AJTCCM VOL. 28 NO. 4 2022 157
RESEARCH
Background. COVID‑19 disease, a pandemic for more than twoyears, has major morbidity and mortality related to pulmonary
involvement. Chest radiography is the main imaging tool for critically ill patients. As the availability of arterial blood gas analysis is
limited in theLevelI and II healthcare centres, which are major partners in providing healthcare in resource‑limited times, we planned
the present study.
Objective. To assess the role of chest radiography in predicting the need for oxygen/ventilator support in critically ill COVID‑19 patients.
Methods. is hospital‑based, retrospective study included 135 patients who needed oxygen/ventilator support and had optimal‑quality
chest radiographs at admission. All the chest X‑rays were evaluated and a severity score was calculated on a predesigned pro forma. Statistical
evaluation of the data obtained was done using appropriate tools and methods.
Results. Males outnumbered females, with a mean age of 54.35 ± 14.49years. More than 72% of patients included in our study needed
ventilator support while the rest needed oxygen support. ere was a signicant statistical correlation between the chest radiograph
severity score and SPO2/PaO2 levels in our study. Using a cut‑o value >8 for the chest radiograph severity score in predicting the need
for ventilator support in a Covid‑19 patient, the sensitivity, specicity and accuracy was 85.7%, 92.5% and 89.5%, respectively.
Conclusions. Chest radiography remains the mainstay of imaging in critically ill COVID‑19 patients when they are on multiple life‑support
systems. ough arterial blood gas analysis is the gold standard tool for assessing the need for oxygen/ventilator support in these patients,
the severity score obtained from the initial chest radiograph at the time of admission may also be used as a screening tool. Chest radiography
may predict the need for oxygen/ventilator support, allowing time for patients to be moved to an appropriate‑level healthcare centre, thus
limiting morbidity and mortality.
Keywords. Chest radiography, ventilator, COVID‑19.
Afr J Thoracic Crit Care Med 2022;28(4):157‑162. https://doi.org/10.7196/AJTCCM.2022.v28i4.248
e COVID‑19 disease caused by severe acute respiratory syndrome
coronavirus II emerged in Wuhan, China, in December 2019 before
becoming a global pandemic in March 2020.[1‑3] ough both chest
X‑ray (CXR) and chest computed tomography (CCT) play a great role
in the diagnosis of COVID‑19 disease at all levels of severity, CXR
remains the primary imaging modality.[1,3] COVID‑19 disease varies
clinically from symptom‑free to severely ill patients, with pneumonia
and death in a signicant number.[4]
The gold standard tool for diagnosis of COVID‑19 is reverse
transcription‑polymerase chain reaction (RT‑PCR) of swab samples
from the nasopharynx and oropharynx.[4] Non‑ambulatory, critically
ill patients on life‑support devices cannot usually undergo CCT owing
to many constraints. But CXR in such cases is not only cost eective but
also saves time, provides quick results and can be repeated.[1] Hence,
portable chest radiography has been considered as the investigation of
choice in critically ill patients by the American College of Radiology
and the Society of oracic Radiology.[3,5] Many institutes around the
world are using portable chest radiography at triage level to assess
severity of disease.[6]
Variable radiographic findings and degrees of lung parenchymal
involvement have been observed on early chest radiographs in Covid‑19
disease. Even though the overall sensitivity of CXR in diagnosing
COVID‑19 disease is only 69%, it nevertheless offers significant
advantages in analysing clinical outcome at the hospital doorstep.[3]
In severely ill Covid‑19 patients, the only investigating tool being
utilised to establish the need for ventilatory support is arterial blood
gas (ABG) analysis, which is not only an invasive and time‑intensive
procedure but also is of limited availability in peripheral areas, with
Level I and II healthcare centres requiring highly skilled sta. Hence,
in the present study, we assessed the utility of CXR in predicting the
need for ventilatory support in severe COVID‑19 disease that will help
the clinician to start ventilatory support early and, in addition to other
advantages of CXR, also avoid signicant risk of contact with patients’
secretions and blood products.
