevalence of each criterion and the ϕ correlation coefficient (for dichotomous data) was calculated between all pairs of predictors to identify any predictors with high correlations that may result in multicollinearity problems. To examine the predictive performance of each of the sepsis screening blocks to identify children with sepsis, a separate multivariable logistic regression analysis for each screening block was initially undertaken and the ORs and 95% CIs were reported for each criterion within each block. The discriminative ability of each block to predict the outcome was assessed using the area under the receiver operating characteristic curve (AUC).

To create the final prediction model, all 32 predictors were initially included in a logistic regression model and were iteratively removed using backward elimination, based on the largest reduction in the Akaike information criterion (AIC).28 The final model was obtained when removing predictors no longer resulted in a reduction in the AIC. This final model was then internally validated using 10-fold cross-validation. Calibration of the predicted probabilities obtained through cross-validation was assessed by a calibration plot. The performance of the model was evaluated by the AUC, and the sensitivity, specificity, positive likelihood ratio (PLR) and negative likelihood ratio (NLR) of the model were reported at different risk thresholds, corresponding to Youden’s index, a sensitivity of 0.9 and a sensitivity of 0.85. Risk thresholds corresponding to high sensitivity rates were selected, as it is recommended that screening tests should prioritise sensitivity to reduce the false negative rate. A sensitivity analysis incorporating demographic and ED admission characteristics of gender, triage category, hospital category and admission time as adjustment variables in the final model was undertaken and evaluated by the AUC.

Given previously reported challenges in defining paediatric sepsis,29 we performed two sensitivity analyses on the final model where we tested the model’s performance when using the outcomes sepsis-associated organ dysfunction (children with sepsis as defined above, who met criteria for organ dysfunction within 24 hours of presentation to ED, as per the 2005 International Pediatric Sepsis Definition Consensus Conference30 and septic shock (children with sepsis as defined above, who met criteria for septic shock within 24 hours of presentation to ED.30

All modelling was performed using R (V.4.1.1)31 and the code used to create the models are available online ( As this is a secondary analysis of data obtained from a quality improvement initiative, an a priori sample size calculation was not undertaken; therefore, all statistical comparisons are to be interpreted with caution. 95% CIs for all ORs were reported alongside the corresponding statistic in place of p values.32 33

Reporting follows the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis standards.34



Five hundred and twenty-three (15.1%) out of 3473 children screened on the pathway during the intervention period met criteria for sepsis (table 1). The children with sepsis were younger, had lower weight, had higher priority triage categories, were more likely to be treated at the quaternary site and had higher likelihood of, meningitis, abdominal, urinary tract infection and sepsis without a source compared with children classified as no sepsis.

Table 1

Demographics and clinical characteristics of the children classified as ‘no sepsis’ and ‘sepsis”

Assessment of sepsis screening blocks

Of the 32 criteria assessed on the sepsis screening tool, the Sepsis Indicators had the highest prevalence overall, with the most prevalent criteria being history of fever and parental concern (table 2 and online supplemental efigure 1). Pairwise ϕ correlation coefficients predominantly ranged between weakly negative (ϕ=−0.25) to weakly positive (ϕ=0.25), indicating no strong relationships between the variables (online supplemental efigure 2). Eight out of 496 pairwise correlations had low-level positive correlation (0.25<ϕ<0.5) but no correlations exceeded 0.5, indicating absence of multicollinearity. When evaluating each of the sepsis screening blocks, the Severe Illness Features were the most predictive of sepsis, followed by the Sepsis Indicators, and Sepsis Risk Factors, while the Moderate Illness Features were the least predictive (table 3, figure 1, online supplemental etable 1). The performance of the 32-item tool to predict sepsis, as measured by the AUC, was 0.80 (95% CI 0.78 to 0.82). The final model also maintained a similar predictive performance when demographic and ED admission variables were included (AUC 0.81; 95% CI 0.79 to 0.83) (online supplemental etable 2) with minimal change to the effect size for each individual criterion.

