Comparison of Pneumonia Severity Indices, qCSI, 4C-Mortality Score and qSOFA in Predicting Mortality in Hospitalized Patients with COVID-19 Pneumonia

This is a retrospective and observational study on 1511 patients with SARS-CoV-2, who were diagnosed with COVID-19 by real-time PCR testing and hospitalized due to COVID-19 pneumonia. 1511 patients, 879 male (58.17%) and 632 female (41.83%) with a mean age of 60.1 ± 14.7 were included in the study. Survivors and non-survivors groups were statistically compared with respect to survival, discharge, ICU admission and in-hospital death. Although gender was not statistically significant different between two groups, 80 (60.15%) of the patients who died were male. Mean age was 72.8 ± 11.8 in non-survivors vs. 59.9 ± 14.7 in survivors (p < 0.001). Overall in-hospital mortality was found to be 8.8% (133/1511 cases), and overall ICU admission was 10.85% (164/1511 cases). The PSI/PORT score of the non-survivors group was higher than that of the survivors group (144.38 ± 28.64 versus 67.17 ± 25.63, p < 0.001). The PSI/PORT yielding the highest performance was the best predictor for in-hospital mortality, since it incorporates the factors as advanced age and comorbidity (AUROC 0.971; % 95 CI 0.961–0.981). The use of A-DROP may also be preferred as an easier alternative to PSI/PORT, which is a time-consuming evaluation although it is more comprehensive.


Introduction
Severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2), known as the "COVID-19 pandemic", is globally associated with high mortality. SARS-CoV-2 has infected over 364,191,457 million people in the world, causing over 5,631,457 million casualties [1]. COVID-19 usually exhibits mild or moderate (81%) clinical appearance, even though 14% of cases are severe and 5% are critical. According to the Chinese Center for Disease Control and Prevention report, including 72,314 COVID-19 cases, the overall mortality rate was reported as 2.3%. This rate was 14.8% for patients older than 80 years, and 49% in critical cases [2].
COVID-19 is a multisystemic viral disease which may show varying clinical presentations such as asymptomatic disorders, injuries associated with organs as liver, heart, kidney, neurological manifestatations, coagulopathy, sepsis, septic shock and multiple organ dysfunction [3][4][5]. Among the above mentioned problems, especially acute respiratory failure, enrolled in the study were over the age of 18, and none of them had been taken to the intensive care unit (ICU). They were managed in accordance with the COVID-19 treatment protocol of the Turkish Health Ministry [47]. The research was approved by the ethics committee of the University of Health Sciences, Bakırköy Dr. Sadi Konuk Training and Research Hospital (approval number 2021/91), and was conducted following the principles of the Declaration of Helsinki.
Retrospectively, medical records were obtained from the hospital electronic database. All the parameters relevant for scoring the disease severity were measured and recorded. A standardized form was used for data collection, which included demographics, past medical history, underlying chronic diseases, vital signs, the severity of admission, laboratory findings and chest computed tomography (CT) scans results. In addition, definitive outcomes (death, discharge or ICU admission) were obtained from the hospital information system.
Patients with COVID-19 were categorized into two groups (non-severe and severe illness) according to National Institutes of Health (NIH) classification based on disease severity [48]. Severe cases were defined as having at least any one of the following criteria: respiratory distress (>30 breaths/min); oxygen saturation < 94% at rest; arterial partial pressure of oxygen (PaO 2 )/fraction of inspired oxygen (FiO 2 < 300 mmHg), or lung infiltrates > 50%. Radiologist-evaluated chest CT scans were classifed into three categories, mild, moderate and severe involvement [49]. Patients with no radiological involvement were excluded in our study. Only moderate and severe cases were included in our study. Patients who had incomplete data or those who had pneumonia arising from other pathogens were all excluded from the study. Another criterion for exclusion was pregnancy. After the exclusion, 1511 patients (older than 18 years) who were hospitalized with the diagnosis of COVID-19 pneumonia were included in the study. The patient flowchart can be seen in Figure 1. hospitalized at Prof. Dr. Murat Dilmener Emergency Hospital, pandemic 3rd level, in Istanbul, from 1 September 2020 to 31 December 2020. The diagnosis of COVID-19 was determined on the basis of World Health Organization (WHO) guidelines [46]. All cases enrolled in the study were over the age of 18, and none of them had been taken to the intensive care unit (ICU). They were managed in accordance with the COVID-19 treatment protocol of the Turkish Health Ministry [47]. The research was approved by the ethics committee of the University of Health Sciences, Bakırköy Dr. Sadi Konuk Training and Research Hospital (approval number 2021/91), and was conducted following the principles of the Declaration of Helsinki. Retrospectively, medical records were obtained from the hospital electronic database. All the parameters relevant for scoring the disease severity were measured and recorded. A standardized form was used for data collection, which included demographics, past medical history, underlying chronic diseases, vital signs, the severity of admission, laboratory findings and chest computed tomography (CT) scans results. In addition, definitive outcomes (death, discharge or ICU admission) were obtained from the hospital information system.
Patients with COVID-19 were categorized into two groups (non-severe and severe illness) according to National Institutes of Health (NIH) classification based on disease severity [48]. Severe cases were defined as having at least any one of the following criteria: respiratory distress (>30 breaths/min); oxygen saturation <94% at rest; arterial partial pressure of oxygen (PaO2)/ fraction of inspired oxygen (FiO2 < 300 mmHg), or lung infiltrates >50%. Radiologist-evaluated chest CT scans were classifed into three categories, mild, moderate and severe involvement [49]. Patients with no radiological involvement were excluded in our study. Only moderate and severe cases were included in our study. Patients who had incomplete data or those who had pneumonia arising from other pathogens were all excluded from the study. Another criterion for exclusion was pregnancy. After the exclusion, 1511 patients (older than 18 years) who were hospitalized with the diagnosis of COVID-19 pneumonia were included in the study. The patient flowchart can be seen in Figure 1.

