Real-world evaluation of a computed tomography-rst triage strategy for suspected coronavirus disease 2019 in emergency outpatients in Japan: an observational cohort study

Background: Coronavirus disease 2019 (COVID-19) continues to spread worldwide. Because of the absence of reliable rapid diagnostic systems, patients with COVID-19 symptoms are suspected of disease. Computed tomography (CT) in patients with suspected COVID-19 may be reasonable for triaging, and CT-rst triage strategies have been proposed. However, clinical evaluation of a CT-rst triage protocol is lacking. The aim of this study is to investigate the real-world e�cacy and limitations of a CT-rst triage strategy in patients with suspected COVID-19. Methods: This was a single-center cohort study evaluating outpatients with suspected COVID-19 who underwent a medical examination at Yokohama City University Hospital and who were prospectively registered between 9 February and 5 May 2020. We treated patients according to the CT-rst triage protocol. CT �ndings were classi�ed into �ve categories according to the COVID-19 Reporting and Data System (CO-RADS). With the CT-rst triage protocol, patients with a suspicious clinical history, symptoms, or suspicious �ndings on chest CT were allocated to the COVID-19 suspected group. The primary outcome was e�cacy of the CT-rst triage protocol for outpatients with suspected COVID-19. We conducted additional analyses of the isolation time of outpatients with suspected COVID-19 and reached �nal diagnoses. Results: In total, 108 outpatients with suspected COVID-19 were examined at our hospital. Forty-eight patients (44.9%) were categorized as CO-RADS 1, 26 patients (24.3%) as CO-RADS 2, 14 patients (13.1%) as CO-RADS 3, 6 patients (5.6%) as CO-RADS 4, and 13 patients (12.1%) as CO-RADS 5. One patient was excluded because of pregnancy. Using the CT-rst triage protocol, 48 (44.9%) patients were suspected of having COVID-19. Nine patients (18.8%) in this group were positive for severe acute respiratory syndrome coronavirus 2 using polymerase chain reaction; no patients in the group not suspected of having COVID-19 were diagnosed with COVID-19 during follow up. The protocol signi�cantly shortened the duration of isolation for the not-suspected versus the suspected group (70.5 vs. 1037.0 minutes, P < .001). Conclusions: Our CT-rst triage protocol was acceptable for triaging outpatients with suspected COVID-19. This protocol will be helpful for appropriate triage, especially in areas where polymerase chain reaction is limited.


Introduction
2][3] COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has affected more than 4.6 million individuals, and has caused over 300,000 deaths to date. 4 The severity and case fatality rate of COVID-19 differs from country to country, depending on the health care system and health policies. 5,6Generally, the excess burden surpasses healthcare capacity for patients in intensive care and causes medical collapse, which may trigger an increase in the death rate.
Several studies have revealed factors associated with poor prognosis such as older age, underlying comorbidities, clinical symptoms, and laboratory ndings. 7Because hospitalized patients constitute a high-risk population, hospital outbreaks are among the most dangerous situations.Therefore, adequate use of disposable personal protective equipment (PPE) is recommended. 8Because of a global shortage of PPE, there are many reports of hospital outbreaks of COVID-19 and infection of medical staff.To prevent such outbreaks in hospitals, it is advisable to separate services for patients with suspected COVID-19. 5 reduce consumption of PPE and implement appropriate isolation, triage to distinguish patients who are less likely to have COVID-19 from those with suspected COVID-19 is necessary. 9,10However, there are no established protocols for managing patients with suspected COVID-19 who present to the emergency department.Because reliable rapid diagnostic systems are lacking, patients with symptoms related to COVID-19 such as fever, fatigue, respiratory symptoms, headache, and taste or olfactory disorders are widely treated as suspected COVID-19.Several triage protocols and their e ciency have been reported. 9,11,12However, because COVID-19 symptoms are not speci c to that disease, there might be many differential diagnoses in patients with suspected COVID-19.3][14] Generally, chest X-ray is useful to diagnose pneumonia; however, it is unsuitable for COVID-19.Because the characteristics of chest radiological images in COVID-19 mainly appear in the interstitium, revealing ground-glass opacities (GGO), it is di cult to diagnose COVID-19 using chest X-ray alone, especially in mild cases.Presently, computed tomography (CT) in patients with COVID-19 can be considered a reasonable method for triaging patients with COVID-19.However, consensus has not been reached regarding clinical evaluation of triage protocols using CT. 15 COVID-19 has affected 17,018 individuals and caused 903 deaths in Japan as of 3 June, 2020. 16kohama City is the second largest city in Japan.Since February 2020, we have treated patients with COVID-19 at Yokohama City University Hospital, a tertiary-level hospital, owing to the outbreak on the "Diamond Princess" cruise ship docked in Yokohama Bay. 17,18A select team (Team COVID-19) was deployed for the treatment of patients with moderate to severe COVID-19 and with suspected COVID-19.One focus of Team COVID-19 is on preventing in-hospital infections and medical staff infections, with reference to protocols that were previously launched during the epidemic of a novel type of in uenza in 2002 to 2003.Team COVID-19 comprises 10 doctors (mainly from the internal medicine and surgery departments) and engages with the outpatient clinic and hospitalized patients with suspected COVID-19 in a 24-hour rotation schedule.At our hospital, a protocol adopting CT as the rst-line examination, the "CT-rst triage protocol," has been used in the management of patients with suspected COVID-19.We adopted the CT-rst triage protocol for several reasons.There are not many patients with COVID-19 in Japan; thus, polymerase chain reaction (PCR) testing of individuals with no or mild symptoms might lead to many false-negative results because of low sensitivity of the PCR test. 19,20Furthermore, in some regions, such as Wuhan and Italy, many people with mild symptoms rushed to a hospital to receive PCR testing, leading to the collapse of medical services and subsequently, many in-hospital infections.To avoid such events and keep hospitals functioning normally, adopting the CT-rst triage protocol was reasonable in Japan.To verify the e cacy and limitations of this protocol, we focused on the characteristics of outpatients with suspected COVID-19 using the CT-rst triage protocol.To our knowledge, this is the rst detailed report using real-world clinical data regarding a CT-rst triage protocol used for patients with suspected COVID-19.