Table 2

Criteria assessed on the paediatric sepsis screening tool, and the prevalence of each criteria in children with sepsis and without sepsis

Figure 1

Receiver operating characteristic (ROC) curves are shown for the models for each of four blocks contained in the sepsis screening tool, and for the final model, validated using 10-fold cross-validation.

Table 3

ORs and 95% CIs are shown for each of the models evaluating the sepsis screening blocks and the final model derived through backward elimination

Derivation of final model

Of the 32 criteria, 15 were removed through backward elimination, and one (parental/healthcare worker concern as a Moderate Illness Feature) was removed prior to analysis because it was already represented by the Sepsis Indicator criteria. The final model consisted of 16 features (table 3): healthcare worker concern, ill appearance (‘looks sick’), altered behaviour, age <3 months, reduced immune defence, chronic disease, need for supplementary oxygen to maintain saturations, respiratory distress, tachycardia, hypotension, increased lactate, altered level of consciousness measured by the Alert Voice Pain Unresponsiveness (AVPU) scale, non-blanching rash, hypothermia, prolonged capillary refill ≥3 s and cold extremities. The strongest associations (OR >2) were found for age <3 months, immunocompromised status, hypotension, increased lactate, non-blanching rash, hypothermia and cold extremities (table 3, online supplemental etable 1 and 2). The final model had better discrimination than the four separate blocks (AUC 0.80; 95% CI 0.78 to 0.82) (table 3, figure 1), and retained similar predictive performance as the full 32-item model.

Model performance and sensitivity analyses

The calibration plot displayed good agreement between predicted and observed probabilities (online supplemental efigure 3). To reach a sensitivity of 90% (95% CI 87% to 92%), the final model reached a specificity of 51% (95% CI 49% to 53%) with a PLR of 1.83 (95% CI 1.75 to 1.92) and an NLR of 0.20 (95% CI 0.15 to 0.26) (online supplemental etable 3). Sensitivity analyses on children with sepsis-associated organ dysfunction (n=210, 6.0%; AUC 0.84; 95% CI 0.81 to 0.87) and septic shock (n=146, 4.2%; AUC 0.84; 95% CI 0.81 to 0.88) confirmed the main findings (online supplemental etable 4).


In this statewide prospective study, a paediatric sepsis screening tool performed well to identify children with sepsis in the ED. We identified the most discriminative criteria available from rapid patient history and vital sign assessments and derived a more parsimonious screening tool which yielded comparable predictive performance to the full 32-item screening tool. The tool has been implemented successfully across a range of paediatric and mixed EDs, and in principle can be digitalised for sites with electronic health records. The study thereby addresses a gap highlighted by the 2020 Surviving Sepsis Campaign paediatric guidelines, relating to the lack of evidence on best approaches to screening and challenges surrounding successful implementation.6 Due to the risk of selection bias, and considering the sample size, independent validation including patient-centred endpoints will be required.

A body of observational evidence demonstrates clear benefits stemming from the implementation of protocolised care bundles for children with sepsis, such as improved process measures, shorter intensive care unit (ICU) length of stay or lower mortality.1 7–14 Yet, most studies did not attempt to untangle effects related to improved recognition from those related to timely delivery of treatment bundles. Due to the non-specific nature of presenting signs and symptoms of children with sepsis,35 screening tools must be evaluated carefully, to enable recalibration and to avoid causing trigger or alarm fatigue.

The criteria associated with sepsis included risk factors based on patient history (such as young infants or immunosuppressed patients), healthcare worker concern and features indicating likely organ dysfunction. Lactate as a rapidly available point of care laboratory value emerged as one of the strongest sepsis predictors, concurring with previous studies,36–38 in addition to respiratory, cardiovascular and neurological criteria. Our findings are supported from a secondary analysis of a large trial in African children with infection, which revealed the importance of respiratory, cardiovascular and neurological signs.39 In agreement with our findings, Romaine et al demonstrated that a modified score based on age-adjusted heart rate, respiratory rate, capillary refill and AVPU predicted ICU admission in febrile children.40 Of note, most paediatric early warning scores (PEWS) incorporate similar features to our derived model.41 42 However, the development and validation of PEWS primarily focused on sick children in ED rather than specifically on children with sepsis. Finally, the performance of our model is comparable to a weighted model to predict hypotensive septic shock in children which was derived from 2464 ED encounters, including 11.4% with shock, and which contains several similar features to our model, such as blood pressure, temperature, respiratory rate and cancer as underlying disease.43

