1. Introduction

Since December 2019, COVID-19 disease caused by severe acute respiratory syndrome coronavirus (SARS-CoV-2) has evolved into a global pandemic, resulting in 14.9 million related deaths. Although the majority of patients with COVID-19 cases exhibit only clinical symptoms of upper respiratory tract infection, progression to severe disease, and even death is still common. In May 2023, the WHO officially declared that the pandemic no longer constitutes a “Public Health Emergency of International Concern,[1] but COVID-19, as an acute respiratory infectious disease, still has a lot of experience to be learned in terms of prevention, control and treatment, with a view to applying it to similar infectious diseases.

Chongqing is one of the most populous and aging cities in China, and the SRAC-COV-2 strain circulating between 2022 and 2023 is one of the most frequently seen strains in China. The spread of COVID-19 in this city can be considered as a microcosm of the COVID-19 pandemic in China.[2] In response to the pandemic, Chongqing banned the movement of people in November 2022, but resumed in December 2022. Chongqing’s highly mobile and aging population has led to a sharp increase in the number of COVID-19 patients during this period, resulting in a large demand for hospitalization. On the other hand, during the SARS-CoV-2 epidemic, China adopted some preventive and curative measures that were not used in other countries, which are of special significance for the study of COVID-19 epidemics and the prevention and treatment of similar acute respiratory infections. For example, China values traditional Chinese medicine (TCM) treatment, which is recommended in all versions of COVID-19 diagnosis and treatment protocols. Depending on the condition of COVID-19 patients, ranging from mild to critical, different types of TCM have been established, including TCM decoctions and proprietary Chinese medicines, for home treatment of patients with early SARS-CoV-2 infection and for hospitalization of patients with severe or more severe disease.[3] The effectiveness of the Chinese version of the treatment regimen in hospitalized critically ill patients with COVID-19, particularly the impact of different treatments and interventions on disease progression, remains under-explored.

In this study, a multicenter retrospective cohort of hospitalized patients was established with confirmed COVID-19 cases in Chongqing from November 1, 2022 to December 22, 2022. Among them, severe and critically ill patients were confirmed and diagnosed by the city’s expert group in accordance with the Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia (Trial Version 10).[4] We collected demographic and sociological indicators and clinical data from patients, developed descriptive and inferential statistical models, analyzed the effects of different treatments, and elucidated key factors associated with mortality outcomes in critically ill patients. Previous studies have shown that comorbidities and advanced age are independent risk factors for mortality in patients with COVID-19.[5–7] This study has come up with some new conclusions that will change many previous perceptions.

2. Methods 2.1. Study population and data collection

Patients with severe pneumonia with COVID-19 who were hospitalized in the intensive care units (ICUs) of 25 tertiary hospitals in Chongqing between November 1, 2022 and December 22, 2022 were included. The patient was confirmed based on the International Classification of Diseases version 10 (ICD-10) diagnostic code (U07.1) and a positive SARS-CoV-2 PCR test or positive serology test for COVID-19 antibodies. Critically ill patients were identified by a panel of experts from the Health Commission after they were diagnosed. Demographic sociological information, medical history, symptoms, drug lists, laboratory results, vital signs, and other physiological variables were collected from HIS. Given the de-identified nature of the data and the fact that it was a retrospective study, ethical review of the study as well as written informed consent was waived by the Ethics Committee of the Affiliated Central Hospital of Chongqing University in accordance with the policies of the National Health and Wellness Commission.

2.2. Drug treatment program

All patients receiving pharmacological treatment were implemented according to the COVID-19 treatment guidelines issued by the Chinese National Health Commission. TCM therapy: Huashi Baidu granules, a well-known formula of TCM, was warmed orally or nasally fed, with 10 g twice daily for 7 to 14 days; Antiviral therapy: intravenous ambavizumab and lomivizumab, alone or in combination, both at a dose of 1000 mg/day; if necessary, intravenous COVID-19 human immunoglobulin, no more than 5 doses in total. In addition, vaccination statistics were available for patients who received at least 1 dose of vaccine prior to admission, of which SARS-COV-2 inactivated vaccine, recombinant protein vaccine, and adenovirus vector vaccine were documented.