Statistical Analysis
All statistical analyses were performed in commercially available SPSS software v.22 (Statistical Package for the Social Sciences Inc., Chicago, IL, USA). Patient characteristics were summarized using descriptive statistics (mean, standard deviation, median, frequency, percentages, minimum, maximum, Q1-Q3). The conformity of the quantitative variables to the normal distribution was tested with the Shapiro-Wilk test and graphical examinations. Student t-test was used for the comparison of normally distributed quantitative variables between two groups, and the Mann-Whitney U test was used for comparisons between two groups of non-normally distributed quantitative variables. Categorical variables were compared using the Chi-square test and Fisher's exact test. Diagnostic screening tests (sensitivity, specificity, positive predictive value and negative predictive value) and ROC (area under the curve of the receiver operating characteristic) analysis were used to determine the cutoff value for all nine scoring systems. ROC Curve analysis and Binomial exact test were used to determine in-hospital mortality prediction. Multivariate logistic regression analysis was performed to identify other risk factors on mortality. A p-value < 0.05 was accepted as statistically significant.

Comparison of Basic Clinical Characteristics between the Two Groups
1511 patients, 879 male (58.17%) and 632 female (41.83%) with a mean age of 60.1 ± 14.7, were included in the study. Although gender difference was not statistically significant, 80 (60.15%) of the patients who died were male. Mean age was 72.8 ± 11.8 in non-survivors vs. 59.9 ± 14.7 in survivors (p < 0.001). 386 patients (25.54%) had only a single comorbidity, while 693 (45.86%) had two or more. The most common comorbidity was hypertension (48.04%, 726/1511), followed by diabetes mellitus (33.28%, 503/1511) and coronary artery disease (14.36%, 217/1511). The number of comorbidities was higher in patients who died (78.19% vs. 42.74%, p < 0.001). Hypertension, coronary artery disease, atrial fibrillation (AF), congestive heart failure and cerebrovascular disease were significantly more common in non-survivors than in survivors (p < 0.001). We observed a difference between cohorts in terms of malignancy and chronic kidney diasease (p = 0.01), while no such difference existed for diabetes, dyslipidemia, COPD and asthma. Ultimately, the non-survivor group was older and had more comorbidities than the survivor group (p < 0.001 for both). Respiratory rate, oxygen saturation by pulse oximetry (SpO 2 ) under oxygen support, supplemental oxygen requirement and heart rate were significantly higher in non-survivors.
General information and baseline characteristics of the patients are shown in Table 1. Laboratory findings, CT scores, disease severity status and outcomes of the patients are given in Table 2.

