key: cord-0867989-lpbshboj authors: Sverzellati, Nicola; Ryerson, Christopher J; Milanese, Gianluca; Renzoni, Elisabetta A; Volpi, Annalisa; Spagnolo, Paolo; Bonella, Francesco; Comelli, Ivan; Affanni, Paola; Veronesi, Licia; Manna, Carmelinda; Ciuni, Andrea; Sartorio, Carlotta; Tringali, Giulia; Silva, Mario; Michieletti, Emanuele; Colombi, Davide; Wells, U Athol title: Chest x-ray or CT for COVID-19 pneumonia? Comparative study in a simulated triage setting date: 2021-02-11 journal: Eur Respir J DOI: 10.1183/13993003.04188-2020 sha: 44839b459c58018aee80fb8e09086189b03b6e45 doc_id: 867989 cord_uid: lpbshboj INTRODUCTION: for the management of patients referred to respiratory triage during the early stages of the SARS-CoV-2 pandemic, either chest radiograph (CXR) or computed tomography (CT) were used as first-line diagnostic tools. The aim of this study was to compare the impact on triage, diagnosis and prognosis of patients with suspected COVID-19 when clinical decisions are derived from reconstructed CXR or from CT. METHODS: we reconstructed CXR (r-CXR) from high-resolution CT (HRCT) scan. Five clinical observers independently reviewed clinical charts of 300 subjects with suspected COVID-19 pneumonia, integrated with either r-CXR or HRCT report in two consecutive blinded and randomised sessions: clinical decisions were recorded for each session. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and prognostic value were compared between r-CXR and HRCT. The best radiological integration was also examined to develop an optimised respiratory triage algorithm. RESULTS: interobserver agreement was fair (Kendall's W=0.365; p<0.001) by r-CXR-based protocol and good (Kendall's W=0.654; p<0.001) by CT-based protocol. NPV assisted by r-CXR (31.4%) was lower than that of HRCT (77.9%). In case of indeterminate or typical radiological appearence for COVID-19 pneumonia, extent of disease on r-CXR or HRCT were the only two imaging variables that were similarly linked to mortality by adjusted multivariable models CONCLUSIONS: the present findings suggest that clinical triage is safely assisted by CXR. An integrated algorithm using first-line CXR and contingent use of HRCT can help optimise management and prognostication of COVID-19. Despite worldwide efforts to halt its transmission, severe acute respiratory syndrome coronavirus type 2 (SARS-CoV-2) has affected more than 30 million individuals and caused nearly 1 million deaths as of late September 2020 [1, 2] . After the initial outbreak, most countries have prepared their healthcare systems to face the pandemic. Although highly desirable, global and shared preparedness planning has faced political, institutional, social, environmental, and technological challenges [3, 4] . A recent International survey reported substantial heterogeneity in the diagnostic approach to coronavirus disease 2019 (COVID-19) pneumonia within and among countries and continents [4] . To date, molecular testing is used in both symptomatic or asymptomatic subjects with risk of contamination. However, the use of imaging, particularly the choice of imaging technique, is still a matter of debate [5] [6] [7] [8] . Molecular and imaging testing are helpful in different aspects of the disease and their integration should be driven by dynamic protocols to be adapted as knowledge of the disease improves. The use of real-time reverse transcription-polymerase chain reaction (RT-PCR) was adapted to the massive needs by shortening of reaction time (and thus reporting time), yet it is still challenged by a substantial proportion of false negatives [9] . Conversely, imaging can show signs of pneumonia in patients with negative RT-PCR but clinically suspected COVID-19, thus offering a potential role in supporting rapid decision making [9, 10] . The use of imaging was thus recommended for patients who present at triage with moderate to severe features of COVID-19 pneumonia regardless of RT-PCR results [7] . Different imaging approaches are available and have been discussed in the recent COVID-19 literature, including chest X-ray (CXR) and computed tomography (CT). There is still no international consensus upon the integrated use of CXR or CT for clinical assessment and management of subjects with suspected COVID-19 pneumonia [4] [5] [6] [11] [12] [13] [14] [15] [16] [17] . Most concerns are focused on the accuracy of these tests, individual resources, and risk of infection for radiographers and other healthcare employees. However, the scientific debate lacks important evidence on the impact of CXR and CT on triage decisions and patient care. In this study, we sought to use a post-processing imaging technique to retrospectively reconstruct CXR from CT scan, and compare the impact of these two imaging tools on the initial clinical triage, diagnosis and prognosis of patients with suspected COVID-19. The study population