key: cord-0956633-gnzwe2z2
authors: Satici, Celal; Demirkol, Mustafa Asim; Altunok, Elif Sargin; Gursoy, Bengul; Alkan, Mustafa; Kamat, Sadettin; Demirok, Berna; Surmeli, Cemile Dilsah; Calik, Mustafa; Cavus, Zuhal; Esatoglu, Sinem Nihal
title: Performance of Pneumonia Severity Index and CURB-65 in Predicting 30-day mortality in patients with COVID-19
date: 2020-06-14
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2020.06.038
sha: bad51ad1f212aa903e8fb9a208331599d6b53388
doc_id: 956633
cord_uid: gnzwe2z2

Abstract Objective The aim of the study was to analyze the usefulness of the CURB-65 and pneumonia severity index (PSI) in predicting 30-day mortality in patients with COVID-19 and to identify other factors associated with higher mortality. Methods A retrospective study was performed at a pandemic hospital in Istanbul, Turkey and 681 laboratory-confirmed patients with COVID-19 were included. Data on characteristics, vital signs and laboratory parameters were recorded form electronic medical records. We used receiver operating characteristic analysis to quantify the discriminatory abilities of the prognostic scales. Univariate and multivariate logistic regression analyses were performed to identify other predictors of mortality. Results Higher CRP levels were associated with an increased risk for mortality (OR:1.015, 95% CI 1.008 to 1.021, p < 0.001). The PSI performed significantly better than the CURB-65 (AUC: 0.91, 95% CI 0.88-0.93 vs AUC:0.88, 95% CI:0.85-0.90; p = 0.01) and the addition of CRP levels to PSI did not improve the performance of PSI in predicting mortality (AUC: 0.91, 95% CI 0.88-0.93 vs AUC:0.92, 95% CI:0.89-0.94; p = 0.29). Conclusion In a large group of hospitalized patients with COVID-19, we found that PSI performed better than CURB-65 in predicting mortality. Adding CRP levels to PSI did not improve the 30-day mortality prediction.

The novel coronavirus disease 2019 , caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), has become a major health concern worldwide. According to the World Health Organization, as of May 31, 2020, there were 5.934.936 confirmed cases and 367.166 deaths (WHO, 2020) . Respiratory failure is the leading cause of mortality in patients with COVID-19 (Ruan et al. 2020) . Myocardial injury, kidney or liver injury, and multi-organ dysfunction are among the other complications leading to death . Several prognostic factors such as older age, male gender, presence of comorbidities and smoking have been found to be associated with severe disease or death Zheng et al. 2020 ). In addition, deceased patients were more likely to have leukocytosis, lymphopenia and higher levels of lactate dehydrogenase, C-reactive protein (CRP) , elevated neutrophil-tolymphocyte ratio , interleukin (IL)-6 (Aziz et al. 2020), troponin and D-Dimer ).

Turkey has a comprehensive public healthcare system and all residents have received medical treatment free of charge in public and private hospitals during the COVID-19 outbreak. According to the Health Ministry Guideline, any suspected case who is over 50 years old or has any comorbidity should be hospitalized irrespective of vital signs, laboratory results and computed tomography (CT) findings (Bilim Kurulu, 2020) Thus, a large proportion of patients with COVID-19 meet criteria for admission as an inpatient. That might lead to over-hospitalization, resulting in many problems such as psychological disturbances, lack of sleep and accidental falls (Zuk et al. 2003 , Hitcho et al. 2004 The CURB-65 and Pneumonia severity index (PSI) are widely used in predicting 30-day mortality in community acquired pneumonia (Shah et al. 2010) . The CURB-65 has been also found to be J o u r n a l P r e -p r o o f useful to predict 14-day mortality in hospital-acquired pneumonia (Oktariani et al. 2019 ). However, they have not been studied in patients with COVID-19. A simple predictive tool would be useful to estimate the risk of 30-day mortality and to stratify patients with COVID-19 to high or low risk for poor outcome at the time of hospital admission.

