If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
A systematic review of the discrimination and absolute mortality predicted by the National Early Warning Scores according to different cut-off values and prediction windows
NEWS reliably identifies patients who are unlikely to die within 24 h.
•
Prediction of mortality after 24 h by NEWS is unreliable.
•
Many hospitalized patients presenting with a low NEWS die within 30 days.
•
NEWS reliably identifies patients who need immediate attention.
Abstract
Background
Although early warning scores were intended to simply identify patients in need of life-saving interventions, prediction has become their commonest metric. This review examined variation in the ability of the National Early Warning Scores (NEWS) in adult patients to predict absolute mortality at different times and cut-offs values.
Method
Following PRISMA guidelines, all studies reporting NEWS and NEWS2 providing enough information to fulfil the review's aims were included.
Results
From 121 papers identified, the average area under the Receiver Operating Characteristic curve (AUC) for mortality declined from 0.90 at 24-hours to 0.76 at 30-days. Studies with a low overall mortality had a higher AUC for 24-hour mortality, as did general ward patients compared to patients seen earlier in their treatment. 24-hour mortality increased from 1.8% for a NEWS ≥3 to 7.8% for NEWS ≥7. Although 24-hour mortality for NEWS <3 was only 0.07% these deaths accounted for 9% of all deaths within 24-hours; for NEWS <7 24-hour mortality was 0.23%, which accounted for 44% of all 24-hour deaths. Within 30-days of a NEWS recording 22% of all deaths occurred in patients with a NEWS <3, 52% in patients with a NEWS <5, and 75% in patient with a NEWS <7.
Conclusion
NEWS reliably identifies patients most and least likely to die within 24-hours, which is what it was designed to do. However, many patients identified to have a low risk of imminent death die within 30-days. NEWS mortality predictions beyond 24-hours are unreliable.
Early warning scores (EWS) were originally proposed by Morgan et al. to help identify patients who need immediate life-saving interventions, but not to predict their outcomes [
]. EWS performance is most frequently reported as its ability to discriminate those patients who will develop an outcome from those who will not. Discrimination can be quantified by the area under the Receiver Operating Characteristic (ROC) curve (AUC). This test is most useful in the early evaluation of a score and may be dependant on patient characteristics and disease spectrum [
]. From a clinical perspective, especially if the discriminatory value is already high, small variations in the AUC between different EWSs may be unimportant. Moreover, odds ratios and other statistical manipulations may deflect attention from the variation in absolute mortality rates reported by different studies. What may matter most clinically is the score's performance at different time prediction windows, the absolute risk of death predicted, and what other factors might influence the accuracy of this prediction.
The original National Early Warning Score (NEWS) developed by the Royal College of Physicians in the United Kingdom [
]. Unlike Morgan's original EWS it was not based on expert opinion, but on the analysis of a huge database of vital signs collected over several years at Portsmouth Hospitals NHS Trust. From this data the Vitalpac™ Early Warning Score (ViEWS) was derived, and subsequently slightly modified for operational reasons so that it could be promoted as NEWS throughout the NHS. ViEWS was originally designed to identify sick patients likely to die from any cause within 24-hours and not to include age as a predictor variable [
]. These changes were not based on data, but on expert opinion. Although NEWS and NEWS2 are similar and intimately connected, it cannot be assumed that they are interchangeable, or that one is superior to the other [
A comparison of the ability of the National Early Warning Score and the National Early Warning Score 2 to identify patients at risk of in-hospital mortality: a multi-centre database study.
The predictive discrimination of a score is likely dependant on the outcomes measured, the length of the prediction window used, and how many effective interventions the patients studied received. Studies that have compared different scores have often used prediction windows, patient populations and outcomes that none of the scores tested were originally intended for [
]. The risk of death within 24-hours, arguably the most useful trigger for immediate escalation of care, occurs rarely in most patients. Therefore, ‘composite outcomes’ that include interventions such ICU admission are frequently used to compare NEWS with other early warning systems, which now include complex algorithms derived by logistic regression or machine learning from large electronic medical record (EMR) data sets [
]. However, transfer to higher levels of care, such as ICU, will vary according to how different hospitals configure their care, as well the number of ICU beds and other resources available. Therefore, to ensure it could compare the performance of NEWS across as many different healthcare settings as possible this review was confined to the prediction of death.
