Article Text

Download PDFPDF

Early warning scores: a health warning
  1. Kirsty Challen1,
  2. Damian Roland2,3
  1. 1Emergency Department, Lancashire Teaching Hospitals NHS Trust, Royal Preston Hospital, Preston, UK
  2. 2Paediatric Emergency Medicine Leicester Academic (PEMLA) Group, Leicester Hospitals, Leicester, UK
  3. 3SAPPHIRE Group, Health Sciences, Leicester University, Leicester, UK
  1. Correspondence to Dr Kirsty Challen, Lancashire Teaching Hospitals NHS Trust, Emergency Department, Royal Preston Hospital, Sharoe Green Lane, Preston PR2 9HT, UK; kirstychallen{at}


Early warning scores are frequently used in UK adult emergency departments (EDs) and are gaining traction in paediatric emergency care. Like many innovations with inherent face validity, they have great appeal to clinicians, managers and commissioners. However, it is important to ensure unintended consequences and balancing measures are mitigated. We review the background to their development and introduction in the ED, the evidence for their usefulness, their limitations in our field and areas for further research.

  • risk management
  • clinical
  • emergency care systems, emergency departments
  • emergency department management
  • paediatrics, paediatric emergency medicine

Statistics from


Early warning scores (EWS) are a routine feature of emergency department (ED) practice in the UK as use of the National Early Warning Score (NEWS), published by the Royal College of Physicians of London, is now a requirement,1 despite concerns with its use in this clinical environment.2 Even before this, over half of UK EDs were using an EWS to trigger senior review of particular patients.3 Queensland also has a statewide ED ‘Adult Deterioration Detection System’,4 and we have reports of EWS use in EDs in South Australia and South Carolina. Anecdotal polling of personal contacts revealed only occasional encounters with EWS in US and Canadian EDs.

We do not aim to present here a systematic review of EWS or track and trigger systems (TTS) as these exist both for adults and children5 ,6 but wish to highlight their limitations in the ED setting.

Where did EWS come from?

EWS initially developed following retrospective reviews of care preceding unplanned admission to intensive care units (ICUs), where a recurrent theme was that of well-documented physiological deterioration over many hours that was either not recognised or not acted upon.7 Similarly, the 2006 Confidential Enquiry into Maternal and Child Health (CEMACH) report ‘Why children die’ identified failure to recognise severity of illness in children as a significant remediable factor in paediatric deaths and recommended “a standardised and rational monitoring system with imbedded early identification systems for children developing critical illness—an Early Warning Score”.8

Early trials of TTS incorporating an EWS and a mandated response demonstrated a reduction in complication rates, particularly in surgical patients,9 although these results were not universally replicated.10 These early results prompted widespread enthusiasm for EWS. Guidance from the National Institute for Health and Care Excellence (an arms-length non-governmental body sponsored by the English Department of Health that aims to reduce practice variation) on the management of the acutely ill adult in 2007 defined various types of TTS (table 1) and recommended their adoption.11

Table 1

National Institute for Health and Care Excellence categorisation of track and trigger systems

The NEWS currently mandated for adults in the UK is a modification of the VitalPAC EWS (ViEWS) developed on a large Acute Medical Unit data set.12 Sadly the working group developing NEWS did not include any emergency physicians and were unable to locate any relevant literature relating to ED patients.

There is as yet no universal EWS for children, and although multiple versions have been developed at local levels13 ,14 with anecdotal reports of introduction into many EDs, direct evidence of their benefit is lacking for their utility even at ward level.6

How well researched is EWS in the ED?

Previous systematic reviews of EWS have related to their use in inpatient settings; in both adults and children, these demonstrated limited sensitivity and heterogeneity of trigger criteria.5 ,6 To examine the current state of play, we conducted a brief scoping review of both the paediatric and adult literature relating to EWS in the ED. We identified 16 publications since 2006 in adults; 2 of these were in non-English journals (1 German, 1 Chinese). Of these, only six were prospective observational studies; two (both from Hong Kong) derived a new score, while seven more examined tools (mostly the Mortality in ED Sepsis score) designed for the ED. The paediatric literature included 7 full papers and 11 abstracts; 7 were prospective studies and 8 examined a score designed for ED use (4 of these assessed the Paediatric Observation Priority Score tool). Full details are given in table 2. The following two sections will explore the implications of this literature.