2.3. Definitions 2.3.1. Respiratory support.

Oxygen is administered (face mask) when PaO2/FiO2 < 300 mm Hg, whereas trans-nasal high-flow oxygen therapy or noninvasive ventilation should be given when PaO2/FiO2 < 200 mm Hg, and tracheal intubation and invasive mechanical ventilation should be considered when PaO2/FiO2 < 150 mm Hg. Given the atypical clinical presentation of hypoxemia in some cases, PaO2/FiO2 attainment should not be an indication for endotracheal intubation and invasive mechanical ventilation, but should be assessed in real time in the context of the patient’s clinical presentation and organ function.[4] The critically ill patients involved in this study had a rapid progression of disease, so this study was based on the clinician’s judgement of the mode of respiratory support required for the progression of the patient’s disease as an influencing factor and entered into a multivariate model for the study.

2.3.2. Signature variables.

VIF or GVIF generally cannot reasonably assess the unintercepted bimodal logistic regression used in this study. Therefore, to adjust the dimensionality of the confidence spheroid, GVIF[1/(2*df)] is used to evaluate the covariance of the model, where “df” is the degree of freedom associated with the term. The primary outcome was in-hospital mortality, defined as death on the day of hospitalization or discharge. The worst clinical severity is defined as the most severe type achieved by the patient in hospital according to the diagnostic and clinical typing criteria in the “Diagnostic and Therapeutic Protocol for SARS-CoV-2 Infections (10th Edition Trial).”

2.4. Statistical analysis

Overall patient characteristics and characteristics of patients in each group were demonstrated by descriptive statistics. The study population were categorized into serve and critical groups according to whether or not they reached the highest type, and into surviving, and into survival and death groups according to whether the patients had died on discharge from the hospital.

To assess the differences between patient subgroups, continuous variables such as age and length of each hospitalization were summarized by mean (standard deviation), and median [minimum, maximum], and compared using the Wilcoxon rank sum test; categorical variables such as whether treatments such as TCM, antiviral medicines, respiratory supportive therapy, tracheotomy, etc, were reported as numbers (proportions) and compared using Pearson’s chi-square test or Fisher’s exact test. Variables that were statistically significant and those deemed necessary by clinical experts were included in the multifactorial model, with inclusion and exclusion thresholds of 0.1 and 0.15, respectively. All variables were analyzed using a no-intercept model, with the confidence ellipse dimension adjusted to assess covariance by calculating GVIF[1/(2*df)]. Area under curve (AUC) with 95% confidence region plus Delong test was used for model performance evaluation and AUC in the 4 intervals of <0.7, 0.7–0.8, 0.8–0.9, and >0.9 were considered as poor, moderate, good, and excellent for the accuracy, respectively.[8] Time from T1 to ICU discharge was analyzed for ICU survival and results are presented by Kaplan–Meier survival curves. All tests were 2-sided and a P value of <.05 was considered significant. Statistical analyses were performed using R software version 4.3.0.

3. Results

From November 1 to December 21, 2022, there were a total of 9093 confirmed COVID-19 cases in Chongqing, including 4607 males and 4486 females, with an average age of 43.89 years, with the oldest patient being 104 years old and the youngest patient being 3 days old. These patients included 7650 mild cases, 1224 moderate cases, 154 severe cases, and 65 critical cases, with 2.41% of the cases being severe or above. The geographical, sexual, conditions, and age distribution of these cases were counted (Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/L903), indicating that severe cases mainly came from 9 districts in the main city of Chongqing, and the top 3 prevalence rates were 7.64%, 6.3%, and 5.75%, respectively (Table S1, Supplemental Digital Content, https://links.lww.com/MD/L896).

Finally, among the 9093 COVID-19 patients, 189 patients with severe disease or above who were admitted to the ICU and had complete case data were included in this study, including 131 severe cases and 58 critical cases. There were 136 males and 53 females. At the time of first admission, 37 patients were asymptomatic, 21 were mild, 50 were moderate, 68 were severe, and 13 were critical. During hospitalization, 76 patients (40.2%) were treated with Huashi Baidu granules and 133 patients (70.4%) with antiviral drugs. In the end, 16 patients died, with a case fatality rate of 8.47%, of which 58 critically ill patients had a mortality rate of 20.69%. Of the 16 deaths, 7 were females and 9 were males, with an average age of 77.7 years. In addition, the deaths were all underlying diseases, including 8 cases of cardiovascular disease, 4 cases of cerebrovascular disease, 6 cases of respiratory disease, 3 cases of metabolic disease, and 1 case of tumor, and 62.5% (10/16) received at least 1 dose of vaccine.

