Our study provides new insights into prevalence, clinical characteristics and predictors of mortality in COVID-19 ICU patients with neurological manifestations. Overall, 12.7% of all COVID-19 ICU patients admitted to ICUs of the study centers during a 17-month period developed neurologic manifestations. The most common ones were encephalopathic, cerebrovascular and neuromuscular disorders, with delirium, AIS, ICH and CIP/CIM being the most prevalent diagnoses. There was hardly any evidence of COVID-related encephalitis. Overall, the occurrence of any cerebrovascular disorder was the strongest predictor for death. Cerebrovascular complications were associated with an unfavorable outcome, i.e., ICH with a 6.1-fold and AIS with a 3.9-fold increase in in-hospital mortality.
In comparison with other studies in this field, our study is the only one to focus on neurologic manifestations of COVID-19 in the ICU exclusively, while other studies evaluated broader cohorts, used different definitions of disorders and frequently did not involve neurologists for assessment. Most studies not limited to ICU patients reported higher frequencies of neurological manifestations [7, 9, 11, 15], which might be explained by inclusion of milder and/or rather non-specific symptoms like headache, anosmia or fatigue. As many ICU patients present with an altered mental status, some neurologic symptoms may have stayed undetected. The baseline and clinical characteristics of the PANDEMIC cohort are quite similar to previously described COVID-19 ICU cohorts not focusing on neurologic manifestations [1, 3, 7, 20, 21]. Hence, we assume that our results are transferable to other ICU populations. The prevalence as reported here is in line with a report by Frontera et al. describing neurological manifestations in 13.6% of all hospitalized patients . Although their study involved 78% (3,504/4,492) non-ICU patients, they captured a similar spectrum of manifestations compared to our study. In another prospective registry study focusing on non-ICU COVID patients (4.1% ICU patients), using similar categories of neurological manifestations (cerebrovascular diseases, encephalopathy, seizures and meningoencephalitis), the authors calculated a prevalence of 12.9% . A retrospective study by Helms et al.  on COVID-19-ICU patients found neurologic manifestations, mostly agitation, delirium, or other types of encephalopathy, in 67% (39/58) of patients after withdrawal of sedation. The differences might be related to different timepoints of assessment as well as to the fact that only ARDS patients were assessed and no non-neurologic “control” population; hence, the rate can hardly be regarded a prevalence. Older patients with more severe COVID-19 disease seemed to have an increased risk of neurologic manifestations. This is in line with findings of higher SOFA scores by Frontera et al. in patients with neurologic manifestations and the study of Kleineberg et al. who found more severe neurologic complications in patients reaching the critical or complicated LEOSS stage of disease (complicated stage: supplemental oxygen necessary; critical stage: use of mechanical ventilation, dialysis and/or catecholamines) [7, 9].
The total number of cerebrovascular events was comparable to previously reported ICU COVID-19 cases, but we found a higher prevalence of AIS in our ICU cohort (3.5%) compared to previous studies in broader cohorts (0.04–1.9%) [7, 9, 12, 23, 24]. A large population-based study in Sweden found a 2.1- to 6.2-fold increased risk for AIS among COVID-19 patients . In the large European LEOSS-registry, a correlation between increased prevalence of AIS and COVID-19 stage of disease was demonstrated, with COVID-19-associated coagulopathy as a hypothesis . However, previous studies in critically ill patients with diseases other than COVID-19 such as sepsis and ARDS, found similar rates of AIS [25,26,27,28,29,30,31,32]. Hence, the proposed mechanisms such as pro-coagulatory inflammation or endotheliitis on the complication AIS remains questionable, as well as the order of events.
Concerning hemorrhagic stroke, the prevalence we found (3.0%; with ICH 1.8%, and SAH 1.2%) was higher compared to most previous studies in broader hospitalized COVID-19 cohorts [33,34,35,36]. Yet, Kleineberg et al.  reported a prevalence of even 5% in patients at the critical stage. In their and our cohort ECMO and higher aPTT values were associated with a higher risk of ICH, which appears plausible . A study based on a propensity matched non-COVID cohort could not find an increased rate in hemorrhagic stroke among COVID patients . Moreover, ICH rates in other ICU cohorts including ECMO are in the same range. Hence, again, it remains unclear whether COVID-19 is an independent risk factor for hemorrhagic stroke or whether this is a consequence of the aggressive ICU treatment [18, 25, 26, 32].
