Early prognostication of COVID-19 to guide hospitalisation versus outpatient monitoring using a point-of-test risk prediction score

Abstract

Introduction
Risk factors of adverse outcomes in COVID-19 are defined but stratification of mortality using non-laboratory measured scores, particularly at the time of prehospital SARS-CoV-2 testing, is lacking.

Methods
Multivariate regression with bootstrapping was used to identify independent mortality predictors in patients admitted to an acute hospital with a confirmed diagnosis of COVID-19. Predictions were externally validated in a large random sample of the ISARIC cohort (N=14 231) and a smaller cohort from Aintree (N=290).

Results
983 patients (median age 70, IQR 53–83; in-hospital mortality 29.9%) were recruited over an 11-week study period. Through sequential modelling, a five-predictor score termed SOARS ( S pO2, O besity, A ge, R espiratory rate, S troke history) was developed to correlate COVID-19 severity across low, moderate and high strata of mortality risk. The score discriminated well for in-hospital death, with area under the receiver operating characteristic values of 0.82, 0.80 and 0.74 in the derivation, Aintree and ISARIC validation cohorts, respectively. Its predictive accuracy (calibration) in both external cohorts was consistently higher in patients with milder disease (SOARS 0–1), the same individuals who could be identified for safe outpatient monitoring. Prediction of a non-fatal outcome in this group was accompanied by high score sensitivity (99.2%) and negative predictive value (95.9%).

Conclusion
The SOARS score uses constitutive and readily assessed individual characteristics to predict the risk of COVID-19 death. Deployment of the score could potentially inform clinical triage in preadmission settings where expedient and reliable decision-making is key. The resurgence of SARS-CoV-2 transmission provides an opportunity to further validate and update its performance.

Full text

Original research
Early prognostication of COVID-19 to guide
hospitalisation versus outpatient monitoring using a
point- of- test risk predictionscore
Felix Chua ,1,2 Rama Vancheeswaran,3 Adrian Draper,4 Tejal Vaghela,5
Matthew Knight,3 Rahul Mogal,3 Jaswinder Singh,6 Lisa G Spencer ,7
Erica Thwaite,8 Harry Mitchell,3 Sam Calmonson,3 Noor Mahdi,3 Shershah Assadullah,3
Matthew Leung,3 Aisling O’Neill,3 Chhaya Popat,3 Radhika Kumar,3
Thomas Humphries,7 Rebecca Talbutt,7 Sarika Raghunath,7 Philip L Molyneaux ,1,2
Miriam Schechter,5 Jeremy Lowe,5 Andrew Barlow3
Respiratory infection
To cite: ChuaF,
VancheeswaranR, DraperA,
etal. Thorax Epub ahead of
print: [please include Day
Month Year]. doi:10.1136/
thoraxjnl-2020-216425
►Prepublication history
and additional material is
published online only. To view
please visit the journal online
(http:// dx. doi. org/ 10. 1136/
thoraxjnl- 2020- 216425).
For numbered affiliations see
end of article.
Correspondence to
Dr Felix Chua, Interstitial Lung
Disease Unit, Department of
Respiratory Medicine, Royal
Brompton and Harefield NHS
Foundation Trust, London SW3
6NP, UK; f. chua@ rbht. nhs. uk
FC, RV and AD contributed
equally.
Received 19 October 2020
Revised 17 January 2021
Accepted 18 January 2021
© Author(s) (or their
employer(s)) 2021. No
commercial re- use. See rights
and permissions. Published
by BMJ.
ABSTRACT
Introduction Risk factors of adverse outcomes in
COVID-19 are defined but stratification of mortality using
non- laboratory measured scores, particularly at the time
of prehospital SARS- CoV-2 testing, is lacking.
Methods Multivariate regression with bootstrapping
was used to identify independent mortality predictors in
patients admitted to an acute hospital with a confirmed
diagnosis of COVID-19. Predictions were externally
validated in a large random sample of the ISARIC cohort
(N=14 231) and a smaller cohort from Aintree (N=290).
Results 983 patients (median age 70, IQR 53–83;
in- hospital mortality 29.9%) were recruited over an
11- week study period. Through sequential modelling,
a five- predictor score termed SOARS (SpO2, Obesity,
Age, Respiratory rate, Stroke history) was developed to
correlate COVID-19 severity across low, moderate and
high strata of mortality risk. The score discriminated
well for in- hospital death, with area under the receiver
operating characteristic values of 0.82, 0.80 and 0.74
in the derivation, Aintree and ISARIC validation cohorts,
respectively. Its predictive accuracy (calibration) in both
external cohorts was consistently higher in patients
with milder disease (SOARS 0–1), the same individuals
who could be identified for safe outpatient monitoring.
Prediction of a non- fatal outcome in this group was
accompanied by high score sensitivity (99.2%) and
negative predictive value (95.9%).
Conclusion The SOARS score uses constitutive and
readily assessed individual characteristics to predict
the risk of COVID-19 death. Deployment of the score
could potentially inform clinical triage in preadmission
settings where expedient and reliable decision- making
is key. The resurgence of SARS- CoV-2 transmission
provides an opportunity to further validate and update
its performance.
INTRODUCTION
Rapid and accurate prediction of the probability of
adverse clinical outcomes is central to the manage-
ment of global outbreaks of infection.1–3 Stratifica-
tion by predicted risk, most commonly for death,
can support clinical judgement and potentially assist
clinicians in community settings to decide how
urgently to refer patients to hospital. Used appro-
priately, predictive scores can also help inform
treatment- related decision- making. The pandemic
caused by SARS- CoV-2 lends itself to predictive
modelling by having a large at- risk population and
a high adverse event rate including death.4
Although the recent incidence of COVID-19 has
decreased in some parts of the world at the time
of writing, many countries are already experiencing
a ‘second wave’ of new cases.5–8 An increase in
incident cases in many localities is already evident
in the UK. It is widely anticipated that viral trans-
mission will continue to surge in the months
ahead, particularly with the onset of winter in the
northern hemisphere. Not all patients infected
with SARS- CoV-2 will require hospitalisation but
even among those who initially experience mild
symptoms, a sizeable proportion remain at risk
of subsequent life- threatening clinical decline.
The availability of a practical prehospital predic-
tive tool to triage patients for safe discharge to an
outpatient (virtual) monitoring system versus direct
Key messages
What is the key question?
►Can patients with COVID-19 be risk stratified
in the prehospital setting without laboratory-
measured data?
What is the bottom line?
►A five- predictor risk prediction score
(SOARS) based on demographic and clinical
characteristics can quickly and reliably identify
COVID-19- positive patients who have a
low probability of mortality for outpatient
monitoring and management.
Why read on?
►Information from the prognostication of SARS-
CoV-2- infected individuals early in their illness
can be used to guide clinical decision- making
with respect to the level of subsequent care.
1ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
admission to hospital for observation or treatment would be
highly advantageous.
Reliable prediction tools to differentiate between levels and
sites of clinical care already exist and have been successfully
implemented in prehospital practice. For example, both the
CURB-65 and the CRB-65 scoring systems for the assessment
of community- acquired pneumonia include recommendations
for out- of- hospital care.9 10 Recent research by the ISARIC- 4C
Consortium has provided an accurate tool to similarly prognos-
ticate for COVID-19- attributed death in hospitalised patients
but its reliance on laboratory- measured indices limits its applica-
bility outside the institutional environment.11 Prognostic evalu-
ation of individuals with suspected SARS- CoV-2 infection at the
time of diagnostic testing is potentially achievable but has yet to
be examined.
Our objective was to develop and evaluate an easy- to- apply
and accurate prognostic score to predict mortality and aid early
clinical decision- making by identifying patients infected with
SARS- CoV-2 who might benefit from an urgent hospital assess-
ment. To develop the initial risk score, we used multivariate
logistic regression to explore the relationships between a large
panel of candidate predictors and COVID-19 death. Iterative
modelling resulted in a pragmatic predictive score based on five
widely available patient variables. The scoring of patients against
these selected predictors permitted three distinct risk classes
to be defined. The performance of the score was then assessed
against two validation cohorts—a large subgroup of the ISARIC
study patients and a smaller single- hospital cohort, the latter to
better reflect local population characteristics and practice.
