INFECTIOUS DISEASES, https://doi.org/10.1080/23744235.2020.

1784457
2020; VOL. 52,
NO. 10, 698–704

ORIGINAL ARTICLE

Predictive value of National Early Warning Score 2

(NEWS2) for intensive care unit admission in patients
with SARS-CoV-2 infection

Anna Gidaria , Giuseppe Vittorio De Socioa , Samuele Sabbatinib and Daniela Franciscia
a
Department of Medicine, Clinic of Infectious Diseases, ‘Santa Maria della Misericordia’ Hospital, University of Perugia, Perugia,
Italy; bDepartment of Medicine, Medical Microbiology Section, University of Perugia, Perugia, Italy

ABSTRACT
Background: From January 2020, Coronavirus disease 19 (COVID-19) has rapidly spread all over the world. An early assess-
ment of illness severity is important for the stratification of patients. We analysed the predictive value of National Early
Warning Score 2 (NEWS2) for intensive care unit admission (ICU) in patients with Severe Acute Respiratory Syndrome-
Coronavirus-2 (SARS-CoV-2) infection.

Methods: Data of 71 patients with SARS-CoV-2 admitted from 1 March to 20 April 2020, to the Clinic of Infectious Diseases
of Perugia Hospital, Italy, were retrospectively reviewed. NEWS2 at hospital admission, demographic, comorbidity and clin-
ical data were collected. Univariate and multivariate analyses were performed to establish the correlation between each
variable and ICU admission.

Results: Among 68 patients included in the analysis, 27 were admitted to ICU. NEWS2 at hospital admission was a good
predictor of ICU admission as shown by an area under the receiver-operating characteristic curve analysis of 0.90 (standard
error 0.04; 95% confidence interval 0.82–0.97). In multivariate logistic regression analysis, NEWS2 was significantly related
to ICU admission using thresholds of 5 and 7. No other clinical variables included in the model were significantly correlated
with ICU admission.

A NEWS2 threshold of 5 had higher sensitivity than a threshold of 7 (89% and 63%). Higher specificity, positive likelihood
ratio and positive predictive value were found using a threshold of 7 than a threshold of 5.

Conclusions: NEWS2 at hospital admission was a good predictor for ICU admission. Patients with severe COVID-19 were
correctly and rapidly stratified.

KEYWORDS ARTICLE HISTORY CONTACT

COVID-19 Received 11 May 2020 Anna Gidari
SARS-CoV-2 Revised 30 May 2020 [email protected]
NEWS2 Accepted 13 June 2020 Department of Medicine, Clinic of Infectious
ICU Diseases, University of Perugia, Piazzale Lucio
National Early Warning Score 2 Severi 1, Perugia, 06132, Italy