comprised patients who had been evaluated with chest highresolution computed tomography (HRCT) scan by the COVID-19 respiratory triage of the University Hospital of Parma, which is located in one of the most affected areas in Northern Italy. In brief, patients were screened for symptoms (e.g. fever and dyspnea) and oxygen saturation. Patients with moderate to severe pulmonary involvement (e.g. oxygen saturation 95%) underwent HRCT scan. Given the turnaround times for SARS-CoV-2 testing results (e.g. ranging from 2 to >48 hours), a presumptive diagnosis based on clinical-radiological findings was considered for swift decision making such as discharge, recommendation for home quarantine, or hospitalization in different hospital areas including either dedicated COVID-19 pavilions or non-COVID-19 wards. Details on clinical triage were previously reported [8] . The study derivation cohort was built up by including 300 patients who consecutively underwent HRCT at the Parma triage from 29 th February to 7 th March 2020, as follows: the first 200 patients consecutively admitted to the triage with moderate-severe respiratory clinical impairment were mixed up with the first 100 patients consecutively admitted with mild respiratory clinical impairment (e.g. oxygen saturation 96-98%, hyperthermia, tachypnea) (Fig. 1 ). The addition of this subgroup ensured that the full spectrum of COVID-19 severity was evaluated. The study results were externally validated in a cohort of 104 patients (validation cohort) consecutively evaluated at a neighbouring hospital during the same time frame (Piacenza, Italy), which adopted a similar diagnostic protocol (see also supplementary material). This retrospective study was approved by the referring local Review Board. Informed consent was obtained from the study patients. Details on CT scanners and HRCT technique are reported in the supplementary material. HRCT scans allow for various post-processing reconstruction algorithms, including Average Intensity Projection (AIP) [18] . AIP images represent the average of each component attenuation values encountered by the X-ray beam through an object. By manipulating the slab thickness of the coronal AIP mages, it was possible to obtain images analogous to frontal CXR, hereafter called reconstructed (r)-CXR (online videoclip, Fig. 2, 3, 4) . Further illustrations, videos and technical details of the conversion from HRCT to r-CXR imaging, as well as evaluation of r-CXR consistency with standard CXR are reported in the supplementary material. HRCTs were prospectively scored by a senior chest radiologist (NS, 16-year experience in imaging of interstitial lung disease). He recorded individual HRCT abnormalities and graded the HRCTs into four diagnostic categories, as follows: normal, alternative diagnosis (to be specified), indeterminate, or typical for COVID-19 pneumonia [8] . The total extent of pulmonary disease was scored to the nearest 5%. In keeping with the visual scoring of the HRCT, two radiologist observers (CM and AC with 11-and 4-year experience in chest imaging, respectively) recorded individual r-CXR abnormalities and graded the r-CXR as follows: normal, alternative diagnosis (to be specified), indeterminate, or typical for COVID-19 pneumonia. The total extent of pulmonary disease on r-CXR was evaluated through an overall visual impression, using a four-point-scale: 0) no parenchymal abnormality; 1) <20%; 2) 20-50%; 3) >50%. Interobserver discrepancies for both diagnostic categories and disease extent were resolved by consensus. Clinical data for each patient were jointly reported in data sheets by a consultant anesthesiologist (AV) and a radiologist (GM) Table 1 . These data were assembled into individual clinical charts that were given for simulation of clinical management to each clinical observer of this study. RT-PCR test results were not included in the clinical charts in order to simulate an environment where turnaround times were long. Five clinical observers from three different countries participated in the study. These comprised three physicians (AV, IC, FB) who worked at a HRCT-based triage, and two physicians (PS, EAR) whose hospital protocol proposed a CXR-based triage. Details on the study observer characteristics are given in the supplementary material. Each clinical observer was asked to independently read twice the full set of clinical charts according to two different settings, as follows: 1) The first review was designed to simulate r-CXR-based integrated clinico-radiological protocol: each clinical observer reviewed the data sheet and r-CXR report for each patient. Each clinical observer was asked to provide a clinical decision according to one of the following options: a. discharge, b. hospitalization in non-COVID-19 area, c. home quarantine, d. hospitalization in COVID-19 area, e. further work-up by chest HRCT. 2) The second