key: cord-253962-ug7yflxh authors: Huang, Dong; Wang, Ting; Chen, Zhu; Yang, Huan; Yao, Rong; Liang, Zongan title: A novel risk score to predict diagnosis with Coronavirus Disease 2019 (COVID‐19) in suspected patients: A retrospective, multi‐center, observational study date: 2020-06-08 journal: J Med Virol DOI: 10.1002/jmv.26143 sha: doc_id: 253962 cord_uid: ug7yflxh BACKGROUND: The aim of the study was to explore a novel risk score to predict diagnosis with COVID‐19 among all suspected patients at admission. METHODS: This was a retrospective, multi‐center, observational study. The clinical data of all suspected patients were analyzed. Independent risk factors were identified via multivariate logistic regression analysis. RESULTS: Finally, 336 confirmed COVID‐19 patients and 139 control patients were included. We found nine independent risk factors for diagnosis with COVID‐19 at admission to hospital: epidemiological exposure histories (OR:13.32, 95%CI 6.39‐27.75), weakness/fatigue (OR:4.51, 95%CI 1.70‐11.96), heart rate <100 beat/min (OR:3.80, 95%CI 2.00‐7.22), bilateral pneumonia (OR:3.60, 95%CI 1.83‐7.10), neutrophil count ≤6.3×10(9)/L (OR: 6.77, 95%CI 2.52‐18.19), eosinophil count ≤0.02×10(9)/L (OR:3.14, 95%CI 1.58‐6.22), glucose ≥6 mmol/L (OR:2.43, 95%CI 1.04‐5.66), D‐dimer ≥0.5 mg/L (OR:3.49, 95%CI 1.22‐9.96), and C‐reactive protein <5 mg/L (OR:3.83, 95%CI 1.86‐7.92). As for the performance of this risk score, a cut‐off value of 20 (specificity: 0.866, sensitivity: 0.813) was identified to predict COVID‐19 according to ROC curve and the area under the curve (AUC) was 0.921 (95%CI: 0.896‐0.945, p<0.01). CONCLUSIONS: We designed a novel risk score which might have a promising predictive capacity for diagnosis with COVID‐19 among suspected patients. This article is protected by copyright. All rights reserved. predicting risk score Since December 2019, an increasing number of coronavirus disease 2019 (COVID-19) cases were identified all over the world for last few months. 1, 2 So far more than one million patients have been diagnosed with COVID-19 worldwide. It is estimated that the overall mortality now is about 5.7% globally. 3 The elders and patients with comorbidities often develop to acute respiratory distress syndrome (ARDS), shock or organ failure, and finally yield poorer clinical outcomes. 4 Respiratory failure, immunosuppression, as well as systemic infection and inflammation are already recognized as main clinical characteristics of COVID-19 patients. 5 This outbreak has caused enormous adverse impacts around the world. The epidemiologic situation is still severe but medical system capacity is limited at present. As a result, it is necessary to improve hospital management and stratification of patients as early as possible. This article is protected by copyright. All rights reserved. The real-time reverse transcription-PCR(RT-PCR) has competent ability to detect virus and is the most reliable diagnostic method for COVID-19 now. However, given the incidence of RT-PCR false-negative results, shortage of PCR kit, and possible delayed diagnosis due to timeconsuming process of RT-PCR, 6 an early efficient identification of confirmed COVID-19 patients is important for early diagnosis and treatments. It can also help decrease the risks of spread of viral infection. Previous few studies about the differentiation of confirmed COVID-19 and suspected cases had some limitations, including relatively small sample size, insufficient clinical utility, etc. The current study is conducted aiming to explore the potential early risk factors, and to develop a risk score used for predicting the probability of diagnosis among all suspected COVID-19 patients at early stage. This was retrospective, multi-center, observational study on patients admitted into twenty-six COVID-19 designated hospitals from January 21 to February 7,2020, in Sichuan province, China This article is protected by copyright. All rights reserved. This study was conducted in accordance with the amended Declaration of Helsinki. This study was approved by the West China Hospital of Sichuan University Biomedical Research Ethics Committee (No. 2020-272). Written informed consent was waived because of the urgent need to collect clinical data and retrospective observational design. All patient data were anonymously recorded to ensure confidentiality. Two doctors reviewed the medical records of all patients independently. Any disagreement was resolved through the third doctor and team discussion until consensus reached. All patients enrolled in this study were diagnosed with confirmed or This article is protected by copyright. All rights reserved. fluorescence RT-PCR. Meanwhile, the suspected patients with finally RT-PCR negative results were included into control group. To be specific, if patients had received at least two RT-PCR tests taken at least 24-hour apart and the results were all negative, they were included into control group. There was no exclusion criterion. Baseline data, including demographic characteristics, comorbidities, basic vital signs, symptoms and signs, chest computed tomography (CT) scan images and laboratory examinations data were retrospectively collected from electronic medical records. These laboratory examinations were all recorded within 24 hours after admission to hospitals. Continuous variables were