Comparison of Laboratory Tests and Chest CT Scans between the Two Groups
Patients had decreased lymphocyte, platelet, albumine and elevated neutrophil count, N/L ratio, C-reactive protein (CRP), urea, creatinine, LDH, ferritin, D-dimer, INR, troponin I levels in non-survivors (p < 0.001). In addition to that, glucose, ALT, AST and fibrinogen levels were higher in non-survivors (p values, respectively 0.003, 0.01, 0.04 and 0.01). There were no significant differences in procalcitonin and hematocrit levels between two groups. Concurrently, deceased patients had significantly higher disease severity status and chest CT scores than the survivors (p < 0.001). Additionally, 97.75% (130/133) of deceased patients had severe disease status.

Score Distribution
The PSI/PORT score of the non-survivor group was higher than that of the survivor group (144.38 ± 28.64 versus 67.17 ± 25.63, p < 0.001). Similarly, all the other eight scores were also found to be higher in non-survivors than survivors (p < 0.001). As seen in Table 3, prognostic scores were higher in deceased patients.
When the mortality rates were investigated in terms of pneumonia scores, the mortality rate in the high-risk group was found to be 6.88% (104 cases) for those having a NEWS2 score of 7 or higher. The mortality rate reduced to 6.22% (94 cases) for patients having Expanded CURB-65 scores of 4 or higher, and to 5.69% (86 cases) for those classified as PSI/PORT Class V. When the same analysis was performed in low-risk groups, there was no mortality among all 872 patients categorized as Class II with regard to PSI/PORT, and for patients having 4C Mortality score between 0 and 3, only a single mortality was observed.

Outcomes in Two Groups
In our study, in-hospital mortality was found to be 8.8% (133/1511 cases), and overall ICU admission was 10.85% (164/1511 cases). While the total number of patients who died in ward and ICU was found to be 133, the remainder (1378) were discharged. 114 patients died in ICU, and 19 died while they were being treated in the ward. It was found that the mean hospitalization stay of the deceased patients was longer (11.27 vs 14.52 days, p < 0.001), and the frequency of admission to the ICU was higher (85.71% vs 3.62%, p < 0.001).
Backward stepwise logistic regression test was performed on 32 parameters predicting mortality including age, number of comorbidities, presence of certain comorbidities (hypertension, coronary artery disease, AF, congestive heart failure, cerebrovascular disease, chronic kidney disease, malignancy), SpO2, O2 support, heart rate, neutrophil-to-

Outcomes in Two Groups
In our study, in-hospital mortality was found to be 8.8% (133/1511 cases), and overall ICU admission was 10.85% (164/1511 cases). While the total number of patients who died in ward and ICU was found to be 133, the remainder (1378) were discharged. 114 patients died in ICU, and 19 died while they were being treated in the ward. It was found that the mean hospitalization stay of the deceased patients was longer (11.27 vs. 14.52 days, p < 0.001), and the frequency of admission to the ICU was higher (85.71% vs. 3.62%, p < 0.001).