review was designed to simulate HRCT-based integrated clinico-radiological protocol: after two days from the first review completed, each clinical observer started the second of data sheet including HRCT report for each patient. Each clinical observer was asked to provide a clinical decision according to one of the following options: The study observers were informed that this triage setting was supposed to simulate a pandemic environment with high influx of subjects with suspected COVID-19 pneumonia, and that the clinical decision could be expressed in the absence of any resource constraints. The study analysis compared the frequency of each clinical decision category and its consistency between "r-CXR-based protocol" and "HRCT-based protocol", by means of both intraand inter-observers analysis. Either Chi-squared or McNemar test were used to compare r-CXR and HRCT diagnostic categories and RT-PCR results. Radiological data were compared between r-CXR and HRCT by the weighted-kappa coefficient to evaluate interobserver agreement. Kendall W test was used to evaluate the overall interobserver clinical decision agreement by r-CXR-based or HRCT-based protocol. Further details are provided in the supplementary material. Spearman's rank correlation coefficient disease was used to evaluate the correlation between r-CXR-extent and HRCT-extent of disease. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of both r-CXR and HRCT diagnostic categories were calculated against RT-PCR, by grouping normal to alternative diagnosis and indeterminate to typical for COVID-19 pneumonia diagnostic categories, respectively. Unadjusted and multivariable logistic regression analyses were used to identify the contribution of clinical and radiological variables to mortality prediction. Multivariable models included age, sex, duration of symptoms at triage, and a comorbidity score of 0-4, obtained by summing the presence (1 point for each) of individual comorbidities consistently associated with a poor outcome in previous reports (diabetes, hypertension, cardio-vascular disease, obesity). The classification performances of the models were evaluated by the area under the receiver operating characteristic curve (AUC). Based on these models, we constructed a user-friendly five-point scale for both CT and CXR, integrating age and extent of disease on imaging (the only two variables emerging as strong independent determinants of mortality). These scales were externally tested in the validation cohort. P values of less than 0.05 were considered to indicate statistically significant differences. Analyses were performed using STATA software (STATA Version 14; Computing Resource Centre, Santa Monica, CA). The derivation cohort included 300 patients (188 men and 112 women, 66.8 15.8, age range 23.1-97.6 years). Clinical and laboratory characteristics are summarized in Table 1 The characteristics of both derivation and validation cohort are summarized in Table 1 . The clinical decision according to either r-CXR-based protocol or HRCT-based protocol is summarized in Tables 2 and 3. The overall interobserver agreement by r-CXR-based protocol was fair (Kendall's W = 0.365; p<0.001), and it improved to good by HRCT-based protocol (Kendall's W = 0.654; p<0.001). This was true even after stratification for categories of oxygen saturation levels (Supplementary Table 1 ). Of note, opposite trends in the agreement across the range of prespecified oxygen saturation levels were observed: the lower the oxygen saturation the worse the agreement by r-CXR and the better the agreement by HRCT-based protocol (Supplementary Table 1 ). Table 2 ) by r-CXR ranged from poor (0.17 -Obs 1 /Obs 5 ) to moderate (0.50 -Obs 2 /Obs 4 ), and it ranged from fair (0.23 -Obs 1 /Obs 5 ) to good (0.75 -Obs 2 /Obs 4 ) by HRCT-based protocol. Intra-observer agreement on the clinical decision by either r-CXR or HRCT ranged from fair (0.37) to good (0.71) (Supplementary Table 3 ). In the r-CXR round, further work-up by HRCT scan was requested in 8% to 46% of patients by the study observers. Once the HRCT was provided, the most frequent decision in this group was COVID hospitalization (41.7% to 73.2%) ( Table 3 ). In particular, HRCT scan was requested by at least one observer for a total of 224 (74.7%) patients, while a management decision without request of HRCT (only clinico-r-CXR findings) was expressed by all observers in 76 (25.3%) patients (Supplementary Table 4 ). In the latter subgroup, the interobserver agreement was similar between r-CXR (Kendall's W = 0.763; p<0.001) and HRCT (Kendall's W = 0.725; p<0.001). When r-CXR was considered sufficient for the decision making by all the observers, patients had higher oxygen saturation, and reported more frequently a history of social contact with COVID-19 infected individuals, as compared to patients for whom HRCT was requested by any