categorized for further analysis. The threshold value of each continuous variable was determined by the clinically relevant cut-off value, or upper limit or lower limit of normal range. Two doctors completed the data collection independently. The primary outcome is diagnosis of COVID-19. The score for each independent risk factor was assigned as integer value close to the regression coefficient. The total risk score of each patient is the sum of each single score. To assess the relationship between the risk score and diagnosis, we did the receiver-operating characteristics (ROC) curve and reported Area under the ROC curve (AUC). The optimal cut-off point of the risk score was based on the Youden's index of ROC curve while sensitivity and specificity were reported. P<0.05 was considered statistically significant. A total of 475 patients who met the inclusion criteria were retrospectively enrolled in the study. The patients in our study consisted of 264 males and 211 females. They had a median age of 40 years old (IQR 30-52 years old). Totally 252(53.1%) patients had epidemiological exposure histories. The most common comorbidities were hypertension This article is protected by copyright. All rights reserved. (n=66, 13.9%) and diabetes (n=39, 8.2%). The most common symptoms were fever (n=314, 66.1%), productive cough (n=170, 35.8%) and dry cough (n=166, 34.9%). The chest CT scan of 448(94.3%) patients showed abnormal signs. Some abnormal laboratory test results were also found, such as lymphocytopenia (n=181, 38.1%) and elevated C-reactive protein (CRP) (n=139, 29.3%). There were 336(70.7%) patients finally confirmed with COVID-19. And 265 (78.9%) patients had mild cases at admission. The median age of confirmed patients was higher than that of control group (43 years vs 34 years, P<0.01). About 39.9% of confirmed patients has at least one comorbidity. However, that of control patients was significantly lower Table 1 . The factors with a P value less than 0.10 in Table 1 (Table 2) Thus, the detailed risk score was calculated and formed. (Table 3) The number of patients, sensitivity and specificity of each cut-off point was shown in Table 4 . A cut-off value of 20 (specificity: 0.866, sensitivity: 0.813) was identified to predict according to ROC curve and area under the curve (AUC) was 0.921 (95%CI: 0.896-0.945, p<0.01) (Figure 1 ). To our knowledge, this is the first predictive tool used for predicting the possibility of diagnosis with COVID-19 among all suspected patients at admission to hospital. We found nine independent risk factors for diagnosis with COVID-19 and the score of each indicator varies from 2 to 13 points. The total score varies from 2 to 45 points for each patient. Higher total score represents increased probability of COVID-19. patients. 9 However, some other results are not identical between our study and above studies. For example, the differences of incidences of fever between two groups are significant in above studies but not significant in our study. We speculate that this result is associated with disease types of control group. In current study, most of RT-PCR test negative patients were finally diagnosed with influenza or bacterial pneumonia, which could also result in fever. However, we could not perform further comparison due to lack of original data of above studies. Our study has larger sample size and includes more laboratory indexes, and explore independent risk factors after multivariate analysis. This article is protected by copyright. All rights reserved. Compared with control group, confirmed patients had lower occurrence of rhinorrhea, but higher rate of diarrhea, nausea and vomiting. It is consistent with previous studies, which demonstrated SARS-Cov-2 need to bind to the angiotensin converting enzyme 2 (ACE-2) receptor for cell entry. 10 Most of ACE-2 receptor located in gastrointestinal tract and lower airway. Therefore, it is understandable that most of symptoms in COVID-patients are nonspecific. The rates of hypertension and diabetes were both higher in confirmed cases, which was found in previous studies 11 and current study. However, they are not independent risk factor. The interference of ACE2 increasing drugs, deficient sample size of hypertension and diabetes patients might be reasons. It is unexpected that the increased level of serum glucose is an independent risk factor, even after adjusting for diabetes history. However, we should be cautious about this finding. The 95%CI is near to non-significance (1.04-5.66). Meanwhile, the uses of corticosteroid are also not clear before admission. And oxidative stress could also affect levels of glucose. We also found normal heart rate is a risk factor in our score. The exact effects of coronavirus on heart are not completely clear so far. Generally, tachycardia correlates with fever and is common in community-acquired pneumonia, 12 which accounted for the majority of control cases. The median heart rates are similar between our study and previous studies about COVID -19. 4 There are also some indicators, such as comorbidities, symptoms, and alanine aminotransferase, which were significantly different between two groups in univariate analysis, but not in multivariate analysis. The This article is protected by copyright. All rights reserved. chronic comorbidities and cough are also risk factors for influenza. 