Several limitations of this study need to be considered which limit the generalisability of the findings. First, data were only available on those patients where the screening tool was applied, and we did not have data on all ED admissions to estimate the degree of selection bias. While the indicator criteria (ie, triggers to screen) were designed to capture most children evaluated for infection, the diagnostic performance related to a clinician deciding to apply the screening tool to a patient was not assessed. Therefore, we cannot rule out that broad application of the screening tool to all comers in ED may yield substantially variable performance. In addition, we did not have data on microbiological results (such as blood cultures) in the ED population to compare against. This challenge is magnified as there is no accepted gold standard of sepsis and there is inherent variability of ‘clinician-diagnosed sepsis’.44 Second, the average acuity in the cohort was low, with only 1% admitted to intensive care and three deaths. As a result, our model may not have been powered to detect certain potentially relevant features that had low prevalence in our sample, and the study was not powered to analyse the impact on patient-centred outcomes. Third, although the primary outcome of clinician-diagnosed sepsis represents an intention-to-treat population,25 26 45 clinicians were not blinded to the sepsis tool and the use of the tool may have influenced the diagnosis. To account for paediatric sepsis coding practices,29 46 we performed sensitivity analyses which showed that the model performed well to identify children with sepsis-associated organ dysfunction, and with septic shock, respectively. Fourth, the information captured by the screening tool was static, whereas clinicians may integrate dynamic information such as progression during initial ED observation, or response to treatment in their decision-making.19 Finally, although the screening tool was tested in over 3000 paediatric patient encounters across 12 institutions, our findings require external validation. In particular, there is urgency to adapt and validate sepsis screening support tools for low-income and middle-income settings in the future.

In conclusion, the findings from this multicentre study demonstrate reasonable diagnostic performance of a systematic screening tool to identify children with sepsis in the ED. While independent validation in other cohorts is required, our findings lend support for the use of systematic screening for sepsis as recommended by the Surviving Sepsis Campaign. Future studies should investigate whether such tools assisting clinicians in ruling in or ruling out children in need of immediate treatment will result in improved patient outcomes.

Data availability statement

No data are available. The code used to implement the models is available online ( Data inquiries should be addressed to the corresponding author. Approval of data release will be governed by the University of Queensland, and the Queensland Sepsis Collaborative.

Ethics statements

Patient consent for publication

Ethics approval

Ethical approval for this study was granted by the Children’s Health Queensland Human Research Ethics Committee including waiver of individual consent (HREC/18/QRCH/167).


The authors would like to thank Queensland Health for the initiation and conduct of the Sepsis Breakthrough Collaborative, the Queensland Statewide Sepsis Collaborative authorship group, the participating Emergency Department clinicians, and the families and children involved.


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.


