Identifying COVID-19 patients’ clinical phenotypes based on characteristics and comorbidities, as well as their differences, helps in terms of clinical care and potential crises. Our goal is to use Latent Class Analysis (LCA) to identify COVID-19 patient profiles based on demographics, symptoms, and comorbidities, and evaluate their correlation with ICU admission, hospitalization duration, and mortality in a cross-sectional study. We included hospitalized patients with positive SARS-CoV-2 tests in two referral hospitals in the south of Iran between January 2020 and July 2021. Data from 7,968 patients were evaluated in two time-specific intervals, known as “stratum,” based on the point at which the COVID-19 fourth peak and vaccinations were nearly initiated. After applying LCA, three classes identified the most at-risk individuals, including elderly women with hypertension, diabetes, and heart disease, and elderly men with heart and lung conditions, compared to two classes of healthier, younger individuals. Hospital stays were longer in the first stratum compared with the second (P-value = 0.043). ICU admission during the first stratum was significantly higher than in the second stratum (P < 0.001) and lower in classes one and two in the first stratum compared to class three (reference class). Mortality rates were not significantly different between strata (P = 0.054). Hospital stays and mortality rates significantly differed between the classes in each first and second stratum (P < 0.001). Using LCA to investigate COVID-19 patients has provided valuable insights into different patterns and outcomes of the disease. Patient variables, including age, gender, and comorbidities, were associated with greater mortality and ICU admission, which can assist in resource allocation and vaccination planning.