Table 2

Results of literature review

What is the purpose of EWS in ED care?

In analysis of the progress in establishing critical care outreach, the UK NHS Modernisation Agency made a number of comments about the utility and limitations of TTS (table 3).49 These include the suggestion that EWS can ‘red flag’ critical illness and secure help for sick patients.

Table 3

Utility and limitations of track and trigger systems (TTS)

It would be reasonable to suppose that most clinicians would hope to use a score that identified high-risk patients in order to focus beneficial interventions towards these patients, and to this end the EWS would be the afferent limb of a system that also included a skilled team as the efferent limb (figure 1). A preliminary study from Rees and Mann piloted the use of a physiological ‘patient at risk’ score to trigger automatic intensive care review in the ED, although only 3 of the 30 patients reviewed were actually admitted to the ICU.50 McGillicuddy et al's51 much larger before-and-after trial of immediate tannoy-based physician attendance to a patient who met specific physiological criteria at triage showed reductions in time to physician evaluation (median 11 min vs 21 min) and first therapeutic intervention (median 26 min vs 58 min). It is unclear, however, how the ‘standard care’ before the tannoy system prioritised patients.

Figure 1

Principle underlying early warning scores.

One overlooked aspect of the benefit of EWS in emergency care is improving communication, especially in respect of creating a common language between EDs and admitting teams. Communication tools such as SBAR52 lend themselves very well to having a common format to discuss the acuity of a patient. With preventable deaths more likely to occur on the wards than in the ED,53 it would seem sensible to maximise the transfer of relevant information at the time of transfer.

Can EWS do this?

Various groups, identified from prior systematic reviews and our scoping study, have examined the predictive value of EWS in the adult ED. Only one of these, however, addressed the whole population of admitted patients in the manner suggested by the NHS Modernisation Agency; Heitz et al24 found that the maximum modified early warning score in the ED had an area under the receiver operating characteristic curve (AUROC) of 0.73 for the prediction of death or ICU admission.

Other studies have evaluated EWS in specific patient subgroups; Corfield et al18 identified 2003 adult patients with sepsis presenting to Scottish EDs and found NEWS to have an AUROC (c-statistic) of 0.7 for prediction of 30-day mortality and 0.67 for admission to intensive care with 2 days. Christensen et al examined use of a modified EWS at triage in a Danish ED and found that a Bispebjerg EWS of >5 at triage identified 63% of patients who died or were admitted to ICU within 48 h. These were, however, patients who had already been identified by nursing gestalt as ‘red’ and therefore in need of immediate or acute treatment.23 In 790 medical patients in a South African ED, Burch et al29 found that increasing scores on an older EWS, which did not include points for oxygen saturations or supplemental oxygen, were associated with increased hospital mortality (p for trend <0.001). However, there was no calculation of sensitivity or specificity for any specific cut-off scores.

The diversity of potential end points for ED-based warning scores and systems has challenged both the development and research of such tools. This is especially true in paediatric practice where mortality is very low but physiological variation extremely common at the point of triage.54 In paediatric practice, there is developing evidence that bespoke systems may allow early identification of risk of admission38 ,38 but previous studies to identify the most critically ill may lack specificity if applied to all presenting patients.47

What can't EWS do?

EWS were developed on the premise that early identification and intervention in the deteriorating patient could reduce mortality. There is limited evidence for this in the adult literature; the landmark cluster randomised trial of a medical emergency team (MET), MERIT, found an increase in calls to the emergency team (from 3.1 to 8.7 per 1000 admissions), but no reduction in cardiac arrest calls (1.31 vs 1.64 per 1000 admissions), unplanned intensive care admissions (4.19 vs 4.68 per 1000 admissions) or unexpected deaths (1.06 vs 1.18 per 1000 admissions).55 This may reflect earlier findings that MET calls often identified dying patients and resulted in the implementation of a DNAR order.56 Alam et al's57 systematic review of the effects of introduction of EWS demonstrated mixed outcomes; of the six studies assessing hospital mortality, two reported statistically significant mortality reduction, two a trend towards reduction and two no change. However, all these studies were on limited sites, in specific clinical areas and of a before-and-after design, which limits their generalisability to the ED.