Objectives
Our objectives were (i) to evaluate the spectrum of chest
radiographic ndings in predicting the need for ventilator support
Role of chest radiograph in predicting the need for ventilator
support in COVID‑19 patients
G Patnayak,1 MD; R Rastogi,1 MD, PGDip, MSK US; L Khajuria,1 MB BS, MD Resident; A Mohan,1 MD; N Jain,1 MD;
R Varshney,2 MD; V K Singh,3 MD; V Pratap,1 MD; S Pathak,1 MD, DNB; A Jain,1 MD; K Duggad,1 MD
1 Department of Radiodiagnosis, Teerthanker Mahaveer Medical College & Research Center, Moradabad, Uttar Pradesh, India
2 Department of Emergency Medicine, Teerthanker Mahaveer Medical College & Research Center, Moradabad, Uttar Pradesh, India
3 Department of Internal Medicine, Teerthanker Mahaveer Medical College & Research Center, Moradabad, Uttar Pradesh, India
Corresponding author: R Rastogi (eesharastogi@gmail.com)
158 AJTCCM VOL. 28 NO. 4 2022
RESEARCH
in COVID‑19 patients; (ii) to evaluate the role of chest radiographs
in conjunction with demographic details in determining the need for
early ventilator support in COVID‑19 patients; and (iii) to correlate
chest radiographic ndings with SPO2/PaO2 in COVID‑19 patients
needing early oxygen inhalation.
Materials and methods
This hospital‑based, retrospective, cross‑sectional study was
conducted on 135 severely ill patients admitted to the intensive care
unit of our institution over a period of 9months, following approval
from the institutional ethics committee using strict inclusion and
exclusion criteria.
Inclusion criteria: Patients with COVID‑19 infection showing
positivity on the RT‑PCR test who needed oxygen/ventilatory support
and were positive with available chest radiograph.
Exclusion criteria: Patients with poor‑quality chest radiographs,
or who were pregnant.
e ndings of the rst CXR taken at the time of admission were
used in our study. All CXRs were reported on by a radiologist with at
least 3years of experience in CXR interpretation. e CXR was divided
into 3 zones (upper, middle and lower) and scoring was given from
0‑3 depending on the percentage of involvement by consolidation
or ground‑glass opacity in that respective zone, i.e.ascore 0 for no
involvement, score 1 for one‑third (<33%) involvement, 2 for one‑
third to two‑thirds (34‑66%) involvement, and 3 for two‑thirds (66%)
involvement. An additional score of 1 point was given for the presence
of pleural eusion. Hence a maximum score of 10 could be given to
each lung and a score of 20 for both (Fig.1).
Appropriate statistical tests were applied after collection of
data in a predesigned pro forma. Sensitivity, specicity, positive
predictive value, negative predictive value and accuracy of values
were calculated.
Observations and analysis (Figs2‑6)
Table1 shows baseline characteristics among the study subjects. e
majority of the subjects were male (60%). Mean age of the patients
was 54years. e CXR severity score was >8 in 104 cases out of 135
subjects, with 98 needing a ventilator and 37 oxygen support. Mean
SPO2 was 86% and mean PaO2 was 79.88.
Table2 and Fig.7 describe the CXR severity score ndings among
the cases and their correlation with SPO2 and PaO2. Among the cases
having an SPO2 level of 91.35 ± 1.082 and PaO2 level of 88.33 ± 1.826,
the reported CXR severity score was <8 whereas cases having an SPO2
level of 85.69 ± 4.950 and PaO2 level of 78.88 ± 6.515, reported a CXR
severity score >8. e dierence was statistically signicant (p<0.01).
Table3 describes oxygen support and ventilator support for ndings
among the cases and the correlation with mean CXR severity score.
Among the patients using oxygen support, the mean CXR severity
score was 5.62 whereas patients on ventilator support had a mean
CXR severity score of 12.04. e comparisons of oxygen support and
ventilator support with mean chest radiograph severity scores were
statistically signicant (p<0.01).
