Article Text


Utility of a single early warning score in patients with sepsis in the emergency department
  1. Alasdair R Corfield1,
  2. Fiona Lees2,
  3. Ian Zealley3,
  4. Gordon Houston4,
  5. Sarah Dickie5,
  6. Kirsty Ward2,
  7. Crawford McGuffie5,
  8. on behalf of the Scottish Trauma Audit Group Sepsis Steering Group
  1. 1Emergency Department, Royal Alexandra Hospital, Paisley, UK
  2. 2Information Services Division, National Services Scotland, Glasgow, Scotland, UK
  3. 3Radiology Department, Ninewells Hospital, Dundee, Scotland, UK
  4. 4Anaesthetic Department, Crosshouse Hospital, Kilmarnock, Scotland, UK
  5. 5Emergency Department, Crosshouse Hospital, Kilmarnock, Scotland, UK
  1. Correspondence to Dr Alasdair R Corfield, Emergency Department, Royal Alexandra Hospital, Paisley PA2 9PN, UK; a.corfield{at}


Background An important element in improving the care of patients with sepsis is early identification and early intervention. Early warning score (EWS) systems allow earlier identification of physiological deterioration. A standardised national EWS (NEWS) has been proposed for use across the National Health Service in the UK.

Aim To determine whether a single NEWS on emergency department (ED) arrival is a predictor of outcome, either in-hospital death within 30 days or intensive care unit (ICU) admission within 2 days, in patients with sepsis.

Methods Data were collected over a 3-month period as part of a national audit in 20 EDs in Scotland. All adult patients who were admitted for at least 2 days or who died within 2 days were screened for sepsis criteria. Patients with  systemic inflammatory response syndrome criteria were included. An EWS was calculated based on initial physiological observations made in the ED using the NEWS.

Results Complete data were available for 2003 patients. Each rise in NEWS category was associated with an increased risk of mortality when compared to the lowest category (5–6: OR 1.95, 95% CI 1.21 to 3.14), (7–8: OR 2.26, 95% CI 1.42 to 3.61), (9–20: OR 5.64, 95% CI 3.70 to 8.60). This was also the case for the combined outcome (ICU and/or mortality).

Conclusions An increased NEWS on arrival at ED is associated with higher odds of adverse outcome among patients with sepsis. The use of NEWS could facilitate patient pathways to ensure triage to a high acuity area of the ED and senior clinician involvement at an early stage.

  • Resuscitation, Clinical Care
  • Triage
  • Death/Mortality
  • Emergency Department
  • Intensive Care

Statistics from


Sepsis is defined as a systemic inflammatory response syndrome, provoked by an infection.1 ,2 Sepsis continues to be a major cause of morbidity and mortality in developed and developing countries,3 ,4 and is a significant burden on healthcare systems in these countries.5 An important element in improving the care of patients with sepsis is early identification and early intervention, which has been shown to improve outcomes.6 ,7

Physiological deterioration often precedes clinical deterioration as patients develop critical illness.8 ,9 Recognition of this led to the development of the concept of the patient at risk and medical emergency teams.10–12 Early warning score (EWS) systems were then further developed to allow earlier identification of physiological deterioration. By assigning numerical values to various physiological parameters, a composite score can be assigned to a patient, allowing early identification of patients who are at risk of critical illness. EWS were initially established to assist in the management of patients in the general ward setting.13 ,14 More recently there has been interest in the use of EWS in the emergency department (ED) with reports from single centres,15 ,16 although their use in the ED remains controversial.17 The use of a standard single national EWS (NEWS) across the National Health Service (NHS) in the UK has been recommended to improve patient care.18

In this study, we aimed to evaluate an EWS in a national cohort of patients with sepsis presenting to EDs, to determine whether a single EWS in the ED was a useful predictor of outcome, either death or intensive care unit (ICU) admission.


Study population

Data were collected over a 3-month period between March and May 2009 as part of the Scottish Trauma Audit Group (STAG) Sepsis Audit. Twenty of the 25 mainland district general and teaching hospital EDs in Scotland participated in the audit (Appendix 1).

Adult patients (>16 years) attending as an emergency were identified prospectively from ED or admission unit records. Hospital information systems were then interrogated to ascertain whether the patient had an inpatient stay of at least 2 days. Patients who died within the first 2 days, and who therefore may have been omitted from data collection, were identified retrospectively using General Register Office Scotland records. Patients who had an obviously non-infective cause for attendance such as acute cardiac ischaemia, trauma or stroke were excluded. This process identified 27 046 patients who required case note review in order to determine the presence of ‘sepsis’ criteria.

