Abstract
OBJECTIVES: Effective early risk stratification of COVID-19 pneumonia patients in emergency departments (EDs) is crucial, especially in resource-limited settings. Common clinical scores (National Early Warning Score version 2 [NEWS2], quick Sequential Organ Failure Assessment [qSOFA], and CRB-65) were not developed for COVID-19 and may inadequately predict short-term mortality. This study aimed to evaluate the prognostic performance of these scores and to develop and internally validate a novel score-the Chiang Mai COVID-19 Pneumonia 7-Day Mortality Prediction Score (CCP-7). METHODS: We conducted a retrospective cohort study of patients aged ≥16 years presenting to a tertiary ED in Northern Thailand with confirmed COVID-19 pneumonia between January 2020 and December 2023. Demographic, clinical, and laboratory data were extracted from electronic records. The predictive accuracy of NEWS2, qSOFA, CRB-65, and the newly developed CCP-7 score for 7-day mortality was assessed using a regression framework, the area under the receiver operating characteristic curve (AUROC), calibration, and DeLong tests for AUROC comparisons. RESULTS: Among 735 patients included, the 7-day mortality rate was 5.2%. Four variables-respiratory rate greater than 30/min, altered mental status, abnormal white blood cell count, and thrombocytopenia-were independently associated with mortality and incorporated into the CCP-7. The CCP-7 score demonstrated superior discrimination (AUROC, 0.83; 95% confidence interval [CI]: 0.76-0.90) compared to CRB-65 (0.80, 95% CI: 0.74-0.86), NEWS2 (0.77, 95% CI: 0.68-0.85), and qSOFA (0.64, 95% CI: 0.52-0.75). DeLong tests showed no statistically significant differences between CCP-7 and CRB-65 or NEWS2. At a cutoff of ≥3 points, CCP-7 achieved 57.1% sensitivity and 89.4% specificity. CONCLUSIONS: CCP-7 is a simple and context-appropriate tool for predicting 7-day mortality in patients with COVID-19 pneumonia. Although it showed numerically higher discrimination than other scores, the differences were not statistically significant. Its reliance on routine ED parameters makes it particularly suited for rapid risk stratification in low-resource settings, but external validation is essential.