  • Twitter @theamandaharley

  • Collaborators Clinical Excellence Queensland: Mr Michael Rice, Director, Patient Safety and Quality Improvement Service (PSQIS) ; Prof Balasubramanian Venkatesh, Medical co-lead, Queensland Sepsis Program ; Ms Kate Weller, Manager (Queensland Paediatric Sepsis Program (QPSP), CHQ; Ms Trina Maturanec, Project Manager, Healthcare Improvement Unit (HIU) ; Ms Lyndell Redpath Manager, PSQIS); Mr Robert Seaton, Principal Data Analyst, PSQIS; Ms Donna Mason, Principal Project Officer, PSQIS; Ms Anna Bell, Principal Project Officer, PSQIS; Dr Paul Lane, Medical co-lead, Queensland Sepsis Program, Mr Naitik Mehta, Principal Project Officer, PSQIS; Mr Vikrant Kalke, Principal Project Officer, PSQIS; Mr Scott Taylor, Principal Project Officer, PSQIS; Mr Mathew Ames, Consumer representative-adults; Ms Mary Steele, Consumer representative – paediatrics. Queensland Children’s Hospital:A/Prof Luregn Schlapbach, Paediatric Intensive Care Unit (PICU) Staff Specialist; Ms Amanda Harley, Paediatric Sepsis Clinical Nurse Consultant ; Dr Adam Irwin, Infectious Disease Staff Specialist, present medical co-lead QPSP; Ms Nicolette Graham, Senior Antimicrobial Stewardship Pharmacist; Dr Fiona Thomson, Emergency Department Staff Specialist; Mr Kieran Owen, Sepsis Clinical Nurse; Ms Kirsten Garrish, Emergency Department Clinical Nurse; Ms Emma Sampson, Emergency Department Clinical Nurse; Ms Meagan O’Keefe, Senior Social Worker; Dr Sainath Raman, PICU Staff Specialist, deputy lead QPSP; A/Prof Debbie Long, PICU Nurse Researcher.Gold Coast University and Robina Hospital: Dr Shane George, Emergency Department Staff Specialist; Prof Keith Grimwood, Infectious Diseases Staff Specialist; Dr Christa Bell, Emergency Department Staff Specialist; Dr Megan King, Emergency Department Consultant; Ms Bethany Semple, Clinical Nurse; Ms Claire Adams, Clinical Nurse; Ms Josea Brown, Clinical Nurse; Ms Louise Maloney, Clinical Nurse; Mr Jack Cross, Antimicrobial Stewardship (AMS) Pharmacist; Ms Louise Caire, Nurse Educator (Paediatric Emergency); Dr Nathan Dryburgh, Senior Emergency Medicine Consultant; Dr Claire Stanford, Emergency Department Consultant; Ms Kasey Calvert, Emergency Department Nurse; Dr Nathan Watkins, Emergency Medicine Consultant.Sunshine Coast University Hospital/Nambour Hospital:A/Prof Paula Lister, Director of Paediatric Intensive Care; Dr Scott Schofield Emergency Department Staff Specialist; Dr Damian Abbott, Senior Emergency Medicine Consultant; Dr Clare Thomas, Paediatric Staff Specialist; Mr Liam De Jong, Clinical Nurse; Ms Esther Bentley, Clinical Nurse. Cairns Hospital: Dr Lambros Halkidis, Emergency Department Staff Specialist; Dr Cheryl Bird, Emergency Department Staff Specialist; Mr Matthew Smith, CNC; Ms Pia Alexander, Clinical Nurse; Ms Laura Davidson-West, Registered Nurse. Rockhampton Hospital: Dr Titiosibina Ebenezer Adegbija, Emergency Department Staff Specialist; Mr James Jenkins, Nursing Director; Dr Farana Khan, Consultant Paeditrician; Ms Alice Brandt, Emergency Department Nurse Educator; Ms Bree Walker, Director Patient Safety; Ms Andrea McLucas, Intensive Care Unit Nurse Educator. Bundaberg Hospital: Dr Adam Philip Michael, Emergency Department Staff SpecialistDr Yulia Sugeng, Emergency Department Staff Specialist; Ms Mirandah Crossett, Senior Clinical Pharmacist; Ms Hannah Clune-Purcell, Senior Clinical Pharmacist; Dr Terry George, Director of Emergency; Ms Samantha Hoole, Emergency Department Nurse Educator; Ms Candice Bauer, Registered Nurse; Ms Moya Zunker, Facilitator – Clinical Governance. Redland Hospital: Dr John Sutherland, Director of Emergency Medicine; Dr Douglas Gordon Thomas, Director