Study design and setting
This was a single-center cohort study aiming to evaluate outpatients with suspected COVID-19 who received medical examination at Yokohama City University Hospital, prospectively registered between 9 February, 2020 and 5 May, 2020.

Ethical approval
The Institutional Review Board at Yokohama City University Hospital approved this study (approval number B200200047).Consent for participation in this study was obtained from all patients after explaining the clinical study by verbal or a description of how to opt out.(https://www.yokohamacu.ac.jp/amedrc/ethics/ethical/fuzoku_optout.html).Participants Among all patients who visited our outpatient clinic, those who exhibited COVID-19-like symptoms such as fever (over 37.5℃), fatigue, respiratory symptoms, headache, and taste or olfactory disorders, were identi ed and instructed to visit our separate clinic as outpatients with suspected COVID-19.
We analyzed the following baseline patient characteristics: age, sex, history of overseas travel, contact history with a con rmed COVID-19 case, and underlying comorbidities (chronic pulmonary disease, diabetes, hypertension, chronic renal failure, cardiovascular disease, cerebrovascular disease, and malignancy).We also evaluated patients' existing symptoms such as fever, fatigue, respiratory symptoms, headache, and taste or olfactory disorders.

CT-rst triage protocol
First, patients complaining of fever, fatigue, respiratory symptoms, headache, and taste or olfactory disorder were redirected to the separate suspected COVID-19 outpatient clinic of the emergency room described above.From that time, patients were isolated from other patients and treated as if they had COVID-19.Second, a nurse in the outpatient clinic equipped with PPE brie y checked patients' vital signs and their present histories.Then, chest CT was conducted.Our hospital was equipped with three CT scanning suites.We set one of these as a dedicated suite for patients with suspected or con rmed COVID-19.Patients with suspected COVID-19 were separated from other patients in the transport corridors and elevators to the CT suite.An 80-row CT scanner (Aquilion Prime, Canon Medical Systems, Otawara, Tochigi, Japan) was used in this study.The tube voltage setting was 120 kVp and automatic exposure control was applied.The average CT dose index volume (CTDIvol) was 9.4 ± 4.1 mGy.These settings were not specialized for CT screening of COVID-19 pneumonia, and a reduction in radiation exposure was not applied.Patients were scanned during full inspiratory breath hold; however, some patients with severe dyspnea were scanned without performing a breath hold.Two-pattern CT images were reconstructed with a slice thickness/increment of 5.0mm/5.0mmand 0.5 mm/0.5 mm.Coronal and sagittal multiplanar reconstruction images were also acquired.All CT scans were immediately interpreted by trained radiologists.During regular business hours, all CT scans were interpreted and full reports were created by board-certi ed radiologists from the Japan Radiological Society (10 radiologists with 6-26 years' experience in chest radiology).Scans were also immediately checked, and necessary information on the possibility of COVID-19 pneumonia was reported to COVID-19 doctors after hours by the 10 boardcerti ed radiologists and 9 fellows from the Department of Radiology.If necessary, fellows contacted board-certi ed radiologists for their opinions.CT ndings were classi ed into ve categories according to the COVID-19 Reporting and Data System (CO-RADS). 