METHODS
Study design and characteristics of the derivation cohort
All individuals aged 18 or older who tested positive for
SARS- CoV-2 nucleic acid by real- time reverse transcriptase PCR
between 1 March and 16 May 2020 after presenting to the emer-
gency department (ED) at Watford Hospital, West Hertfordshire
NHS Hospitals Trust were prospectively recruited. Baseline
clinical characteristics and investigation results were collected
according to a prespecified protocol. Patients were either
referred to the virtual hospital (VH) for outpatient monitoring
or admitted to a medical ward.
Laboratory, physiologic and radiographic data
All laboratory tests were performed as part of routine clinical
care. Nasopharyngeal mucosal swabs for rRT- PCR were couri-
ered to the regional UK Public Health England laboratory.
Baseline vital observations included all the parameters recom-
mended by the National Early Warning Score.12 Chest radio-
graphs acquired in ED were collated and scored at the end of the
recruitment period.
Location and level of care
After presentation, patients who were clinically judged to have
mild illness were referred to the VH for subsequent moni-
toring. To avoid missing early clinical deterioration in the post-
assessment period, they were observed for up to 24 hours in
hospital. Patients who remained admitted after the first 24 hours
but who did not require additional respiratory support beyond
wall- based oxygen were managed on designated medical wards.
Where clinically indicated, continuous positive airway pressure
(CPAP) was provided on such wards or on the intensive care unit
(ICU); intubation and mechanical ventilation were undertaken
on the ICU.
Identifying predictors of death in the derivation cohort
The primary outcome of the study was in- hospital death.
TRIPOD (Transparent Reporting of a multivariable predic-
tion model for Individual Prognosis or Diagnosis) recommen-
dations were followed for multivariate model evaluation and
reporting.13 Seventy- five baseline clinical and non- clinical vari-
ables were initially collected based on their reported association
with COVID-19 and analysed by univariate and multivariate
logistic regression with bootstrap resampling.14 15 Of these, vari-
ables with numerically small ORs or a p value of >0.05 were not
included in the final analysis. Candidate predictors of death were
assessed for potential clustering effects and missing at random
values were addressed by multiple imputation with chained
equations (MICE),16 with 10–20 random draws to account for
data variability.
Development and external validation of the clinical risk score
The large external cohort comprised a randomly selected
subpopulation of the ISARIC 4C derivation population
(N=20 000 provided; 14 231 with complete data for scoring).
The primary data of these individuals were submitted by 260
hospitals across England, Scotland and Wales to the prospective
ISARIC WHO Clinical Characterization Protocol UK (CCP- UK)
study.11 We also tested our score against a smaller population
of SARS- CoV-2- positive cases from Aintree Hospital, Liverpool
(N=303 provided; N=290 with complete data for scoring) as a
single- setting validation control.
In the initial stages, a preliminary score comprising 12 inde-
pendent predictors of death, including care home residency, was
developed; to enable external validation against the ISARIC
cohort (which did not include residential data), care home status
was excluded as a variable to yield an 11- predictor score. Its
ability to discriminate for in- hospital mortality was assessed by
the area under the receiver operating characteristic (AUROC).
From this score, a condensed version comprising five clinical
predictors was developed for prehospital application. Mortality
cut- points at each risk level were assessed to define mild,
moderate and high risk classes, followed by determination of
positive and negative predictive values, as well as sensitivity and
specificity thresholds. Model performance was further assessed
by calibration using a graphical representation of the Hosmer-
Lemeshow ‘goodness- of- fit’ test to depict agreement between
the expected (predicted) and observed (actual) outcome across
the entire COVID-19 severity range in both validation cohorts.17
The summary relationship between the dependent variable
(death) and different levels of disease severity in the external
ISARIC population was expressed as McFadden’s R2.18
Statistical analysis
Categorical variables were expressed as frequency (%), with
significance determined by the χ2 test. Continuous variables
were expressed as median (IQR) or mean (SD) and analysed
by the t- test, Kruskal- Wallis or Mann- Whitney U test, as appro-
priate. ORs were assessed as unadjusted and adjusted values with
respect to in- hospital death, the latter determined by multivar-
iate regression with bootstrapping of 1000 resamples. We used
this method as internal validation to improve statistical infer-
ence by deriving a better estimate of the sampling distribution.
Bootstrapping involves randomly drawing repeat samples from
the core dataset to calculate SEs and CIs for the final regres-
sion analysis. MICE was used to generate valid estimates of
randomly missing values in the derivation model. A p value
of <0.05 was considered statistically significant. All statistical
2ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
analyses including risk modelling calculations were performed
using STATA, V.16 (Stata, Texas, USA).
RESULTS
Baseline demographic and clinical characteristics of the
derivation cohort
Nine hundred eighty- three patients (52.5% male) confirmed as
SARS- CoV-2 rRT- PCR positive were recruited over the 11- week
study period. Five patients remained in hospital at the time of
data cut- off on 31 May 2020. The median age of the cohort was
70 (IQR 53–83; range 23–99); median age was lowest in the
virtual hospital pathway (53; IQR 43–67) and highest among
hospitalised patients who did not receive CPAP (77; IQR 61–86)
(p<0.001) (online supplemental appendix 1).
The most common comorbidities were hypertension (48.4%),
pulmonary disease (30.0%), cardiac disease (26.6%), diabetes
mellitus (23.6%), chronic kidney disease (CKD; 20.0%)
and dementia (15.4%). Obesity, defined as body mass index
(BMI)>30, was present in 24.7% (243) of the cohort and is asso-
ciated with an unadjusted OR for death of 1.40 (95% CI 1.18
to 2.68, p<0.05).
Overall, 294 out of 983 (29.9%) patients died in hospital, the
vast majority (97.3%) aged 50 or older. The mortality rates of
different age brackets in the cohort (compared with the ISARIC
and Aintree validation cohorts) are shown in online supplemental
appendix 2. The univariate OR for death increased with rising
age, and was 14.86 (95% CI 6.89 to 32.04) for those aged 70–79
and 20.87 (95% CI 9.93 to 43.86) for those aged 80 or older
(table 1). When stratified by maximal levels of care, mortality
rate was lowest in the VH (1.8%) and highest in the ICU group
(62.1%) (p<0.001) (online supplemental appendix 3).
White Caucasian ethnicity constituted 77.3% (760/983) of
the whole cohort, while Asian, Black and other minor ethnici-
ties (BAME) represented 16.5%, 4.5% and 1.7%, respectively.
Overall, white ethnicity was associated with the highest propor-
tion of non- survivors (85.0%); in comparison, patients of Asian
(OR 0.57, 95% CI 0.38 to 0.85, p<0.01) or black (OR 0.39,
95% CI 0.17 to 0.88, p<0.05) background in this cohort had
lower ORs for death from COVID-19. The proportion of non-
survivors within each ethnic group was also highest in white
(32.9%), followed by Asian (21.6%), Black (15.9%) and other
minority groups (17.6%) (p<0.01). Of note, white patients were
significantly older by median age (74, IQR 58–85) compared
with Asian (57, IQR 46–71; p<0.0001) or black (58, IQR
50–72; p<0.001).
Care home residency (204/983; 20.8%) was more common
among non- survivors (p<0.001) and was associated with
an unadjusted OR for death of 3.14 (95% CI 2.28 to 4.32,
p<0.001). Based on data from 644 patients aged 65 or older,
the univariate OR of frailty for death was 2.52 (95% CI 1.73 to
3.69; p<0.001) in those with a group 1 frailty score and 2.56
(95% CI 1.62 to 4.06; p<0.001) in those with a group 2 frailty
score (table 1).