observer (Supplementary Table 4 ). r-CXR vs. HRCT r-CXR and HRCT data are summarized in Table 1 . Interobserver analysis data is given as supplementary data. Eighty-five r-CXR were Unadjusted associations between mortality and core clinical and imaging variables in patients with r-CXR findings compatible with COVID-19 pneumonia are summarized in Table 4 . Two models examined mortality against the extent of disease on either r-CXR or HRCT, using the same 4-point categorical grading system (Table 4 ). In the logistic regression model with r-CXR, mortality was associated with extent of disease (OR=2.38; 95%CI 1.61-3.50; p<0.001). In the logistic regression model with HRCT, mortality was associated with extent of disease (OR=2.62; 95%CI 1.67-4.10; p<0.001). Both r-CXR and HRCT associations were robust with inclusion in models of age (also strongly linked to mortality), sex, duration of symptoms, and comorbidity score (Supplementary Table 5) . A further third model was built to explore a two-point grading system for risk stratification: the extent of disease was classified by contingent categories as either limited (r-CXR extent<20%; or r-CXR 20-50%, with HRCT extent 0-50%) or extensive (r-CXR extent >50%; or r-CXR 20-50%, with HRCT extent >50%). The distinction between limited and extensive disease was strongly associated with mortality (OR=5.24; 95%CI 2.69-10.22; p<0.001) and was robust with adjustment for age, sex, and comorbidity score (Supplementary Table 5 ). For each of the three abovementioned models, AUC values fell minimally when multivariable analysis was confined to age and extent of disease on imaging (Supplementary Table 5 ). Therefore, simplified scoring systems for r-CXR and HRCT were constructed based on these two variables. Age was categorized as <60 (score 0), 60-74 (score 1), and >74 (score 2) years, dividing the cohort into approximate thirds (OR=2.79; 95%CI 1.80-4.31; p<0.001). Imaging extent scores were categorized as <20% (score 0); 20-50% (score 1); >50% (score 2) (Fig. 2, 3, 4 ). Age and imaging scores were summed to provide five point scales (0-4) for age/r-CXR and age/HRCT scoring systems. Mortality in relation to the age/r-CXR and age/HRCT scores are shown in Fig. 5 and Supplementary Table 6 . In a logistic regression model, the age/r-CXR score was associated with mortality (OR=2. In this simulated COVID-19 pandemic triage for a high influx of patients with mostly moderate-to-severe clinical features suspicious of COVID-19 pneumonia, the use of r-CXR for morphology and extent of pneumonia allowed fair interobserver agreement for clinical management, however r-CXR with moderate extent displayed limited yield that could be assisted by integration by HRCT. The interobserver agreement considerably increased when using HRCT, which lead to an increase in the number of recommended hospitalizations. This observation is of particular relevance as the observers were from different countries, used various COVID-19 triage strategies in their daily practice, and had different subspecialty interests. The major source of disagreement for the r-CXR-based protocol was the frequency of the requested HRCT scans, which ranged from 8% to 46% of cases. When HRCT work-up was not recommended by any of the study observers, the agreement was similar to that obtained with the HRCT-based protocol. Although we observed a tendency in considering the r-CXR sufficient in patients with milder pulmonary dysfunction, abnormal r-CXR and high exposure risk, the variety in requesting HRCT scans had still no obvious reasons (e.g. age, disease extent on r-CXR etc.). Such observation emphasizes the need to define the clinical indications to CT scanning when the triage workflow relies on CXR as the first imaging modality. In fact, poorly defined criteria may unreasonably increase the number of CT scans after CXR, with potential detrimental effects on the workflow and healthcare worker safety. In order to evaluate the clinical implications of the two imaging modalities, the sensitivity, the specificity, and the prognostic value of r-CXR and HRCT were compared. The levels of sensitivity for r-CXR were in keeping with previous findings [9, 16, 17, 19] . Sensitivity, specificity, and PPV of r-CXR were marginally lower than HRCT, whereas major discrepancy was found for NPV due to the remarkable proportion of false negative r-CXRs. Any optimization process of COVID-19 triage protocols should consider mortality data and resource utilization predictions [20] . First, the proportion of deaths among patients with normal r-CXR was higher (12%) as compared to normal HRCT (5%). Such a discrepancy was also evident by comparing r-CXR-and HRCT-based prognostic scoring systems ( Figure 5 ). Moreover, the distinction between typical COVID-19 pneumonia and indeterminate findings was not worthwhile for both r-CXR and HRCT as they were associated with similar