Clinical severity and mortality predictors in COVID-19 intensive care patients: CTSS and CO-RADS

Aim: Chest computed tomography (CT) plays an important role in the diagnosis of coronavirus infection disease 2019 (COVID-19) in patients with negative polymerase chain reaction (PCR) test but with clinical findings. The aim of this study was to determine whether the disease can predict clinical severity and/or mortality with CO-RADS and/or CTSS in intensive care COVID-19 patients. Material and Methods: In the study retrospectively, COVID-19 intensive care patients with PCR positive and chest CT between 23 March - 31 December 2020 were included. CTs were evaluated by two independent radiologists without providing the clinical information of the patients. CO-RADS and CTSS were calculated for each CT, and pathological features were recorded. Demographic, clinical characteristics and mortality rates of the patients were recorded. Patients were divided into three groups [mild (nasal/mask oxygen), severe (noninvasive mechanichal ventilator (NIMV) or high flow nasal oxygen (HFO)), critically severe (invasive mechanichal ventilation (IMV))] according to the clinical severity of COVID-19. Mortality and clinical severity markers were determined by logistic regression analysis. Results: Four hundred seventy three patients were included in the study. Patients were divided into three groups according to clinical severity, mild (34.7%), severe (11.8%), and critically severe (53.5%). The mean CTSS of all patients was 19.58 and the rate of patients in the CO-RADS 5 group was 50.7%. The mortality rate was 41.2%. APACHE II score and CTSS were predictors of clinical severity; age, female gender and CO-RADS were found as mortality predictors. The CO-RADS cut-off value predicting mortality was 5. Ground glass appearance was the most common pathological finding with a rate of 84.4%.Receiver operating characteristic (ROC) curves were drawn for mortality markers CO-RADS and APACHE II, and the area under the curve (AUC) values were 0.580 and 0.881, respectively. AUC was found to be 0.697 in the ROC curve drawn for CTSS, which is a clinical indicator of severity. The mortality cut-off value was found to be 16.5 with 77% sensitivity and 79% specificity for the APACHE II score (LR:3.7). The clinical severity cut-off value was found to be 18.5, with 61% sensitivity and 66% specificity for the CTSS. Conclusion: CO-RADS can be used to predict mortality and CTSS can be used to predict clinical severity in COVID-19, which are radiological-based scoring systems.


Introduction
Coronavirus disease-2019 (COVID-19), which is caused by Severe acute respiratory virus -2 (SARS-CoV-2), has been continuing all over the world since December 2019 [1].COVID-19, which occurs most frequently with respiratory symptoms, can create different clinical situations from flu-like symptoms to respiratory failure.
The diagnosis of COVID-19 is confirmed by the real time-polymerase chain reaction (rt-PCR) method in the nasopharyngeal swab sample.However, even with clinical compatibility, the PCR test may be negative.Although the diagnosis of COVID-19 is confirmed by PCR testing, COVID-19 pneumonia is detected by radiological imaging.The cause of hypoxemia requiring hospitalization in the intensive care unit (ICU) in COVID-19 is lung involvement.Chest X-ray evaluated in two dimensions may be insufficient to show lung pathology.On the other hand, chest CT is another imaging method in which the lung is evaluated in three dimensions and gives more detailed information.Therefore, chest CT may be one of the best indicators of the clinical severity, morbidity and mortality of COVID-19.Chest CT is scored for COVID-19 with the COVID-19 Reporting and Data System (CO-RADS) scoring system reported by The Radiological Society of Netherlands (NVvR) (Table 1) [2].In addition, CT severity score (CTSS), which is a semiquantitative scale, is another scoring method used to show disease severity.

Statistical Analysis
Statistical analyzes of the data obtained in the study were performed using the "SPSS for windows 26.0" Statistical Package Program.Continuous variables were expressed as mean±SD.The conformity of the numerical data to the normal distribution was evaluated with the Shapiro-Wilk test, and then the One-Way ANOVA test or t test was used to compare the numerical data with the normal distribution, and the result was evaluated according to the equality of variances.The Kruskal-Wallis or Mann-Whitney U test was used to compare the numerical data that did not fit the normal distribution.
Categorical data were given as numbers and percentages.
Pearson Chi-square test was used to compare categorical data.
Logistic regression analysis was performed to detect mortality predictors.AUC values were calculated by plotting ROC curves for APACHE II and CORADS scores, which are predictors of mortality.In addition, the cut-off values of the scores were found.P<0.05 was considered significant.