The median time from symptom onset to presentation for the
derivation cohort was 6 days (IQR 2.0–11.0), with no difference
between survivors and non- survivors. The four most common
reported symptoms were fever (61%), breathlessness (57.9%),
cough (52.8%) and myalgia (21.7%). Tachypnoea (respiratory
rate>24/minute) and hypoxia (SpO2≤92% on ambient air)
were evident in 35.9% and 31.4% of patients, respectively, and
is associated with crude OR for death of 2.15 (95% CI 1.63
to 2.84, p<0.001) and 3.74 (95% CI 2.73 to 5.12, p<0.001),
respectively. C reactive protein>50 mg/L was more frequently
documented in non- survivors (76.7% vs 58.5%; p<0.001) and
associated with an unadjusted OR for mortality of 2.40 (95%
CI 1.73 to 3.32, p<0.001). Lymphopenia was similarly more
common in non- survivors (44.7% vs 29.1%; p<0.001), with
an unadjusted OR for death of 1.97 (95% CI 1.47 to 2.64;
p<0.001). A baseline chest radiograph (CXR) was available in
91% (895/983) patients; abnormalities in ≥4 radiographic zones
were evident in 338 (37.8%) of cases and was associated with
increased mortality on univariate analysis (OR 1.89, 95% CI
1.42 to 2.52, p<0.001).
Multivariate regression for independent risk factors of
mortality
The bootstrapped multivariate regression analysis included
the whole derivation cohort of 983 patients comprising 689
(70.1%) survivors and 294 (29.9%) non- survivors with complete
or multiply imputed values for data. Older age, CKD stage 5,
baseline hypoxia, elevated BMI, tachypnoea, leucocytosis and a
history of stroke were identified as the strongest independent
predictors of mortality (table 1). In particular, age ≥70 had
the highest OR for death over the other individual variables.
Following multivariate regression, care home residency no
longer independently predicted mortality.
Iterative modelling to construct an 11-predictor and
5-predictor risk prediction scores
ORs and their respective p values from the multivariate logistic
regression model were used to identify constituent variables for
developing risk prediction scores. We began by constructing an
initial 11- predictor score that ranged from 1 to 18 points, using
data from 770 patients in the derivation cohort with a complete
dataset (online supplemental appendix 4). The lowest score of
1 point reflected the KDIGO (kidney disease improving global
outcomes) categorisation of CKD.19 In this score, the correla-
tion between increasing COVID-19 severity and in- hospital
mortality followed a linear dose–response relationship, particu-
larly between a score of 3 (below which no deaths occurred) and
12 (above which no patient survived), and an accuracy (AUROC)
of 0.84 (figure 1). A shorter five- predictor score based solely
on clinical parameters and scaled from 0 to 8 points was also
developed (table 2; figure 1). This short score, abbreviated as
SOARS (SpO2, Obesity, Age, Respiratory rate, Stroke history),
demonstrated an AUROC of 0.82 and was retained for further
evaluation as a practical prehospital risk stratification tool.
External validation and performance of the SOARS score
The performance metrics of the long 11- predictor and 5- predictor
(SOARS) scores were assessed by their ability to discriminate
for in- hospital mortality against both the ISARIC and Aintree
validation cohorts (table 3). The longer score showed higher
discrimination in the Aintree (AUROC 0.87) than the ISARIC
cohort (0.77). The SOARS score had a slightly lower AUROC
against both cohorts, namely, 0.80 (Aintree) and 0.74 (ISARIC).
In comparison, the performance of other scores based solely on
different cut- offs of age were associated with inferior discrimina-
tory ability. Comparison of some of the main population param-
eters between the derivation and both external cohorts is shown
in online supplemental appendix 5.
The mortality rate at each level of the SOARS score in the
derivation and both validation cohorts are shown in table 4. For
increased applicability, the SOARS score results were further
categorised into three risk classes: low (SOARS 0–1), moderate
(SOARS 2) and high (SOARS≥3) (table 5). Between 2.3% and
3ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
Table 1 Risk factors of mortality in the derivation cohort (N=983)
Univariate OR
P value
Likelihood ratio
χ2 of five final
predictors
Multivariate OR
P value(95% CI) (95% CI)
Age (years) 2177.27
<50 1.00 1.00
50–59 1.30 (1.28 to 7.02) 0.01 2.39 (1.49 to 3.84) <0.001
60–69 6.10 (2.73 to 13.66) <0.001 2.15 (1.32 to 3.50) <0.01
70–79 14.86 (6.89 to 32.04) <0.001 7.40 (4.67 to 11.74) <0.001
≥80 20.87 (9.93 to 43.86) <0.001 10.73 (6.82 to 16.90) <0.001
Male sex (vs female) 1.23 (0.94 to 1.62) 0.137 –
Ever smoked (vs never smoked) 2.14 (1.52 to 3.01) <0.001 2.51 (1.91 to 3.29) <0.001
Ethnicity (vs White)
Asian 0.57 (0.38 to 0.85) 0.006 1.44 (0.71 to 2.94) 0.301
Black 0.39 (0.17 to 0.88) 0.024 0.51 (0.16 to 1.64) 0.257
Symptoms (vs none)
Breathlessness 0.80 (0.59 to 1.07) 0.135 –
Fever 0.84 (0.62 to 1.14) 0.26 –
Cough 0.61 (0.45 to 0.82) 0.001 0.67 (0.41 to 1.09) 0.108
Myalgia 0.41 (0.28 to 0.63) <0.001 0.79 (0.44 to 1.41) 0.423
Headache 0.51 (0.26 to 0.97) 0.039 1.70 (0.64 to 4.51) 0.286
Clinical parameters
SpO2 (≤92% on air) 3.74 (2.73 to 5.12) <0.001 31.87 2.69 (1.80 to 4.01) <0.001
Respiratory rate (>24/min) 2.15 (1.63 to 2.84) <0.001 158.21 2.12 (1.35 to 3.32) 0.001
Systolic BP (≤90 mm Hg) 2.19 (0.96 to 5.03) 0.064 –
BMI (>30) 1.40 (1.03 to 1.90) 0.033 11.13 2.18 (1.46 to 3.20) <0.001
Frailty (vs not frail, CFS 0–4)
1 (CFS 5–6) 2.53 (1.73 to 3.70) <0.001 1.26 (0.82 to 1.96) 0.294
2 (CFS 7–9) 2.56 (1.62 to 4.06) <0.001 1.28 (0.77 to 2.13) 0.341
Residency in care home (vs own home) 3.14 (2.28 to 4.32) <0.001 1.38 (0.90 to 2.11) 0.137
Peripheral blood markers
CRP (>50 mmol/L) 2.40 (1.73 to 3.32) <0.001 1.42 (0.87 to 2.31) 0.16
Total white cell count
≤4 × 109/L 0.93 (0.56 to 1.55) 0.79 –
>11 × 109/L 2.04 (2.09 to 4.15) <0.001 1.76 (1.18 to 2.61) <0.01
Lymphocytes (<0.7 × 109/L) 1.97 (1.47 to 2.64) <0.001 1.67 (1.17 to 2.37) <0.01
Chronic kidney disease stage (vs 1, eGFR≥90 mL/min/1.73 m2) 1.00
2 eGFR 60–89 2.35 (1.53 to 3.62) <0.001 0.91 (0.49 to 1.69) 0.769
3 eGFR 30–44 and 45–59 6.09 (3.90 to 9.50) <0.001 1.46 (0.78 to 2.71) 0.234
4 eGFR 15–29 8.88 (4.82 to 16.34) <0.001 1.78 (0.82 to 3.85) 0.145
5 eGFR<15 15.73 (6.85 to 36.14) <0.001 3.82 (1.43 to 10.26) <0.01
CXR (≥4 zones affected, vs no abnormal zones) 1.89 (1.42 to 2.52) <0.001 1.73 (1.22 to 2.46) <0.01
Medications (≥5 different) 2.59 (1.93 to 3.48) <0.001 0.90 (0.53 to 1.52) 0.695
Co- morbidities (vs none)
Dementia 3.53 (2.46 to 5.08) <0.001 1.54 (0.97 to 2.44) 0.066
CVA/stroke 3.50 (2.32 to 5.28) <0.001 44.98 1.92 (1.18 to 3.11) <0.01
Cardiac disease 2.84 (2.13 to 3.80) <0.001 1.15 (0.70 to 1.91) 0.579
Cancer 2.60 (1.77 to 3.83) <0.001 1.85 (0.91 to 3.75) 0.091
Hypertension 2.46 (1.85 to 3.26) <0.001 1.10 (0.65 to 1.88) 0.717
Diabetes mellitus 1.40 (1.02 to 1.91) 0.036 1.00 (0.58 to 1.75) 0.981
Anxiety/psychosis 1.20 (0.84 to 1.71) 0.318 –
Lung disease 1.15 (0.86 to 1.55) 0.35 –
Out of the starting 75 variables, those with numerically small ORs or a p value of >0.05 following univariate regression were not included in table1.
