Frequency and burden of neurological manifestations in hospitalized patients with COVID-19: ndings from a large Brazilian cohort

Scientic data regarding the prevalence of COVID-19 neurological manifestations and prognosis in Latin America countries is still lacking. Therefore, the study aims to understand neurological manifestations of SARS-CoV 2 infection in the Brazilian population and its association with patient outcomes, such as in-hospital mortality. This study is part of the Brazilian COVID-19 Registry, a multicentric COVID-19 cohort, including data from 37 Brazilian hospitals. For the analysis, patients were grouped according to the presence of self-reported vs. clinically-diagnosed neurological manifestations and matched with patients without neurological manifestations by age, sex, number of comorbidities, hospital, and whether or not patients ha neurological underlying disease. with clinically dened neurological syndromes presented a poorer prognosis for the disease when compared to matched controls.


Introduction
The coronavirus disease  pandemic has affected millions of people worldwide. Clinical signs of upper respiratory tract infection such as nasal congestion and cough, alongside systemic symptoms like fatigue and fever usually precede lung involvement. 1 Besides the severity of respiratory symptoms, risk factors associated with worse clinical outcomes include, older age, male sex, baseline comorbidities (e.g. diabetes mellitus, chronic kidney disease, cerebrovascular disease, hypertension and obesity), and abnormal laboratory biomarkers. 2,3 COVID-19 can also evolve with cardiac, renal, ophthalmologic, skin, and other manifestations. Several reports have described a series of neurological manifestations associated with COVID-19. 4 Both peripheral and central nervous systems may be affected, with a wide range of symptoms, signs and syndromes. 4,5,6 Importantly, the presence of neurological signs and/or syndromes has shown to be associated with ve times higher risk of in-hospital death in a large cohort, even when adjusting for site, age, sex, race, and ethnicity, but this nding has not been consistent across studies. 7,8 Additionally, some manifestations are related to persistent disability, potentially associated with long-term care needs and high health, social, and economic costs. 9 Despite the epidemiological and clinical relevance of the matter, data on the prevalence of those manifestations and their prognosis in Latin American patients is still lacking Therefore, this study aimed: (i) to characterize the spectrum of neurological manifestations among Brazilian patients hospitalized with COVID-19; and (ii) to investigate the potential association between neurological manifestations and clinical outcomes, speci cally in-hospital mortality.

Study design and subjects
This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 10 This was an urgent public health research study in response to a Public Health Emergency of International Concern.
Patients were selected from the Brazilian COVID-19 Registry, a prospective multicenter cohort project with 37 participant hospitals in 17 cities from ve Brazilian states (Minas Gerais, Pernambuco, Rio Grande do Sul, Santa Catarina, São Paulo).
Details of the cohort were published elsewhere. 11 The study was approved by the National Commission for Research Ethics (CAAE 30350820.5.1001.0008). Individual informed consent was waived by the National Commission for Research Ethics owing to the pandemic situation and the use of deidenti ed data, based on medical chart review only.
The cohort study included consecutive patients with con rmed COVID-19 diagnosis according to World Health Organization guidance who were hospitalized in one of the participating centers from March to September 2020. 12 For the present study, patients aged 18 or older were selected and categorized according to the presence or not of neurological manifestations upon hospital presentation, as de ned below. Patients who developed rst COVID-19 symptoms while hospitalized for other conditions were not included in this analysis.

Data collection
Study data were collected by trained hospital staff using Research Electronic Data Capture (REDCap) tools. 13 Medical records were reviewed to collect data on patients' demographic and clinical characteristics, including age, sex, pre-existing medical conditions and home medications; COVID-19 symptoms at hospital presentation; clinical assessment upon hospital admission, third and fth admission days; laboratory, imaging, electrocardiographic data; inpatient medications, treatment and outcomes.