The Surviving Sepsis Campaign Guidelines6 defined sepsis as ‘infection, documented or suspected’ and the presence of some of a number of general variables indicating physiological derangement. Patients were included in the audit if they had: (a) a suspicion or confirmation of infection within 2 days of attendance; and (b) two or more of the following physiological derangements: temperature >38.3°C or <36°C; heart rate >90 bpm; respiratory rate >20/min; white cell count of >12 000/μl or <4000/μl or >10% immature forms; acutely altered mental status; systolic blood pressure <90 mm Hg; and blood glucose >7.7 mmol/l (in the absence of diabetes).

Data collection

A total of 5285 patients fulfilled the entry criteria. Data were collected retrospectively by local audit coordinators at each hospital on a variety of demographic, physiological, process and outcome variables using a standardised proforma. Where available, patient observations taken on attendance were recorded. All patients were followed to discharge or death. The difficulties encountered in obtaining and extracting data from case notes were such that a pragmatic decision was taken to stop data collection in July 2010.

Of the 5285 patients identified, complete data were collected for 3890 (74%). Age, gender, length of stay, critical care attendance and outcome were recorded for all patients who met the sepsis criteria (N=5285). For each of these variables, the sample population (N=3890) was representative.19 For the purposes of this analysis, only patients who presented with or developed signs of sepsis prior to leaving the ED were included (N=2489). In this sample of 2489 patients, patients were excluded if they did not have a full set of observations made as part of their first set of observations. This resulted in a final sample size of 2003 patients (figure 1)

Figure 1

Patient study inclusion pathway. ED, emergency department.

Definition of NEWS

The NEWS contains six physiological parameters (table 1), each of which is assigned a value of between 0 and 3 along with an additional parameter for supplemental oxygen, which scores 0 or 2. The score for each of the seven parameters is summed to calculate the NEWS which may range between 0 and 20; the higher the score, the greater the deviation from normality. The NEWS is based on previous similar scoring systems.11 ,12 ,15

Table 1

National early warning score

Observations taken on attendance were used to calculate the NEWS. For some analyses patients were divided into four categories based on their total score: 0–4, 5–6, 7–8, 9–20. This analysis grouping was based on the distribution of NEWS scores to give four approximately equal sized groups for comparison.

In order to assess the effect of age on all endpoints an age adjusted NEWS was also calculated (+0 points for <50 years, +2 points for 50–70 years, +3 points for >70 years).20

The NEWS uses an AVPU (alert, voice, pain, unresponsive) score to define the patient's level of consciousness. Where an AVPU score was unavailable, the Glasgow Coma Score (GCS) was considered to be an acceptable alternative (GCS 15=A, GCS <15=V,P,U).


Primary endpoints were ICU admission within 2 days of attendance and 30-day mortality (in hospital). A combined endpoint of ICU admission and/or mortality was also assessed.

Statistical analysis

All analyses were carried out using SPSS V.17.0 for MS Windows.

Differences between medians were tested using the Mann–Whitney U test. ORs for each endpoint were estimated using logistic regression, with NEWS group as the independent variable and age as a continuous covariate. Receiver operating characteristic (ROC) curves plotting sensitivity (true positives) against 1-specificity (false positives) were used to measure the accuracy of NEWS and age adjusted NEWS in predicting endpoint. The associated ROC area under the curve was also calculated for each endpoint.

Statistical significance was set at p<0.05 for all analyses. All results are reported along with their associated 95% CIs and p values.

All data were anonymised prior to submission to the STAG central team at the Information Services Division (ISD). The Caldicott Guardian for each participating Health Board was informed of the audit and the processes involved. ISD, as part of NHS National Services Scotland, has adopted a policy to protect personal information. Personal health information is held securely, and managed according to data protection regulations.


A total of 2003 patients were available for analysis; 949 (47%) were male and 1054 (53%) female. The median age of patients was 72 years, with no significant difference in age between men and women. The median NEWS for all patients was 7; there was no significant difference between men and women. The distribution of NEWS recorded for patients on attendance is illustrated in online supplementary figure S3.

Differences in age and NEWS for each endpoint are shown in table 2. Patients who were admitted to the ICU within 2 days of attendance had a median age of 61 and were significantly younger than those who were not (61 vs 72, p<0.05). ICU patients also had a significantly higher NEWS than the non-ICU group (9 vs 6, p<0.05). Patients who died within 30 days were significantly older than those who did not (77 vs 70, p<0.05) and had a higher NEWS (9 vs 6, p<0.05).