The hope that an EWS in the ED can reduce mortality is based on the premise that deaths in the ED or soon after admission are preventable. This assumption has not been widely explored; when Lu et al58 examined 210 cases where death occurred within 24 h after ED admission (excluding post cardiac arrest patients and those with terminal cancer), only 32 were deemed to have been preventable (25.8%). Even this is higher than Hogan et al's53 much larger study of 1000 deaths in English acute hospitals, where only 5.2% were judged to have been preventable; of these 19 were identified to have had a problem in their assessment in the ED.

Although EWS can identify the patient with deranged physiology, it cannot differentiate between specific illnesses such as sepsis versus pancreatitis or chronic obstructive pulmonary disease versus cardiogenic pulmonary oedema. There are also clearly limitations in using EWS in isolation as a triage tool as some clearly time-critical presentations such as ST elevation myocardial infarction or acute CVA may present with normal physiology. In emergency and acute paediatrics, there is a national drive to improve recognition of serious bacterial illness; however, the majority of patients presenting with potential features of serious bacterial illness such as pyrexia have self-limiting viral infections. The search for the very small proportion of children with septicaemia may not be the best use of an EWS system. This confusion between triage, illness identification and early warning systems may be one reason why no standard tool has been introduced nationally. Furthermore, it is not clear what criteria should be used in children to determine abnormal HR and RR values.59 Given the large number of children discharged home, a more appropriate utilisation of EWS for children may be identification of safe discharge rather than identification of illness.60

What else do we need to know?

A number of issues around EWS in the ED still need to be addressed. Cuthbertson identified four questions for users of EWS in 2008,61 and they remain applicable today:

  1. Does the EWS I use utilise and suitably weight early signs of deterioration such as HR and RR?

  2. Does the EWS I use avoid giving disproportionate weight to late signs of deterioration such as BP?

  3. What is the diagnostic accuracy of the EWS I use in the populations in which I use it and can it be improved?

  4. What is the optimal cut point for the EWS I use and do I use this as the trigger for activating a response?

In order to answer these questions, clinicians need to consider the purpose for which the EWS is being used; if it is at the front door, is it any better than the existing triage system or than unstructured assessment by the triage clinician? Fullerton et al62 found that use of an EWS in the prehospital environment added little to the judgement of the treating paramedic in identifying the critically ill patient. As Foëx and Jones eloquently commented in an editorial on the Fullerton paper, the value of a diagnostic test is in fact whether it affects patient outcome.63

We need to know whether EWS can reliably provide information we need in the ED, such as need for time-critical interventions, safety of discharge or potential workload (either nursing or medical) related to the patient stay. None of these has as yet been addressed adequately. This should be remedied by robust assessment of NEWS as it is implemented in UK EDs. Equally, the patient safety aspect of potentially improved communication using a standardised system should be formally assessed.


It seems that EWSs are likely to remain an expected safety measure, certainly for the foreseeable future. We would, however, encourage ED clinicians to assess how best they can be applied in our environment. Although an EWS can flag the patient deteriorating towards death, this is not necessarily the patient who can benefit the most from emergency intervention64 and much more evidence needs to be developed before we can assume that EWS will reliably prioritise ED patients. Additionally, the higher prevalence of seriously sick and injured patients in an ED will cause problems if a tool developed for lower-acuity areas such as wards is applied with no modification of trigger thresholds as the false negative rate (alarm triggered but no new intervention required) will rise.

In order for a TTS in any setting to be of value, the identification of a patient at risk by the afferent limb needs to be matched by an efferent limb intervention. Although studies of rapid response teams have shown that patients who benefited did so from critical care interventions,65 this would generally be already available in an ED without need for specific calling criteria.

In summary, it is not clear that an EWS will add any value to the processes and staff expertise already present in the ED.


View Abstract


  • Twitter Follow Kirsty Challen at @kirstychallen and Damian Roland at @damian_roland

  • Contributors Both authors contributed to the literature searching and manuscript drafting.

  • Competing interests None declared.

  • Provenance and peer review Commissioned; externally peer reviewed.

  • Data sharing statement Original literature searches available from the authors.

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.