AbstractIntroductionCOVID‐19 has been threatening global public health since 2019. To address the overwhelming caseload, several tools were developed to predict prognosis and aid triage of critically ill patients for intensive care. Currently, there is a lack of local data on the validity of such tools.ObjectiveThe objective of this study is to identify the predictors of intensive care unit (ICU) mortality in COVID‐19 patients in Hong Kong and externally validate the different scoring systems.MethodsA retrospective cohort analysis of patients admitted to the HKEC ICUs from 1 January 2022 to 30 April 2022 was performed. We collected data on patient demographics, vaccination status, laboratory parameters, and clinical outcomes including need for organ support and mortality. Clinical severities were estimated by Sequential Organ Failure Score, 4C Mortality Score, COVID‐Gram score, and Acute Physiology and Chronic Health Evaluation (APACHE) IV score based on the original studies. Comparison between individual scoring systems' performance on hospital mortality was conducted and summarized.ResultsIn these four months, 137 patients with COVID‐19 admitted to ICUs of Ruttonjee Hospital and Pamela Youde Nethersole Eastern Hospital were recruited. 64 patients (46.7%) were admitted for COVID‐19 related respiratory failure, among which mortality was 66.7%. The overall hospital and ICU mortality were 21.9% and 13.1%, respectively. Invasive mechanical ventilation (IMV) (OR 3.221, p 0.034), high flow nasal cannula (HFNC) oxygen therapy (OR 3.107, p 0.039), and higher APACHE IV score (OR 1.043, p < 0.001) were independent predictors of hospital mortality using multivariate analysis. The scoring systems had good performance in mortality prediction in our population. The APACHE IV score (AUROC 0.79, 95% CI 0.698–0.894) and 4C Mortality Score (AUROC 0.751, 95% CI 0.657–0.844) outperformed other scoring systems in predicting hospital mortality.ConclusionIn patients with COVID‐19, the use of IMV or HFNC and APACHE IV score were independent risk factors for hospital mortality. The APACHE IV and the 4C Mortality Score were most useful in our population for predicting ICU and hospital mortality.