3.1. Risk factors for COVID-19 progression to criticality 3.1.1. Univariate analysis for risk of critical illness.

Here, we explored the factors influencing the progression to critical in patients with COVID-19 by univariate analysis. After excluding 13 patients who were already critically ill at the time of diagnosis, a total of 176 severe patients were included in the study and then divided into the “not progression to critical” group and the “progressing to critical” group according to the highest type of progression. The difference analysis was performed and the results, as shown in Table 1, showed that 131 severe patients did not progress to critical and 45 patients progressed to critical, containing 126 males and 50 females. The youngest age in the population was 2 months and the oldest 101 years, with a mean age of 71 years and a median age of 75 years, but the percentage of patients in different age ranges did not differ significantly between the 2 groups (P = .08).

Table 1 - Univariate analysis of patients with severe COVID-19 progressing to critical illness. Characteristics Total No progressing Progressing P (N = 176) (N = 131) (N = 45) General characteristics  Male 126 (71.6%) 89 (67.9%) 37 (82.2%) .067  Age (yr) >.900   Mean (SD) 71.7 (17.6) 71.7 (17.6) 71.5 (17.9)   Median [min, max] 75.0 [0.02, 101] 76.0 [0.02, 100] 74.0 [10.0, 101]  Range (yr) .800   0–18 4 (2.3%) 3 (2.3%) 1 (2.2%)   18–60 28 (15.9%) 23 (17.6%) 5 (11.1%)   60–80 79 (44.9%) 58 (44.3%) 21 (46.7%)   80– 65 (36.9%) 47 (35.9%) 18 (40.0%) Conditions  Vaccinated 88 (50.0%) 63 (48.1%) 25 (55.6%) .800  PaO2/FiO2 (mm Hg) <.001   <300 52 (29.5%) 50 (38.2%) 2 (4.4%)   <200 81 (46.0%) 65 (49.6%) 16 (35.6%)   <150 43 (24.4%) 16 (12.2%) 27 (60.0%)  Clinical severity at first day .001   Asymptomatic 37 (21.0%) 25 (19.1%) 12 (26.7%)   Mild 21 (11.9%) 14 (10.7%) 7 (15.6%)   Moderate 50 (28.4%) 31 (23.7%) 19 (42.2%)   Severe 68 (38.6%) 61 (46.6%) 7 (15.6%) Underlying diseases  Cardiovascular disease 116 (65.9%) 88 (67.2%) 28 (62.2%) .500  Cerebrovascular disease 36 (20.5%) 26 (19.8%) 10 (22.2%) .700  Respiratory disease 89 (50.6%) 65 (49.6%) 24 (53.3%) .700   COPD 29 (16.5%) 21 (16.0%) 8 (17.8%) .800  Metabolic disease 77 (43.8%) 58 (44.3%) 19 (42.2%) .800  Tumor 8 (4.5%) 5 (3.8%) 3 (6.7%) .400  Dialysis 13 (7.4%) 11 (8.4%) 2 (4.4%) .500 Treatments  Respiratory support <.001   OT 88 (51.7%) 81 (61.8%) 7 (15.6%)   NIRS 45 (23.9%) 17 (13.0%) 28 (62.2%)   IMV 43 (24.4%) 33 (25.2%) 10 (22.2%)  Tracheotomy 31 (23.9%) 12 (9.2%) 19 (42.2%) <.001  ECMO 7 (4.0%) 4 (3.1%) 3 (6.7%) .400  TCM therapy 73 (41.5%) 61 (46.6%) 12 (26.7%) .019  Antiviral drug 127 (72.2%) 96 (73.3%) 31 (68.9%) .600  Antiviral drug combination .400   None 49 (27.8%) 35 (26.7%) 14 (31.1%)   Single 114 (64.8%) 88 (67.2%) 26 (57.8%)   Multiplex 13 (7.4%) 8 (6.1%) 5 (11.1%)   CRRT 16 (9.1%) 11 (8.4%) 5 (11.1%) .500  Length of hospital stay (d) 11 (8, 17) 11 (8, 16) 14 (9, 19) .065  Length of ICU stay (d) 9 (5, 14) 8 (5, 12) 14 (8, 19) <.001  ICUdischarge–T1 (d) 13 (9, 17) 12 (9, 17) 15 (11, 21) .039  Death 16 (9.1%) 4 (3.1%) 12 (27%) <.001

P-values for indicators of significance are shown in bold.

