This retrospective, observational cohort study enrolled moderate to critical illness COVID-19 patients admitted to isolation rooms from April 1 to June 30, 2022, at Taichung Veterans General Hospital. Patients aged <18 years and pregnant were excluded. The study was approved by the Taichung Veterans General Hospital Human Studies Committee.

COVID-19 patients were defined by a positive SARS-CoV-2 infection test result, confirmed by real time reverse transcriptase–polymerase chain reaction (RT-PCR) assay from nasal and pharyngeal swabs. According to COVID-19 severity definitions established by the Taiwan CDC and American College of Emergency Physicians (ACEP), moderate illness was defined as having signs and symptoms of lower respiratory disease or abnormal imaging and a SpO2 ≥ 94% at room temperature at sea level. Severe illness was defined as a SpO2 < 94% at room temperature at sea level, a PaO₂/FiO₂ < 300 mmHg, a respiratory frequency > 30 breaths/minute or having lung infiltrates > 50%. Critical illness was defined as respiratory failure, septic shock, and/or multiple organ dysfunction. The index day (day 0) was the day when a SARS-CoV-2 RT-PCR test confirmed positive, with patients followed until death in hospital or hospital discharge, whichever occurred first.

Medical chart reviews were conducted by infectious disease doctors using a standard case report form. We recorded each patient’s age, gender, body mass index (BMI) and preexisting comorbidities, including cancer, CVD, diabetes, CKD, CLD and chronic liver disease. Obesity was defined as a calculated BMI ≥ 30 kg/m2. The Charlson comorbidity index (CCI), used to estimate the risk of death from comorbid diseases, was used for calculations.15 We also recorded each patient’s history of COVID-19 vaccinations. In Taiwan, four types of vaccines, Moderna, AstraZeneca (AZ), Pfizer-Biontech (BNT) and the MVC-COV1901 COVID-19 Vaccine, were available during the study period. Groups of respiratory support were stratified from non-oxygen use/nasal cannula, to simple mask/non-rebreathing mask/high-flow nasal cannula (HFNC), and bilevel positive airway pressure (BiPAP)/mechanical ventilators. Time to remdesivir or dexamethasone was identified as the onset of COVID-19 symptoms to medication administration.

The demographic data regarding the survivors and non-survivors were analyzed. Descriptive statistics for patient and hospital characteristics were calculated using either mean (standard deviation (SD)), median (range or interquartile range (IQR)), or frequency count (percentage). Continuous variables were analyzed by the Wilcoxon rank-sum test and categorical variables by the Chi-square test or Fisher’s exact test. All tests were two-sided, and a p value of <0.05 was considered to be statistically significant. The Kaplan-Meier method was used to determine the cumulative survival rate, category by CCI score, cancer and diabetes. Logistic regression analysis was used to determine independent predictors of hospital mortality. Model 1 was adjusted for age, gender, severity of COVID-19, cancer, CVD, diabetes, CKD, CLD, chronic liver disease, obesity, history of vaccination, remdesivir use, dexamethasone use and the cycle threshold (Ct) value at day 0. Model 2 was adjusted for age, gender, severity of COVID-19, obesity, CCI ≥ 6, history of vaccination, remdesivir use, dexamethasone use and Ct value at day 0. Variables differing between survivors and non-survivors with a p value < 0.5, according to the Chi-square test or Fisher’s exact test, or the Wilcoxon rank-sum test, were entered into the multivariable regression model. Results were presented as odds ratios (ORs) with 95% confidence intervals (CI). Two-way analysis of variance (ANOVA) was used to examine the interaction between the independent risk factors for hospital mortality, including the onset of remdesivir and dexamethasone use time, history of vaccinations, cancer and diabetes. Statistical analysis were performed using SPSS software (Version 22.0).

A total of 838 patients were assessed for eligibility, with 196 children (23.39%) and 21 pregnant women (2.5%) being excluded. 416 patients were excluded due to mild COVID-19 illness. Ultimately, 205 patients experiencing moderate to critical COVID-19 illness were enrolled in the final analysis (Figure 1). Amongst them, 60 patients (29.27%) expired and 145 (70.73%) survived, with the mortality rate within the moderate to critical illness COVID-19 patients being 29.5%.