The aim of this systematic review was to determine and describe the AUC of NEWS and NEWS2 for death at different prediction time windows and/or the absolute mortality above and below defined thresholds or ‘cut-off’ values. The review examined the variation in all the published peer reviewed reports of NEWS and NEWS2 performance, plus additional unpublished data available to the authors. Studies that only reported composite outcomes were excluded.
2. Methods
2.1 Study design
All studies reporting NEWS and NEWS2 with mortality outcome data were reviewed using the Preferred Reporting Items for Systemic Reviews and Meta-Analyses (PRISMA) guidelines [
All studies reporting NEWS and NEWS2 and providing enough information to determine at different prediction time windows the AUC for death or the absolute mortality above and below defined cut-off values (Table 1). All study designs containing the requisite data were eligible for inclusion, including abstracts. However, studies reporting only composite outcomes were excluded.
Table 1Inclusion and exclusion criteria.
Subject
Inclusion Criteria
Exclusion Criteria
Date
Any time
Any time
Geographical location
Any
Any
Language
English
Non-English language, as translation was not feasible
Participants
Adult patients
Children (age 15 or less)
Obstetrics
Peer review or
Peer reviewed journals
Editorials
Type of Publication
Peer reviewed reviews
Letters
Abstracts
Newspaper articles
Conference abstracts
Books
Reports, unless peer reviewed
Theses
Dissertations
Non-peer reviewed articles
Grey literature
No mention of ‘National Early Warning Score’ in title or abstract
Reported outcomes
AUC of NEWS and NEWS2 for death at different prediction time windows and/or the absolute mortality above and below defined thresholds or ‘cut-off’ values
Odds ratios
Composite outcomes
Setting
Any setting providing data that fulfilled the reported outcomes
Maternity
Obstetrics
Paediatrics
Study Design
Any study providing data that fulfilled the reported outcomes
The search was limited to peer reviewed studies reporting adult patients from any speciality, excluding obstetrics. The authors also had access to unpublished supplemental data from two Society for Acute Medicine Benchmarking Audits [
], and to the data collected from August 2016 to January 2021 by a previously reported ongoing quality improvement project in a low resource Ugandan hospital [
A prospective, observational study of the performance of MEWS, NEWS, SIRS and qSOFA for early risk stratification for adverse outcomes in patients with suspected infections at the emergency department.
], was completed on 12 April 2021. [email protected] was used to search MEDLINE, CINHAL and BMJ databases. In addition, PubMed, Paperchase, and the Cochrane Library were searched independently. References in review articles were also manually searched for relevant studies.
Preliminary searches confirmed that the term ‘NEWS’ was often misinterpreted as a source of information and could not be used. A full text search using the term ‘National Early Warning Score’ identified over 15,000 irrelevant papers. However, when ordered by [email protected] to relevance, limiting the term ‘National Early Warning Score’ to either the abstract or a full text search yielded a comparable number of papers. Therefore, the final search strategy over all search engine platforms was limited to titles and abstracts.
The search terms used were ‘National Early Warning Score’, ‘Acute’, ‘Emergency’, ‘Mortality’, ‘Survival’, ‘Death’ and ‘Died’. Boolean operators were used to construct the search (Abstract:(National Early Warning Score)) and (Abstract:(National Early Warning Score) AND ((Mortality) OR (Survival) OR (Died) OR (Death)) AND ((Acute OR Emergency)). The search was limited to studies written in English. Although the search was not time limited, NEWS only dates from 2012.
2.4 Study selection
Papers were collated using the reference manager ProQuest® RefWorks. Duplicate studies were removed. All titles and abstracts were independently reviewed by both authors. Shortlisted studies were retrieved and independently reviewed by both authors.
2.5 Data extraction
Data pertaining to authors, year of publication, country of origin, clinical setting, sample size (number of sets of patient observations, not patients), NEWS, its precursor ViEWS, NEWS2, threshold cut-off values, and mortality were extracted by both authors. Mortality data included the time between an observation and death (i.e., the prediction window).
2.6 Quality assessment and risk of bias of included studies
As many studies as possible were included, to ensure that any factors that might bias the performance of NEWS and/or NEWS2 were examined; these included patient age, in-study mortality, study size, location, clinical setting, and the prediction windows and cut-off values reported.