Table 2

Median age and NEWS by sex, admission to ICU within 2 days, 30-day outcome and combined endpoint (ICU and/or mortality)

Absolute numbers in each NEWS category are given (table 3) and age adjusted ORs were estimated for each NEWS category (table 4). Each rise in NEWS category was associated with an increased risk of mortality when compared to the lowest category (0–4): (5–6: OR 1.95, 95% CI 1.21 to 3.14), (7–8: OR 2.26, 95% CI 1.42 to 3.61), (9–20: OR 5.64, 95% CI 3.70 to 8.60). This was also the case for the combined outcome (ICU and/or mortality): (5–6: OR 1.72, 95% CI 1.14 to 2.60), (7–8: OR 2.17, 95% CI 1.45 to 3.25), (9–20: OR 5.78, 95% CI 4.02 to 8.31). Patients with a NEWS of 5–6 were not associated with an increased risk in ICU admission when compared to those with a NEWS of 0–4, but patients with a NEWS of 7–8 or 9–20 were: (7–8: OR 2.01, 95% CI 1.02 to 3.97), (9–20: OR 5.76, 95% CI 3.22 to 10.31).

Table 3

Number and percentage of patients admitted to ICU within 2 days, patients who died within 30 days, and patients who were admitted to ICU and/or died, by NEWS group

Table 4

Age adjusted ORs for each NEWS category for admission to ICU within 2 days, 30-day mortality, and combined endpoint (ICU and/or mortality)

Patients aged 50–70 years were significantly more at risk of dying within 30 days than patients aged <50 (OR 5.38, 95% CI 2.56 to 11.29), as were patients aged >70 (OR 9.42, 95% CI 4.60 to 19.32).

ROC curves plotting sensitivity (true positives) against 1-specificity (false positives) were used to measure the accuracy of NEWS and age adjusted NEWS in predicting each endpoint (figure 2). With regard to ICU admission, adjusting the NEWS for age decreased the area under the curve from 0.67 to 0.61. Adjusting for age had little effect on the combined endpoint of ICU/mortality (0.71 vs 0.70). When using the NEWS to predict 30-day mortality, the area under the curve was increased from 0.70 to 0.73 by adjusting for age, but this increase was not significant.

Figure 2

National early warning score (NEWS) and age adjusted NEWS receiver operating curve (ROC) for: (A) admission to intensive care unit (ICU) within 2 days; (B) 30-day mortality; (C) combined endpoint (ICU and/or mortality).

NEWS ROC characteristics for the combined endpoint of ICU and/or mortality are presented in table 5. The positive predictive value illustrates that 27% of patients with a NEWS of ≥7 were admitted to the ICU within 2 days and/or died within 30 days. For a NEWS of ≥9 this rose to 35%.

Table 5

NEWS receiver operating curve (ROC) characteristics for combined endpoint (ICU and/or mortality)


Systems for the triaging of patients are well established in emergency medicine. The concept of triage dates back to the Napoleonic wars, and at its core is the assessment of a patient by an experienced person to determine the priority and timescale of treatment and intervention based on the severity of their condition. Over the years the concept of triage has become more scientific and a variety of tools utilising physiological data and algorithms have been developed to assist with effective triage.

It is well recognised from large national audits that existing systems failed to recognise or respond appropriately to early signs of critical illness.21 ,22 This in turn may lead to significant morbidity or mortality for the patient. Many such patients exhibit physiological derangement hours before their deterioration was detected by clinical staff and hospital systems.9 Evidence suggests that a large proportion of ward based patients receive substandard care prior to ICU admission and that 20–40% of such admissions are potentially avoidable.23 This number has decreased over the last 15 years, but remains significant.24

In response to such findings, EWS have become increasingly prevalent across medicine globally over the past two decades to identify the deteriorating patient and to activate an appropriate response, the so called ‘track and trigger’ systems. Within the UK, there are many such systems in place across the NHS but most lack rigorous evaluation and validation.25 A recent postal survey also highlighted that there is great variability in how such EWS are utilised in EDs across the NHS.26