outcomes. This suggests that, in a pandemic situation, indeterminate CXR or CT findings likely represent signs of COVID-19 pneumonia. Hence, after grouping indeterminate and typical COVID-19 categories, we found that the diagnostic categories had fairly major prognostic significance for both r-CXR and HRCT. This finding is important as the diagnostic categories used in the present study substantially overlap both CO-RADS categories and those proposed by the Radiologic Society of North America [8, 21] . Our findings do not suggest that HRCT extent scoring adds greatly to prognostic evaluation, over and above CXR scoring, and this is concordant with the need to first level approach by CXR to optimize use of radiology resources and maximize patient and healthcare employee safety. However, an alternative strategy -HRCT complement when CXR extent findings are intermediate -allowed categorization into limited and extensive disease with a major prognostic separation. The proposed contingent staging algorithm -where r-CXR grades 1 (<20%) and 3 (>50%) are accepted and HRCT is used to adjudicate on r-CXR 2 (21-50%) -showed the highest prognostic value. Separation into a higher risk/lower risk dichotomy may have major clinical value but we suggest that this observation requires further exploration if CT is to be integrated into routine prognostic evaluation in selected cases, when CXR findings are not definitive. The higher number of deaths in subjects with normal r-CXR still represents the main limitation of this staging system. In patients dying despite a r-CXR grade of 1, the HRCT extent grade was either intermediate or extensive in two thirds of cases. These data suggest that a first level approach by CXR will require some improvement in terms of sensitivity. Otherwise, first level approach by HRCT would provide major confidence in clinical decision and prognostication, therefore it might considered in the context of local logistics that minimize risk of contamination (e.g. dedicated scanner in proximity of the triage rooms). This study has some limitations. It is worth emphasizing that our observations were derived from a pre-peak endemic environment (e.g. when RT-PCR demand was beyond capacity), while the triage process is being adjusted as the endemic changes. Nevertheless, the study findings may still be helpful in epidemic scenarios, the increased availability of the lab test results within a few hours could further mitigate the CXR limitations (e.g. lower diagnostic accuracy as compared to CT), encouraging its use. Moreover, the reported heterogeneity in the request of HRCT scans among the study observers when first interpreting a CXR might be overrated as a simulated triage setting does not likely reflect the routine clinical context, where multidisciplinary discussion (e.g. between radiologists and triagers) could support other clinical decisions with better integration of available data. The CXR was derived from HRCT data given the impossibility of running a two armrandomized-controlled trial between CXR and CT in the COVID-19 setting, yet we tested this approach against original CXR (supplementary materials). In conclusion, this study showed that r-CXR findings were often regarded as not sufficiently informative by clinicians in a COVID-19 pandemic triage setting. This observation suggests that chest CT should be considered after CXR in a substantial percentage of patients with suspected COVID-19 pneumonia, thus potentially causing detrimental effects in the absence of pre-defined diagnostic work-up criteria. Nevertheless, the present study findings suggest that clinicians could rely on positive CXR showing low or high extent of pneumonia, whereas intermediate extent by CXR should be complemented by CT for optimal classification into high and low risk group. Data collected for the study will not be made available to others. The study was not funded. the observer compared the r-CXR and CXR side-by-side and recorded adjunct signs that were visible on CXR (the standard of reference in use for clinical practice). This process was structured on a case-by-case basis to avoid the intra-observer bias potentially generated by an independent scoring of the experimental tool and the standard of reference. The consistency of detection for each individual finding was given as the ratio between its detection on r-CXR and that on standard CXR, as follows: consolidation (10/12), reticular opacities (5/7), ground glass (7/8). No pleural effusion was observed. The observer was also asked to provide her/his visual impression on the quality of r-CXR images, using the standard CXR as standard of reference. As compared to the standard CXR, the r-CXR was classified as follows: very similar (31/42), slightly different (7/42), different (2/42), remarkably different (2/42). r-CXR were all considered very similar to corresponding standard CXR (Fig. 1s, and 2s) . Non-contrast HRCT was performed with either a 128-slice scanner (SOMATOM Definition Edge, Siemens Healthineers, Erlangen, Germany) or a 16-slice mobile scanner on truck (SOMATOM Emotion, Siemens Healthineers, Erlangen, Germany). HRCT images were acquired with the patient in the supine position during end-inspiration breath-hold, without intravenous contrast material. The acquisition parameters were set at 110-120 kVp, 80 reference mAs, pitch 0.9-1. The extent of combined GGO and consolidation was visually scored at the nearest 5% on the whole lungs. The distribution was described as follows: a) axial distribution: predominantly peripheral (within the outer third of the lung), predominantly central, or mixed; b) cranio-caudal distribution: predominantly upper (above the carina), middle (between the carina and the right inferior pulmonary vein) or lower (below the inferior pulmonary vein) (1); c) bilateral or unilateral involvement; d) lobar involvement was accounted over 6 lobes (lingula considered as a single lobe). Description of the pattern was also tabulated into the various HRCT categories of our local COVID-19 protocol (2). These categories aimed to define disease severity by encompassing both morphology and extent of parenchymal findings, as follows: 1) non-COVID-19 findings, 2) findings indeterminate for COVID-19, either because of differential or overlapping disease, 3) typical pattern COVID-19, including different combinations of GGO and consolidations (from fluffy consolidation to organized pneumonia and their overall extent ( Table 1) . The interobserver variability was tested against a second radiologist observer (M.S., with 9- year experience in imaging of interstitial lung disease). The study observers ranged in age from 45 to 55 years, and their clinical experience ranged in age from 12 to 20 years. Three of the observers were pulmonologists with long -time ( Among the 100 patients with oxygen saturation 96-98%, fever (72%) and cough (58%) were most frequent symptoms (dyspnea 14%). Cardiovascular disease (43%), and chronic pulmonary disease (8%) were the most frequent comorbidities. 12 patients (12%) died after a mean time of 12 days. In the 73 patients with oxygen saturation ≤ 95%, fever (94%) and dyspnea (62%) were the most common symptoms (cough 48%). The majority of the patients (34/73, 46%) showed severe pulmonary involvement (oxygen saturation 88-92%); critical pulmonary involvement (oxygen saturation <88%) was observed in 21 (29%) patients, while moderate pulmonary involvement (oxygen saturation 95-93%) in 18 (25%) cases. The most frequent comorbidities were cardiovascular disease (68.5%), diabetes (18%), and cancer (18%). Death occurred in 29 (40%) patients in a mean time of 9 days. Among 31 patients with oxygen saturation >95%, fever (87.5%) and dyspnea (62.5%) were most frequent symptoms (cough 25%). The most common comorbidity was cardiovascular disease (54%), followed by diabetes (21%), cancer (17%), and kidney disease (17%). During a mean time of 12 days, death occurred in 8 (26%) patients. r-CXRs were obtained from corresponding HRCT scans using the same method described for the study cohort, with a different PACS system tool (Syngo.plaza, Siemens, Erlangen, Germany). There was good (k w = 0.74; 95% CI: 0.67-0.81) interobserver agreement for the r-CXR The lack of information from the observers about the reasons for requesting HRCT represents a study limitation that only became obvious at the analysis stage. The most likely explanation for the variable request of HRCT scans is the variability in the degree of mismatch between CXR findings and clinical data that would trigger the need for the most informative imaging diagnostic test to support challenging clinical decisions. However, the limitations of CXR with respect to CT are still largely unknown in the evaluation of suspected COVID-19 pneumonia. In a study evaluating the frequency and distribution of CXR findings in COVID-19 positive patients, only one out of four (25%) subjects had false negative CXR as compared to CT (3). In our cohort, the proportion of false negative r-CXRs was more than half of negative r-CXRs. Such discrepant findings may be related to several factors such as the larger study cohort, the high proportion of subjects with dominant ground-glass opacification on HRCT in this group of subjects with falsenegative r-CXRs (86.7%), and potential diagnostic limitations of the r-CXR technique. Indeed, the higher frequency of abnormalities (mostly typical for COVID-19 pneumonia) of variable extent on HRCT reports would explain the higher frequency of hospitalizations among the observers clinical decisions at the HRCT-based protocol round. The five point age/imaging scales provided very similar prognostic information for r-CXR and HRCT evaluation, which was reproduced in the validation cohort. These findings