2.7 Analysis
Data were extracted to Microsoft Excel (version 2018). Average AUC and mortality rates were calculated for all observations at different prediction windows and cut-off values. AUCs were compared according to the method of Hanley and McNeil [
]. An AUC of 0.5 suggests no ability to discriminate patients who are going to die from those who will survive, 0.7 to 0.8 is acceptable discrimination, 0.8 to 0.9 is excellent, and more than 0.9 is outstanding [
The search returned 1323 articles, including 745 duplicates. Of the 578 original studies, 295 were considered eligible for a full review, of which 121 from 28 countries were suitable for inclusion (Fig. 1). Authors to three of these papers [
] provided unpublished data, yielding 122 sets of data for analysis. Most were of either consecutive or a convenient sample of patients. One study compared a cross-sectional sample of patients who did not die in hospital with those who died within 24-hours of NEWS measurement [
Evaluation of the feasibility and performance of early warning scores to identify patients at risk of adverse outcomes in a low-middle income country setting.
] that reported an AUC for 24-hour mortality of 0.65 was excluded from the final analysis because it used imputation so extensively for missing data.
Fig. 1flow chart of literature search process and eligible studies identified. API = any paper that simply reported and assessment procedure, policy or implementation of an early warning score without outcomes data provided.
Prognostic accuracy of the quick sequential organ failure assessment score and national early warning scores in mortality rate of the non-traumatic patients.
A comparison of the ability of the National Early Warning Score and the National Early Warning Score 2 to identify patients at risk of in-hospital mortality: a multi-centre database study.
], and most were from the UK and the USA (Supplemental Figure 1). Most studies (96 articles) reported the performance of NEWS only or its precursor ViEWS (5 articles); 16 were NEWS2 only, four NEWS and NEWS2, and one NEWS2 and ViEWS; 102 studies only reported the first observation recorded, and only 11 studies reported all observations recorded while in hospital (Supplemental Table 1). Observations were made on patients described as hospital admissions in 38.5% of studies, as patients in emergency departments in 25.4%, emergency admissions in 16.4%, pre-hospital in 10.7%, acute medical unit or medical assessment unit in 4.9%, intensive care units in 2.5%, and as transplant patients in 1.6%. Fifty-one studies included all patients, 22 were limited to patients with sepsis or suspected sepsis, 12 to medical patients only, 11 to patients with pneumonia or respiratory illness, and 10 to patients with COVID-19 or suspected COVID-19 (Table 2). Overall patient age and hospital length of stay could only be approximated as they were not consistently reported in all studies. Citation and further details of the 122 included studies are provided in Supplemental Table 1.
Table 2Observations made according to where or what type of patient they were recorded on. Observations were made on patient reported as hospital admissions in 37.7% of studies, as in emergency departments in 25.4%, as emergency admissions in 16.4%, as pre-hospital in 10.7%, as in medical assessment units in 5.7%, in intensive care units in 2.5%, and on transplant patients in 1.6%. Fifty-one studies included all patients, 22 were limited to patients with sepsis or suspected sepsis, 12 to medical patients only, 11 to patients with pneumonia or respiratory illness, and 10 to patients with COVID-19 or suspected COVID-19. Other miscellaneous studies were limited to patients specified by illness severity [
Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department.
Comparison of national early warning score, rapid emergency medicine score and acute physiology and chronic health evaluation II score for predicting outcome among emergency severe patients.
Zhonghua Wei Zhong Bing Ji Jiu Yi Xue.2017; 29: 1092-1096
Prognostic accuracy of the Hamilton Early Warning Score (HEWS) and the National Early Warning Score 2 (NEWS2) among hospitalized patients assessed by a rapid response team.
Comparison of the National Early Warning Score+Lactate score with the pre-endoscopic Rockall, Glasgow-Blatchford, and AIMS65 scores in patients with upper gastrointestinal bleeding.
). AMU/MAU = acute medical or medical assessment unit. PRE-HOSPITAL = prior to hospital transfer, recorded before entering or in the ambulance en route to hospital.
3.2 Performance of news across different prediction windows
The discrimination of NEWS for death has been reported for several prediction windows. We found 74 published reports and one unpublished report from which the average AUC could be estimated, ranging from 89 to 2,759,469 observations (Supplemental Table 2a). The total number of observations for prediction windows ranged from over 5 million for 24-hour mortality, 1,888,535 for in hospital death and 101,680 for death within 30-days. The overall AUC averages decline from 0.897 at 24-hours, to 0.762 by 30 days. The range of AUC reported by each study increased with the length of the prediction window, ranging from 0.801 – 0.910 for 24-hour mortality to 0.610 – 0.910 for 30-day mortality (Fig. 2). Similar declines in AUC for death as the time after observation increased were observed for NEWS2 (Supplemental Table 2b) and ViEWS (Supplemental Table 2c).