BMI, body mass index; BP, blood pressure; CFS, Clinical Frailty Scale; CRP, c- reactive protein; CVA, cerebrovascular accident; CXR, chest radiograph; eGFR, estimated glomerular filtration rate.
4ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
3.2% of patients scoring 0 or 1 (low risk) in each of the three
cohorts died due to COVID-19, whereas 2.3%–6.3% of those
scoring 2 (moderate risk) failed to survive to discharge. Overall,
9 out of every 10 deaths in each of the derivation and validation
cohorts had a SOARS score of 3 or greater. Within this broad
range of increasing scores, the highest proportion of deaths was
encountered at SOARS score 4 in both the derivation cohort and
the ISARIC validation cohort (28.3% and 34.5%, respectively)
and at SOARS score 5 in the Aintree validation cohort (30.9%).
Sensitivity thresholds calculated for the larger validation
(ISARIC) cohort showed that the low risk class (SOARS 0–1)
comprised 16.6% of patients with an in- hospital mortality of
5.4%, sensitivity of 97% and negative predictive value (NPV)
of 94.6% (table 6). By comparison, 13.2% of patients were clas-
sified as moderate risk; their mortality rate was 14.5%. When
combining the low and moderate risk groups (SOARS 0–2), the
sensitivity reduces to 90.7% and the NPV to 90.5%. The high-
risk group comprised 70.2% of the validation cohort who scored
across a wide range of SOARS (scores 3–8). The specificity for
a prediction of death was thus more variable, from 58.2% for a
SOARS score of 3 to 99.8% at the other end of the scale when
the SOARS score was 7. Only one patient scored 8; they survived
to discharge.
Reliability of the risk estimates in the ISARIC cohort,
modelled as calibration or goodness- of- fit between expected
(predicted) and observed outcomes using SOARS, showed a cali-
bration slope of 0.70, calibration- in- the- large (CiTL) 0.02 and
an expected- to- observed (E:O) ratio of 0.990 (figure 2). Cali-
bration was slightly improved in the Aintree cohort with a slope
of 0.80, CiTL −0.16 and E:O ratio of 1.06. A LOWESS (locally
weighted scatterplot smoothing algorithm) curve was generated
to show differences between these outcomes in both cohorts.
The plot characteristics suggested that the model, while demon-
strating good concordance, had greater predictive accuracy in
low- risk to moderate- risk patients whose predicted probability
of mortality was under 40%. Conversely, overestimation of
mortality risk was evident in patients in the high- risk group.
DISCUSSION
We show that prognostic evaluation of a small panel of base-
line clinical and demographic characteristics of patients with
COVID-19 enables their subsequent risk of in- hospital death to
be quantified across three strata of risk. Findings were obtained
by applying the five- predictor SOARS (SpO2, Obesity, Age,
Respiratory rate, Stroke history) score to a large random sample
of the ISARIC cohort and a smaller single- hospital cohort
from Aintree, both with individual- level data. Our objective
was to enable risk stratification to be undertaken early, ideally
prehospitalisation (eg, in the community), during the encounter
Figure 1 (A) In- patient mortality stratified according to the
11- predictor (SOARS) scores; (B) In- patient mortality stratified according
to the 5- predictor (SOARS) scores.
Table 2 SOARS score (five predictors; range: 0–8 points) for
predicting in- hospital COVID-19 death
Predictor Points
SpO2
>92% on air 0
≤92% on air 1
Obesity (BMI>30)
Absent 0
Present 1
Age (years)
<50 0
50–59 1
60–69 2
70–79 3
≥80 4
Respiratory rate
≤24/min 0
>24/min 1
Stroke/CVA
Absent 0
Present 1
CVA, cerebrovascular accident; SOARS, SpO2, Obesity, Age, Respiratory rate, Stroke
history.
Table 3 Discriminatory performance (area under the receiver
operating characteristic; AUROC) of different risk stratification models
for predicting COVID-19 in- hospital mortality
Cohort 11- predictor
5- predictor
(SOARS)
Age
≥80
Age
70–79 Age ≥50
Derivation 0.84 0.82 0.73 0.76 0.73
Aintree
(validation)
0.87 0.80 0.78 0.69 0.57
ISARIC
(validation)
0.77 0.74 0.68 0.70 0.63
SOARS, SpO2, Obesity, Age, Respiratory rate, Stroke history.
5ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
with COVID-19- suspected individuals. This role is not met
by currently available prediction tools that rely on laboratory
measurements.
The SOARS score discriminated well for COVID-19 mortality
and its simplicity obviated the need for complex calculations. It
also retained good predictive accuracy in two external validation
cohorts, with performance metrics that were primarily reflected
in its high negative predictive values for mortality among patients
with the lowest risk scores (0 or 1). This characteristic is consis-
tent with a high accuracy for predicting a non- fatal outcome in
its key target group, namely, individuals with milder COVID-19.
Patients who score SOARS 0 or 1 could be discharged home
with advice to re- establish urgent contact if their symptoms
worsened. Patients stratified as moderate risk (SOARS 2) could
be virtually monitored with a predefined plan for care escala-
tion if specific thresholds relating to deteriorating symptoms or
self- recorded SpO2 were triggered. Patients in the high- risk class
(score ≥3) are highly likely to be symptomatic and would, in
all probability, be referred directly to the ED for hospital- based
management. Thus, the target individuals for the SOARS score
are those with a low or moderate risk of COVID-19 mortality.
Our data concur with other reports that advanced age is the
strongest predictor of death from COVID-19.20–23 The increase
in mortality in patients within the derivation cohort who were
in or beyond their seventh decade of life was reflected in the
magnitude of their respective adjusted ORs for in- hospital death,
namely, 7.4 (aged 70–79) and 10.7 (aged ≥80). Even so, predic-
tive models generated with age as the lone variable showed
poorer discriminatory ability than the SOARS score.
The early development of physiological abnormalities in
COVID-19 does not always result in timely clinical presenta-
tion. In our study, two measures of acute physiological perturba-
tion proved to be important predictors of COVID-19 mortality:
hypoxia and tachypnoea. Although persistent hypoxia is more
common in non- survivors of COVID-19, its relationship with
tachypnoea remains incompletely understood.24 25 Fewer than
half of patients with COVID-19 who present to hospital with
decreased oxygen saturation report experiencing subjective
breathlessness.26–29 One reason for this observation might be the
so- called ‘silent hypoxia’ where a blunted symptomatic percep-
tion of the effects of hypoxaemia is apparent even when low
arterial oxygen tension is evident.30 This phenomenon may be
responsible for delays in seeking clinical attention. Such danger
could be mitigated by accurate risk assessment including the
measurement and tracking of SpO2 in patients who are deemed
to not require immediate hospitalisation. The absence of oxygen
determination in the CURB-65 score has been cited as limiting
its utility in stratifying patients with COVID-19 for outpatient
management.31
The SOARS score was constructed with data from hospitalised
patients as the very low adverse event rate among non- admitted
cases (eg, in the VH pathway) curtailed the development of a
prognostic tool. This issue has previously been highlighted in the
context of CRB-65 where low event rates in community studies
of pneumonia made predictive inferences difficult to conclude.9
Other scores that have been used in COVID-19 studies have
either not been designed for this disease or have relied heavily
on laboratory- measured data.32–35
Table 4 Mortality in the derivation and validation cohorts at different levels of SOARS
Score
Derivation cohort
(N=821; deaths=258)
Aintree validation cohort
(N=290; deaths=94)
ISARIC validation cohort
(N=14 231; deaths=4319)
No of deaths
(n/N)
Mortality rate
(%)
Proportion of
deaths (%)
No of deaths
(n/N)
Mortality rate
(%)
Proportion of
deaths (%)
No of deaths
(n/N)
Mortality rate
(%)
Proportion
of deaths
(%)
01/69 1.4% 0.4% 1/12 8.3% 1.1% 34/833 4.1% 0.8%
15/94 5.3% 1.9% 2/23 8.7% 2.1% 94/1529 6.1% 2.2%
26/102 5.9% 2.3% 3/37 8.1% 3.2% 273/1879 14.5% 6.3%
329/124 23.4% 11.2% 11/60 18.3% 11.7% 650/2577 25.2% 15.1%
473/206 35.4% 28.3% 24/72 33.3% 25.5% 1490/4013 37.1% 34.5%
563/117 53.9% 24.4% 29/48 60.4% 30.9% 1170/2402 48.7% 27.1%
658/80 72.5% 22.6% 16/30 53.3% 17.0% 571/939 60.8% 13.2%
722/28 78.6% 8.5% 6/6 100.0% 6.4% 36/58 62.1% 0.8%
81/1 100.0% 0.4% 2/2 100.0% 2.1% 0/1 0.0% 0.0%
The proportion of deaths at each score is defined as the number of deaths at that score divided by the total number of deaths in that particular cohort.