The Royal College of Physicians of London, in its 2007 report ‘Acute medicine: the right person, in the right setting—first time’, recognised that a standardised EWS, used across the NHS could provide a step change in improving clinical outcomes in people with acute illness.27 A standardised NHS-wide NEWS would have advantages with regard to reliability of completion and promoting continuity throughout the patient's journey. However given the size of the NHS and diversity of work which is undertaken, any such tool is unlikely to perform well across all patient groups. This may give rise to false positives, or patients being falsely reassured by false negatives. There is some evidence that the use of EWS can predict need for hospital admission and mortality in ED patients.28 ,29 However, Roland and Coates suggested, in a recent commentary, that any such system proposed for the ED should be validated in ED patients.17 Other scoring systems such as the mortality in emergency departments sepsis (MEDS) score have been suggested as useful predictors of outcome in the ED.28 Although these scoring systems have a higher sensitivity, they rely on some laboratory investigations which necessitate a delay in obtaining the results. The NEWS also has the advantage that it is proposed to be used universally; therefore a diagnosis for the patient is not required in order to use the tool.

This study has several potential limitations. Among the group of patients selected for this analysis (N=2489), only 2003 could be assessed due to missing observations on attendance. This missing information was generally only one or two of the six required physiological data points, but prevented an accurate calculation of a NEWS value. No information is available for patients who attended and were discharged within 2 days of attendance. This group should by virtue of the fact that they are discharged within 2 days have a much lower incidence of significant illness. Our data only included ICU admission within 2 days, so we are unable to comment on patients who may have been admitted to the ICU later in their hospital admission. However, for our target group of ED patients at presentation it could be argued that ICU admission more than 2 days after ED attendance is less linked to features of illness present at initial ED presentation.

We only collected information on in-hospital mortality. No attempt was made to follow-up patients after discharge from hospital, so any patients who were discharged and died at home within 30 days are not included in our data. Our study did not record any information on patients’ co-morbidity.

The data presented in this study show there is some promise for the use of a single EWS in the ED, when applied to a large cohort of patients with a potentially serious condition. Among patients who have sepsis, a single EWS of ≥7 in the ED indicates a 27% chance of requiring admission to the ICU within 48 h and/or death within 30 days. At this level, an argument can be made for mandating senior ED clinical review for all these patients. In addition, there could also be an argument for mandatory review by a critical care outreach team, regardless of ultimate destination. Our study only looked at patients with sepsis, so the generalisability to other serious conditions is unknown. However, given that sepsis is a common condition with potential significant morbidity and mortality but that also has a heterogeneous presentation, suggests that this approach may be more widely applicable and this potential warrants further research.

This concept also lends itself to extension to prehospital care and ambulance services. Most ambulance services routinely collect the physiological data required to calculate an EWS score, and indeed some ambulance services have incorporated this into electronic patient record forms. An agreed EWS score of greater than a specific level could be used as a trigger for ambulance service pre-alert of a receiving ED.

Point of care lactate testing has also been shown to be feasible in the ED.30 Serum lactate is recognised as an independent predictor of mortality in sepsis31 and there may be potential for combining EWS systems and point of care to further improve diagnostic accuracy for patients at risk of adverse outcomes in sepsis.32

Appendix 1: Participating hospitals

  1. Aberdeen Royal Infirmary, Aberdeen

  2. Ayr Hospital, Ayr

  3. Crosshouse Hospital, Kilmarnock

  4. Dumfries & Galloway Royal Infirmary, Dumfries

  5. Glasgow Royal Infirmary, Glasgow

  6. Hairmyres Hospital, East Kilbride

  7. Inverclyde Royal Hospital, Greenock

  8. Monklands Hospital, Airdrie

  9. Ninewells Hospital, Dundee

  10. Perth Royal Infirmary

  11. Queen Margaret Hospital, Dunfermline

  12. Raigmore Hospital, Inverness

  13. Royal Alexandra Hospital, Paisley

  14. Royal Infirmary of Edinburgh

  15. Southern General Hospital, Glasgow

  16. Stirling Royal Infirmary

  17. Victoria Infirmary, Glasgow

  18. Victoria Hospital, Kirkcaldy

  19. Western Infirmary, Glasgow

  20. Wishaw General Hospital


View Abstract

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.

    Files in this Data Supplement:


  • Collaborators Scottish Trauma Audit Group Steering Committee.

  • Contributors All authors have made substantial contributions to all of the following: (1) the conception and design of the study, or acquisition of data, or analysis and interpretation of data; (2) drafting the article or revising it critically for important intellectual content; and (3) final approval of the version to be submitted.

  • Competing interests None.

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

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.