13 It has been reported 2-11% of patients with COVID-19 had liver comorbidities and 14-53% cases had liver injury 14 , which was consistent with our results. Similarly, severe COVID-19 cases often have higher rates of liver dysfunction. Dysregulated innate immune response, immunocompromised status and cytokine storm might be the reasons. However, other researchers suspect that COVID-19-induced hepatic damage is a clinical distraction and it is not necessary for physicians to excessively focus on indicators of liver injury. 15 The viral control is most important and major issue during treatment of COVID-19. Based on our results, we believed these parameters had relatively limited effects on the identification of COVID-19 from suspected patients. However, our findings remain to be confirmed in the future. It's worth noting that some changes in white blood cell counts was inconsistent with previous studies. It is reported that coronavirus might mainly act on lymphocytes and cause lower percentages of lymphocytes, monocytes and eosinophils. 16 Accepted Article such a short time. Moreover, about 23% (15/66) of patients who was mild at admission would progress to severe COVID-19 during hospital stay. 18 In the current study most of patients were mild at admission. It is acknowledged that decrease of white blood cell counts were more common in patients with severe diseases. The serum level of procalcitonin was normal and similar between two groups. However, the level of C-reactive protein was significantly higher in control group. Less than 5 mg/L C-reactive protein is even an independent risk factor for COVID- 19 . It has been demonstrated that pneumonia caused by 2009 H1N1 influenza alone had significantly lower C-reactive protein level than mixed bacterial and viral infection pneumonia. 19 Additionally, patients with Middle East respiratory syndrome coronavirus (MERS-CoV) also often have lower C-reactive protein level among patients with acute febrile illness. 20 To sum up, all indicators in this novel risk score are easy to get at admission to hospital. ROC analysis suggests it is promising for the risk stratification among suspected patients. We believed this study reflected the "real world" situation, to some degree. It is crucial for physicians to differentiate COVID-19 from other similar diseases because it is highly infectious. This risk score might help physicians make appropriate decisions about early diagnosis and treatments. Furthermore, it might become a suitable supplement to RT-PCR and help researchers reveal detailed pathophysiological mechanisms of COVID-19 in future. This article is protected by copyright. All rights reserved. Nevertheless, there are still several limitations in the study. First, it was a retrospective observational study, unavoidable subjective selection bias existed. Second, the sample size was relatively small, and the number of patients was not equal between groups. Third, drugs and therapies before admission might have disturbed our results. Forth, we did not explore the variation trend of laboratory examinations in few days after admission among suspected patients. Further well-designed, multi-center studies with better comparability are warranted to update this risk score. We found a novel risk score, which is based on nine easy-to-get parameters in clinical practice. It has a promising predictive capacity for diagnosis with COVID-19 among all suspected patients. Our findings need to be confirmed in further studies. Tables Table 1. Comparisons of clinical characteristics between confirmed patients and control group. Overall ( Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia First Case of 2019 Novel Coronavirus in the United States Real estimates of mortality following COVID-19 infection Comorbidity and its impact on 1590 patients with Covid-19 in China: A Nationwide Analysis Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan Positive rate of RT-PCR detection of SARS-CoV-2 infection in 4880 cases from one hospital in New coronavirus pneumonia prevention and control program Differences between COVID-19 and suspected then confirmed SARS-CoV-2-negative pneumonia: a retrospective study from a single center A comparative study on the clinical features of COVID-19 pneumonia to other pneumonias SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor Are patients with hypertension and diabetes mellitus at increased risk for COVID-19 infection? Community-Acquired Pneumonia in Adults: Diagnosis and Management Liver injury in COVID-19: management and challenges COVID-19 and the liver: little cause for concern Dysregulation of immune response in patients with COVID-19 in Wuhan, China Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan Host susceptibility to severe COVID-19 and establishment of a host risk score: findings of 487 cases outside Wuhan Role of procalcitonin and C-reactive protein in differentiation of mixed bacterial infection from 2009 H1N1 viral pneumonia. Influenza Other Respir Viruses Differential Cell Count and CRP Level in Blood as Predictors for Middle East Respiratory Syndrome Coronavirus Infection in Acute Febrile Patients during Nosocomial Outbreak Bilateral pneumonia OR: odds ratio None. Bilateral pneumonia 4Heart rate (beat/min) (<100) 4 C-reactive protein, mg/L (<5) 4Eosinophil count, ×10 9 /L (≤0.02) 3Glucose, mmol/L (≥6) 2