SOARS, SpO2, Obesity, Age, Respiratory rate, Stroke history.
Table 5 COVID-19 mortality risk stratification based on SOARS score
Risk class
(score level)
Derivation cohort
(N=821 with full data; deaths=258)
Aintree validation cohort
(N=290 with full data; deaths=94)
ISARIC validation cohort
(N=14 231; deaths=4319)
Mortality by risk
class (%)
Proportion of
deaths by risk
class (%)
Mortality by risk
class (%)
Proportion of
deaths by risk
class (%)
Mortality by risk
class (%)
Proportion of deaths
by risk class (%)
Low (0–1) 6/163 (3.7%) 6/258 (2.3%) 3/35 (8.6%) 3/94 (3.2%) 128/2362 (5.4%) 128/4319 (3.0%)
Moderate (2) 6/102 (5.9%) 6/258 (2.3%) 3/37 (8.1%) 3/94 (3.2%) 273/1879 (14.5%) 273/4319 (6.3%)
High (≥3) 246/556 (44.2%) 246/258 (95.3%) 88/218 (40.4%) 88/94 (93.6%) 3917/9990 (39.2%) 3917/4319 (90.7%)
SOARS, SpO2, Obesity, Age, Respiratory rate, Stroke history.
6ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
Our multivariate regression model was bootstrapped to
reduce overfitting but not penalised prior to external valida-
tion. In common with other severity scores for COVID-19, we
dichotomised several continuous data parameters which may
have potentially obscured non- linear effects between predictors
and outcome, contributing to the difference in AUROC values
between our derivation and validation cohorts and between
both validation cohorts.11 32 34 Other prediction systems, notably
CURB-65 and the Pneumonia Severity Index for pneumonia
similarly categorise some of their score parameters.10 36 We also
used in- hospital mortality as an unambiguous disease- related
primary outcome rather than 30- day or 60- day mortality. The
better performance of the SOARS score in the smaller Aintree
validation cohort compared with the much larger ISARIC
cohort may have been due to its more homogeneous case- mix.
This comparison suggested that, on balance, the simplicity of a
prehospital risk prediction tool, provided it retained acceptable
accuracy, may outweigh any minor diminution of its perfor-
mance arising from improved practicality.
Other limitations in the study include the occurrence of
missing information despite prospective data collection. The use
of multiple imputation to estimate missing values for multivariate
regression and the availability of nearly 85% of observations for
constructing the risk stratification rule helped to mitigate against
underestimating their role. The modest sample size of our deri-
vation cohort was dictated by the incident caseload during the
pandemic. However, selective sampling of the pandemic timeline
was avoided by including all COVID-19 cases from the initial
rise to the subsequent decline in new case numbers over the
11- week study period. Finally, reduced score calibration at the
high- risk end suggests that SOARS may overestimate the prob-
ability of death in the highest risk cases. However, the principal
objective of this score was to enhance frontline decision- making
in patients with a low predicted risk of mortality at a time when
demand for in- patient resources is likely to be high.
In summary, prognostication using the SOARS score can be
undertaken concomitantly with SARS- CoV-2 diagnostic testing
to inform clinical triaging, including decisions about the place-
ment of the patient for ongoing care. Analysis of the ISARIC
validation cohort in this study showed that between 16.6% and
29.8% (those scoring up to SOARS 1 or 2, respectively) could
potentially have avoided admission provided a safe alternative
to hospitalisation was in place. Prospective studies of SOARS
implementation will enable the score to be calibrated against
other independent cohorts of patients with COVID-19 to
examine its performance under conditions that may be unique
to different localities. Such an opportunity may soon present
itself if SARS- CoV-2 transmission continues to increase in the
UK and beyond.
Author affiliations
1Interstitial Lung Disease Unit, Department of Respiratory Medicine, Royal Brompton
and Harefield NHS Foundation Trust, London, UK
2National Heart and Lung Institute, Imperial College London, London, UK
3Respiratory Medicine, West Hertfordshire Hospitals NHS Trust, Watford, UK
4Respiratory Medicine, St. George’s Hospital, London, UK
5Information Governance, West Hertfordshire Hospitals NHS Trust, Watford, UK
6Radiology, West Hertfordshire Hospitals NHS Trust, Watford, UK
7Respiratory Medicine, Aintree site, Liverpool Hospitals NHS Foundation Trust,
Liverpool, UK
8Radiology, Aintree site, Liverpool Hospitals NHS Foundation Trust, UK, Liverpool, UK
Twitter Matthew Knight @mjknight0380 and Andrew Barlow @Andyatfrogmore
Acknowledgements We are grateful to the healthcare teams whose efforts in the
clinical field were fundamental to this work. We would like to thank all the patients
involved for their vital contributions. We are very grateful to the ISARIC Coronavirus
Clinical Characterisation Consortium (ISARIC- 4C) Investigators, in particular, J
Table 6 Sensitivity analysis of the SOARS score for predicting mortality in the ISARIC validation cohort
Score cut- off
Patients
N (% of total)
True
positive
True
negative
False
positive
False
negative Sensitivity Specificity PPV NPV Cumulative mortality (%)
>0 833 (5.9%) 4284 799 9114 34 99.2% 8.1% 32.0% 95.9% 34/833 (4.1%)
>1 2362 (16.6%) 4190 2234 7679 128 97.0% 22.5% 35.3% 94.6% 128/2362 (5.4%)
>2 4241 (29.8%) 3917 3840 6073 401 90.7% 38.7% 39.2% 90.5% 401/4241 (9.5%)
>3 6818 (47.9%) 3267 5767 4146 1051 75.7% 58.2% 44.1% 84.6% 1051/6818 (15.4%)
>4 10 831 (76.1%) 1777 8290 1623 2541 41.2% 83.6% 52.3% 76.5% 2541/10 831 (23.5%)
>5 13 233 (93.0%) 607 9522 391 3711 14.1% 96.1% 60.8% 72.0% 3711/13 233 (28.0%)
>6 14 172 (99.6%) 36 9890 23 4282 0.8% 99.8% 61.0% 69.8% 4282/14 172 (30.2%)
>7 14 231 (100.0%) 0 9912 1 4318 0.0% 100.0% 0.0% 69.7% 4318/14 231 (30.3%)
SOARS, SpO2, Obesity, Age, Respiratory rate, Stroke history.
Figure 2 Calibration accuracy of the SOARS (SpO2, Obesity, Age,
Respiratory rate, Stroke history) score on external validation cohorts.
7ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

Respiratory infection
Kenneth Baillie (lead investigator) and Malcolm G Semple (chief investigator) for
providing access to data for the external validation of our model.