! 2020 Society for Scandinavian Journal of Infectious Diseases

 INFECTIOUS DISEASES 699

Introduction It has been demonstrated that high scores of NEWS/

NEWS2 at hospital admission are associated with deteri-
On 11 March 2020, following the rapid escalation of
oration of patients and poor clinical outcomes [8,9].
the new Severe Acute Respiratory Syndrome
The aim of this study was to evaluate NEWS2 at hos-
Coronavirus 2 (SARS-CoV-2), the World Health
pital admission of patients with COVID-19 as a predictor
Organisation declared Coronavirus disease 19 (COVID-
of ICU admission.
19) as a pandemic [1].
From February 2020, the Italian National Health
Service had to face a rapid increase of patient admis- Materials and methods
sions for COVID-19, many with severe respiratory failure. Patients and data collection
In Italy, the disease had an overall mortality of 12.6% [2].
Rapid national response was required, especially build- Consecutive adults with COVID-19 admitted to the Clinic
ing new dedicated clinical wards and, above all, increas- of Infectious Diseases of Santa Maria Della Misericordia
ing equipment and intensive care (ICU) unit beds [3]. Hospital of Perugia, Italy, were recruited. We retrospect-
Clinicians had to be aware that some patients rapidly ively analysed prospectively collected data of patients
deteriorate and require intensive care. Early assessment admitted with SARS-CoV-2 infection from 1 March to 20
of illness severity is of primary importance to stratify April 2020. We included all adult patients (age
patients and assure correct medical management [4]. "18 years) with an established diagnosis of SARS-CoV-2
The National Early Warning Score (NEWS) was devel- infection [10]. All data were obtained from electronic
oped to standardise measurement and evaluation of medical records.
physiological parameters in acutely ill patients. NEWS is Collected data were analysed anonymously. The study
used to identify and monitor patients at risk of serious was approved by the institutional ethics committee
clinical deterioration. The score can be used in prehospi- (Ethics Committee of the Umbria Region). All participat-
tal assessment and in emergency departments, and as a ing subjects provided oral informed consent to clinical
surveillance system for all hospitalised patients [5]. data collection that was transcribed in medical records.
NEWS includes respiratory rate, oxygen saturation, need Data on demographics, comorbidities and clinical
for supplemental oxygen, body temperature, blood pres- presentation were obtained for each patient. Charlson
sure, heart rate and level of consciousness (alert, verbal, Comorbidity Index and NEWS2 at hospital admission
pain, unresponsive, AVPU). NEWS2 is an update of NEWS were calculated using medical records. Exclusion criteria
which also includes a dedicated oxygen scale for were: patient not eligible for intensive care unit admis-
patients with hypercapnic respiratory failure and the sion, death before intensive care specialist evaluation,
addition of ‘new confusion’ as ‘C’ to the AVPU score and lack of data needed to calculate NEWS2. The pri-
which became ACVPU. It identifies need for urgent clin- mary outcome was ICU admission within 21 days from
ical action in patients with a score of 5 or above [6]. In hospital admission.
the additional implementation guidance of 2020, the
Royal College of Physicians recommends four trigger lev-
Statistical analysis
els for clinical alert and consequent response:
We assessed the potential of NEWS2 thresholds of 5 and
! Null score (0): monitor patient every 12 hours. 7 to predict ICU admission by the corresponding sensi-
! Low score (1–4): monitor patients every 4–6 hours. tivity and specificity, positive and negative predictive
! A single red score (3 in a single parameter): rapid value, positive and negative likelihood ratio, and accur-
clinical evaluation to establish the cause. acy. Each value is reported with 95% confidence inter-
! Medium score (5–6): patients should initially be moni- vals (CI).
tored hourly because the score indicates a potentially The relationship between NEWS2 scores and ICU
serious acute clinical deterioration. Assessment is rec- admission was evaluated by correlation analysis. The
ommended within 1 hour. test was chosen after verifying that data followed a nor-
! High score (7 or above): monitor patients every mal distribution.
30 minutes initially and complete the assessment ICU admission was evaluated by receiver operating
within 30 minutes [7]. characteristic (ROC) curve analysis, describing areas
700 A. GIDARI ET AL.