Table4 describes oxygen support and ventilator support ndings
among the cases and their correlation with CXR severity score using
a cut‑o value of 8. Among the subjects having <8 CXR severity score,
31 subjects required oxygen support and no subjects required ventilator
support, whereas among the subjects having >8 CXR severity score, 6
subjects required oxygen support, and 98 subjects required ventilator
support. This comparison of CXR severity score with oxygen and
ventilator support using a cut‑o value of 8 was statistically signicant
(p<0.01).
Pleural eusion
Lower
zone
Middle
zone
Upper
zone
Fig.1. Schematic diagram and corresponding chest radiograph showing
arbitrary division of three lung zones and percentage of involvement by
consolidation or ground-glass opacity of that respective zone, i.e.score 1
for one-third (33%) involvement (shown by green colour), 2 for one-third
to two-thirds (34-66%) involvement (blue colour) and 3 for two-thirds
(66%) involvement (shown by orange colour). A score of point one is
given for presence of pleural eusion for each side (shown by sky-blue
colour).
AJTCCM VOL. 28 NO. 4 2022 159
RESEARCH
Fig.2. Chest radiograph showing opacity involving two-thirds (67%)
area of bilateral middle and lower lung zones with obscuration of
bilateral costophrenic angles, representing chest radiograph severity
score of 14. is patient required ventilatory support.
Fig.3. Chest radiograph showing opacity involving more than 33% but
less than 66% area of the bilateral middle lung zones and more than
two-thirds (67%) area of bilateral lower lung zones without obscuration
of bilateral costophrenic angles, representing chest radiograph severity
score of 10. is patient required ventilatory support.
Fig.4. Chest radiograph showing opacity involving less than 33% area of
the right upper lung zone and more than 33% but less than 66% area of
right lower lung zone with right-sided pleural eusion, representing chest
radiograph severity score of 4. is patient did not require ventilatory
support.
Fig.5. Chest radiograph showing opacity involving more than 33% but
less than 66% area of the right upper lung zone, less than 33% area of the
le upper lung zone and more than two-thirds (67%) area of the bilateral
middle and lower lung zones with obscuration of bilateral costophrenic
angles, representing chest radiograph severity score of 17. is patient
required ventilatory support.
160 AJTCCM VOL. 28 NO. 4 2022
RESEARCH
Fig.8 shows a ROC curve analysis of CXR severity score with an area
under the curve of 0.973 with a cut‑o value >8 having a sensitivity of
85.7%, specicity of 92.5% and accuracy of 89.5%.
Discussion
e COVID‑19 pandemic placed unforeseen burdens on healthcare
and highlighted the role of CXR in management, admission and
predicting outcomes, especially with limited resources.[3,4] In our
study, we evaluated the role of initial CXR in COVID‑19 disease
in emergency settings. e CXR severity score on the initial chest
radiograph can be used as a predictor for ventilatory support.[3]
Other criteria such as old age, obesity, diabetes and hypertension‑
like comorbidities are important factors in predicting ventilatory
support.[3]
In the present study, most of the subjects were male (60%) and their
mean age was 54 ± 14years. Most of the patients with high CXR
severity scores were >50years of age. In a study by Hanley etal.,[7] out
of 325 hospitalised and COVID‑19‑positive patients, 63% were male,
with 65years as the mean age of the study group. In another study
by Toussie etal.,[3] out of 338 COVID‑19‑positive patients, 62% were
male, with a median age of 39years.
Hanley etal.[7] in their study, also stated that out of 325 patients,
non‑invasive ventilation was required in 9%, intubation and ICU
admission were required in 14% during admission, and 69% were
discharged without intubation.
In our study, patients with higher SPO2 (91.35 ± 1.082) and PaO2
(88.33 ± 1.826) levels had lower CXR severity scores, i.e. <8 and vice
versa. is dierence was statistically signicant (p<0.01).