Fig. 2The area under the Receiver Operating Characteristic curve (AUC) for mortality at different times (i.e., predictions windows up to 30 days) after NEWS recorded. Results of individual studies are shown as open circles, solid triangles show average result for each prediction window if more than one study was performed. Complete data is provided in Supplemental Table 2a; data for NEWS2 and ViEWS in Supplemental Tables 2b and 2c.
3.3 Comparison of auc for 24-hour mortality of news with news 2 and views (Table 3)
One unpublished and 15 published studies reported the discrimination of NEWS for death within 24-hours; the average AUC for a total of 5,637,622 observations was 0.897 (95% CI 0.895 – 0.899; range 0.801 to 0.910). Five papers reported that 269,008 measurements of ViEWS had an average AUC for death within 24-hours of 0.886 (95% CI 0.877 – 0.895, range 0.873 – 0.908), and three papers reported an average AUC of 0.848 (95% CI 0.811 – 0.885, range 0.840 – 0.861) for 9717 measurements of NEWS 2 (Table 3). The smaller the number of observations made, the greater the range of AUC reported; all studies with more than 50,000 observations had an AUC ≥0.880 (Supplemental Figure 2).
Table 3the area under the Receiver Operating Characteristic curve (AUC) for 24-hour mortality from all eligible studies for NEWS, NEWS2 and ViEWS. Overall results of approximate average age, total number of observations and average AUC are shown in bold. 95% CI = 95% confidence interval.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
Efficacy of prehospital National Early Warning Score to predict outpatient disposition at an emergency department of a Japanese tertiary hospital: a retrospective study.
The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
3.4 Potential modifiers of the AUC of news for 24-hour mortality (Table 4)
Most observations of NEWS were made either on UK patients throughout their hospital stay (2,245,778 observations) or in emergency departments patients in the USA (2,759,469), but the AUC for 24-hour mortality was little influenced by where, geographically, it was measured (Table 4). However, the AUC of NEWS for 24-hour mortality was lower in studies with a small number of observations and only changed slightly by patient age. The AUC for 24-hour mortality of observations made throughout hospitalization are slightly higher than for 24-hour mortality after the first observation made. The 24-hour mortality AUC may be lower if observations are made prior to hospital admission, or in the emergency department, or in patients admitted as emergencies or in ICU, compared with observations made on any patient admitted to hospital. The overall 24-hour mortality of patients with NEWS observations ranged from 0.3% to 10.2%, and the AUC for 24-hour mortality decreased slightly as 24-hour mortality increased. The average AUC for 24-hour mortality reported for observations on patients with a 24-hour mortality ≤0.6% was 0.910 (95% CI 0.906 – 0.914), compared with 0.866 (95% CI 0.857 – 0.875) for observations on patients with a 24-hour mortality ≥1% (Table 4).
Table 4Potential modifiers of the AUC of NEWS for 24-hour mortality. AUC = the area under the Receiver Operating Characteristic curve (AUC) for 24-hour mortality, 95% CI = 95% confidence interval.
3.5 The relationship of AUC of news to in-hospital mortality and underlying illness (Table 5)
After studies that reported NEWS AUC for 24-hour mortality, the next largest set of observations (1,890,093) were from studies that reported AUC for in-hospital mortality. Only 15 of these 35 studies also reported hospital length of stay with averages ranging from 2.6 to 20 days (Supplemental Table 2a). Although the length of hospital stays could only be surmised, the AUC for in-hospital mortality decreased as the absolute in-hospital mortality observed increased. Also, the AUC for in-hospital mortality of patients with respiratory illness, sepsis or suspected sepsis was significantly lower (p <0.0001) than for other patients (Table 5)
Table 5The AUC of NEWS for in-hospital mortality according to the absolute in-hospital mortality observed and patients’ underlying illness. AUC = the area under the Receiver Operating Characteristic curve (AUC) for in-hospital mortality, 95% CI = 95% confidence interval. ED = emergency department, ICU = intensive care unit.
Comparison of early warning scoring systems for hospitalized patients with and without infection at risk for in-hospital mortality and transfer to the intensive care unit.
Comparison of the National Early Warning Score+Lactate score with the pre-endoscopic Rockall, Glasgow-Blatchford, and AIMS65 scores in patients with upper gastrointestinal bleeding.
Comparison of early warning scoring systems for hospitalized patients with and without infection at risk for in-hospital mortality and transfer to the intensive care unit.
Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
The most commonly used disease severity scores are inappropriate for risk stratification of older emergency department sepsis patients: an observational multi-centre study.
Prognostic accuracy of the quick sequential organ failure assessment score and national early warning scores in mortality rate of the non-traumatic patients.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
Validation of modified early warning score using serum lactate level in community-acquired pneumonia patients. The National Early Warning Score-Lactate score.
Evaluation of the risk prediction tools for patients with coronavirus disease 2019 in wuhan, china: a single-centered, retrospective, observational study.
The utility of the rapid emergency medicine score (REMS) compared with SIRS, qSOFA and NEWS for Predicting in-hospital Mortality among Patients with suspicion of Sepsis in an emergency department.
The most commonly used disease severity scores are inappropriate for risk stratification of older emergency department sepsis patients: an observational multi-centre study.
3.6 Range of absolute 24-hour mortalities observed at different news cut-off values (Table 6)
One unpublished and 12 published studies provided enough information to determine the absolute mortality within 24-hours for different cut-off values of NEWS, from which the 24-hour mortality rate for patients with values above and below the cut-off could be determined. Although the ranges of 24-hour mortality of patients with NEWS points greater or equal to the cut-off are wide (Table 6), their average values trend upwards from 0.8% for a cut-off ≥1 point to 7.8% for a cut-off ≥7 points (Fig. 3). In contrast, the 24-hour mortality for patients with NEWS values below all cut-off values only increased from 0.04% for a cut-off ≥1 point to 0.23% for a cut-off ≥7 points (Table 6). Although patients with NEWS <7 points only had a 24-hour mortality of 0.23%, these deaths accounted for 44.3% of deaths within 24-hours, whereas the deaths of patients with a NEWS <3 points made up 8.9% of all 24-hour deaths (Fig. 4).
Table 6Range of absolute 24-hour mortalities observed at different NEWS cut-off values. Overall results of approximate average age, total number of observations and average values are shown in bold. IQR = inter-quartile range. SD = standard deviations.
Glucose as an additional parameter to National Early Warning Score (NEWS) in prehospital setting enhances identification of patients at risk of death: an observational cohort study.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
Glucose as an additional parameter to National Early Warning Score (NEWS) in prehospital setting enhances identification of patients at risk of death: an observational cohort study.
Efficacy of prehospital National Early Warning Score to predict outpatient disposition at an emergency department of a Japanese tertiary hospital: a retrospective study.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
Investigating the discriminative value of Early Warning Scores in patients with respiratory disease using a retrospective cohort analysis of admissions to Nottingham University Hospitals Trust over a 2-year period.
A comparison of the ability of the National Early Warning Score and the National Early Warning Score 2 to identify patients at risk of in-hospital mortality: a multi-centre database study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
Glucose as an additional parameter to National Early Warning Score (NEWS) in prehospital setting enhances identification of patients at risk of death: an observational cohort study.
Fig. 324-hour mortality above and below NEWS cut-off thresholds ranging from 1 to 7 points. Note that the most frequently used cut-off value reported in the literature is ≥5 points. Full data is shown in Table 6.
Fig. 4The proportion of all deaths within 24-hours of patients below NEWS cut-off thresholds ranging from 1 to 7 points, and the proportion of all patients equal or above each cut-off. Although patients with <7 NEWS points only had a 24-hour mortality of 0.23%, these deaths accounted for 44.3% of deaths within 24-hours, whereas the deaths of patients with a NEWS <3 points made up 8.9% of all 24-hour deaths.
3.7 Absolute mortalities observed at different news cut-off values and prediction windows
The mortality for patients at different time intervals after an observation for NEWS cut-off values ≥1 to ≥15 points were determined from 24-hours to 30 days and at hospital discharge. The range of mortality rates above cut-off values varied greatly, but less for mortality rates below each cut-off (Supplemental Tables 3a-3 g). The most frequently reported cut-off values were ≥3, ≥5 and ≥7. Within 30 days 9.4% of patient with a NEWS ≥3, 14.9% with a NEWS ≥5 and 20.6% with a NEWS ≥7 points had died (Fig. 5). In contrast, for patients with NEWS <3 points, the average mortality remained below 1% throughout their hospital stay, and for 72 h for patients with a NEWS <5; the mortality of patients with a NEWS <7 remained below 2% for up to 5 days. Less than 4% of patients had died within 30 days of a NEWS observation of <7 points. Nevertheless, 21.8% of all deaths within 30 days of an observation occurred in patients with a NEWS <3, 51.9% in patients with a NEWS <5, and 74.5% in patient with a NEWS <7 points.