under curves with 95% CI and comparing to the null where they received invasive respiratory support. The
hypothesis (area ¼ 0.5) [11–13]. The probability value (p) median time between hospitalisation and ICU admission
of the ROC curve was assessed using the Mann–Whitney was 3.0 days (interquartile range 1.5–5.5 days).
U test [13]. Demographic, comorbidity and clinical data are listed
To assess if other variables might be correlated to ICU in Table 1. Mean age was 64 ± 14 years (mean ± SD), and
admission, multivariate logistic regression analysis was 66% were male. Hypertension, dyslipidemia and type 2
performed. Variables selected, based on clinical rele- diabetes mellitus were the most frequent comorbidities
vance, were gender, NEWS2, Charlson Comorbidity Index (40%, 15% and 10% respectively). Subcategorising based
and hypertension. Before multivariate analysis, univariate on NEWS2 highlighted small but not significant differen-
analysis was performed applying Chi-square or ces between ICU and non-ICU patients.
Mann–Whitney U-tests as appropriate. Multivariate logis- At admission, PiO2/FiO2 ratio averaged 268 ± 118 mmHg
tic regressions were performed for NEWS2 with thresh- (mean ± SD). Fifty percent of patients with NEWS2 "5 and
olds of 5 and 7. The results of multivariate logistic "7 had a PiO2/FiO2 >200 mmHg, 45% and 39%, respect-
regression are given as odds ratio with 95% CI and ively, had between 100 and 200 mmHg. Only two patients
p value. had a ratio <100 mmHg, and both had a NEWS2 "7.
Statistical analyses were performed using Prism In the entire cohort, 88% presented radiological findings
Graphpad 7 software. A p < .05 was considered of pneumonia compatible with COVID-19. Figure 2(A)
significant. shows the distribution of NEWS2: patients subsequently
transferred to the ICU had a median score of 8 (IQR 6-8)
and those who were not had a median score of 3 (IQR
Results
1–5), (p < .0001).
As shown in Figure 1, during the study period, 71 NEWS2 showed a normal distribution in patients
patients were hospitalised for SARS-CoV-2. Three admitted to the ICU according to Kolmogorov–Smirnov
patients were excluded from the analysis. One was normality test. NEWS2 was positively correlated to ICU
excluded due to insufficient data to calculate NEWS2 at admission (r ¼ 0.91, 95% CI 0.70-0.97, p < .0001)
admission. The other two patients had NEWS2 of 7 and (Figure 2(B)).
9, and presented with severe acute disease that led to The prognostic accuracy of NEWS2 for prediction of
death. The first patient had advanced neoplastic disease ICU admission is summarised in Table 2. Using the
and was not eligible for ICU admission, while the second threshold of 5, we obtained higher sensitivity than with
died before intensive care specialist evaluation. the threshold of 7 (89% and 63% respectively). A higher
Therefore, 68 patients were included in the analysis. specificity was attained with the threshold of 7 (98%)
Among these, 27 patients were admitted to ICU (38%) compared to the threshold of 5 (69%). The threshold of

Figure 1. Patients selection flow chart. COVID-19: Coronavirus disease 2019; NEWS2: National Early Warning Score 2; ICU: Intensive
Care Unit.
 INFECTIOUS DISEASES 701

Table 1. Demographics, comorbidities and clinical presentation.

Total population NEWS2 " 5 (N 38) NEWS2 " 7 (N 18)
Total ICU Non-ICU ICU Non-ICU ICU Non- ICU
N 68 27 41 24 14 17 1
Age:
Average, years (range) 64 (31–93) 65 (48–83) 63 (31–93) 65 (47–82) 68 (31–90) 66 (48–83) 69
SD, (years) 14 9 17 9 18 10
Sex:
Male (%) 66 81 56 79 57 76 100
Chronic lung disease (%) 7 7 7 8 4 0 0
Chronic renal impairment (%) 1 0 2 0 7 0 0
High blood pressure (%) 40 41 39 46 50 41 0
Chronic ischaemic cardiomyopathy (%) 6 7 5 4 14 6 0
Type 2 diabetes mellitus (%) 10 7 12 8 21 6 0
Dyslipidemia (%) 15 7 20 4 21 0 100
PiO2/FiO2 at admission (N 64)a: N 64 N 27 N 37 N 24 N 14 N 17 N1
>200 (%) 67 48 81 42 64 50 0
100-200 (%) 30 44 19 50 36 39 100
<100 (%) 3 7 0 8 0 11 0
Radiological findings of pneumonia (%) 88 100 80 100 93 100 100
Abbreviations: NEWS2: National Early Warning Score 2; SD: standard deviation; ICU: intensive care unit.
a
Blood gas analysis of 4 patients was not performed.

Table 2. Prognostic accuracy of NEWS2 for ICU admission using

two different threshold values: 5 and 7 (N ¼ 68).
NEWS2 " 5 NEWS2 " 7
Sensitivity, % (95% CI) 89 (70.8–97.7) 63 (42.4–80.6)
Specificity, % (95% CI) 66 (49.4–79.9) 98 (87.1–99.9)
Positive likelihood ratio (95% CI) 2.6 (1.7–4.1) 25.9 (3.7–182.8)
Negative likelihood ratio (95% CI) 0.2 (0.1–0.5) 0.4 (0.2–0.6)
Positive predictive value, % (95% CI) 63 (52.3–72.8) 94 (70.6–99.2)
Negative predictive value, % (95% CI) 90 (75.2–96.4) 80 (70.9–86.8)
Accuracy, % (95% CI) 75 (63–84.7) 84 (72.9–91.6)
Abbreviations: NEWS: National Early Warning Score; CI: confidence interval.

