Demographic Characteristics
The demographic characteristics of vaccinated and non-vaccinated patients are presented in Table 1. Notable differences were observed between vaccinated and non-vaccinated patients in terms of age (95% CI [1.048; 1.102]), virus variant (95% CI [2.09; 33.00]), pre-existing diseases (95% CI [2.87; 27.14]), immunodeficiency (95% CI [1.96; 12.97]), type 2 diabetes (95% CI [1.29; 6.43]), hypertension (95% CI [2.12; 7.96]), oncological disease (95% CI [2.59; 18.68]), cardiac disease (95% CI [2.50;10.26]), vascular disease (95% CI [1.48-9.27]), pulmonary disease (95% CI [1.56; 6.85]), neurological disease (95% CI [1.58; 12.18]), and organ transplant (95% CI [2.85; 70.90]). Patients in the vaccinated group were older than patients in the non-vaccinated group. In all cases, vaccinated patients were suffering from pre-existing diseases more often than non-vaccinated patients.
There were six instances of pregnancy in the non-vaccinated group and none in the vaccinated group, rendering a statistical comparison unfeasible.
Virus variant, pre-existing disease, immunodeficiency, oncological disease, vascular disease, neurological disease, organ transplant, and pregnancy are rare characteristics in our dataset and are thus not suitable as confounding variables in a large multiple logistic regression model. As a result, we utilized age, type 2 diabetes, hypertension, cardiac disease, and pulmonary disease as confounders in a comprehensive model and conducted a sub-analysis for virus variant, pre-existing disease, immunodeficiency, oncological disease, vascular disease, neurological disease, organ transplant, and pregnancy. This was undertaken to ascertain if the effect remained even when the confounding variable was removed.
Vaccination effect on incidence of ICU admissions and mortality rate
No difference was observed in mortality between vaccinated and non-vaccinated COVID-19 patients before adjusting for confounding variables (RR: 0.72 95% CI [0.36; 1.31]). Upon adjusting for confounding variables such as age, type 2 diabetes, hypertension, cardiac disease, and pulmonary disease, vaccination exhibited a minor protective effect on mortality (RR: 0.46, 95% CI [0.13; 0.99]) (Table 2, Figure 2).
Before adjusting for confounding variables a significant protective effect of vaccination on ICU admissions was observed in hospitalized COVID-19 patients (RR: 0.60 95% CI [0.40; 0.83]). The NNT was merely 4 patients. Upon adjusting for confounding variables such as age, type 2 diabetes, hypertension, cardiac disease, and pulmonary disease, the protective effect of vaccination on ICU admissions persisted (RR: 0.59, 95% CI [0.28; 0.89]) (Table 2, Figure 2).
Given the impossibility of adjusting for the rare characteristics within the study population, we conducted the same analysis in subgroups excluding patients with the omicron variant, healthy subjects, immunodeficiency, oncological disease, vascular disease, neurological disease, organ transplant, and pregnancy. For the mortality outcome, the results remained consistent, except for the groups excluding healthy subjects, oncological disease, vascular disease, and pregnancy, where the significant effect of vaccination disappeared even after adjusting for confounding effects. However, in these subgroups, the RRs were similar, only the CIs were wider (Supplementary Table 1, Supplementary Figure 1).
For the ICU admissions outcome, the results remained significant in all subgroup analyses (Supplementary Table 2, Supplementary Figure 2).
While these sub-analyses permit a statement about the general population, we cannot make a statement about the effect of COVID-19 vaccination for patients with these rare characteristics.
Patients hospitalized with COVID-19 who were admitted to the ICU were marginally younger than those who were not admitted to the ICU (OR: 0.98, 95% CI [0.96; 0.99]). However, this effect did not persist when adjustments were made for vaccination status (OR: 0.99, 95% CI [0.97; 1.01]) (Table 3).
To gain further insight, we opted to assess the impact of vaccination in younger and older patients separately. Younger patients (< 60 years) exhibited fewer ICU admissions when they were vaccinated compared to non-vaccinated patients (RR: 0.64, 95% CI [0.32; 0.99]), whereas older patients (> 60 years) did not display a significant effect of the vaccination on ICU admissions (RR: 0.68, 95% CI [0.35; 1.20]). There was no discernible effect of vaccination on mortality, neither in the younger group (< 60 years) (RR: 0.643, 95% CI [0.11; 1.45]) nor in the older group (> 60 years) (RR: 0.84, 95% CI [0.33; 2.54]) (Table 4, Figure 3).