In this study, we aimed to assess whether CURB-65 or PSI is useful tool to predict 30-day mortality and to identify other factors that are associated with higher mortality in patients with COVID-19.

We performed a retrospective cohort study at Gaziosmanpasa Research and Training Hospital, University of Health Sciences, Istanbul, Turkey. Our hospital has been working as a pandemic hospital since the outbreak began.

Our study was conducted in line with the Declaration of Helsinki. The local institutional ethics committee approved the study protocol (Ethics approval number: 59/05.2020) and the requirement for written informed consent was waived by our ethics committee.

The first case was reported on March 11, 2020 in Turkey. Management strategies have been revised and updated during the outbreak. As favipiravir treatment has become a suggested therapeutic option for COVID-19 patients with severe pneumonia on April 2, 2020, we retrospectively enrolled the patients who have been diagnosed with COVID-19 pneumonia at our center between April 2, 2020 and May 1, 2020. All patients over 18 years old with COVID-19 confirmed by PCR on nasopharyngeal swab who were hospitalized in our hospital were included in the study. Pregnant patients were excluded.

Based on the Health Ministry guideline, any suspected case who was over 50 years old, or had any comorbidity including cardiopulmonary disease, diabetes mellitus, hypertension, chronic renal disease, immunosuppressive conditions or malignancy, or with tachycardia (pulse >125/min), tachypnea (respiratory rate >22/min), hypotension (<90/60 mm Hg) or hypoxemia (Spo2 <93%) have been hospitalized. (Bilim Kurulu, 2020) Severe cases were defined as those with any of the followings; (1) respiratory distress (>30 breaths/min), (2) oxygen saturation lower than <90% at rest, or (3) arterial partial pressure of oxygen/fraction of inspired oxygen ≦300mmHg (Bilim Kurulu, 2020).

Demographic characteristics, comorbidities, presenting symptoms, triage vitals including fever, blood pressure, respiratory rate, oxygen saturation at rest, heart rate, and initial laboratory parameters and time to death were collected from electronic medical records.

Our primary outcome was 30-day mortality defined as documented death from any cause during hospitalization or within 30 days of admission to our emergency department. The CURB-65 and PSI scores at hospital admission were calculated as shown in Table 1 and Table 2 . The CURB-65 scores range from 0 to 4. Having 0-1 scores indicate a low risk for mortality whereas 2 or higher scores are associated with higher mortality ( Table 1 ). The PSI scores are classified as group I, II, III, IV and V. Patients are stratified into two levels of risk groups: PSI low risk (group I-III) and high risk (group IV-V) ( Table 2) .

Protocol issued by the Turkish Ministry of Health (Bilim Kurulu, 2020). The recommended J o u r n a l P r e -p r o o f hydroxychloroquine regimen for all hospitalized patients was a loading dose of 400 mg twice on day 1, followed by 400 mg daily for 4 additional days. In addition, azithromycin at a dose of 500 mg on day 1 and then 250 mg daily for 4 more was also used cautiously with QT interval monitoring. Favipiravir was initiated in patients with severe pneumonia or in those with ongoing fever despite hydroxychloroquine and/or azithromycin treatment at a loading dose of 1600 mg twice on day 1, followed by 600 mg twice a day for 4 additional days. Tocilizumab was used at a dose of 8mg/kg in patients with elevated inflammatory markers and ongoing hypoxemia despite favipiravir treatment. In case of inadequate clinical response, a second dose of tocilizumab was considered within 24-48 hours after the initial dose. Prophylactic dose of enoxaparin has been initiated in all patients unless there was a contraindication. Therapeutic dose of enoxaparin was used in the following conditions; severe pneumonia, D-dimer level  1000 ng/mL, body mass index  40 kg/m 2 , and acute venous thromboembolism.