of Paediatrics; Dr David Van der Walt, Emergency Department Staff Specialist; Ms Jessica Hulme, Clinical Facilitator; Ms Kerrie Burke, Clinical Nurse Consultant, Patient Safety. Redcliffe Hospital: Ms Helena Cooney, Sepsis Clinical Nurse Consultant; Dr Doug Morel, Emergency Department Staff Specialist; Ms Louise O’Riordan, Director Safety and Quality; Ms Sophie Paviour, AMS Pharmacist. Ipswich Hospital: Dr Samantha Fairless, Emergency Department Staff Specialist; Dr Ian Brandon, Senior Medical Officer; Ms Megan Bool, Clinical Nurse Consultant; Ms Victoria Bates, Quality Co-ordinator; Ms Amy Fryer, CNC – Dept Emergency Medicine; Ms Rachael Wiedman, Antimicrobial Stewardship Pharmacist. Logan Hospital: Dr Nandini Choudary, Emergency Department Staff Specialist; Dr Shalini Arora, Emergency Department Staff Specialist; Dr Ben Lawton, Emergency Department Staff Specialist; Ms Jo Farrell, Clinical Nurse Consultant. Hervey Bay Hospital: Dr Penelope Prasad, Emergency Department Staff Specialist; Dr Rudesh Prasad, Emergency Department Staff Specialist; Dr Chukwuemeka Nwufoh, Emergency Department Clinical Director; Dr Shamin Family, Emergency Department Senior Medical Officer (SMO); Dr Peter Stevenson, Paediatrician; Ms Amy Kim, Pharmacist; Dr Amy Wain, Emergency Department SMO; Ms Laura O’Connor, Registered Nurse; Mr Timothy Butters, Registered Nurse. The Prince Charles Hospital: Mr Peter Kennedy, Clinical Nurse; Dr Hanh Pham, Emergency Department Staff Specialist; Dr Rajeev Jarugula, Staff Specialist Emergency Medicine; Dr Suzanne Royle, Paediatrician; Ms Tanya Mountford, Paediatric Clinical Nurse Consultant. Caboolture Hospital: Dr Maya Aoude, Emergency Department Staff Specialist; Dr Miron Kazi, Emergency Department Staff Specialist; Dr Bautista Morales, SMO Paediatrics; Ms Sara Blundell, Emergency Department Clinical Nurse; Ms Natasha Willmett, Emergency Department Clinical Nurse; Mr Dion Zunker, AMS Pharmacist; Ms Nicola Farrell, AMS Pharmacist. Mackay Hospital: Dr Frans Pretorius, Clinical Director of Surgery; Dr Kerri Winstanley, Emergency Department Paediatric SMO; Ms Kathleen Cox, Emergency Department Pharmacist; Ms Louise McGrath, Nurse Unit Manager; Ms Karen Smith, Clinical Nurse Consultant. The University of Queensland: A/Prof Kristen Gibbons, Senior Epidemiologist, Child Health Research Centre, Faculty of Medicine; Dr Patricia Gilholm, Data Scientist, Child Health Research Centre, Faculty of Medicine; Mr Endrias Ergetu, Data Analyst, Child Health Research Centre, Faculty of Medicine; Ms; Rachel Treadwell, Student Nurse/Midwifery; Ms Tahlia Van Raders, Student Nurse/Midwifery; Ms Jessicah Minogue, Master of Philosophy Student.

  • Contributors PG and LJS designed the study, performed the main analyses, wrote the first draft and finalised the manuscript. LJS accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish. KG contributed to study design, supervised all analyses, and contributed to the writing of the manuscript. PL, AH, AI, SR and MR contributed to study design, oversaw data collection, contributed to interpretation of analyses and contributed to writing of the manuscript.

  • Funding This study was supported by a grant from the Children’s Hospital Foundation Brisbane, Australia (LJS). LJS was supported by a Practitioner Fellowship from the National Health and Medical Research Council of Australia (NHMRC) and by a grant from the Children’s Hospital Foundation. AI is supported by an NHMRC Investigator Fellowship (APP1197743). The Queensland Sepsis Collaborative was funded by Clinical Excellence Queensland.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

  • 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.