21The CO-RADS was used to assess suspected pulmonary involvement in COVID-19 on a scale from 1 (very low) to 5 (very high), as follows: CO-RADS 1, normal CT or non-infectious etiology (e.g., lung tumor, lung brosis); CO-RADS 2, infectious etiology that is not compatible with COVID-19 (e.g., infectious bronchiolitis, bronchopneumonia, lobar pneumonia); CO-RADS 3, equivocal ndings for COVID-19 (e.g., homogeneous extensive GGO); CO-RADS 4, typical CT ndings of COVID-19 (similar to CO-RADS 5) but showing some overlap with other pneumonia; CO-RADS 5, typical CT ndings of COVID-19 (e.g., multifocal GGO with or without consolidation in lung regions close to visceral pleural surfaces).Representative images in each category are shown in Figure 1.Finally, patients were allocated to a "COVID-19 suspected" group or a "COVID-19 less likely" group.Patients with a suspicious clinical history (contact with an infected person and overseas travel history), suspicious ndings on chest CT scan (CO-RADS 3 to 5), or who were suspected of having COVID-19 according to their clinical symptoms that were di cult to explain by the underlying disease were included in the COVID-19 suspected group.Patients with neither a suspicious clinical history nor chest CT features suggestive of COVID-19 (CO-RADS 1 to 2) were included in the COVID-19 less likely group.If patients were diagnosed as less likely to have COVID-19, isolation and PPE support were discontinued, and detailed physical and laboratory examinations were performed.However, if patients were diagnosed with COVID-19 suspected, isolation and PPE support were continued until a negative PCR test result was obtained for the patient.A PCR test for SARS-CoV-2 was performed only in COVID-19 suspected group patients.All samples obtained after 11:00 on the previous day were immediately submitted for PCR testing each day, and the results were obtained at 15:00 on the same day.

Outcomes
The primary outcome was the e cacy of the CT-rst triage protocol in outpatients with suspected COVID-19.We assessed all results of PCR for SARS-CoV-2 and the nal diagnoses of patients.Patients who did not undergo PCR testing were followed for up to 14 days until con rmed clinically negative for COVID-19.
Secondary outcomes were the duration of isolation among all patients and safety of the medical staff and hospitalized patients.Isolation duration was de ned as the time that PPE was required or that patients were isolated from medical staff and other patients to prevent COVID-19 infection.For the COVID-19 suspected group, the time to a PCR-con rmed negative result was applicable for determining the duration of isolation.The isolation duration of PCR-positive patients was excluded from the results for isolation duration in the COVID-19 suspected group because COVID-19-positive patients remained in isolation until their treatment was nished.However, in the COVID-19 less likely group, the time to a CTcon rmed negative result was applicable to determine the isolation duration.The patient data were retrospectively examined using medical records.

Statistical analysis
To account for biases, we performed the analysis after the study period was complete.Results are presented as mean ± standard deviation for quantitative data and as frequency (percentage) for categorical data.The Student's t-test was used for continuous data.A P value <.05 was considered statistically signi cant.All statistical analyses were performed using Prism 7.9 J for Windows (GraphPad Software, Inc., San Diego, CA, USA).

