Predictive performance of emergency department-specific variables on COVID-19 pneumonia.

Tan, Hann-Yee; Yeo, Mathew; Tay, Xin-Ying; Fung, Michael; Kumar, Ranjeev; Ooi, Say-Tat; Amirah, Lina; Ubeynarayana, Chalani Udhyami; Mao, Desmond

INTRODUCTION The majority of patients with COVID-19 infection do not progress to pneumonia. We report emergency department (ED)-specific variables and evaluate their predictive performance on diagnosis of pneumonia, intensive care unit (ICU) admission and death. METHODS This was a retrospective, single-centre cohort study of confirmed COVID-19 patients admitted to a Singapore tertiary hospital. Primary outcome was diagnosis of COVID-19 pneumonia. Secondary outcomes were ICU admission and/or death. Multivariate logistic regression was used to analyse the predictive performance of ED-specific variables. Accuracy of continuous variables was measured by area under receiver operating characteristic (ROC) curve. RESULTS 294 patients were included. Patients with pneumonia were older (52.0 years, p < 0.001) and had higher C-reactive protein (CRP; 33.8 mg/L, p < 0.001). Patients with indeterminate chest radiograph (CRX) findings were at risk of pneumonia vs. patients with normal CRX (37.5% vs. 4.3%, p < 0.001). Patients admitted to ICU were older (60.0 years, p < 0.001) and had higher CRP (40.0 mg/L, p < 0.001). Diagnosis of COVID-19 pneumonia was associated with ICU admission and death (30.0% vs 0.39%, p < 0.001). Multivariate logistic regression analysis showed that age (aOR 1.07, p = 0.049), CRP (aOR 1.05, p = 0.006) and CRX findings (aOR 50.00, p < 0.001) had increased odds of pneumonia. ROC curve analysis showed that CRP of 23.3 mg/L was the optimal cut-off for predicting pneumonia. CONCLUSION Older age, higher CRP and CRX findings are associated with COVID-19 pneumonia, ICU admission and death. Prospective studies should be undertaken to validate these findings.