IMV = invasive mechanical ventilation, contains: ventilation orotracheal intubation and endotracheal intubation via tracheotomy; NIRS = noninvasive respiratory support, contains: high flow nasal cannula (HFNC), noninvasive positive pressure ventilators (NIPPV), Helmet NIPPV; OT = oxygen therapy, contains: nasal cannula (NC), nonrebreathing mask (NRM), and high oxygen (Hi-Ox) mask; T1 = the day of the first positive SARS-CoV-2 test.

There were no statistically significant differences between the 2 groups in terms of the SARS-COV-2 vaccination rate, the history of underlying diseases (e.g., cardiovascular disease, cerebrovascular disease, respiratory disease, metabolic disease, tumor, and history of dialysis), and the use of antiviral drugs. Statistically significant differences were found between the groups in terms of initial disease phenotype, oxygenation index, respiratory support mode, whether or not tracheotomy required, and the use of TCM (P < .05). For example, the percentage of patients with PaO2/FiO2 levels <150 was 12.1% and 60% in the nonprogressing and progressing groups, respectively (P < .001). In addition, 46.6% of the patients in the nonprogressing group used TCM adjuvant therapy, compared with only 26.7% in the progressing group (P = .019). Moreover, a significantly higher proportion of patients in the progressive group required tracheotomy (42.2% vs 0.9%, P < .001). Compared with the nonprogressing group, the average time from initial diagnosis to ICU discharge in the progressing group was 3 days longer (P = .039), and the ICU mortality rate was 23.9% higher (P < .001). Correlation analysis of statistically significant variables in univariate analyses showed a correlation between oxygenation index, respiratory support mode, and tracheostomy or not (R > 0.8) (Fig. S2, Supplemental Digital Content, https://links.lww.com/MD/L904).

3.1.2. Multivariate analysis for risk of critical illness.

We established multivariate logistic regression models to evaluate patients whose phenotype had not yet reached severe disease at the time of first diagnosis to determine the degree of influence of different factors on their subsequent progression to severe or critical disease. The expert diagnosis conclusion (severe or critical) was tested for covariance with variables such as gender, condition at first diagnosis, oxygenation index, respiratory support mode, tracheotomy, and TCM therapy. Different respiratory support modes correspond to different oxygen index levels, which is making the 3 factors of oxygenation index, respiratory support mode, and tracheotomy show collinearity (Table S2, Supplemental Digital Content, https://links.lww.com/MD/L897). Therefore, 3 multivariate regression models A, B, and C were developed to eliminate the interference of collinearity, and the ability of the models to predict whether a patient progresses to critical illness in the ICU was observed separately (Table S3, Supplemental Digital Content, https://links.lww.com/MD/L898). Finally, model A achieved the highest AUC (0.847) while possessing the lowest Akaike information criterion and Bayesian information criterion values. Multivariate analysis and subgroup forest analysis according to this model (Fig. 1A) showed that treatment with TCM was an independent protective factor for reducing the risk of ICU COVID-19 critical illness (OR = 0.20, P = .018). Persistent decline in patient oxygen index and the need for respiratory ventilation to support vital signs were independent risk factors for disease progression, with an average increase in critical illness of 26% (OR = 3.53, P < .01) for each higher level of ventilation device (OR = 3.53, P < .01). The more severe the symptoms at the time of initial diagnosis, the risk of progression to critical illness in the ICU decreases accordingly (OR = 0.36, P < .01).

Figure 1.:

Multivariate analysis and risk prediction for severe COVID-19 progressing to critical illness and death in the ICU: (A) random forest plot of meaningful variables for progressing to critical; (B) ROC curve for critical risk prediction; (C) random forest plot of meaningful variables for death; (D) ROC curve for mortality risk prediction.

In summary, the multivariate model we screened had an AUC of 0.847 (95% CI: 0.778–0.917) for predicting the progression of ICU severe care to critical care, and could achieve a specificity of 82.4% when the sensitivity was 77.8%, with a Yoden index of 0.602 (Fig. 1B).

