COVID-19 has shown a broad clinical spectrum, ranging from asymptomatic to mild, moderate, and severe infections. Many symptoms have already been identified as typical of COVID-19, but few studies show how they can be useful in identifying clusters of patients with different severity of illness. This interpretation may help to recognize the different profiles of symptoms of COVID-19 expressed in a population at certain time. The aim of this study was to identify symptom-based clusters of hospitalized patients with severe acute respiratory illness by SARS-CoV-2 in Brazil. The clusters were evaluated based on sociodemographic characteristics, admission to the Intensive Care Unit (ICU), use of respiratory support, and outcome. The Multiple Correspondence Analysis (MCA)-based cluster analysis was applied to symptoms presented before admission. Pearson’s chi-square test was used to compare the proportions of symptoms between the clusters and to examine differences in the calculated rates for the following variables - sex, age group, race, Brazilian region, use of respiratory support, admission to the ICU and outcome. Three COVID-19 clusters with distinct symptom profiles were identified by MCA-based cluster analysis. Cluster 1 had the mildest severity profile, with the lowest frequencies for most symptoms investigated. Cluster 2 had a severe respiratory profile, with the highest frequencies of patients with dyspnea, respiratory discomfort and O2 saturation< 95%. Cluster 2 was also the most prevalent in all Brazilian regions and had the highest percentages of patients who used invasive respiratory support (27.4%) (p-value<0.001), were admitted to the ICU (42.6%) (p -value<0.001) and died (39.0%) (p-value<0.001). Cluster 3 had a prominent profile of gastrointestinal symptoms. The study identified three distinct COVID-19 clusters based on the symptoms presented by patients with severe acute respiratory illness by SARS-CoV-2, but without distinction in their prevalence in the Brazilian regions.