are also consistent with prior studies showing that the visual score of disease extent on CXR is independently predictive of outcome (e.g intubation, or mortality) (4-6). The advantage of this simplified approach is that other variables that had prognostic value on unadjusted analysis did not add materially to the accuracy of prognostic evaluation, quantified using ROC values. In particular, the comorbidity score was associated with mortality when examined in isolation, as observed in other cohorts, but was no longer significant when age and disease extent were taken into account. This likely reflects the association between comorbidities and age, as the comorbidity score remained significant when age was excluded from the multivariable analysis. Given the impossibility of running a two arm-randomized-controlled trial between CXR and CT in the COVID-19 setting, we sought to retrospectively compare the two modalities by artificially reconstructing a bidimensional image from coronal HRCT scans that was very similar to a standard bedside CXR. The AIP -a post-processing algorithm available on most CT workstations and dicom viewers -easily and rapidly allowed for r-CXR images that showed individual COVID-related findings very similar to those observed on standard CXR. Yet, the r-CXR may be less accurate of standard upright CXR that is obtained for subjects with suspicious COVID-19 pneumonia in the pandemic setting. Nevertheless, the good interobserver agreement for the interpretation of r-CXR, the levels of sensitivity and specificity in keeping with standard CXR, as well as the significant prognostic value of both diagnostic categories and disease extent on r-CXR further substantiate the utility of this surrogate tool. Furthermore, any potential limitations related to this postprocessing technique was mitigated by the fact that the study triagers were informed that r-CXR reports were identical to those achievable from standard CXR. However, either r-CXR or supine standard CXR showed bilateral consilidations that were confirmed on HRCT (C). Management of Critically Ill Adults With COVID-19 World Health Organization: Director-General's opening remarks at the media briefing on COVID-19 -29 The Lancet Infectious D. Challenges of coronavirus disease 2019 The use of imaging in COVID-19-results of a global survey by the International Society of Radiology ACR Recommendations for the use of Chest Radiography and Computed Tomography (CT) for Suspected COVID-19 Infection Askin tumor: four case reports and a review of the literature The Role of Chest Imaging in Patient Management during the COVID-19 Pandemic: A Multinational Consensus Statement from the Fleischner Society Integrated Radiologic Algorithm for COVID-19 Pandemic Correlation of Chest CT and RT-PCR Testing in Coronavirus Disease 2019 (COVID-19) in China: A Report of 1014 Cases Sensitivity of Chest CT for COVID-19: Comparison to RT-PCR The role of CT for Covid-19 patient's management remains poorly defined COVID-19) Outbreak: What the Department of Radiology Should Know Novel Screening and Triage Strategy in Iran During Deadly Coronavirus Disease 2019 (COVID-19) Epidemic: Value of Humanitarian Teleconsultation Service Policies and Guidelines for COVID-19 Preparedness: Experiences from the University of Washington Diagnosis, Prevention, and Treatment of Thromboembolic Complications in COVID-19: Report of the National Institute for Public Health of the Netherlands Diagnostic Performance of Chest X-Ray for COVID-19 Pneumonia During the SARS-CoV-2 Pandemic in Lombardy, Italy Chest X-ray for predicting mortality and the need for ventilatory support in COVID-19 patients presenting to the emergency department Computed tomography diagnosis of acute appendicitis: advantages of reviewing thin-section datasets using sliding slab average intensity projection technique Frequency and Distribution of Chest Radiographic Findings in COVID-19 Positive Patients Triage of Scarce Critical Care Resources in COVID-19 An Implementation Guide for Regional Allocation: An Expert Panel Report of the Task Force for Mass Critical Care and the Radiological Society of North America Expert Consensus Statement on Reporting Chest[ZERO WIDTH SPACE] CT Findings Related to COVID-19. Endorsed by the Society of Thoracic Radiology, the American College of Radiology, and RSNA CT Features of Coronavirus Disease 2019 (COVID-19) Pneumonia in 62 Patients in Wuhan, China Integrated Radiologic Algorithm for COVID-19 Pandemic Frequency and Distribution of Chest Radiographic Findings in COVID-19 Positive Patients Clinical and Chest Radiography Features Determine Patient Outcomes In Young and Middle Age Adults with COVID-19 COVID-19 outbreak in Italy: experimental chest X-ray scoring system for quantifying and monitoring disease progression Model-based Prediction of Critical Illness in Hospitalized Patients with COVID-19