Clinical characteristics within the moderate to critical COVID-19 patients are shown in Table 1. The median age was 73.0 years for survivors and 78.5 years for non-survivors. Non-survivors were more likely to be experiencing critical illness and also have a higher proportion of cancer, diabetes and CVD. The median CCI score was significantly higher in non-survivors (7.5) compared with survivors (5.0). Only 21.7% of non-survivors received ≥ 3 doses of the COVID-19 vaccinations. With respiratory support, 40.0% of non-survivors required Bi-PAP or mechanical ventilator respiratory support. For COVID-19 treatment, oral antiviral agents included nirmatrelvir/ritonavir (Paxlovid) and Molnupiravir, with 16 patients (16/205, 7.8%) receiving Paxlovid and 31 patients (31/205, 15.12%) receiving Molnupiravir in the survivor group. In the non-survivor group, only 14 patients (14/205, 6.82%) received Molnupiravir, with no patients receiving Paxlovid. The onset of remdesivir and dexamethasone use time was more likely to be more than 2 days for non-survivors. An empirical antibiotic was used in 198 patients (198/205, 96.59%), 97.24% of survivors, and 95% of non-survivors. The overall length of hospital stay was 14.91 (7-20) days, 15.31(10-20) days for survivors, and 14.23 (1-58) days for non-survivors.

The cumulative survival rates during admission are shown in Figure 2a, Figure 2b and Figure 2c. The cumulative survival rate amongst patients with a CCI score ≥ 6 was significantly lower than those with a CCI score < 6 (P < 0.001) (Figure 2a). The cumulative survival rate amongst patients with cancer was significantly lower than those without cancer (P = 0.002) (Figure 2b). The cumulative survival rate amongst patients with diabetes was significantly lower than those without diabetes (P = 0.044) (Figure 2c).

Multivariable analysis for COVID-19 mortality risk factors is shown in Table 2. Critical illness of COVID-19 (Model 1: OR=6.55; 95% CI, 2.61–16.43; Model 2: OR=6.13; 95% CI, 2.09–17.97), underlying comorbidities of cancer (OR=2.35; 95% CI, 1.47–3.76), diabetes (OR=2.32; 95% CI, 1.23–4.37), chronic liver disease (OR=3.22; 95% CI, 1.01–10.29), and a CCI score ≥ 6 (OR=7.77; 95% CI, 2.70–22.35) were significantly associated with a higher risk of mortality. Patients who never received any vaccination (Model 1: OR=3.22; 95% CI, 1.07–9.66; Model 2: OR=3.57; 95% CI, 1.12–11.36), had an onset of remdesivir use time ≥ 2 days (Model 1: OR=6.04; 95% CI, 2.81–12.99; Model 2: OR=7.89; 95% CI, 1.34–16.40), and had an onset of dexamethasone use time ≥ 2 days (Model 1: OR=4.30; 95% CI, 1.88–9.83; Model 2: OR=4.47; 95% CI, 1.34–22.39) were all independent risk factors for mortality in both Models 1 and 2.

Table 3 shows the association between cancer, diabetes and the onset of remdesivir use time for mortality rates. An onset of remdesivir use time (≥ 2 days) was associated with an increased mortality rate, particularly amongst patients with cancer/diabetes, compared to those without cancer/diabetes (p for interaction = 0.046/0.049). Table 4 shows the association between cancer, diabetes and the onset of dexamethasone use time for mortality rates. An onset of dexamethasone use time (≥ 2 days) was associated with a higher mortality rate in patients with cancer. In addition, the impact was significant amongst patients with diabetes, compared to those without diabetes (p for interaction = 0.042). Table 5 shows the association between cancer, diabetes and COVID-19 vaccines. Compared to patients who received ≥ 3 doses of COVID-19 vaccinations, those who received < 3 doses were associated with a higher mortality rate, particularly amongst patients with cancer/diabetes, compared to those without cancer/diabetes (p for interaction = 0.038/0.048).

Since our study was retrospective and mainly used data from existing medical records, it was anonymized when analyzed, protecting the participants' privacy and not imposing additional risks or burdens on them. Therefore, the ethics committee (Taichung Veterans General Hospital Human Studies Committee) waived the requirement for informed consent. (approval No. CF22345B) on Sep 29, 2022.

No potentially identifable human images or data is presented in this study.

Not applicable

All data generated or analyzed during this study are included in this published article.

None received.

The authors would like to thank all the patients for participating in this study and the staff of the COVID-19 Care Unit in Taichung Veterans General Hospital.