While overall COVID-19 mortality has declined over the course of the pandemic, largely due to vaccination campaigns and improved treatment strategies, it remains unclear whether these advancements have translated into improved survival for critically ill COVID-19 patients on the ICU. Our study directly addresses this knowledge gap by specifically focusing on ICU patients, a population at persistently high risk of mortality despite evolving management approaches. By analyzing ICU admissions across two distinct pandemic periods, we assessed whether 30-day mortality in critically ill COVID-19 patients decreased over time, independent of changes in disease severity or treatment protocols.

Notably, despite improved clinical experience and the widespread availability of vaccines in period 2 (from June 2021 to January 2022), mortality rates remained unchanged between the two periods, suggesting that critical illness from COVID-19 continues to carry a substantial fatality risk, irrespective of broader public health improvements. This finding contrasts with epidemiological studies that have shown a significant reduction in overall COVID-19-associated mortality over time⁹. These studies attributed this trend mainly to factors such as the global vaccination campaign and novel treatment options, which mainly contributed to preventing the occurrence of a severe to critical course of COVID-19^18,19,20. However, our data compellingly show that these factors did not impact mortality once COVID-19 progressed into a critical course. Therefore, we must acknowledge that vulnerable subgroups of patients remain prone to fatal complications caused by SARS-CoV-2 infections. As this finding may initially seem controversial, our focus was on a detailed description of the two cohorts to investigate whether there were any specific factors contributing to this observation.

A priori, we expected and hypothesized that patients in the later phase (period 2) of the pandemic would exhibit lower mortality rates also for critical COVID-19 itself due to increased clinical experience and clearer evidence-based treatment approaches for managing these patients^21,22. This expectation is strengthened by the fact that patients in our study demonstrated higher disease severity both at the time of admission and throughout their ICU stay during the early phase (period 1), as evidenced by significantly higher SOFA scores and Murray Lung Injury Scores at admission and their peak levels. Additionally, a greater proportion of patients in the early phase required vasopressors upon ICU admission compared to those in the later period. And finally, time between the onset of symptoms and ICU admission also differed significantly, being shorter in the later phase of the pandemic. Here, two factors might be responsible for this observation. First, it is known that the dynamics of the disease changed, with a shorter incubation period as the pandemic progressed^23,24. Additionally, this may reflect a successful strategy for early detection and increased vigilance regarding COVID-19 symptoms²⁵. On the other hand, it is important to consider that during the early phase of the pandemic, there was significant concern about shortages in ICU capacity^4,5. Consequently, hospital and ICU admissions were often delayed, which aligns with the observation that patients in the early phase of the pandemic exhibited greater disease severity upon ICU admission, with higher SOFA scores and more frequent need for vasopressors. Nevertheless, this did not translate into an association with 30-day survival.

So, at first glance, our results appear contradictory. One would expect a reduction in disease severity and mortality, particularly given the widespread availability of the vaccine, but our findings did not support this expectation. However, our study’s observation of consistent 30-day ICU mortality rates across different periods aligns with the French study, which also reported stable ICU mortality despite changes in treatment protocols²⁶.

Regarding immunization, a nuanced interpretation is required. While vaccination was not available during the early phase and population immunity was low, the situation changed in December 2020 with the introduction of vaccines²⁷. Interestingly, 81% of patients in the later period of our study (June 2021 to January 2022) remained unvaccinated. Meanwhile, by mid of 2021, less than 40% of the general population in Germany and Austria were still not vaccinated²⁸. Therefore, our cohort shows a dramatic overrepresentation of unvaccinated patients among critically ill COVID-19 cases. When examining the subgroup of non-immunized, our study found that unvaccinated patients in period 2 were significantly younger than vaccinated patients, raising the question of whether this shift reflects the protective impact of vaccination in preventing severe disease among higher-risk populations or is a result of targeted immunization policies²⁹. The demographic shift towards younger, unvaccinated patients in later periods mirrors findings from other studies, indicating that vaccination efforts initially prioritized older and high-risk populations, thereby reducing their representation among critically ill patients in subsequent waves^30,31. Thus, the overrepresentation of younger, unvaccinated patients in period 2 aligns with broader epidemiological trends showing that COVID-19 ICU admissions became increasingly concentrated in non-immunized individuals as vaccination programs progressed. By mid-2021, a substantial proportion of the general population in Germany and Austria had received at least one vaccine dose. However, our study cohort demonstrates that the vast majority of critically ill COVID-19 patients in ICUs remained unvaccinated, emphasizing that vaccine hesitancy and lack of immunization were major risk factors for severe disease requiring ICU admission.