Fig. 5Mortality at different times after NEWS observation above and below cut-off values of 3, 5 and 7 points.
4.1 General interpretation of the results in the context of other evidence
The overall risk of death within 24-hours for most patients reported is low, and NEWS reliably identifies those patients who are the least likely to die within 24-hours. After 24-hours the discrimination of NEWS for subsequent death declines, with an ever-widening range of reported results, which by 30 days after a NEWS observation is from an AUC of 0.61 to 0.91. Although patients with low NEWS scores make up a significant fraction of all in-hospital deaths, these patients appear to have a reduced risk of death for some time, which implies they have a degree of clinical stability.
This review found that patients with a NEWS <3 points, on average, only have a 0.07% chance of dying within 24-hours. It has been suggested that measuring a complete set of vital signs in these patients more frequently than once a day is not required [
A protocolised once a day modified early warning score (MEWS) measurement is an appropriate screening tool for major adverse events in a general hospital population.
]. However, these “low risk” patients still need some form of ongoing monitoring as they comprise 8.9% of all deaths within 24-hours. The best way to anticipate deterioration in these patients remains unclear. Unlike patients with NEWS <4 points, all patients with a NEWS <3 points will have normal mental status. Nearly three quarters of all patients were observed to have a NEWS <3 points, and these patients accounted for 9% of all deaths within 24-hours and 16% of all in-hospital deaths. No study reported a mortality above 0.35% within 10 days of a NEWS <3 being observed, and in most studies their risk of in-hospital death remained below 1%. However, one study in Thailand of patients with suspected sepsis [
] reported patients with a NEWS <3 points had an in-hospital mortality ranging from 12% to 21%. Therefore, although there may be specific patient cohorts with normal or near normal vital signs who are likely to die while in hospital, very few of them are likely to die within 24-hours.
The discrimination of NEWS was not significantly influenced by patient age or the country it was measured in. However, it did decline as in-hospital mortality rates increased, as has been observed with another EWS [
The prognostic value of national early warning scores (NEWS) during transfer of care from community settings to hospital: a retrospective service evaluation.
Efficacy of prehospital National Early Warning Score to predict outpatient disposition at an emergency department of a Japanese tertiary hospital: a retrospective study.
]. A possible explanation for both these observations may be the availability of life-saving interventions and when and where they were or could be provided. Early measurements were probably made before any treatment was provided, whereas later measurements were more likely after everything possible had been done, and little more could save the patient. This hypothesis is supported by the report of Hwang et al. [
] who found that the AUC of NEWS for 7-day mortality of patients with sepsis was 0.680 when measured at triage, 0.777 after resuscitation with fluid and vasopressors, and 0.888 on exiting the emergency department. Others have also shown the discrimination of NEWS to be lower on arrival to hospital than when repeated later [
Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department.
]. Similarly, our finding that the discrimination of NEWS2 for 24-hour mortality is marginally inferior to NEWS should be viewed with caution because only three NEWS2 studies of 9,717 patients fulfilled our inclusion criteria, and all reported emergency patients early in their treatment. Therefore, although NEWS or NEWS2 may not accurately predict the mortality of conditions for which there are effective treatments, such as sepsis [
] was not adhered to as the aim of this review was to determine from all the available literature, biased or otherwise, the range of discrimination for mortality and the mortality rates associated with different levels of NEWS at different prediction windows, in different patient populations. It intentionally only examined absolute mortality and did not consider relative risks or odds ratios and did not distinguish between those deaths that were preventable and those that were not. Our literature search was confined to papers on NEWS or NEWS2 and did not specifically search for ViEWS. For pragmatic reasons the final search strategy over all search engine platforms was limited to titles and abstracts. It is possible that papers that otherwise fulfilled our inclusion criteria but only referred to NEWS in the body of their text were missed. As our search ended on April 12th, 2021, it only included publications reporting the performance of NEWS in COVID-19 patients published up to that date.
4.3 Limitations of the evidence included in the review
The information available made it difficult to determine and compare the age and length of hospital stay for patients in different studies. Although most studies validated NEWS at a single point of time, commonly using the first observation recorded [
], treatments given before the observation were poorly documented. We could only determine patient diagnoses in those studies where they were specified, and we have no information on the quality of care provided. As authors often used the same data in different publications, we made every effort to ensure that the results for AUC, specific prediction windows and cut-off values were only used once.