Results
A there was a difference between the groups (p=0.002).There was a difference between the three groups in terms of CTSS mean and CO-RADS stratification (p<0.001)(Table 2).The CO-RADS cut-off value predicting mortality was 5. When the patients were grouped according to the determined cut-off value (CO-RADS <5 and CO-RADS=5), CTSS and mortality rates were higher in the group with CO-RADS=5 (p<0.001 and p=0.015, respectively) (Table 3).In Chest CT, ground glass was the most common pathological finding with a rate of 84.4% in all patients.There was a statistical difference between the groups in terms of ground glass, air bronchogram, pleural effusion, crazy paving, pleural thickening, bronchial enlargement, and lymphadenopathy (LAP), and these findings were most common in the critically severe group (Table 4).
While APACHE II score and CTSS are clinical severity predictors; age, female gender and CO-RADS-category 5 were found as mortality predictors (Tables 5 and 6).ROC curves were drawn for mortality predictors CO-RADS and APACHE II, and the area under the curve (AUC) value was 0.580 and 0.881,respectively (Figures 1 and 2).AUC was found to be 0.697 in the ROC curve drawn for CTSS, which is a clinical severity predictor (Figure 3).The mortality cut-off value was found to be 16.5 with 77% sensitivity and 79% specificity for the APACHE II score (LR:3.7).
The clinical severity cut-off value was found to be 18.5 with 61% sensitivity and 66% specificity for CTSS (LR:1.8).
119   the need for intubation [8].The difference of this study; all patients were admitted to the ICU and the validity of CT scores was demonstrated in predicting the course of the clinic in the ICU and mortality.
In CO-RADS stratification, 7 different stratifications are made between 0-6.According to the CO-RADS stratification, RT-PCR is classified as positive for SARS-CoV-2 Category-6.In fact, all patients in this study had a positive PCR test and thus could be included in Category-6.However, since the PCR test results of the patients were not reported to the radiologists who evaluated the CTs, they were excluded from Category-6 stratification.Accordingly, 50.7% of the patients were in the CO-RADS Category-5 group, that is, they had typical COVID-19 findings.9.1% were in the Category-4, that is, Suspicious for COVID-19 group.14.6% were Category-3; features compatible with COVID-19 but also other disease, and 18.6% were in the Category-2; typical for other infection but not COVID-19 group.
Therefore, 93% of all patients had lung involvement.Hence, patients were often admitted to the ICU due to hypoxemia and/or respiratory failure.Although 33 patients in Category-1 without lung involvement do not have primary respiratory failure or need for oxygen; were admitted to the ICU for reasons related to other system involvements of COVID-19 such as myocarditis, myocardial infarction, cerebrovascular event.
CO-RADS Category-5 was higher in severe and critically severe group than mild group.This is a finding that supports the clinical findings of patients with radiologically typical CT for COVID-19 may be more severe.Although CO-RADS Category-5 was the highest in the severe group, no mortality was observed in this group.The reasons for this are; it can be explained by the younger mean age of the patients in this group, the lower mean APACHE II score, which is a predictor of mortality, and no IMV requirement.In patients whose symptom onset times could not be determined, the time between symptom onset and CT performed may also be shorter in the mild group.That's because the sum of CO-RADS Category 1, 2, and 3 (53.7%)was highest in the mild group.Previous studies have also reported that chest imaging might be negative in the earlier phase of COVID-19 due to it has not involved the lung paranchyma yet [7].
In this study, we found the mortality cut-off value for CO-RADS to be CO-RADS=5.patients.The fact that the mean CTSS and mortality were significantly higher in the group with CO-RADS=5 indicates that the lung involvement rate of COVID-19 is also higher in this group.The mean of CTSS, another scoring system, was the lowest in the mild group.In the severe and critically severe groups, the CTTS mean was similar.Although all patients were positive for PCR, with CTSS, which is a semi-quantitative method that shows lung involvement rates, the mean CTSS scores were higher in these two groups, where the disease was more severe, hypoxemia and oxygen support systems were needed more, such as HFO, NIMV or IMV.In addition, CTSS, which is a clinical severity predictor in our study, was reported to have the strongest positive correlation with the clinical status of patients in the study of Mruk et al. [9].
In the study of Lieveld et al in which they used CO-RADS and CTSS, the mean CTSS of patients admitted to the ICU was determined as 14.8 [8].In this study, the mean CTSS of all patients (19.58) and clinical severity cut-off value (18.5) were found higher than the study of Lieveld et al [6].The reasons for this are; it may be that all the patients in our study were PCR positive and had a higher sample size (88 vs 473 ICU patients).At the beginning of the pandemic, Yang et al. study reported similar results to our study.In the reported study, the optimal CTSS threshold for identifying severe COVID-19 was 19.5 [10].In the study of Bellos et al, the CTSS of patients with ICU admission was 12.6 [11].The small number of patients in this previous study may be the reason for having a lower CTSS mean than our study.Abbasi et al compared survived and deceased COVID-19 patients and reported a CTSS of 14.5 in the deceased group [12].Further, this value was lower than the mean CTSS of all patients in our study.The differences between the two studies; Abbasi et al. may have carried out their studies with a small number of patients at the beginning of the pandemic (February-March 2020) and the population in their study included patients admitted not only to the ICU but also to the hospital [12].In the study, which included only COVId-19 patients in the intensive care unit, CTSS was reported as >15 as a mortality predictor [13].
It has been reported in the literature that peripheral ground glass and consolidation are the most common CT findings of COVID-19 [14].In this study, the most common lesions were ground glass and consolidation areas, and the results were similar.
The difference of this study from the others is that only the PCR positive COVID-19 patients followed in the ICU were included and their radiological findings were compared.
Cause studies comparing the radiological features of PCR positive and negative patients for CO-RADS have been reported in the literature [15].For CTSS, it has been reported that there is a significant relationship between emergency to hospital admission, ICU admission and 30-day mortality [6].In this study, both CO-RADS and CTSS were scored, and which parameter predicted mortality and clinical severity of the disease was evaluated separately.
In the study of Zayed et al, it was reported that both CO-RADS and CTSS can predict severe COVID-19 [8].In our study, CO-RADS, which predicts the mortality of COVID-19 patients in intensive care, and CTSS, which predicts clinical severity, were found to be two different radiological-based scoring systems.
As a mortality predictor, the AUC in the ROC curve plotted for APACHE II was greater than the AUC calculated for CO-RADS.
While the APACHE II score, which has been used as a mortality predictor for a long time, is calculated with clinical and laboratory parameters, CO-RADS includes only radiological findings.
Therefore, the AUC determined for CO-RADS may not be as high as the AUC of APACHE II.It was found as the clinical severity predictor of CTSS, which calculates how much of each segment of the lung is affected together with the radiological findings.
Although COVID-19 causes disease in all age groups and in both genders, previous studies in the literature have reported that male gender and elderly individuals are more affected by the disease or have the disease more severely [16,17].In this study, similar to the literature data; the ratio of male patients was predominant, and the mean age of the critically severe group was the highest.
Limitations of the study; a single-center, retrospective study, with one CT evaluated only by a radiologist.Therefore, the compatibility between radiologists could not be evaluated.In addition, the time between the onset of symptoms and the time the CT was performed could not be determined.