Contributors RV, AB, MK and RM developed the clinical algorithm and supervised
patient management. FC, RV and AD conceived and designed the investigational
plan. FC drafted the manuscript with contributions of intellectual content from RV,
AD, LGS, PLM and AB. RV, TV, MK, RM, JS, LS, ET, HM, SC, NM, SA, ML, AO, CP, RK,
TH, RT, SR, MS and JL collected the data at respective sites. AD, FC and RV examined
the data and undertook statistical analyses. All authors approved the final version of
the manuscript for submission. RV is guarantor and attests that all named authors
and contributors meet authorship criteria and that no others meeting such criteria
have been omitted.
Funding The authors have not declared a specific grant for this research from any
funding agency in the public, commercial or not- for- profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval Ethical approval was provided by Stanmore Research Ethics
Committee, London, England (IRAS ID: 283888). The study is registered by the
National Health Service Health Research Authority under the reference 20/
HRA/2344.
Provenance and peer review Not commissioned, externally peer reviewed.
No part of this work has been written by a medical writer or published in printed
or electronic form. Some of the findings of this study have been accepted for
presentation at the British Thoracic Society Winter Meeting 2020 to be held in
February 2021, at which point the abstract will be published in printed format in a
supplement of Thorax. A copy of the originally submitted manuscript was uploaded
to the medRixv preprint website; https:// doi. org/ 10. 1101/ 2020. 10. 19. 20215426.
Data availability statement Data are available upon reasonable request.
Deidentified participant data may be requested from the corresponding author
following publication of the study (ORCID identifier: 0000-0001-7845-0173)
This article is made freely available for use in accordance with BMJ’s website
terms and conditions for the duration of the covid-19 pandemic or until otherwise
determined by BMJ. You may use, download and print the article for any lawful,
non- commercial purpose (including text and data mining) provided that all copyright
notices and trade marks are retained.
ORCID iDs
FelixChua http:// orcid. org/ 0000- 0001- 7845- 0173
Lisa GSpencer http:// orcid. org/ 0000- 0003- 3558- 992X
Philip LMolyneaux http:// orcid. org/ 0000- 0003- 1301- 8800
REFERENCES
1 Challen K, Goodacre SW, Wilson R, etal. Evaluation of triage methods used to select
patients with suspected pandemic influenza for hospital admission. Emerg Med J
2012;29:383–8.
2 PL H, Chau PH, PSF Y. A clinical prediction rule for clinical diagnosis of severe acute
respiratory syndrome. Eur Resp J;200:474–9.
3 Wynants L, Van Calster B, Collins GS, etal. Prediction models for diagnosis and
prognosis of covid-19 infection: systematic review and critical appraisal. BMJ
2020;369:m1328.
4 Intensive Care National Audit and Research Centre. ICNARC reports on COVID-19 in
critical care: England, Wales and Northern Ireland. Available: https://www. icnarc. org
[Accessed 4 Dec 2020].
5 Dighe A, Cattarino L, Cuomo- Dannenburg G, etal. Response to COVID-19 in South
Korea and implications for lifting stringent interventions. BMC Med 2020;18:321.
6 Chen B, Zhong H, Ni Y. Epidemiological trends of coronavirus disease 2019 in China.
Front Med 2020;7:250.
7 Center for Strategic & International Studies Southeast Asia national response to
Covid-19 tracker. Available: https://www. csis. org/ programs/ southeast- asia- program/
southeast- asia- covid- 19- tracker-0 [Accessed 15 Nov 2020].
8 Cacciapaglia G, Cot C, Sannino F. Second wave COVID-19 pandemics in Europe: a
temporal playbook. Sci Rep 2020;10:15514.
9 McNally M, Curtain J, O’Brien KK, etal. Validity of British thoracic Society guidance
(the CRB-65 rule) for predicting the severity of pneumonia in general practice:
systematic review and meta- analysis. Br J Gen Pract 2010;60:e423–33.
10 Lim WS, van der Eerden MM, Laing R, etal. Defining community acquired pneumonia
severity on presentation to hospital: an international derivation and validation study.
Thorax 2003;58:377–82.
11 Knight SR, Ho A, Pius R, etal. Risk stratification of patients admitted to hospital with
covid-19 using the ISARIC who clinical characterisation protocol: development and
validation of the 4C mortality score. BMJ 2020;370:m3339.
12 National Early Warning Score (NEWS)2. Royal College of Physicians, London.
Available: https://www. rcplondon. ac. uk/ projects/ outputs/ national- early- warning-
score- news-2 [Accessed 30 May 2020].
13 Collins GS, Reitsma JB, Altman DG, etal. Transparent reporting of a multivariable
prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD
statement. Ann Intern Med 2015;162:55–63.
14 Shipe ME, Deppen SA, Farjah F, etal. Developing prediction models for clinical use
using logistic regression: an overview. J Thorac Dis 2019;11:S574–84.
15 Leisman DE, Harhay MO, Lederer DJ, etal. Development and reporting of prediction
models: guidance for authors from editors of respiratory, sleep, and critical care
journals. Crit Care Med 2020;48:623-633.
16 Resche- Rigon M, White IR. Multiple imputation by chained equations for
systematically and sporadically missing multilevel data. Stat Methods Med Res
2018;27:1634–49.
17 Hosmer DW, Lemesbow S. Goodness of fit tests for the multiple logistic regression
model. Commun Stat Theory Methods 1980;9:1043–69.
18 Hu B, Shao J, Palta M. Pseudo- R2 in logistic regression model. Statistica Sinica
2006;16:847–60.
19 Kidney Disease, Improving Clinical Outcomes. KDIGO 2012 clinical practice guideline
for the evaluation and management of chronic kidney disease. Available: https://
kdigo. org/ wp- content/ uploads/ 2017/ 02/ KDIGO_ 2012_ CKD_ GL. pdf [Accessed 1 Jun
2020].
20 Williamson EJ, Walker AJ, Bhaskaran K. OpenSAFELY: factors associated with
COVID-19 death in 17 million patients. Nature 2020;584:430–4.
21 Mikami T, Miyashita H, Yamada T, etal. Risk factors for mortality in patients with
COVID-19 in New York City. J Gen Intern Med 2020. doi:10.1007/s11606-020-
05983-z. [Epub ahead of print: 30 Jun 2020].
22 Feng Y, Ling Y, Bai T, etal. COVID-19 with different severities: a multicenter study of
clinical features. Am J Respir Crit Care Med 2020;201:1380–8.
23 Onder G, Rezza G, Brusaferro S. Case- Fatality rate and characteristics of patients dying
in relation to COVID-19 in Italy. JAMA 2020;323:1775–6.
24 Xie J, Covassin N, Fan Z, etal. Association between hypoxemia and mortality in
patients with COVID-19. Mayo Clin Proc 2020;95:1138–47.
25 Wu C, Chen X, Cai Y, etal. Risk factors associated with acute respiratory distress
syndrome and death in patients with coronavirus disease 2019 pneumonia in Wuhan,
China. JAMA Intern Med 2020;180:934–43.
26 Richardson S, Hirsch JS, Narasimhan M, etal. Presenting characteristics, comorbidities,
and outcomes among 5700 patients hospitalized with COVID-19 in the new York City
area. JAMA 2020;323:2052.
27 Guan W, Ni Z, Hu Y. For the China medical treatment expert group for
Covid-19. clinical characteristics of coronavirus disease in China. N Engl J Med
2020;382:1708–20.
28 Zhou F, Yu T, Du R, etal. Clinical course and risk factors for mortality of adult
inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. Lancet
2020;395:1054–62.
29 Huang C, Wang Y, Li X. Clinical features of patients infected with 2019 novel
coronavirus in Wuhan, China. Lancet 2020.
30 Tobin MJ, Laghi F, Jubran A. Why COVID-19 silent hypoxemia is Baffling to physicians.
Am J Respir Crit Care Med 2020;202:356–60.
31 Nguyen Y, Corre F, Honsel V, etal. Applicability of the CURB-65 pneumonia severity
score for outpatient treatment of COVID-19. J Infect 2020;81:e96–8.
32 Fan G, Tu C, Zhou F, etal. Comparison of severity scores for COVID-19 patients with
pneumonia: a retrospective study. Eur Respir J 2020;56:2002113.
33 Liang W, Liang H, Ou L. Development and validation of a clinical risk score to predict
the occurrence of critical illness in hospitalized patients with COVID-19. JAMA Intern
Med.