Statistical analysis
Categorical data were presented as absolute numbers and proportions, and continuous variables were expressed as medians and interquartile ranges. Fisher Exact test was used to compare the distribution of categorical variables, and the Wilcoxon-Mann-Whitney or Kruskal-Wallis test for continuous variables. In the case of statistically signi cant results in Kruskal-Wallis test, we conducted Dunn's test with Bonferroni correction to determine which groups are different.
As categorizing patients according to the presence of neurological manifestations would lead to groups with different age distribution and given that age is a prognostic factor in COVID-19, propensity score analysis through nearest neighbor matching (within 0.25 standard deviations of the logit of the propensity score, on a scale from 0-1.00) was performed to speci cally investigate the impact of neurological manifestations. Propensity score model was estimated by a logistic regression model using the MatchIt package in R software.
The analysis was conducted in four different subsamples: i) Any neurological manifestation, sex, age, number of comorbidities (hypertension, diabetes mellitus, obesity, coronary artery disease, heart failure, atrial brillation or utter, cirrhosis, chronic obstructive pulmonary disease, cancer, and previous stroke), admitting hospital, and past history of neurological disease ; ii) Any neurological manifestation, sex, age, number of comorbidities and admitting hospital; iii) Any clinically-de ned neurological syndrome, sex, age, number of comorbidities, admitting hospital, and past history of neurological disease; iv) Any clinicallyde ned neurological syndrome, sex, age, number of comorbidities and admitting hospital, without taking into account past history of neurological disease.
Results were considered statistically signi cant if the two-tailed p-value was lower than 0.05. All statistical analysis was performed with R software (version 4.0.2).

Ethics
The study was approved by the National Commission for Research Ethics (CAAE 30350820.5.1001.0008). Individual informed consent was waived by the National Commission for Research Ethics owing to the pandemic situation and the use of deidenti ed data, based on medical chart review only.

Results
This study involved 7,232 patients hospitalized with COVID-19 in Brazil. The median age was 60.0 (47.0-72.0) years and 53.8% were male. Neurological manifestations were present in 38.2% of the patients, being 27.8% exclusively with self-reported neurological manifestations and 9.9% with exclusively clinically-de ned neurological syndrome. Less than 2% (1.2%) of the patients presented both self-reported symptoms and clinically-de ned neurological syndrome. Headache was the most common self-reported neurological symptom (19.3%), followed by ageusia (10.4%) and anosmia (7.4%). Regarding clinically-de ned neurological syndromes, acute encephalopathy was the most commonly diagnosed, affecting 10.5% of patients. Other neurological syndromes were much less frequent (Table 1). When comparing sociodemographic and clinical characteristics in groups with clinically-de ned or self-reported neurological manifestations versus no neurological manifestations, differences emerged (Table 2). Men and younger patients were more likely to present self-reported neurological manifestations, while women and older patients had clinically-de ned neurological syndromes. Patients with clinically-de ned neurological syndromes had a higher prevalence of hypertension, coronary artery disease, heart failure, atrial brillation utter, chronic obstructive pulmonary disease, cerebrovascular disease, epilepsy and dementia when compared to patients with self-reported neurological symptoms and no neurological symptoms. All of those comorbidities, except for coronary artery disease and epilepsy, were found to be signi cantly lower in the group with selfreported symptoms when compared to patients with no neurological manifestations.
Regarding clinical ndings upon hospital admission, the median peripheral oxygen saturation/inspiratory oxygen fraction (SF) ratio and systolic blood pressure median were signi cantly lower among the clinically-de ned neurological syndrome group when compared to patients with no neurological manifestations and the ones with self-reported neurological symptoms. A higher frequency of Glasgow score below 15 was more common in patients with clinically de ned neurological syndrome (81.4%) than the other ones ( Table 2). Patients with clinically-de ned neurological symptoms had lower median hemoglobin level, as well as higher median white blood cell count, lactate, C-reactive protein, urea, creatinine and sodium levels ( Table 2). Data is presented as frequency (%) or median (IQR) * In case patients had a clinically-de ned neurological syndrome and a self-reported symptom, they were analyzed in the clinically-de ned neurological syndrome group. ** P-values for each comparison among individual groups are shown in Table S1.  (Table 3).
When comparing patients with clinical manifestations overall (self-reported or clinically-de ned) to matched controls, there was no difference in mechanical ventilation or intensive care unit requirement, and no difference in the proportion of patients who died (Table S2).