Early vaccination campaigns in Germany and Austria also prioritized high-risk groups, including older adults and individuals with pre-existing conditions, particularly those with malignancies and immunosuppressive disorders. This prioritization likely contributed to the observed age difference, as these individuals were more likely to be vaccinated and still at risk of severe COVID-19 despite immunization. Of note, vaccinated patients who nevertheless developed a critical course of COVID-19 showed a dramatically higher prevalence of oncological comorbidities. Thus, patients suffering from malignancies, particularly haemato-oncological patients, represent a subgroup that seems particularly at risk despite vaccination. The crucial question of whether this is due to reduced vaccine efficacy from inherently low immunocompetence or evoked by other co-factors associated with the malignant disease is beyond the scope of this study. Therefore, the immunological profiles of patients with a higher risk of experiencing critical COVID-19 should be explored in future research to address this important question. Additionally, in the context of future pandemics and public health interventions, this group requires special protection despite the availability of effective vaccines.

Summarizing, our findings suggest that two distinct phenomena shaped ICU admissions in the later phase of the pandemic: (1) a protective shift, where vaccination reduced the likelihood of severe disease in lower-risk populations, leading to fewer ICU admissions in vaccinated individuals, and (2) a selective effect, where vaccinated patients who nevertheless developed critical illness were predominantly those with underlying malignancies or immunosuppression, reflecting their persistent vulnerability despite immunization.

In summary, our data indicate that once COVID-19 progresses to a critical stage, advances in treatment protocols and availability of novel therapeutics have not translated into improved survival rates in critical COVID-19. The persistently high mortality rate in unvaccinated patients, as they are highly comparable to unvaccinated patients in period one, suggests that once crossed an immunological or pathophysiological threshold, elaborate intensive care treatments are still less effective. This is similar to bacterial sepsis and addressing this challenge may require a phenotyping approach or systemic immunological characterization^32,33. Such detailed analysis could uncover critical insights into the mechanisms driving severe outcomes and identify novel targets for personalized therapeutic interventions.

Limitations

This study has some limitations. The generalizability of our findings is constrained by the fact that the study was conducted across a limited number of centers in Germany and Austria, which may not be representative for lower income countries and regions. Additionally, despite efforts to standardize treatment across study centers, variations in the management of critical COVID-19 cases, such as the timing and application of mechanical ventilation and administration of medications, were not fully standardized between the different centers. As this is a retrospective study, the accuracy of our data relies heavily on the quality of documentation in medical records, and potential inaccuracies or missing data could affect the validity of our findings. Furthermore, there may be unmeasured confounding factors, such as socioeconomic status, pre-existing health conditions not captured in the study, and variations in supportive care, that could influence the outcomes observed. In addition, one key limitation is the sample size, particularly the smaller number of patients in period 2 (131 vs. 316 in period 1), which may have reduced statistical power and the ability to detect subtle mortality trends. This imbalance reflects the natural decline in ICU admissions as vaccination rates increased and case severity decreased. Additionally, the modest cohort size limits the generalizability of our findings to broader ICU populations. Nonetheless, our study provides valuable real-world data on persistent ICU mortality in critically ill COVID-19 patients. Finally, we stopped data collection with the emergence of the Omicron variant^15,34. Therefore, assessing the impact of all virus variants following Delta is beyond the scope of this study.