negative likelihood ratios were 0.17 and 0.38, and the

negative predictive values 90% and 80% for the thresh-
olds of 5 and 7, respectively. Accuracy was higher for
the threshold of 7 (84% vs 75%). The ROC curve of
NEWS2 for predicting ICU admission is depicted in
Figure 3. The AUROC curve was 0.90 (standard error, SE,
0.04; 95% CI 0.82–0.97; p < .0001).
Being male was associated with ICU admission
(p ¼ .038). Charlson Comorbidity Index and hypertension
did not differ significantly between patients admitted
and not admitted to the ICU. Table 3 shows the multi-
variate logistic regression analyses. Only high NEWS2
was significantly related to ICU admission using both
Figure 2. National Early Warning Score 2 (NEWS2) of Coronavirus
disease 2019 (COVID-19) patients at hospital admission. (A) Patients the threshold of 5 (OR 18.05; 95% CI 4.80–92.86;
were divided into two groups: intensive care unit (ICU) patients, p < .0001; Hosmer–Lemeshow test was not significant,
who required ICU admission, and non-ICU patients, who did not p ¼ .43) and 7 (OR 77.77; 95% CI 12.31–1619; p ¼ .0001;
require ICU admission. NEWS2 is expressed as median with inter-
quartile range. $p < .0001. (B) Correlation between NEWS2 of
Hosmer–Lemeshow test was not significant, p ¼ .18).
COVID-19 patients at hospital admission and percentages of patients
subsequently admitted to ICU. Pearson r correlation coefficient and
p value are shown in the plot. Discussion
The aim of this study was to assess if high NEWS2 at
7 had a high positive likelihood ratio (25.9) and high hospital admission predicts subsequent ICU transfer of
positive predictive value (94%), while using the thresh- COVID-19 patients. A high NEWS2 showed good predict-
old of 5, we obtained 2.6 and 63%, respectively. The ive value for ICU admission (AUROC: 0.90).
702 A. GIDARI ET AL.

To the best of our knowledge, this is the first study proven. Our results support the use of this score to
to evaluate the performance of NEWS2 in patients with evaluate COVID-19 patients at hospital admission.
COVID-19. Hu et al. recently underlined the importance According to our results, NEWS2 could be useful in
of rapid scoring systems for the clinical evaluation of two ways:
patients with COVID-19. They compared the Modified
Early Warning Score (MEWS) and the Rapid Emergency ! Using a threshold of 5 results in high sensitivity
Medicine Score (REMS) as predictors of mortality in (89%), high negative predictive value (90%) and good
COVID-19. They concluded that REMS, with a high nega- negative likelihood ratio (0.17). It means that a
tive predictive value (96.8%), is useful in stratification of patient with a score <5 has low probability of ICU
critically ill patients [14]. The primary outcome of our admission. We suggest monitoring patients with
study was not mortality but ICU admission. This choice NEWS2 <5 every 4-6 hours or less, following the
was based on: (1) the importance of identifying patients Royal College of Physicians guidelines [6]. We also
who may rapidly deteriorate and need resuscitation recommend strict monitoring of patients with NEWS2
assessment; (2) the epidemic in Italy, as yet, is in an ear- "5 because a considerable proportion may rapidly
lier phase than in China, so most patients are hospital- progress to severe respiratory failure.
ised, and it was not possible to estimate the mortality in ! Using a threshold of 7 results in high specificity
our cohort at this time. (98%), high positive likelihood ratio (26) and high
We have chosen NEWS2 as it is easily and rapidly positive predictive value (94%). This indicates a high
applicable and has shown high predictive capacity for probability of need for later ICU admission. Based on
mortality in other critically ill patients [15,16]. NEWS and
our results, we propose rapid assessment of patients
NEWS2 can also be used in the prehospital setting to
in a sub-intensive care unit with strict monitoring
support ambulance clinician decision making [17].
and preliminary evaluation by an intensive care spe-
The potential benefits of NEWS2 in prehospital stratifi-
cialist to discuss further treatment.
cation of COVID-19 patients are presumable but not yet
Interestingly, the only patient in our cohort with NEWS2
"7 who did not require ICU admission, had very severe
respiratory failure but improved rapidly after treatment
with tocilizumab [18–21].
Our results in patients with COVID-19 are in line with
the recent guidelines published by the Royal College of
Physicians that recommend a greater urgency in man-
agement of patients with NEWS2 "7 than in those with
scores "5 [6].
The main limitations of our study are the retrospect-
ive single-center study design and the small number of
patients. The results need to be confirmed in studies
with higher numbers of patients.
Figure 3. Receiver Operating Characteristic (ROC) curve for patients The main strength of this study is the well-character-
ICU admission using NEWS2 of COVID-19 patients at hospital admis- ized cohort with low missing data and the high predict-
sion. NEWS2 showed strong predictive ability with an area under
ive value of NEWS2 found in COVID-19 patients for ICU
the ROC (AUROC) curve of 0.90 (p < .0001).
admission. Furthermore, we also individuated two