Triage results of the CT-rst triage protocol
In conducting the CT-rst triage protocol, one patient refused chest CT because she was pregnant.The remaining outpatients with suspected COVID-19 in the clinic received chest CT and all images were immediately categorized according to the CO-RADS by specialized radiologists and Team COVID-19 doctors.Forty-eight (44.9%) patients were categorized as CO-RADS 1, 26 (24.3%)CO-RADS 2, 14 (13.1%)CO-RADS 3, 6 (5.6%) CO-RADS 4, and 13 (12.1%)patients were categorized as CO-RADS 5. Finally, 48 (44.9%) patients were included in the COVID-19 suspected group.On chest CT, two-thirds of patients in the suspected COVID-19 group showed features of CO-RADS categories 3 to 5; the remaining patients in the suspected COVID-19 group were classi ed according to their clinical history and other examinations conducted by Team COVID-19 doctors.All patients in the COVID-19 less likely group showed features of CO-RADS categories 1 or 2 and had no clinical history suspicious for COVID-19.The participant ow is shown in Figure 2.
Isolation was continued until patients were diagnosed as either unlikely to have COVID-19 using the CTrst triage protocol or had a con rmed negative PCR test result for SARS-CoV-2.The mean duration of isolation was signi cantly shorter in the COVID-19 less likely group (70.5 ± 41.2 minutes) than in the COVID-19 suspected group (1037.0± 512 minutes) (P < 0.001) (Table 2).

Final clinical diagnosis after CT-rst triage protocol
All patients in the COVID-19 suspected group underwent PCR testing for SARS-Co-V-2.Among 48 patients in this group, 9 (18.8%) were positive for SARS-CoV-2 on PCR and the remainder were negative.One patients with suspected COVID-19 had one negative PCR test result; however, Team COVID-19 doctors strongly suspected COVID-19 because the patient was categorized as CO-RADS 5 according to CT ndings; the patient was retested and de nitively diagnosed with COVID-19 after two positive PCR test results.The others eight of nine patients de nitely diagnosed with COVID-19 had CT features of CO-RADS category 5.However, during the study period, no patients were subsequently diagnosed with COVID-19 in the group less likely to have COVID-19.The nal diagnoses of outpatients with suspected COVID-19 who did not have COVID-19 were respiratory infection (51, 47.7%), other focal infection (22, 20.6%), tumorrelated fever (5, 4.7%), and interstitial pneumonia (4, 3.7%).Seven outpatients (6.5%) had life-threatening conditions that mimicked COVID-19: four with meningitis, one with acute myocardial infarction, and one patient who had pneumonia with venous thrombosis.Cases of life-threating conditions among outpatients with suspected COVID-19 are described in the Supplemental material.A total 31 patients (64.6%) in the COVID-19 suspected group and 20 patients (33.9%) in the COVID-19 less likely group were hospitalized.One patient with suspected COVID-19 died owing to severe COVID-19 causing multiple organ failure, and one patient in the COVID-19 less likely group died owing to acute myocardial infarction (Table 3).

Infection control during the study period
Via separation of outpatients with suspected COVID-19 in our specialized outpatient clinic and using the CT-rst triage protocol, adequate isolation was achieved for these outpatients.Throughout the study period, no medical staff developed fever or other health concerns.In addition, no in-hospital infections occurred, and we were able to use less PPE.