We used descriptive statistics to define variables. Categorical data were reported as proportions and counts and continuous data were presented as median and interquartile range (IQR) unless the data were normally distributed. The sensitivity, specificity, and positive predictive value (PPV) and negative predictive value (NPV) of CURB-65  2 and PSI  4 were calculated by the standard two-by-two tables. Univariate and multivariate logistic regression analyses were performed to identify other independent predictors of 30-day mortality. The variables that are components of the CURB-65 and PSI as well as CURB-65 and PSI themselves were not taken into account in multivariate analysis. We used multiple logistic regression analysis to determine whether the CURB-65, PSI, and/or other independent risk factors predicted 30-day mortality. The discrimination capability of the combination of each prognostic scoring system with other factors J o u r n a l P r e -p r o o f was evaluated in the receiver-operating-characteristic analysis. The areas under the curves (AUC) of the prediction models were compared using the Delong and Clarke-Pearson approach (DeLong et al. 1988 ). P value < 0.05 was accepted as statistically significant. The analyses were computed with IBM SPSS Statistics 23.

A total of 681 patients were included in the study. Mean ± SD age was 56.9 ± 15.7 years and 49% of the patients were female. Three hundred and seventy (54.3%) patients had at least one comorbidity. The most common comorbidity was hypertension, followed by diabetes mellitus, asthma, chronic obstructive lung disease, ischemic heart disease, hyperlipidemia, chronic renal disease and congestive heart failure. The most common clinical presentations were fever (32.5%) and respiratory tract symptoms including cough (71.2%) and dyspnea (27.3%) ( Table 3) .

Among the 681 patients hospitalized with COVID-19, 672 (98.6%) patients had been initially transferred to the ward. Of these, 596 (88.6%) patients were discharged, 74 (11%) patients were transferred to intensive care unit (ICU), and 2 patients died in the ward. Among the 74 patients transferred to ICU, 45 patients died and 29 patients were discharged. Among the 9 patients who were initially transferred to ICU, 8 patients died and 1 patient was discharged ( Figure 1 ).

Overall, 55 (8%) patients died within 30 days of admission to the hospital and the median time from admission to death was 9.5 (IQR: 6-22) days. Deceased patients were older, more hypoxic, tachycardic, tachypneic and hypotensive at admission. They were more likely to have at least one comorbidity. Regarding laboratory parameters, they had higher levels of neutrophil count, blood urea nitrogen, ferritin, CRP, troponin and lower levels of lymphocyte count (Table 3) .

J o u r n a l P r e -p r o o f A total of 550 (80.8%) patients had a CURB-65 score of 0 or 1. Of these, 15 (2.7%) patients died within 30 days. One hundred thirty-one patients (19.2%) had a CURB-65 score of ≥2. Of these, 40 patients (30.5%) died within 30 days. A CURB-65 score of ≥ 2 had a fair discriminatory ability to predict 30-day mortality with a sensitivity of 73%, specificity of 85%, PPV of 31%, NPV of 97% (AUC: 0.79, 95% CI 72 to 86, p <0.001) ( Table 4 ).

One hundred eighty-two patients (26.7%) were in group I, 249 patients (36.6%) were in group II, 136 patients (20%) were in group III, 82 patients (12%) were in group IV and 31 patients (4.7%) were in group V. There were no deaths among the patients in group I. The mortality rate was 2% in group II, 4.4% in group III, 28% in group IV, and 65.6% in group V. The PSI group ≥ 4 had a good discriminatory ability to predict 30-day mortality with a sensitivity of 80%, specificity of 89%, PPV of 39%, NPV of 98% (AUC=0.85, 95% CI 78 to 90, p <0.001) ( Table 4) .

The univariate analysis revealed that levels of ferritin, CRP, troponin and lymphocyte count were associated with 30-day mortality. After multivariate analysis, only elevated CRP values (OR:1.015, 95% CI 1.008 to 1.021, p<0.001) were significantly associated with 30-day mortality. (Table 4) .