Table 3. Multivariate logistic regression analyses.

Multivariate analysis NEWS2 " 5 Multivariate analysis NEWS2 " 7
Outcome: ICU admission OR (95% CI, p) OR (95% CI, p)
Charlson Comorbidity index 0.9 (0.6–1.2, p ¼ .40) 1 (0.7–1.3, p ¼ .77)
Hypertension 0.9 (0.2–3.1, p ¼ .8) 1.5 (0.4–6.3, p ¼ .58)
Sex (male vs female) 3.4 (0.9–14.1, p ¼ .08) 4.5 (1–33, p ¼ .08)
NEWS2 18.05 (4.8–92.9, p <.0001) 77.8 (12.3–1619, p ¼ .0001)
HL¼ 8.1 (p ¼ .43) 10.4 (p ¼ .24)
Abbreviations: NEWS: National Early Warning Score; OR: Odds ratio; CI: confidence interval; HL: Hosmer–Lemeshow test.
 INFECTIOUS DISEASES 703

different thresholds that could help clinicians to cor- pneumonia in Wuhan, China: a descriptive study. Lancet.
rectly classify these patients. 2020;395(10223):507–513.
[5] Royal College of Physicians. National Early Warning Score
(NEWS): standardising the assessment of acute-illness sever-
Conclusions ity in the NHS. 2012. Available from: https://www.rcplon-
don.ac.uk/projects/outputs/national-early-warning-score-
In a pandemic like COVID-19, rapid assessment of news-2.
patients is essential. A fast and easily calculated score, [6] Royal College of Physicians. National Early Warning Score
such as NEWS2 could help clinicians manage acutely ill (NEWS) 2: standardising the assessment of acute-illness
patients. The correlation between NEWS2 and ICU severity in the NHS. 2017. Available from: https://www.
rcplondon.ac.uk/projects/outputs/national-early-warning-
admission makes this scoring system useful in clinical
score-news-2.
practice to correctly and rapidly treat patients with [7] Royal College of Physicians. NEWS2: additional implementa-
severe disease. tion guidance. 2020. Available from: https://www.rcplon-
don.ac.uk/projects/outputs/news2-additional-implementa-
tion-guidance.
Acknowledgement [8] Scott LJ, Redmond NM, Tavar"e A, et al. Association
We would like to thank Prof. Stefano Ricci for support on statis- between National Early Warning Scores in primary care and
tical analysis. clinical outcomes: an observational study in UK primary
and secondary care. Br J Gen Pract. 2020;70(695):
e374–e380.
Author contributions [9] Echevarria C, Steer J, Bourke SC. Comparison of early warn-
ing scores in patients with COPD exacerbation: DECAF and
AG, GDS and DF contributed conception and design of the
NEWS score. Thorax. 2019;74(10):941–946.
study; AG and SS contributed analysis and interpretation of the
[10] World Health Organization. Clinical management of severe
data; AG wrote the first draft of the manuscript; AG, GDS, SS
acute respiratory infection when COVID-19 is suspected.
and DF wrote sections of the manuscript. All authors contributed 2020. Available from: https://www.who.int/publications-
to critical manuscript revision, read and approved the submit- detail/clinical-management-of-severe-acute-respiratory-
ted version. infection-when-novel-coronavirus-(ncov)-infection-is-
suspected.
[11] Fernando SM, Fox-Robichaud AE, Rochwerg B, et al.
Disclosure statement
Prognostic accuracy of the Hamilton Early Warning Score