Discussion
To the best of our knowledge, this is the rst cohort study to systematically investigate COVID-19-related acute neurological manifestations and their impact in a representative sample of hospitalized patients from Brazil/Latin America. Previous Brazilian and Latin American studies have reported cross-sectional case-series, usually focusing on speci c neurological manifestations and investigating pathophysiological processes instead of assessing the whole picture and the prognostic impact. 14,15,16 Our results showed that approximately 40% of the patients admitted to a hospital due to COVID-19 presented self-reported or clinically-diagnosed neurological symptoms and/or syndromes upon hospital presentation. More importantly, presence of clinically-de ned neurological symptoms was associated with worse clinical outcomes, including the need for ICU admission, ventilatory support and death.
The incidence of COVID-19, as well as its complications and mortality rates differ substantially depending on the region/country. These differences seem to be related to various factors, including political decisions regarding social distancing, organization of health care delivery, and epidemiological characteristics of the affected population (e.g., age composition, comorbidities). 17 The COVID-19 pandemic has exerted a particularly devastating impact on Brazil that has exhibited the second highest COVID-19 related mortality number. 18 Therefore, an in-depth analysis of clinical manifestations and factors associated with worse prognosis among Brazilians is of utmost importance.
Previous studies comprising case series and/or cohorts using different de nitions and clinical samples led to very diverse incidence estimates of COVID-related neurological manifestations. To provide more reliable and/or generalizable information on the incidence, type, and outcomes of neurological manifestations among patients a recent systematic review analyzed 350 studies, involving 145,721 patients. 8 When considered only the subanalysis of hospitalized patients, the observed incidence of ageusia in our study was inside the con dence interval ( It has been hypothesized that patients with severe COVID-19 might not be able to provide a clear history regarding smell or taste impairment. 8 This could have impacted our results, especially when taking into account that the overall mortality previously observed in our cohort (22.0%) was observed to be higher than what was observed in other countries. 11 Ageusia and anosmia have been regarded as independent positive prognostic factors of a less severe COVID-19 infection. 19,20 In the aforementioned meta-analysis, patients with severe COVID-19 were less likely than those with mild disease to have decreased smell (OR 0.44, 95% CI 0.28-0.68) and taste (OR 0.62, 95% CI 0.42-0.91). 8 The frequency of headache was much higher in an European multicenter study (44%), but the study was limited for having included not only hospitalized patients, but also medical doctors and nurses with COVID-19 who worked in the participant hospitals and volunteered. 21 Those who volunteered might have had more symptoms than the other patients. Our frequency of dizziness was much lower than this study and the meta-analysis as well (1.4% vs. 7% [95%CI 6 to 9%]). 8 Altogether, our ndings and these studies support that self-reported neurological symptoms are relatively frequent during COVID-19, but the numbers vary signi cantly depending on the sample studied and the severity of the presentation.
Acute encephalopathy was the most common clinically-de ned neurological syndrome (10.5%) in our study, similar to the one reported in the aforementioned systematic review. 8 This review showed that 1 in every 3 hospitalized older patients with COVID-19 had delirium compared to 5% of younger adults. In addition, acute encephalopathy has shown to be a risk factor for mortality after one year of hospitalization and for the development of dementia. 8,22 Actually, the World Health Organization has alerted clinicians about the importance of implementing measures to prevent acute encephalopathy or delirium, as well as its prompt identi cation and management. 23 The Global Consortium Study of Neurologic Dysfunction in COVID-19 (GCS-NeuroCOVID), and the European Academy of Neurology (EAN) Neuro-COVID Registry (ENERGY) worked together producing a joint report from four cohorts 7 , and observed a higher prevalence of neurological manifestations overall. The presence of clinically-de ned syndromes, but not self-reported symptoms, were associated with worse outcomes, i.e. increased risk of in-hospital death, as observed in our study. It is worth noticing that despite meaningful results and aiming at a global representativeness of COVID-19 neurological impact, both GCS-NeuroCOVID and ENERGY cohorts clearly had a skewed composition of developed countries in North America and Europe. 7 Additionally, in some of those cohorts only patients with neurological manifestations were eligible. Therefore, the overall incidence of neurological manifestations was overestimated.