The frequency of non-vascular neurologic manifestations as encephalopathic, epileptic or neuromuscular disorders shows a large variability in the preexisting literature, depending on timepoint of assessment, qualification of the investigator, definitions of disorders and control for confounders. According to our data and in line with previous reports, no specific direct pathomechanism between those manifestations and COVID-19 became obvious [19, 38]. The number of patients with positive PCR detection of the virus in CSF or brain tissue was negligible which supports previous results arguing against SARS-CoV-2-associated meningoencephalitis [7, 39, 40].
Overall ICU mortality in the preexisting literature on COVID-19 varies greatly depending on the geographic location, the period of analysis and the clinical characteristics of the analyzed cohort and ranges between 21 and 100% [41,42,43]. A German study performed in 2020 in ventilated patients reported a mortality of 38.8% . Another earlier study confirmed differences in mortality depending on age, sex and comorbidities . Based on a review article by Misra et al., a higher mortality in COVID-19-ICU patients with neurologic complications can be expected . This was also in line with registry data of 16,225 COVID-19 patients suggesting higher odds for death and worse functional outcome in patients with neurologic manifestations . Total in-hospital mortality in our preselected ICU cohort appears within the range reported in the literature. Possibly, neurologic manifestations in previously published ICU cohorts contributed to mortality, but remained undetected being masked by the clinical course being dominated by respiratory failure. Our data suggest that neurologic manifestations as a predictor of mortality differ considerably. For instance, some patients get diagnosed with CIP/CIM later in their ICU course, often after having survived the critical COVID-19 phase. On the other hand, patients with cerebrovascular complications show a much higher mortality of almost 60%, which is in line with the current literature [9, 11]. Thus, the diverse neurological manifestations in COVID-19 ICU patients may give rise to different modes of prognostication, decision-making, triage and (preventive or symptomatic) treatment.
Unfavorable outcomes at ICU discharge were noticed in a higher number of patients (mortality 36%, only 17.7% could be discharged in home care, mRS 3–5; in 67.9% of the survivors) compared to other studies, reflecting the prediction of worse outcome by some neurological diagnosis. In addition to long-term effects of COVID-19 (post-COVID-19 syndrome), ICU survivors were described to suffer an even lower quality of life due to persistent fatigue, dyspnea, sleep disturbances, and mental health issues . Similarly, post-intensive care syndrome (PICS) has been described after shock, sepsis, hypoxia, ARDS or delirium, significantly increasing the risk of long-term cognitive and physical impairment [47, 48]. Longer-term studies with follow-up and control groups are warranted to shed light on this aspect [49,50,51].
Our study has several limitations. Some of these are related to a pragmatic study design allowing for data collection by ICU physicians during a pandemic with uncertain course. First, and inherent to the ICU cohort, neurologic findings could have been missed due to sedation or other ICU measures, failure to consult a neurologist with ICU experience or difficulties performing adequate diagnostic investigations. Thus, the prevalence of neurologic manifestations might have been underestimated. Another potential compromise regarding prevalence is due to the aspect that the “control” group of those COVID-19 ICU patients without detected neurologic abnormalities was not further characterized and was not included in our regression model, which has certainly caused ascertainment bias. Moreover, the multicenter approach might increase the degree of heterogeneity. However, all sites used a standardized eCRF with precise directives and definitions to minimize variability. In addition, the electronic record was combined with a plausibility check and followed by data clearing. Another possible limitation could be associated with the inclusion period of the study (between April 2020 and September 2021), in which more than one variants of SARS-COV-2 caused at least two waves. Possible differences concerning neurologic manifestations or outcomes of different variants might have been missed. Nevertheless, a subgroup analysis comparing patients included during the first wave (April 2020–August 2021) versus patients included during the second wave (August 2021–September 2021) did not reveal any significant differences regarding prevalence or type of occurred neurological manifestations (data not shown).
Strengths of our study are its prospective nature, the focus on ICU patients as a distinct yet very relevant COVID-19 subgroup, use of a standardized eCRF, employment of consultations by experienced neurointensivists, and a large sample size given that special subgroup yielded by a multicenter approach reflecting a real-world scenario.