NEWS requires the accurate measurement of vital signs and correct calculation [
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
]; only 13 studies (10.7%) explicitly stated that the vital sign measurements to calculate NEWS were recorded electronically and contemporaneously. Moreover, the measurement of respiratory rate is often influenced by subjective bias [
How accurate is the AVPU scale in detecting neurological impairment when used by general ward nurses? An evaluation study using simulation and a questionnaire.
] and the need for supplemental oxygen is subjective and may vary according to opinion, expertise and the availability of oxygen and other resources in different locations and clinical settings [
], and medication may also change NEWS performance; the AUC of NEWS for in-hospital mortality was lower in patients with suspected sepsis on hypertensive medication than those not on medication [
]. Despite these and other possible unconsidered confounders, we found only two studies that reported an AUC for 24-hour mortality <0.83, one used extensive imputation for missing data and was excluded [
Evaluation of the feasibility and performance of early warning scores to identify patients at risk of adverse outcomes in a low-middle income country setting.
4.4 Implications of the results for practice, policy, and future research
Whilst the risk of death increases as the cut-off increases (Fig. 3), the optimal point for intervention is unclear. The available literature we reviewed shows that if a cut-off of ≥7 points is selected, only 4% of patients would trigger an intervention and 44% of patients who die will be missed. Alternatively, a cut-off of ≥1 point will trigger an assessment and/or an intervention in 83% of patients, which may not benefit many of them. NEWS ≥5 points is the most adopted cut-off and 91% of patients are observed to have a NEWS below it; the overall 24-hour mortality of these patients <5 points is only 0.06%, but their in-hospital mortality averaged nearly 3% (range 0.65% to 24.7%). Moreover, a quarter of all deaths within 24-hours and more than 40% of all in-hospital deaths occur in patients with a NEWS <5. Although NEWS was never intended to be used as a diagnostic test for any specific condition, NEWS ≥5 has been recommended as a flag for sepsis [
Superior performance of National Early Warning Score compared with quick Sepsis-related Organ Failure Assessment Score in predicting adverse outcomes: a retrospective observational study of patients in the prehospital setting.
Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement program.
Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement program.
National early warning score at Emergency Department triage may allow earlier identification of patients with severe sepsis and septic shock: a retrospective observational study.
]. Moreover, many life-saving interventions, such as anti-coagulation for pulmonary embolus, thrombolysis for stroke, emergency surgery, rehydration to prevent acute kidney injury etc., should be given as soon as possible, and regardless of the patient's NEWS value.
5. Conclusion
NEWS reliably discriminates between patients who are most and least likely to die within 24-hours. It can, therefore, reliably identify sick patients who may need urgent attention, which is what it was designed to do. However, after 24-hours the prediction of mortality by NEWS declines and becomes unreliable.
The predictive performance of NEWS is likely to be lower if it is recorded before an effective treatment is delivered, and will be highest if no treatment is required, or there is no effective treatment, or if everything possible has already been done. Although NEWS or NEWS2 may be less likely to accurately predict the mortality of patients given urgent effective treatments, this does not mean that they cannot accurately identify patients who need them.
Although patients with a low NEWS score appear to have a reduced risk of death for several days, implying a degree of clinical and physiological stability, many of them die while still in hospital. The best way to anticipate deterioration in these patients remains unclear.
Authors’ contributions
All authors contributed to the preparation of this paper. MH and JK conceived the study and supervised the collection of the data and ensured its accuracy, analysed the data, and drafted the manuscript and critically revised the manuscript for intellectual content. Both authors read and approved the final manuscript and are guarantors of the paper.
Ethical approval
As there was no patient involvement ethical approval was not required for this study.
Declaration of Competing Interest
All costs were borne by the authors. John Kellett is a major shareholder, director, and chief medical officer of Tapa Healthcare DAC. Mark Holland has no potential conflicts of interest.
Acknowledgments
The authors wish to acknowledge access to anonymised unpublished data provided by the Kitovu Hospital Study Group, Kitovu Hospital, Masaka, Uganda, and the organising committee of the Society for Acute Medicine Benchmarking Audit (SAMBA).
National Early Warning Score (NEWS): standardising the assessment of acute illness severity in the NHS.