Conclusion
Thoracic CT has an important role in predicting clinical severity and mortality, as well as confirming the diagnosis in COVID-19 intensive care patients.From two different radiological scorings; CO-RADS can be used to predict mortality and CTSS to predict clinical severity.

(
NIMV), high flow nasal cannula oxygen (HFO), need for invasive mechanical ventilation (IMV), IMV duration length of stay in ICU and mortality rate were recorded.Intensive care patients were stratificated three groups according to the clinical severity of COVID-19; mild (nasal/mask oxygen), severe (NIMV or HFO), critically severe (IMV).Demographic, clinical and radiological characteristics of the groups were compared.Mortality and clinical severity predictors were determined by logistic regression analysis.Further, cut-off values that determine mortality and clinical severity were determined.Receiver operating chareacteristic (ROC) curve was drawn for to predict predictore of clinical severity and mortality and area under the curve (AUC) was calculated Radiological technique Chest CTs were obtained with 2 devices with multidetector-128 slices specially reserved for patients with suspected SARS-CoV-2 (GE Revolution EVO 128 Slice CT Scanner, GE Medical Systems, Milwaukee, WI, USA).During inspiration, shots were taken in the supine position without the use of intravenous contrast material.As CT acquisition parameters, section thickness was chosen as 1.3 mm, pitch factor 0.98, tube voltage average 100 kV, mA 90-300, collimation width 0.625.

Table 1 :
Overview of CO-RADS Categories and the corresponding level of suspicion pulmonary involvement in COVID-19 CO-RADS; COVID-19 Reporting and Data System, RT-PCR; real time-polymerase cahin

reaction Material and Methods
Demographic characteristics of patients (age, gender), APACHE II (Acute Physiological and, Chronic Health Evaluation) score, nasal/mask oxygen, noninvasive mechanical ventilation the critically severe group (23.92) and there were differences between the groups (p<0.001).The length of stay in the ICU was the shortest in the mild group (8.43 days), and 41.2%.Patients were divided into three groups according to clinical severity, mild (34.7%), severe (11.8%), and critically severe (53.5%).There was a significant difference between the mean ages of the groups, and the mean age (73.21) was the highest in the critically severe group (p<0.001).The genders were similar in all three groups.The APACHE II score was highest in

Table 2 :
Comparison of clinical and demographic characteristics by groups APACHE II; Acute Physiological and, Chronic Health Evaluation, IMV; invasive mechanichal ventilation, ICU; intensive care unit, CTSS; CT severity score, CO-RADS; COVID-19 Reporting and Data System,

Table 4 :
Chest APACHE II; Acute Physiological and, Chronic Health Evaluation, CTSS; CT severity score APACHE II; Acute Physiological and, Chronic Health Evaluation, CTSS; CT severity score, ICU; intensive care unit, CO-RADS; COVID-19 Reporting and Data System Figure 1.ROC curve for CTSS as a clinical severity predictor Figure 2. ROC curve for APACHE II score as a mortality predictor to decide on regular ward admission versus ICU admission [5].In line with this information, chest CTs of COVID-19 patients with positive PCR test who required ICU admission were evaluated in this study.CTs taken within 2 days before and after admission to the ICU were scored by two radiologists to standardize the CT performed time.In the literature, studies have been reported on how effective CTs performed in the emergency room are in predicting hospitalization in the service or ICU or in predicting The study of Zayed et al. reported that if