Discussion
We aimed to analyze the utility of the well-known CAP severity indices; CURB-65, Expanded CURB-65, PSI-PORT, NEWS2, MEWS and A-DROP, as well as qCSI 4-C Mortality and qSOFA scores introduced for COVID-19 in predicting mortality and progression to severe disease.
In our study, in-hospital mortality was found to be as 8.8% (133 cases). This result seems to be lower than those in other studies though it stands as a higher value when compared with 0.77% (cases: 11,249,216, deaths: 86,661), which comes from estimations based on national data [1,8,50].
Although no relationship was found between diabetes and mortality, hyperglycemia was found to be more frequent in those patients who died. Similarly, the disease severity and chest CT scores were found to be related with mortality [8, 52,53,[57][58][59].
The higher score at admission have higher risk of ICU care and death in patients with COVID-19 pneumonia. High prognostic scores indicate worse prognosis. All prognostic scores were higher in deceased patients. The mortality in low-risk groups that were designated to manage outside the hospital was 0 in PSI/PORT Class II, 1.73% (21/1208) in A-DROP low risk and 2.33% (26/1114) in CURB-65 low-risk.
When the high-risk groups were investigated the mortality rate was observed to be as 17.35% (42/242) in PSI/PORT Class IV and 74.78% (86/115) in PSI/PORT Class V. 30-day mortality in CAP patients in PSI Class and Mortality in the Pneumonia PORT Validation Cohort group were reported to be 0.6%, 9.3% and 27% for class II, IV and V patients, respectively [30].
Previous studies have already shown that PSI/PORT score was demonstrated to have a strong predictive performance in CAP cases. We have obtained even a higher performance on our COVID-19 pneumonia cohort than those found on the CAP, in which PSI/PORT presented an AUROC of 0.78 (95% CI 0.73-0.82) to predict composite primary endpoint (death within 28 days by any cause, or transfer to ICU) [61] and 0.812 (95% CI 0.673-0.951) to predict 30-day mortality [62].
PSI/PORT score consists of certain parameters such as age, male gender, comorbidities, metabolic abnormalities, tachypneia and hypoxemia which are shown to have a relation with mortality in COVID-19 patients [3,4,8,9,18,19,51,63,64]. The higher performance of PSI/PORT score may be attributed to the fact that the mean age of non-survivors was 72.8 ± 11.8, or that 78.19% (104/133) of them had two or more comorbidities and lower SpO 2 levels compared with survivors. PSI/PORT score may be the best predictor for mortality, since it gives extra credit for such factors as advanced age, clustered comorbidities and hypoxemia. However, PSI/PORT may have some disadvantages as not to include chronic lung diseases as COPD and asthma; only evaluating heart failure in cardiologic parameters; limiting the respiratory rate above 30 per minute; underestimating severe pneumonia in young healthy patients due to absolute age parameters; automatic classification of individuals without any comorbidities over age 50 as Class II; and the fact that clinicians may find considering 19 parameters for score quantification more time-consuming.
Previous studies have shown that the A-DROP scoring system was accurate and clinically useful for assessing the severity of both bacterial and atypical pneumonia [65,66]. Recently, Miyashita et al. reported that a high A-DROP score, indicating severe or extremely severe pneumonia, was associated with a high mechanical ventilation rate or high death rate [67].
A-DROP score is a score which is calculated by adding such parameters as advanced age (male ≥ 70 years, female ≥ 75 years versus > 65 years in CURB-65), and respiratory failure (arterial oxygen saturation ≤ 90% or arterial oxygen pressure ≤ 60 mmHg) parameters to CURB-65. A-DROP score had the highest second rank in predicting in-hospital mortality with 0.929 AUROC (95% CI 0.911-0.948). In contradiction, there are also studies reporting that A-DROP score performs better than the PSI/PORT in predicting the in-hospital mortality of COVID-19 patients AUROCs for A-DROP 0.87 (95% CI 0.84-0.90) vs. PSI 0.85 (95% CI 0.81-0.88) [16] and AUROCs for A-DROP 0.875 (95% CI; 0.822-0.937) vs. PSI 0.873 (95% CI 0.820-0.925) [68]. These findings suggest that involving advanced age [4,8,18], which is related to higher COVID-19 mortality, places A-DROP score in a superior situation compared to CURB-65 in predicting mortality. In our study, the mean age of deceased patients was 72 years, and they had lower SpO 2 levels. Zhou et al. also reported that the mean age of non-survivors with COVID-19 was 69 years [8,18] and Xie et al. reported hypoxemia was independently associated with in-hospital mortality.
