Clinical data from 513 patients supporting: The accuracy of teleradiologists in diagnosing COVID-19 based on a French multicentric emergency cohort

Summary
This metadata record provides details of the
data supporting the claims of the related manuscript: “The accuracy of
teleradiologists in diagnosing COVID-19 based on a French multicentric emergency
cohort”.
The data consist of a single Excel .xlsx
format spreadsheet.
The related study aimed to evaluate the
accuracy of diagnoses of COVID-19 based on chest CT as well as interobserver
agreement between teleradiologists during on-call duty and senior radiologists
in suspected COVID-19 patients.
Patient
selection and study design

At our emergency
teleradiology center, all consecutive patients who were suspected, by a
board-certified emergency physician, to have COVID-19, and who underwent both
chest CT imaging and RT-PCR from 03/13/2020
to 04/14/2020 in 15 French emergency departments, were included. The
examinations took place during on-call duty periods between 6pm and 8.30am on
weekdays and 24 hours a day on weekends. A COVID-19 dedicated workflow was implemented, which consisted of a
standardised COVID-19 CT request form for the requesting physician, structured
and standardised radiological reports, and a systematic review by a senior
radiologist.

CT techniques

Chest CT
examinations were performed using a 16, 64 or 80-detector row CT scanner with a
standardised non-contrast chest CT COVID-19 protocol for all hospitals. If
pulmonary embolism was suspected, a CT pulmonary angiographic protocol with bolus-tracking
intravenous iodine contrast agent administration at a rate of 3–4 mL/sec was
used instead. The on-site requesting physician supervised contrast
administration if needed.

Radiology
interpretation protocol

The teleradiology
interpretation protocol met the current French recommendations for teleradiology
practice. Reports and requests with clinical data for the interpretation of COVID-19
chest CT images were received from partner hospitals at our teleradiology center,
using teleradiology software (ITIS; Deeplink Medical). The images were securely
transferred over a virtual private network (VPN) to a local picture archiving
and communication system for interpretation (PACS; Carestream Health 12).
Images were interpreted by a teleradiologist (TR) in two dedicated emergency reading rooms during the study period. The panel of TRs
consisted of 106 senior radiologists with at least five years of emergency
imaging experience (mean length of practice: seven years) and 45 junior
radiologists (i.e., residents) with between three and five years of
emergency imaging experience (mean length of practice: 4 years). TRs operated
an on-call rota in groups of at least five TRs per night, and the report turn-around times were recorded.

CT
examinations were systematically reviewed within a week after each on-call period by a
senior radiologist (15 senior radiologists; mean length of practice: 12.1 years) who was not involved in the on-call
duty period, blinded to RT-PCR results and the first reader report, and unblinded
to the patient’s medical history.

Clinical
data

Clinical
information was prospectively provided by emergency physicians upon presentation
and was collected using the teleradiology software as a dedicated COVID-19 CT
request form (ITIS; Deeplink Medical). This clinical information included: age;
gender; active smoking; significant medical history; recent medication with
anti-inflammatory drugs; time since onset of symptoms (categorised as: <1
week, 1–2 weeks, ≥2 weeks); oxygen saturation (categorised as: ≥95%, 90–95% and
<90%); dyspnoea; fever (³38°C); cough; asthenia; headache; and ear, nose and throat symptoms. The
RT-PCR results were retrospectively collected from the patients’ electronic
medical records by each partner hospital. The initial RT-PCR was considered as
the standard of reference.

Discrepancies between the RT-PCR results and the score
from the second reading were reviewed by contacting hospitals and investigating
patients’ outcome, in order to determine whether a second chest CT and/or a
second RT-PCR test had been performed.

CT image
analysis

Six common
radiological features were extracted from the specific, structured COVID-19 chest-CT
reports by the first and second independent readers. These features included
the presence of GGO, consolidation, fibrosis (with traction bronchiectasis and
architectural distortions), intralobular reticulations, and extent of
abnormalities (categorised as low [<25%], moderate [25-50%] or high
[>50%]). Additionally, the second reading included an assessment of image
quality (categorised as good, moderate or poor) and the following radiological
features: (a) underlying pulmonary disease (categorised as emphysema, lung
cancer, interstitial lung disease, pleural lesions, bronchiectasis); (b) GGO
pattern (categorised as rounded and non-rounded GGO); (c) consolidation pattern
(categorized as rounded, non-rounded consolidations and subpleural bands); (d)
predominant pattern (categorised as GGO or consolidation); (e) distribution
pattern of lesions (categorised as peripheral predominant, central predominant,
or mixed); (f) bilateral lesions; (g) diffuse lesions (i.e., five lobes
involved); (h) basal predominant lesions; (i) pleural effusion (categorised as
uni- or bilateral); (j) adenomegaly (defined as lymph node with short axis >
10 mm); (k) bronchial wall thickening (further categorised as
lobar/segmental or diffuse); (l) airways secretions; (m) tree-in-bud
centrilobular micronodules; and (n) pulmonary embolism.

Each
reading was categorised using a five-point score, adapted from the
recommendations of the Société Française de Radiologie (SFR): (1) normal; (2)
non-infectious findings; (3) infectious findings but not consistent with
COVID-19 infection; (4) consistent with COVID-19 infection; (5) typical
appearance of COVID-19 infection.

To support the
conclusions of the TRs, all radiologists underwent a two-hour e-learning
session based on reported chest-CT findings associated with COVID-19 from the
literature, which was made publicly available on the 7th April. A private
medical discussion group (PandaLab) was used at the onset of the outbreak, such
that on-call TRs could discuss and share images from their cases with all TRs
who were not on call, prior to completing their reports.
Additional manuscript using this data
The authors of the manuscript described above also have a second manuscript which uses this data and is currently under review. The manuscript is titled: "Practical Clinical and Radiological Models to Diagnose COVID-19 on Chest CT in a French Multicentric Emergency Population", and aims to develop practical models built with simple clinical-radiological features to facilitate COVID-19 diagnosis. When this manuscript is published, it will be added as a Reference on this figshare record.