Abstract
BACKGROUND: Many studies have revealed several risk factors associated with the prognosis of patients with coronavirus disease 2019 (COVID-19), but the risk factors associated with death in critically ill COVID-19 patients still needs to be fully elucidated. Therefore, we analyzed clinical characteristics and laboratory data of ICU patients to identify risk factors associated with COVID-19 death. METHODS: Patients with COVID-19 from the ICU in the Sino-French New City Branch of Tongji Hospital Wuhan, China, between February 4 and February 29, 2020, were enrolled in this study. The final date of follow-up was April 4, 2020. Clinical manifestations, laboratory tests, treatment, and outcome of participants before and during the ICU stay were retrospectively collected and analyzed. RESULTS: A total of 92 patients were admitted or transferred to the ICU from February 4 to February 29, 2020. Compared to survivors, the majority of non-survivors (73.8%) presented with dyspnea. A random forest classifier and ROC curve were used to develop a predictive model. IL-6, D-dimer, lymphocytes, and albumin achieved good performance with AUCs of 0.9476, 0.9165, 0.8994, and 0.9251, respectively, which were consistent with clinical observations, such as inflammation, lymphopenia, and coagulation dysfunction. Combining IL-6 and D-dimer improved the performance of this model with an excellent AUC (0.997). CONCLUSIONS: Mortality in COVID-19 was not rare in critically ill patients. The model that combined IL-6 and D-dimer was valuable for predicting the mortality of patients with COVID-19 with excellent performance. This model needs to be further optimized by adding more indicators and then evaluated with a multicenter study.