34 Satici C, Demirkol MA, Sargin Altunok E, etal. Performance of pneumonia severity
index and CURB-65 in predicting 30- day mortality in patients with COVID-19. Int J
Infect Dis 2020;98:84–9.
35 Liu S, Yao N, Qiu Y, etal. Predictive performance of SOFA and qSOFA for in- hospital
mortality in severe novel coronavirus disease. Am J Emerg Med 2020;38:2074–80.
36 Fine MJ, Auble TE, Yealy DM, etal. A prediction rule to identify low- risk patients with
community- acquired pneumonia. N Engl J Med 1997;336:243–50.
8ChuaF, etal. Thorax 2021;0:1–8. doi:10.1136/thoraxjnl-2020-216425
on March 15, 2021 by guest. Protected by copyright.http://thorax.bmj.com/Thorax: first published as 10.1136/thoraxjnl-2020-216425 on 10 March 2021. Downloaded from

SUPPLEMENTARY MATERIAL
Appendix 1. Baseline clinical characteristics of the derivation cohort
All patients (N = 983)
Cohort
Survivors
(N=689)
Non-survivors
(N=294)
P value
Age, years – median (interquartile range)
70 (53 – 83)
61 (50 – 78)
81 (72 – 87)
<0.0001
Age range – no. (%)
18 – 49 years
176 (17.9)
168 (24.4)
8 (2.7)
<0.001*
50 – 59 years
160 (16.3)
140 (20.3)
20 (6.8)
--
60 – 69 years
151 (15.4)
117 (17.0)
34 (11.6)
--
70 - 79 years
181 (18.4)
106 (15.4)
75 (25.5)
--
≥ 80 years
315 (32.0)
158 (22.9)
157 (53.4)
--
Age – median (IQR) by level of care
Virtual hospital
53 (43 – 67)
53 (42 – 64)
81 (78 – 86)
0.007
Ward
77 (61 – 86)
70 (55 – 82)
84 (77 – 89)
<0.0001
Ward + received CPAP
71 (61 – 75)
61 (58 – 68)
75 (73 – 81)
<0.0001
Intensive Care Unit (ICU)
60 (52 – 67)
59 (50 – 62)
61 (56 – 71)
0.0349
Male sex – no. (%)
516 (52.5)
351 (50.9)
165 (56.1)
0.137
Ethnic background
White
760 (77.3)
511 (74.1)
249 (84.7)
0.003*
Asian
162 (16.5)
127(18.4)
35 (11.9)
--
Black
44 (4.5)
37 (5.4)
7 (2.4)
--
Other
17 (1.7)
14 (2.0)
3 (1.0)
--
Smoking history – no. (%)
Former or current smoker
168 (17.1)
94 (13.7)
74 (25.3)
<0.001
BMI > 30 – no. (%)
243 (24.7)
156 (23.3)
87 (29.8)
0.033
Care Home residency – no. (%)
204 (20.8)
101 (14.7)
103 (35.0)
<0.001
Clinical frailty score – no./total scored (%)
1 – 4
250/644 (38.8)
192/415 (46.3)
58/229 (25.3)
<0.001*
5 – 6
268/644 (41.6)
152/415 (36.6)
116/229 (50.7)
--
7 – 9
126/644 (19.6)
71/415 (17.1)
55/229 (20.0)
--
Symptoms at presentation– no. (%)
Fever (temperature >37.3°C)
508 (61.0)
357(62.2)
151 (58.1)
0.259
Breathlessness
482 (57.9)
342 (59.6)
140 (54.1)
0.135
Cough
440 (52.9)
325 (56.7)
115 (44.4)
0.001
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Myalgia
181 (21.7)
148 (25.9)
33 (21.8)
<0.001
Headache
62 (7.4)
50 (8.8)
12 (4.6)
0.036
Symptom duration, days – median (IQR)
6 (2 – 11)
7 (3 – 11)
5 (2 – 13)
0.454
Vital baseline observations – no./total (%)
Respiratory rate >24/min
294/819 (35.9)
155/562 (27.6)
139/257 (54.1)
<0.001
SpO2 ≤92% (on ambient air)
258/822 (31.4)
125/564 (22.2)
133/258 (51.6)
<0.001
Systolic blood pressure <90 mm Hg
23/811 (2.8)
12/556 (2.2)
11/255 (4.3)
0.086
Pulse rate >120/min
81/818 (9.9)
44/560 (7.9)
37/258 (14.3)
0.004
Laboratory findings – no./total (%)
C-reactive protein >50 mg/L
524/812 (64.5)
306/528 (58.0)
218/284 (76.7)
<0.001
Total white cell count >11 x 109/L
175/891 (19.6)
82/600 (13.7)
93/291 (32.0)
<0.001
Lymphocyte count ≤0.7 x 109/L
304/890 (34.2)
174/599 (29.1)
130/291 (44.7)
<0.001
Chronic kidney disease – no./total (%)
Stage 1 (eGFR ≥ 90 ml/min/1.73 m2)
118/832 (14.2)
94/547 (17.2)
24/285 (8.4)
<0.001*
Stage 2 (eGFR 60-89 ml/min/1.73 m2)
380/832 (45.7)
286/547 (52.3)
94/285 (33.0)
--
Stage 3 (eGFR 30-59 ml/min/1.73m2)
237/832 (28.5)
128/547 (23.4)
109/285 (38.3)
--
Stage 4 (eGFR 15-29 ml/min/1.73m2)
65/832 (7.8)
29/547 (5.3)
36/285 (12.6)
--
Stage 5 (eGFR <15 ml/min/1.73m2)
32/832 (3.9)
10/547 (1.8)
22/285 (7.7)
--
≥4 abnormal CXR zones – no./total (%)
338/895 (37.8)
203/615 (33.0)
135/280 (48.2)
<0.001
Co-morbid conditions – no. (%)
Hypertension
475 (48.4)
288 (41.8)
187 (63.8)
<0.001
Ischaemic heart disease
194 (19.7)
103 (15.0)
91 (31.0)
<0.001
Cardiac failure
33 (3.4)
21 (3.1)
12 (4.1)
0.41
Cardiac arrhythmias
34 (3.5)
20 (2.9)
14 (4.8)
0.144
Diabetes mellitus
232 (23.6)
150(21.8)
82 (28.0)
0.036
Respiratory disease
293 (30.0)
199 (29.1)
94 (32.1)
0.35
Chronic kidney disease
196 (20.0)
102 (14.8)
94 (32.1)
<0.001
Cerebrovascular disease
107 (10.9)
47 (6.8)
60 (20.4)
<0.001
Mental health/behavioural disorders
Dementia
157 (16.0)
73 (10.6)
84 (28.6)
<0.001
Anxiety, depression or both
156 (15.9)
113 (16.4)
43 (14.6)
0.486
Prescribed medications ≥5 – no. (%)
550 (56.1)
340 (49.4)
210 (71.7)
<0.001
Median length of stay – days (IQR)
7 (3.0 – 13.5)
7 (3.0 – 15)
7 (3.0 – 12)
0.564
Abbreviations: BMI – body mass index; CXR – chest radiograph; ED –Emergency Department; GFR – glomerular filtration
rate; IQR – inter-quartile range. P values denote comparisons between survivors and non-survivors. *χ2 test comparing
all subcategories.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Appendix 2. In-patient mortality rate by age bracket in the derivation and validation cohorts
Age group
Derivation cohort
Validation cohorts
(N=983)
ISARIC (N=14,231)
Aintree (N=303)
< 50
4.6%
5.6%
9.7%
50 - 59
12.5%
13.8%
7.9%
60 – 69
22.5%
25.4%
27.0%
70 – 79
41.4%
37.5%
44.1%
≥ 80
49.8%
46.2%
64.9%
Overall mortality
29.9%
30.9%
31.7%
Appendix 3. In-patient mortality rate by level of care in the derivation cohort (N=983)
LEVEL OF MAXIMAL CARE
Whole cohort
VH
Ward
Ward CPAP
ICU
N
983
228
627
41
87
Median age (IQR)
70 (53 – 83)
53 (43 – 67)
77 (61 – 86)
71 (61 – 75)
60 (52 – 67)
Deaths
294/983
(29.9%)
4/228
(1.8%)
216/627
(34.4%)
20/41
(48.8%)
54/87
(62.1%)
% of non-
survivors by
ethnicity at each
level of care
W 3 (75.0%)
A 1 (25.0%)
B 0
OTH 0
W 194 (90.0%)
A 17 (7.9%)
B 4
OTH 1
W 16 (80.0%)
A 3 (15.0%)
B 0
OTH 1
W 37 (68.5%)
A 13 (24.1%)
B 3
OTH 1
The difference between the median age of patients by level of care, expressed as the Chi-square
value with ties, is 150.542 (with 80 degrees of freedom) (P <0.001).