We believe that our study is unique as no other study in recent
medical literature has compared CXR severity score with SPO2 or
PaO2. However, in a study by Baratella etal.,[8] degree of hypoxia
was assessed using the PaO2/FiO2 ratio. Lower baseline PaO2/FiO2
values were reported in critically ill patients than in non‑severely ill
patients. In addition, patients with signicantly lower PaO2/FiO2 ratios
succumbed. Balbi etal.[4] in their study, stated that older patients with
a higher number of comorbidities had lower SpO2 and PaO2/FiO2
values along with severe CXR ndings at the time of admission than
in patients who survived (p <0.001).
In our study, CXR severity score in patients needing oxygen support
was lower (5.62) v. those needing ventilatory support (12.04). is
dierence was statistically signicant (p<0.01). Hanley etal.[7] in their
study, reported a median score of 11.5 for pre‑intubation CXR in an
ICU group with a median score of 9 at the time of admission. Toussie
etal.[3] reported that a CXR severity score ≥3 independently predicted
intubation with a p‑value of 0.002.
In our study, among the subjects having a CXR severity score <8,
31 subjects required oxygen support and no subject required ventilator
support. Among patient with a CXR severity score >8, 98 of 104 cases
required ventilator support while only 6 required oxygen support. is
dierence was statistically signicant (p<0.01). Hanley etal.[7] reported
Fig.6. Chest radiograph showing opacity involving less than 33% area of
the bilateral upper and middle lung zones, more than 33% but less than
66% area of the right lower lung zone and more than two-thirds (67%)
area of le lower lung zone with obscuration of bilateral costophrenic
angles, representing chest radiograph severity score of 11. is patient
required ventilatory support.
Table3. Comparison of support according to mean total
chest radiograph severity score
Support Mean (SD) t‑test p‑value
Oxygen support 5.62 (1.605) 140.12 <0.01*
Ventilator support 12.04 (3.142)
*Statistically signicant.
Table2. Comparison of SPO2 and PaO2 with total CXR
severity score
Total score Statistical characteristic SPO2PaO2
<8 Mean 91.35 88.33
SD 1.082 1.826
>8 Mean 85.69 78.88
SD 4.950 6.515
t‑test ‑ 39.79 28.45
p‑value ‑ <0.01* <0.01*
CXR = chest X‑ray.
*Statistically signicant.
Table1. Baseline characteristics among the study subjects
Variables N=135 %
Male 81 60
Female 54 40
Age inyears, mean (SD) 54.35 (14.49)
Oxygen support 37 27.4
Ventilator support 98 72.6
Chest score, mean (SD) 10.28 (4.01)
<8 31 23
>8 104 77
SPO2, mean (SD) 86.99 (4.98)
PaO2, mean (SD) 79.88 (6.84)
SD = standard deviation
AJTCCM VOL. 28 NO. 4 2022 161
RESEARCH
that patients who had a baseline CXR score of 9‑11.5 were admitted
to ICU and intubated, with a p‑value <0.01. ey also stated that
a similar score was also seen in deceased patients during their
hospitalisation, with a p‑value <0.001. Toussie etal.[3] studied the
severity of opacication on the initial CXR and reported that the
need for hospitalisation or for ventilatory support could be assessed
by severity of opacication. If at least two lung zones were aected,
the patient should be hospitalised and, if opacities were present in
≥3lung zones, the patient would require ventilatory support.
In our study, ROC curve analysis of the CXR severity score with
an area under the curve of 0.973 with a cut‑o value >8 revealed a
sensitivity of 85.7%, specicity of 92.5% and accuracy of 89.5%.
Study limitations
• Firstly, owing to the study’s retrospective nature, observer
bias cannot be avoided. As chest radiographs were done in
the emergency department, reports could have inuenced the
decision of physicians to admit, resulting in overestimating the
relationship between admission and chest radiograph severity.
However, the degree of inuence is unclear, and a previous study[9]
has reported that physicians in the emergency department do
not take chest radiographs as the basis of decision‑making for
admitting community‑acquired pneumonia patients.
• Secondly, because patients were bed‑ridden, most of the CXRs
included in the study were portable, taken in the antero‑posterior
projection, leading to suboptimal evaluation of the lungs.