NEWS2 score, having the respiratory parameters (respiratory rate, oxygen saturation, supplemental oxygen), performed as the third-highest system in predicting mortality. Its lack of a scale to indicate increased oxygen requirement, its insensitivity to hypoxic respiratory failure (type 1) often encountered in COVID-19 cases and its seldom monitoring of hypercapnic respiratory failure (type 2), neglecting the age, comorbidities and organ dysfunctioning, come forward as the disadvantages of NEWS2 [7,11,32,50,69,70].
4C Mortality which was specifically developed for COVID-19 cases, came out to be the fifth rank in our score performance results, although it included parameters as age, inflammatory markers and respiratory measures yet underestimating comorbidities dealing them only by number. 4C Mortality score with AUROC 0.78 (95% CI 0.75-0.81) was reported in the second place after the PSI/PORT score having 0.79 (95% CI 0.77-0.82) AUROC in performance in a recent study [72]. 4C Mortality was found to have a good prognostic value for mortality on patients over the age of 60 (AUROC 0.799; 95% CI 0.738-0.851) and qCSI had a similar effectiveness as reported in (AUROC 0.749; 95% CI 0.685-0.806) [71].
qSOFA Fan et al. has also compared in-hospital mortality of different risk scores systems like our study [16]. Although both studies took A-DROP, CURB-65 and 4C Mortality scores by the same cut off values; we have used a different cut off value as 107 points for PSI/PORT score, which is a mid-value between Class III and IV patients. Fan et al. has found the greatest score as A-DROP; our results suggest that a cut-off value of 107 points with PSI/PORT score yield a more valid score for predicting in-hospital mortality.
In our study, in-hospital mortality was found to be as 8.8% (133 cases). As it was reported to WHO, that there have been 11,249,216 confirmed COVID-19 cases, of which 86.661 were fatal (0.77%) in Turkey during the time interval from January 2020 to January 2022 [1,8,50]. It is suggested that the relative highness of our mortality value, in comparison with the national statistics can be attributed to our cohorts, consisting of moderate and severe cases with an average age of 60.12 ± 14.73. An additional factor might be that the study was conducted in a hospital dedicated entirely to COVID-19 patients. Varying mortality rates were reported in other studies ranging from 2.3% to 36%. The characteristics of the patients included in these studies, different treatment regimens applied and different mortality criteria (e.g., in-hospital, 48 h, 72 h, 7-day or 30-day measures) used might have been the major reasons why they yielded such a wide range of different mortality rates [2,69,71,76].

Limitation
This study has some limitations. It is a single-center study involving only hospitalized patients with moderate and severe disease in a retrospective design. Omicron and delta variants were not included, since the study was conducted during the early stages of the COVID-19 pandemic.

Conclusions
In conclusion, the accuracies of a variety of severity scores to predict in-hospital mortality in COVID-19 pneumonia patients were examined in our study. We found that PSI/PORT score showed the highest accuracy of in-hospital death prediction compared to other widely used CAP-specific and COVID specific score systems, such as qCSI and 4C Mortality. The PSI/PORT with a cut-off value of 107 points yielding the greatest performance was the best predictor for mortality, since it incorporated the factors such as advanced age and comorbidities. On the other hand, determination of the PSI/PORT required a longer time, since it involved the measurement and evaluation of a wide range of clinical parameters. Although all these data have to be validated in newer patient groups involving Cmicron and delta variants; our findings suggest that the use of A-DROP may also be preferred as a practical alternative to PSI/PORT, which is more time-consuming. Informed Consent Statement: Written informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy reasons.