AUCs for 30-day mortality prediction of the CURB-65 alone, PSI alone and PSI with CRP were 0.88 with %95 CI from 0.85 to 0.90 (p <0.001), 0.91 with 95% CI from 0.88 to 0.93 (p <0.001) and 0.92 with 95% from 0.89 to 0.94 (p <0.001), respectively (Figure 2 ). Comparing the AUCs for 30-day mortality prediction of the CURB-65 alone, PSI alone and the model including PSI and CRP levels showed that the two-variable model and PSI alone predicted 30-day mortality significantly better than the CURB-65 alone (p=0.01, p=0.04, respectively). However, J o u r n a l P r e -p r o o f discriminatory abilities of the PSI and the two variable model including PSI and CRP were similar (p=0.29).

In this study, we assessed the abilities of two prognostic scoring systems to predict 30-day mortality and evaluated independent predictive factors of mortality in a large group of patients with COVID-19. The 30-day mortality rate was 8% in our study. The PSI group ≥ 4 showed better sensitivity (80% vs 73%) and specificity (89% vs 85%) but a similar negative predictive value (98% and 97%) in predicting death compared to the CURB-65 score of ≥ 2. Only elevated levels of CRP were independently associated with 30-day mortality. The PSI scores alone and the twovariable model including PSI scores and CRP levels performed better than the CURB-65 scores whereas the PSI scores alone and the two-variable model had similar discriminatory ability in predicting 30-day mortality.

The mortality rate of COVID-19 has been reported as between 11.7% and 28.2%. Cao et al. 2020; Wu et al. 2019; Giacomelli et al. 2020; Huang et al. 2019) . The variation in the mortality rate may be due to the heterogeneity in the patient characteristics, treatment strategies and mortality measures (e.g. in-hospital or 30-day measure).

In this study, our mortality rate was somewhat lower than reported in previous studies although our cohort had similar demographic features and comorbidities compared with previous studies Cao et al. 2020; Wu et al. 2019; Giacomelli et al. 2020; Huang et al. 2019; ) . The hospitalization criteria in Turkey may be a possible explanation for this finding. As we mentioned earlier in Introduction, a considerable number of non-severe patients were hospitalized according to their older age and/or coexisting comorbidities. Thus, our cohort J o u r n a l P r e -p r o o f might represent less severe COVID-19 patients. For instance, the proportion of severe cases at admission was 21.1% in our cohort while this was 63% in the study by Zhou et al., and their mortality was 28% ). On the other hand, a retrospective study including only nonsevere cases at admission showed that 20 (5%) of the 348 patients became severe during hospitalization and 40% of them received only conventional oxygen therapy (Duan et al. 2020 ).

There have been ongoing attempts to develop a prognostic scoring system that can predict a poor outcome for patients with COVID-19 (Wynants et al. 2020) . The CURB-65 scores were found to be significantly higher in deceased patients with COVID-19 ). Liu and colleagues compared the clinical characteristics and outcomes of elderly and young patients with and showed that the PSI scores are higher in the elderly compared to young patients . As far as we know, this is the first study to evaluate the performance of the CURB-65 and PSI in the prediction of mortality. In our study, in predicting 30-day mortality, CURB-65 score of ≥ 2 had a sensitivity of 73% and specificity of 85%, and PSI group ≥4 had a sensitivity of 80% and specificity of 89%. When we analyzed the prognostic scoring systems as continuous variables, we found that PSI scores alone predicted mortality significantly better than CURB-65 scores (p=0.04). Finally, we included CRP levels to PSI scale in order to improve its performance, however, adding CRP levels to PSI scale did not perform better than PSI scores alone. A better discriminatory ability of PSI scale was an expected finding since the PSI scale consists of several parameters such as age, comorbidities, hypoxemia that were found to be associated with increased risk of mortality in patients with COVID-19. More surprising was the finding that CRP levels did not add prognostic information beyond PSI scores alone. However, adding CRP to PSI scale also did not increase the prognostic performance of PSI in hospitalized patients with communityacquired pneumonia (Lee et al. 2011 ).