• As there was an imbalance between the need for ventilatory
support and resources, not all the patients who needed ventilation
had their requirements fullled.
• Readers of the study were not blinded to the severity of outcome.
• Comorbidities might have aected the CXR ndings.
Conclusion
Based on the results of our study, we conclude that the chest radiograph
severity score noted on initial radiography at the time of admission
to the emergency department can be used to evaluate the need for
oxygen/ventilator support in COVID‑19 patients. Patients with a chest
radiograph severity score >8 pose a high chance of requiring ventilatory
support; hence, aggressive management with close monitoring is needed
to reduce overall mortality and morbidity in such patients.
Declaration. None.
Acknowledgements. None.
Author contributions. Each author has contributed significantly at
dierent steps of the study and preparation of the manuscript,
Funding. None received.
Conicts of interest. None.
Table4. Comparison of support according to total chest radiograph severity score
Support
Number of patients
and percentage
Total score Chi‑square p‑value
<8 >8
Oxygen support n31 6 106.58 <0.01*
% 100.0% 5.8%
Ventilator support n0 98
% 0.0% 94.2%
Tot al N31 104
% 100.0% 100.0%
* Statistically signicant.
95
90
85
80
75
70
65
60
Total Score <8 Total Score >8
91.35
85.69
88.33
78.88
SP02AIK (Pa02)
Fig.7. Graphical representation of comparison of SPO2 and PaO2 with
total CXR severity score.
1.0
0.8
0.6
0.4
0.2
0.0
Sensitivity
1 - Specicity
ROC Curve
Fig.8. e ROC curve.
162 AJTCCM VOL. 28 NO. 4 2022
RESEARCH
1. Xiang C, Huang L, Xia L. Mobile chest X‑ray manifestations of 54 deceased patients
with coronavirus disease 2019: Retrospective study. Medicine 2020;13:99(46):e23167
2. Salehi S, Abedi A, Balakrishnan S, Gholamrezanezhad A. Coronavirus disease
2019 (COVID‑19): A systematic review of imaging ndings in 919 patients. Am J
Roentgenol 2020;1:1‑7.
3. Toussie D, Voutsinas N, Finkelstein M, etal. Clinical and chest radiography features
determine patient outcomes in young and middle age adults with COVID‑19.
Radiology 2020;14:201754.
4. Balbi M, Caroli A, Corsi A, etal. Chest X‑ray for predicting mortality and the need for
ventilatory support in COVID‑19 patients presenting to the emergency department.
Eur Radiology 2020;8:1‑4.
5. Xiao N, Cooper JG, Godbe JM, etal. Chest radiograph at admission predicts early
intubation among inpatient COVID‑19 patients. Eur Radiology 2020;13:1‑8.
6. Yasin R, Gouda W. Chest X‑ray ndings monitoring COVID‑19 disease course and
severity. Egyptian J Radiol Nuclear Med 2020;51(1):1‑8.
7. Hanley M, Brosnan C, O’Neill D, etal. Modied Brixia chest X‑ray severity scoring
system and correlation with intubation, non‑invasive ventilation and death in a
hospitalised COVID‑19 cohort. J Med Imaging Radiat Oncol 2022;66(6):761‑767.
https://doi.org/10.1111/1754‑9485.13361.
8. Baratella E, Crivelli P, Marrocchio C, etal. Severity of lung involvement on chest
X‑rays in SARS‑coronavirus‑2 infected patients as a possible tool to predict clinical
progression: An observational retrospective analysis of the relationship between
radiological, clinical, and laboratory data. J Bras Pneumol 2020;46(5):e20200226.
https://doi.org/10.36416/1806‑3756/e20200226
9. Fine MJ, Hough LJ, Medsger AR, etal. e hospital admission decision for patients
with community‑acquired pneumonia. Results from the pneumonia Patient Outcomes
Research Team cohort study. Arch Intern Med 1997;157(1):36‑44.
Accepted 27 September 2022.