Discussion
In this study, we report the evaluation results for the CT-rst triage protocol in outpatients with suspected COVID-19 at our clinic.To our knowledge, these are the rst real-world clinical data regarding a CT-rst triage protocol worldwide.In our experience, most outpatients with suspected COVID-19 did not have COVID-19 (Table 1).From the perspective of conserving medical resources, a reasonable triage protocol is essential.We demonstrated the effectiveness of our CT-rst triage protocol in selected outpatients suspected of having COVID-19 based on their clinical history and chest CT features.In our experience, the CT-rst triage protocol is acceptable for clinical use.Our CT-rst triage protocol did not miss any patients who were ultimately con rmed to have COVID-19 and improved the pre-test probability of PCR testing for SARS-CoV-2 from 8.4% to 18.8% (Table 2).Furthermore, no medical staff or hospitalized patients developed new COVID-19-related symptoms during the study period.The consumption of PPE was also kept to a minimum.
To respond to the threat of infectious diseases, reliable rapid diagnostic technologies like that for in uenza are required.The use of new technologies including arti cial intelligence (AI) 22 , speci c antibody 23 , and immunochromatography 24 has recently been reported.For clinical use, these new technologies clearly require additional studies and concrete evidence.Nevertheless, increasing pre-test probability using AI-based medical interviews and AI analysis of CT scans are promising directions. 25Our hospital is equipped with three CT scanning suites and one is a dedicated suite for patients with suspected or con rmed COVID-19.Therefore, the CT-rst triage protocol is currently feasible for screening suspected cases of COVID-19.Whereas the CO-RADS categories are reliable, owing to the existence of false-negative CT categorization, 21 a protocol combining the clinical history of COVID-19 risk factors and CT features is recommended.Prokop et al. reported that 13.8% (95% con dence interval 9%-18%) of cases were false negatives in their cohort of 840 images. 21We using CO-RADS and patients' clinical histories for diagnosis; there were no patients diagnosed with COVID-19 in the group classi ed as less likely to have COVID-19 during the follow-up period.
One of the merits of our CT-rst triage protocol is the ability to quickly release patients from isolation who are unlikely to have COVID-19 (Table 2).In general, COVID-19 was diagnosed with PCR testing for SARS-CoV-2.However, PCR tests normally require at least several hours, and the PCR test for SARS-CoV-2 diagnosis takes longer because it is performed using several specimens at once.In that sense, the CT-rst protocol is an acceptable method to judge whether patients who are unlikely to have COVID-19 can be released from isolation.
However, in cases of suspected COVID-19, physicians generally tend to do less thorough examinations.This is because several studies have shown that certain medical procedures carry a high risk of viral exposure for medical staff. 26,27Moreover, patients with suspected COVID-19 must be isolated from other patients owing to the risk of infection via aerosols that can remain in the air for several hours. 28,29Therefore, it is easy to fail to reduce risk or narrow the differential diagnosis and delay the start of therapy.A CT scan seems unnecessary for diagnosis in many cases, and the time lost before the start of therapy could be reduced.However, from the perspectives of protecting medical staff, conserving resources, and returning to the normal diagnostic ow, use of the CT-rst triage protocol is acceptable.
We should emphasize that physicians are susceptible to cognitive bias (con rming bias) with respect to the clinical history and CT features of COVID-19. 30,31Some of our patients had life-threatening conditions that may mimic COVID-19.Because the symptoms and chest CT features of COVID-19 are non-speci c, the CT-rst triage protocol overestimates patients with COVID-19. 15,32Our results show that the CT-rst triage protocol is useful for exclusive diagnosis of COVID-19.As mentioned, several life-threatening conditions that mimic COVID-19 were present in our outpatients.Therefore, medical doctors must be aware that patients suspected of COVID-19 may have any one of a number of diseases.Careful assessment of vital signs and medical interviews is important, and wide differential diagnosis and checking for life-threatening diseases are crucial.
There are some limitations in our study.First, this was a retrospective review at a single center.The characteristics of outpatients with suspected COVID-19 may depend on the study region and hospital characteristics.Our hospital is a tertiary-level hospital in Yokohama City and the patients in our hospital had many comorbidities; therefore, patient bias may be present.Furthermore, a hospital's medical protocol depends its medical resources; future validation studies are needed to con rm our results.Second, the number of patients was limited; therefore, further studies among outpatients with suspected COVID-19 are needed to yield more concrete evidence and con rm our ndings.Third, in our study, we attempted to reduce exposure to radiation; however, the radiation dose did not reach the level of low-dose CT.Thus, radiation exposure is a potential limitation of this study.In accordance with the latest articles, we continue to revise our CT protocol to reduce the radiation dose to within a range suitable for triage.Finally, we did not perform SARS-CoV-2 PCR testing in the group of patients classi ed as less likely to have COVID-19.Thus, we might have missed COVID-19 cases in this group because many patients with COVID-19 have mild or no symptoms. 1,33However, no patients were diagnosed with COVID-19 in this group during follow-up, and there were no cases of secondary infection originating from COVID-19.Furthermore, one purpose of establishing a clinic for outpatients with suspected COVID-19 is to separate those suspected of having COVID-19 from other patients and to protect medical staff.In that sense, the CT-rst triage protocol was useful; no medical staff infections or in-hospital infections occurred during the study period.

Conclusion
We demonstrated the e cacy and safety of the CT-rst triage protocol for patients with suspected COVID-19 in a real-world setting.Our CT-rst triage protocol is acceptable for triaging patients with suspected COVID-19.Whereas the CT-rst triage protocol is adequate for triage of COVID-19, there is a risk of delayed diagnosis and the start of therapy for life-threating conditions that mimic COVID-19.Our ndings might improve the accuracy of diagnosis for outpatients with suspected COVID-19 in the present era of COVID-19 and can be helpful for appropriate triage, especially in settings where PCR testing is limited.
Abbreviations AI: cial intelligence

Figures
Figures

Figure 1 Representative
Figure 1