Vaccination effect on various clinical parameters
The effects of COVID-19 vaccination on clinical characteristics of hospitalized COVID-19 patients is presented in Table 5. Figure 4 illustrates a protective effect of COVID-19 vaccination on hospitalized COVID-19 patients, which intensifies with the severity of the disease. After adjusting for confounding variables, it was observed that vaccinated patients hospitalized with COVID-19 experienced less dyspnea (RR: 0.69, 95% CI [0.36; 0.96]), required less oxygen therapy (RR: 0.54, 95% CI [0.22; 0.90]), less non-invasive ventilation (NIV) (RR: 0.59, 95% CI [0.26; 0.98]), less intubation with invasive ventilation (RR: 0.30, 95% CI [0.08; 0.62]), and less extracorporeal membrane oxygenation (ECMO) (RR: 0.29, 95% CI [0.00; 0.84]). Vaccinated patients exhibited fewer chest CT lesions (Cohen’s d: -0.38, 95% CI [-0.74; -0.02]), had fewer ICU admissions (RR: 0.59, 95% CI [0.28; 0.89]), a significant reduction in ICU duration (Cohen’s d: -0.44, 95% CI [-0.67; -0.19]), and decreased mortality (RR: 0.46, 95% CI [0.13; 0.99]). Moreover, vaccinated patients developed sepsis less frequently (RR: 0.42, 95% CI [0.08; 0.94]), had lower PCT values (Cohen’s d: -0.37, 95% CI [-0.73; -0.05]), experienced less coagulopathy (RR: 0.26, 95% CI [0.00; 0.88]), and had reduced AST (Cohen’s d: -0.40, 95% CI [-0.69; -0.12]) and ALT (Cohen’s d: -0.45, 95% CI [-0.73; -0.17]) values. In addition, vaccinated patients demonstrated higher levels of COVID-19 antibodies compared to non-vaccinated patients (RR: 1.38, 95% CI [1.08; 1.95]).
Without adjusting for confounding factors, vaccinated patients required less oxygen (Cohen’s d: -0.088, 95% CI [-0.477; -0.288]), less high-flow oxygen therapy (RR: 0.64, 95% CI [0.32; 0.98]), fewer tracheotomies (RR: 0.45, 95% CI [0.13; 0.96]), had reduced pCO2 (Cohen’s d: -0.38, 95% CI [-0.68; -0.03]), and had increased creatinine (Cohen’s d: 0.55, 95% CI [0.23; 0.83]) and urea values (Cohen’s d: 0.48, 95% CI [0.11; 0.88]). However, these effects lost their significance after adjusting for confounding variables.
To assess the risk factors associated with mortality and ICU admissions in hospitalized COVID-19 patients, we employed a regularized logistic regression using LASSO regression, incorporating all variables. This method represents a novel approach for selecting the most pertinent variables while accounting for autocorrelation and overfitting.
For the prediction of mortality, the model selected the following variables: a positive correlation was observed with type 2 diabetes mellitus, and a negative correlation was noted for vascular disease. Subsequently, these relevant variables were used to conduct a logistic regression between mortality and the patients' characteristic variables, controlling for vaccination status (Table 6). The odds for increased mortality in hospitalized COVID-19 patients were elevated for type 2 diabetes (OR: 2.51, 95% CI [1.07; 5.76]) and were reduced for vascular disease (OR: 0.25, 95% CI [0.03; 0.92]). Both effects persisted after adjusting for vaccination status (ORtype 2 diabetes: 3.02, 95% CI [1.24; 7.34] and ORvascular disease: 0.27, 95% CI [0.04; 0.99]).
For the prediction of ICU admissions, the model selected the following variables: a positive correlation was observed with type 2 diabetes and pregnancy; a negative correlation was noted with immunodeficiency, vascular disease, vaccination status and age. Subsequently, these relevant variables were used to conduct a logistic regression between ICU admission and the patients' characteristic variables, controlling for vaccination status (Table 3). The odds for increased ICU admissions in hospitalized COVID-19 patients were elevated for type 2 diabetes (OR: 2.14, 95% CI [1.01; 4.85]), pregnancy (all of the pregnant women were admitted to the ICU, therefore computing OR and CI was not possible, however, a relationship seems likely) and decreased for immunodeficiency (OR: 0.35, 95% CI [0.13; 0.85]), vascular disease (OR: 0.23, 95% CI [0.07; 0.60]) and older age (OR: 0.98, 95% CI [0.96; 0.99]). After adjusting for vaccination status, the effect of type 2 diabetes (OR: 3.09, 95% CI [1.33; 7.57]) and vascular disease (OR: 0.27, 95% CI [0.08; 0.75]) remained significant, whereas the effect of immunodeficiency (OR: 0.46, 95% CI [0.16; 1.16]) and age (OR: 0.99, 95% CI [0.97; 1.01]) were no longer significant.