No potential conflict of interest was reported by the author(s). (HEWS) and the National Early Warning Score 2 (NEWS2)
among hospitalized patients assessed by a rapid response
team. Crit Care. 2019;23(1):60.
ORCID [12] Pimentel MAF, Redfern OC, Gerry S, et al. A comparison of
Anna Gidari http://orcid.org/0000-0002-6556-6553 the ability of the National Early Warning Score and the
Giuseppe Vittorio De Socio http://orcid.org/0000-0001- National Early Warning Score 2 to identify patients at risk
8774-4843 of in-hospital mortality: a multi-centre database study.
Samuele Sabbatini http://orcid.org/0000-0001-8334-8325 Resuscitation. 2019;134:147–156.
Daniela Francisci http://orcid.org/0000-0001-8752-8278 [13] Hanley JA, McNeil BJ. The meaning and use of the area
under a receiver operating characteristic (ROC) curve.
Radiology. 1982;143(1):29–36.
References [14] Hu H, Yao N, Qiu Y. Comparing rapid scoring systems in
[1] World Health Organization. WHO Director-General’s open- mortality prediction of critical ill patients with novel cor-
onavirus disease. Acad Emerg Med. 2020;27(6):461–468.
ing remarks at the mission briefing on COVID-19. 2020.
[15] Mart"ın-Rodr"ıguez F, Lo
"pez-Izquierdo R, Del Pozo Vegas C,
Available from: https://www.who.int/dg/speeches/detail/
et al. Accuracy of National Early Warning Score 2 (NEWS2)
who-director-general-s-opening-remarks-at-the-media-brief-
in prehospital triage on in-hospital early mortality: a multi-
ing-on-covid-19—11-march-2020. center observational prospective cohort study. Prehosp
[2] Istituto Superiore di Sanit!a. COVID-19 integrated surveil- Disaster Med. 2019;34(6):610–618.
lance: key national data. 2020. Available from: https://www. [16] Mellhammar LT, et al. NEWS2 is superior to qSOFA in
epicentro.iss.it/en/coronavirus/bollettino/Infografica_ detecting sepsis with organ dysfunction in the emergency
27aprileENG.pdf. department. J Clin Med. 2019;8:E1128.
[3] Remuzzi A, Remuzzi G. COVID-19 and Italy: what next?. [17] Shaw J, Fothergill RT, Clark S, et al. Can the prehospital
Lancet. 2020;395(10231):1225–1228. National Early Warning Score identify patients most at risk
[4] Chen N, Zhou M, Dong X, et al. Epidemiological and clinical from subsequent deterioration? Emerg Med J. 2017;34(8):
characteristics of 99 cases of 2019 novel coronavirus 533–537.
704 A. GIDARI ET AL.

[18] Fu B, Xu X, Wei H. Why tocilizumab could be an effective [20] Perricone C, Triggianese P, Bartoloni E, et al. The anti-viral
treatment for severe COVID-19?. J Transl Med. 2020;18(1): facet of anti-rheumatic drugs: lessons from COVID-19. J
164. Autoimmun. 2020;111:102468.
[19] Zhang C, Wu Z, Li JW, et al. The cytokine release syndrome [21] Alhazzani W, Møller MH, Arabi YM, et al. Surviving Sepsis
(CRS) of severe COVID-19 and Interleukin-6 receptor (IL-6R) Campaign: guidelines on the management of critically ill
antagonist tocilizumab may be the key to reduce the mor- adults with coronavirus disease 2019 (COVID-19). Intensive
tality. Int J Antimicrob Agents. 2020;55(5):105954. Care Med. 2020;46(5):854–887.