Since our first aim was to assess the performance of two prognostic scoring systems and to find additional variables that could improve their performance, we did not include variables that are components of these tools in the multivariate analysis. Non-survivor patients had increased levels of CRP, troponin, ferritin, lower lymphocyte counts and higher neutrophil counts compared to survivor patients. After multivariate analysis, elevated CRP levels were significantly associated with increased risk for mortality and this finding was consistent with the previous studies. Elevated CRP levels were also reported to predict progression to severe illness and to correlate with the radiological extent of disease (Duan et al. 2020; Wang et al. 2020) .

Our study has some limitations. First, we did not calculate the prognostic scores prospectively.

However, the hospitals in Turkey has collected the clinical data in a standard format during the outbreak. Regarding laboratory results, other than D-dimer levels there was no missing data as all the laboratory parameters were part of the routine evaluation of all hospitalized patients. Second, among the previously reported risk factors for mortality in COVID-19, our analysis did not take into account the potential risk factors such as body mass index, IL-6 levels and radiological findings.

In conclusion, this single center retrospective study including a large cohort of COVID-19 patients showed that PSI is a powerful tool to predict mortality in patients with COVID-19. It performed significantly better than CURB-65 and the addition of CRP levels to PSI scale did not improve the performance of PSI. During the outbreak, PSI can help physicians to stratify patients at admission.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Table 3 . Comparison of demographic, clinical and laboratory findings between alive and deceased patients Table 4 . Discriminative accuracy of the CURB-65 and PSI in predicting 30-day mortality J o u r n a l P r e -p r o o f 

Clinical Features and Short-term Outcomes of 102 Patients with Corona Virus Disease

Comparing the Areas under Two or More Correlated Receiver Operating Characteristic Curves: A Nonparametric Approach

Predictors of mortality for patients with COVID-19 pneumonia caused by SARSCoV-2: A prospective cohort study

Correlation between the variables collected at admission and progression to severe cases during hospitalization among COVID-19 patients in Chongqing

30-day mortality in patients hospitalized with COVID-19 during the first wave of the Italian epidemic: a prospective cohort study

Characteristics and circumstances of falls in a hospital setting: A prospective analysis

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Albumin and C-reactive protein have prognostic significance in patients with community-acquired pneumonia

Clinical features of COVID-19 in elderly patients: A comparison with young and middle-aged patients

Clinical characteristics of novel coronavirus cases in tertiary hospitals in Hubei Province

Neutrophil-to-lymphocyte ratio as an independent risk factor for mortality in hospitalized patients with COVID-19

CURB 65 score as a predictor of early mortality in hospital-acquired pneumonia

Clinical predictors of mortality due to COVID-19 based on an analysis of data of 150 patients from Wuhan

Validity of pneumonia severity index and CURB-65 severity scoring systems in community acquired pneumonia in an Indian setting

C-reactive protein levels in the early stage of COVID-19

WHO Novel Coronavirus (2019-nCoV) situation report

Risk Factors Associated With Acute Respiratory Distress Syndrome and Death in Patients With Coronavirus Disease

Prediction models for diagnosis and prognosis of covid-19 infection: Systematic review and critical appraisal

An interpretable mortality prediction model for COVID-19 patients

Analysis of 92 deceased patients with COVID-19

D-dimer levels on admission to predict inhospital mortality in patients with Covid-19

Risk factors of critical & mortal COVID-19 cases: A systematic literature review and meta-analysis

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Essential psychological problems of hospitalized patients

II or III versus IV or V. Abbreviations: CI: confidence interval, PPV: positive predictive values, NPV: negative predictive values, AUC: area under curve