Report of a working party. RCP,
London2012(Accessed 9 October 2021, at file:///C:/Users/markd/Downloads/National%20Early%20Warning%20Score%20(NEWS)%20-%20Standardising%20the%20assessment%20of%20acute-illness%20severity%20in%20the%20NHS_0%20(2).pdf)
A comparison of the ability of the National Early Warning Score and the National Early Warning Score 2 to identify patients at risk of in-hospital mortality: a multi-centre database study.
A prospective, observational study of the performance of MEWS, NEWS, SIRS and qSOFA for early risk stratification for adverse outcomes in patients with suspected infections at the emergency department.
Evaluation of the feasibility and performance of early warning scores to identify patients at risk of adverse outcomes in a low-middle income country setting.
Prognostic accuracy of the quick sequential organ failure assessment score and national early warning scores in mortality rate of the non-traumatic patients.
A protocolised once a day modified early warning score (MEWS) measurement is an appropriate screening tool for major adverse events in a general hospital population.
The prognostic value of national early warning scores (NEWS) during transfer of care from community settings to hospital: a retrospective service evaluation.
Efficacy of prehospital National Early Warning Score to predict outpatient disposition at an emergency department of a Japanese tertiary hospital: a retrospective study.
Comparison of the National Early Warning Score (NEWS) and the Modified Early Warning Score (MEWS) for predicting admission and in-hospital mortality in elderly patients in the pre-hospital setting and in the emergency department.
Performance of externally validated enhanced computer-aided versions of the National Early Warning Score in predicting mortality following an emergency admission to hospital in England: a cross-sectional study.
How accurate is the AVPU scale in detecting neurological impairment when used by general ward nurses? An evaluation study using simulation and a questionnaire.
Superior performance of National Early Warning Score compared with quick Sepsis-related Organ Failure Assessment Score in predicting adverse outcomes: a retrospective observational study of patients in the prehospital setting.
Empiric antibiotic treatment reduces mortality in severe sepsis and septic shock from the first hour: results from a guideline-based performance improvement program.
National early warning score at Emergency Department triage may allow earlier identification of patients with severe sepsis and septic shock: a retrospective observational study.
Use of the first National Early Warning Score recorded within 24 h of admission to estimate the risk of in-hospital mortality in unplanned COVID-19 patients: a retrospective cohort study.
A comparison of the ability of the physiologic components of medical emergency team criteria and the U.K. National early warning score to discriminate patients at risk of a range of adverse clinical outcomes.
The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death.
Validation of modified early warning score using serum lactate level in community-acquired pneumonia patients. The National Early Warning Score-Lactate score.
Comparison of the National Early Warning Score+Lactate score with the pre-endoscopic Rockall, Glasgow-Blatchford, and AIMS65 scores in patients with upper gastrointestinal bleeding.
Evaluation of the risk prediction tools for patients with coronavirus disease 2019 in wuhan, china: a single-centered, retrospective, observational study.
Comparison of early warning scoring systems for hospitalized patients with and without infection at risk for in-hospital mortality and transfer to the intensive care unit.
The utility of the rapid emergency medicine score (REMS) compared with SIRS, qSOFA and NEWS for Predicting in-hospital Mortality among Patients with suspicion of Sepsis in an emergency department.
The most commonly used disease severity scores are inappropriate for risk stratification of older emergency department sepsis patients: an observational multi-centre study.
Investigating the discriminative value of Early Warning Scores in patients with respiratory disease using a retrospective cohort analysis of admissions to Nottingham University Hospitals Trust over a 2-year period.
Glucose as an additional parameter to National Early Warning Score (NEWS) in prehospital setting enhances identification of patients at risk of death: an observational cohort study.
Comparison of national early warning score, rapid emergency medicine score and acute physiology and chronic health evaluation II score for predicting outcome among emergency severe patients.
Zhonghua Wei Zhong Bing Ji Jiu Yi Xue.2017; 29: 1092-1096
Prognostic accuracy of the Hamilton Early Warning Score (HEWS) and the National Early Warning Score 2 (NEWS2) among hospitalized patients assessed by a rapid response team.
Decisions in clinical practice, especially in the acute setting where identification of potentially unstable patients is critical, are hard to be made. They rely, or should rely, on the clinical information of the patient, the best available knowledge, the expertise of the operator and the patient's preferences. Clinical prediction tools are often used as a “short-cut” to simplify and standardize such choices, without making errors.
00435-0&id=gr1.jpg){kind=link}
00435-0&id=gr2.jpg){kind=link}
00435-0&id=gr3.jpg){kind=link}
00435-0&id=gr4.jpg){kind=link}
00435-0&id=gr5.jpg){kind=link}