Abbreviations: A – Asian; B – Black; CPAP – continuous positive airway pressure; ICU – Intensive
Care Unit; IQR – interquartile range; OTH – other ethnicity; VH – virtual hospital; W – White.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Appendix 4. 11-predictor score (1 – 18 points) for predicting in-hospital COVID-19 death
PREDICTOR
POINTS
SpO2
> 92% on air
≤ 92% on air
0
1
Obesity (BMI >30)
absent
present
0
1
Age
< 50
50 – 59
60 – 69
70 – 79
≥ 80
0
1
2
3
4
Respiratory rate
≤ 24/min
> 24/min
0
1
Stroke/CVA
absent
present
0
1
Ever smoked
no
yes
0
1
Dementia
no
yes
0
1
CKD stage
1
2
3
4
5
1
2
3
4
5
White cell count >11 x109
no
yes
0
1
Lymphocytes ≤ 0.7 x109
no
yes
0
1
CXR (≥ 4 zones affected)
no
yes
0
1
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Appendix 5. Comparison of key parameters between the derivation and validation cohorts
Derivation cohort
N = 983
Validation (ISARIC)
N = 14,231
P value
Median age (IQR)
Age range (years)
70 – 79
≥ 80
70 (53 – 83)
18.4%
32.5%
73 (59 – 83)
21.9%
34.8%
NS
<0.05
NS
Mortality
29.9%
30.9%
NS
Ethnicity (% of cohort)
White
BAME
77.3%
22.7%
69.0%
31.0%
<0.01
Male sex (% of cohort)
52.5%
55.7%
<0.05
Derivation cohort
N = 983
Validation (Aintree)
N = 303
P value
Median age (IQR)
Age range (years)
70 – 79
≥ 80
70 (53 – 83)
18.4%
32.5%
67 (57 – 77)
22.4%
18.8%
NS
NS
<0.001
Mortality
29.9%
31.7%
NS
Ethnicity (% of cohort)
White
BAME
77.3%
22.7%
95.7%
4.3%
<0.001
Male sex (% of cohort)
52.5%
61.1%
<0.01
Abbreviations: BAME – Black Asian or other Minor Ethnicity Black; IQR – interquartile range
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Appendix 6. Receiver operating curves (ROCs) of the 11-predictor and 5-predictor (SOARS)
scores on the derivation and validation cohorts
Area under the ROC (AUROC) for the 11-predictor score on the derivation cohort (0.84)
Area under the ROC (AUROC) for the 11-predictor score on the ISARIC validation cohort (0.77)
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Area under the ROC (AUROC) for the 11-predictor score on the Aintree validation cohort (0.87)
Area under the ROC (AUROC) for the 5-predictor score on derivation cohort (0.82)
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Area under the ROC (AUROC) for the 5-predictor score on ISARIC validation cohort (0.74)
Area under the ROC (AUROC) for the 5-predictor score on Aintree validation cohort (0.80)
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Appendix. Pre-specified study protocol for clinical data collection
PREDICT COVID Clinical Data Collection protocol
Date: 1st March 2020
Aims
1. To prospectively collect data on all adults (> 18 years) with laboratory-confirmed SARS-CoV-2 infection (COVID-19)
presenting to Watford Hospital, West Hertfordshire NHS Trust, during the first wave of the SARS-CoV-2 pandemic,
including an outcome of in-hospital death or hospital discharge,
2. To develop a prognostic (risk prediction) score using the above derivation data,
3. To construct a practical clinical scoring system for predicting mortality (and risk of morbidity) after validation of
score against external, i.e. independent cohorts of COVID-19 patients from other UK sites,
4. To include and assess outcomes of patients referred to the COVID-19 Virtual Hospital (out-of-hospital monitoring)
in the same NHS Trust,
5. To monitor and characterise surviving patients for up to 12 months from the time of confirmation of SARS-CoV-2
infection, including the domains of psychological, physiological and radiological impairment and recovery.
Primary outcome
In-hospital mortality with minimum 30-day follow-up data.
Secondary outcomes
Longer term mortality and morbidity: radiological, psychological and cardiorespiratory. This will also be assessed
against ongoing health care needs.
Patient inclusion
All adult patients (aged >18 years) with SARS-CoV-2 real-time reverse transcriptase polymerase chain reaction (rRT-PCR)
confirmation.
Completed admission and outcomes at 3 months and 12 months.
Inclusion criteria:
• Readily available patient or clinical characteristic to attending clinicians upon presentation to hospital
(Accident & Emergency department, Acute Medical Receiving Unit)
• Blood markers should be commonly measured and results available for review within the first 24 hours of
admission
• All parameters relating to oxygen supplementation and advanced respiratory support including one or more
of: continuous positive airway pressure (CPAP), bilevel non-invasive ventilation (NIV), high-flow nasal cannula
(HFNC) oxygenation and invasive mechanical ventilation (IMV) needs
Exclusion criteria
All SARS-CoV-2 rRT-PCR negative patients irrespective of clinical suspicion of COVID-19.
All individuals aged <18y.
Selection of candidate variables for initial data collection
Candidate variables were chosen based on knowledge of their potential association/s with SARS-CoV-2 infection and
clinical disease (COVID-19). A systematic literature search was undertaken to identify these variables with respect to
their predictive association for mortality and other adverse outcomes including COVID-19 severity and disease-related
complications such as requirement for critical care and the development of COVID-19-associated acute respiratory
distress syndrome (ARDS).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F

Systematic literature for English language articles in the following search databases: PubMed, EMBASE, WHO Medicus,
Web of Science and Google Scholar (particularly for pre-print publications on medRxiv). Search terms included SARS-
CoV-2; COVID-19; coronavirus; ARDS; pneumonia; sepsis; influenza; risk prediction; risk score; prognosis; validation. No
date restrictions were imposed.
Statistical analysis for derivation and validation modelling
In analysing the data collected from the derivation (West Herts) cohort, categorical variables will expressed as
frequency (%), with significance determined by the Chi-squared test. Continuous variables will be analysed for median
(interquartile range) or mean (standard deviation) outcomes and analysed by the t-test, Kruskal-Wallis or Mann-
Whitney U test, as appropriate. Missing data in the derivation cohort will be expected even with prospective data
collection; missingness of data will be assumed to be at random and handled by multiple imputation by chained
equations (MICE) with at least ten imputations, provided the proportion of missingness for the defined parameter
constitutes no more than 20% of the cohort. Collated data will be subjected to univariate and multivariate logistic
regression in order to determine odds ratios (OR) for in-hospital mortality. The latter will also be internally validated by
bootstrapping using a minimum of 1000 re-samples. Predictor interactions will be analysed by appropriate
methodology such as the likelihood ratio (LR) test comparing broad and narrow (constrained) models.
All performance metrics against external validation cohorts will be analysed using the prediction score and not with the
multivariate regression model. The performance of the derivation model will be assessed for discriminatory ability (area
under the receiver operating characteristic, AUROC) and calibration (graphical representation of Hosmer-Lemeshow
analysis).
All statistical analyses including risk modelling calculations will be performed using STATA, version 16 (Stata Corp.,
Texas, USA).
R Vancheeswaran, F Chua, A Draper, T Vaghela and A Barlow (study design and responsible persons for data analysis
and interpretation)
MARCH 2020
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) Thorax
doi: 10.1136/thoraxjnl-2020-216425–8.:10 2021;Thorax, et al. Chua F