3.2. Risk factor analysis for death in ICU patients with COVID-19 3.2.1. Univariate analysis of mortality risk.

A total of 189 patients with severe COVID-19 hospitalized in the ICU were divided into death group (n = 16) and survival group (n = 173) according to outcome, and univariate difference analysis was performed (Table 2). Among them, the average age of the death group was 77.7 years (median = 80 years), and the average age of the survival group was 70.4 years (median = 75 years). In terms of mortality, the mortality rate was 6.62% (9/136) for males and 13.21% (7/53) for females. There were no statistically significant differences between the 2 groups in terms of gender and distribution of people in different age ranges. Clearly, there was a significant difference in the PaO2/FiO2 levels in the ICU between the death group and the survival group (P < .001), with up to 75% of the dead patients having a blood oxygen index lower than 150, while only 24.3% in the survival group. There was no significant difference in the illness condition of the 2 groups at the day of the first positive test, and even 13 of the critically ill patients did not eventually die, while 45 of the 176 noncritically ill patients eventually progressed to critical illness in the ICU and 12 died (P < .001). After excluding 29 patients whose vaccination records could not be confirmed, 89 patients received at least 1 dose of the vaccine and 10 died, with a mortality rate of 11.2% (10/89), which was even higher than the mortality rate of 8.45% (6/71) in the unvaccinated group, but the difference was not significant. In terms of respiratory support, 91 patients received oxygen therapy, 45 patients received noninvasive ventilation therapy, and 53 patients received invasive ventilation therapy.

Table 2 - Univariate analysis of deaths in patients with severe COVID-19 in the ICU. Characteristics Total (N = 189) Survival (N = 173) Death (N = 16) P General characteristics  Male 136 (72.0%) 127 (73.4%) 9 (56.3%) .200  Age .200  Mean (SD) 71.0 (19.4) 70.4 (19.9) 77.7 (9.34)  Median [min, max] 76.0 [0.0200, 101] 75.0 [0.0200, 101] 80.0 [59.0, 91.0]  Range (yr) .500   0–18 7 (3.7%) 7 (4.0%) 0 (0%)   18–60 28 (14.8%) 27 (15.6%) 1 (6.3%)   60–80 82 (43.4%) 76 (43.9%) 6 (37.5%)   80– 72 (38.1%) 63 (36.4%) 9 (56.3%) Conditions  PaO2/FiO2 <.001  <300 52 (27.5%) 50 (28.9%) 2 (12.5%)  <200 83 (43.9%) 81 (46.8%) 2 (12.5%)  <150 54 (28.6%) 42 (24.3%) 12 (75.0%)  Vaccinated 89 (47.1%) 79 (45.7%) 10 (62.5%) .600 Clinical severity at first day .300  Asymptomatic 37 (19.6%) 34 (19.7%) 3 (18.8%)  Mild 21 (11.1%) 20 (11.6%) 1 (6.3%)  Moderate 50 (26.5%) 42 (24.3%) 8 (50.0%)  Severe 68 (36.0%) 64 (37.0%) 4 (25.0%)  Critical 13 (6.9%) 13 (7.5%) 0 (0%) Worst clinical severity  Critical 58 (30.7%) 46 (26.6%) 12 (75.0%) <.001 Underlying diseases  Cardiovascular disease 123 (65.1%) 115 (66.5%) 8 (50.0%) .200  Cerebrovascular disease 39 (20.6%) 35 (20.2%) 4 (25.0%) .700  Respiratory disease 95 (50.3%) 89 (51.4%) 6 (37.5%) .300  Respiratory failure 33 (17.5%) 33 (19.1%) 0 (0%) .079  COPD 31 (16.4%) 29 (16.8%) 2 (12.5%) >.900  Metabolic disease 80 (42.3%) 77 (44.5%) 3 (18.8%) .046  Tumor 8 (4.2%) 7 (4.0%) 1 (6.3%) .500  Dialysis 15 (7.9%) 15 (8.7%) 0 (0%) .600 Treatments  Respiratory support <.001  OT 91 (48.1%) 87 (50.3%) 4 (25.0%)  NIRS 45 (23.8%) 45 (26.0%) 0 (0%)  IMV 53 (28.0%) 41 (23.7%) 12 (75.0%)  Tracheotomy 42 (28.0%) 37 (23.7%) 5 (75.0%) .4  ECMO 9 (4.8%) 8 (4.6%) 1 (6.3%) .500  TCM therapy 76 (40.2%) 74 (42.8%) 2 (12.5%) .018  Antiviral drug 133 (70.4%) 126 (72.8%) 7 (43.8%) .022 Antiviral drug combination .060  None 56 (29.6%) 47 (27.2%) 9 (56.3%)  Single 119 (63.0%) 112 (64.7%) 7 (43.8%)  Multiplex 14 (7.4%) 14 (8.1%) 0 (0%)  Length of hospital stay (d) 11 (8, 17) 11 (8, 16) 13 (8, 18) .600  Length of ICU stay (d) 9 (5, 14) 9 (5, 13) 12 (6, 18) .200  ICUdischarge–T1 (d) 12 (9, 17) 12 (9, 17) 14 (11, 20) .500