The pooled prevalence of stroke in the aforementioned systematic review of neurological manifestations of COVID-19 was 2% (95% CI 1-2%, but with high heterogeneity, I 2 = 86%), a number ten times higher than the one observed in our cohort (0.2%). This might be partially explained by the fact that our cohort probably included more severe cases of COVID-19 leading to a higher mortality rate. Another hypothesis that could explain this difference is that in our study we analyzed neurological manifestations at hospital admission, meanwhile stroke may be presented during disease course. It is worth mentioning that the diagnosis of stroke can be overlooked in critically-ill patients, especially those requiring sedation for ventilation support. 23 The pathogenesis of these neurological manifestations is still under investigation, and it seems to involve different mechanisms for distinct signs or symptoms. 9 Overall, while a direct role of CNS infection remains controversial, hypoxemia, hypovolemia, in ammatory and/or immune-mediated damage are very likely to play relevant roles. For example, patients with severe COVID-19 probably have delirium because of a 'cytokine storm', i.e. marked increase of circulating levels of proin ammatory cytokines, such as interleukins (IL-1ß, IL-6, IL-10) and tumor necrosis factor (TNF-α). TNF-α can cross the bloodbrain barrier, activating microglia and astrocytes that secrete several mediators able to interfere in neuronal functioning. 24,25,26 In a recent meta-analysis assessing the impact of neurological manifestations on COVID-related mortality and involving 21 studies, the authors observed a higher mortality in this group of patients compared to the mortality among patients with any neurological manifestations in the current analysis (18.3% vs. 27% [95% CI 19-35%]). 8 Some of the studies included in the meta-analysis only assessed clinically-de ned neurological syndromes, what could explain the higher mortality rate. In our study, when assessing patients with any neurological manifestations when compared to matched controls, there were no signi cant differences in the assessed outcomes. Of note, relevant differences emerged when considering the group of patients with clinically-de ned neurological syndromes. When comparing patients with clinically de ned neurological syndromes with matched patients without neurological manifestations, we observed a higher incidence of ICU requirement, septic shock and mortality, regardless of past history of neurological diseases. Importantly, patients with neurological manifestations and a past history of neurological diseases had a higher frequency of mechanical ventilation requirement and acute heart failure compared to the matched controls. This novel information may be useful to clinicians and healthcare managers, alerting to the need of careful neurological follow-up of these patients who may need more intensive clinical care and possibly should be prioritized for an ICU bed.
While this study has several strengths, including its sample size, careful characterization of neurological manifestations, control for multiple confounding variables, and representativeness of multiple Brazilian regions ensuring the diversity of the population studied, it has limitations that must be acknowledged. First, the study is subjected to the drawbacks inherent to data retrospectively obtained from medical record reviews. To minimise that, research staff was extensively trained and the data was subject to periodic auditing to ensure data quality. Another limitation in the analysis of prevalence of neurological manifestations is the fact that we had to exclude patients who were admitted on mechanical ventilation (for being attended rst by the emergency medical service, or being transferred from another institution without any information about self-reported neurological symptoms or clinically de ned neurological syndromes before intubation). Additionally, the pragmatic design of the study implies that it was not possible to control for interexaminer reliability in neurological examination and diagnosis. The severity of self-reported neurological symptoms could not be determined, and relevant information (e.g. neuroimaging results) was not available in all sites. Furthermore, participant hospitals were not randomly selected, and not necessarily representative of the whole healthcare system in Brazil.

Conclusion
In conclusion, our ndings in a large Brazilian cohort corroborate the emerging view that neurological manifestations represent a signi cant risk of morbidity in COVID-19 patients. More importantly, the development of clinically-de ned neurological Supplemental Tables   Table S1. Comparison between groups and p-values