Background: Identification of distinct clinical phenotype can lead to personalized patient care. This study aimed to cluster hospitalized patients with COVID-19 pneumonia using unsupervised machine learning approach.

Methods: This study included hospitalized patients with COVID-19 pneumonia from July to September 2021. The clustering analysis was performed based on clinical and laboratory data. The characteristics for each cluster were identified by standardized mean difference. The risk of intubation and mortality within 90 days were compared between clusters using Cox proportional hazard analysis.

Results: Of 538 hospitalized patients with COVID-19 pneumonia, three clinically distinct clusters were identified. Cluster 1 (N=27) was characterized by male and significant elevation of serum liver enzymes and LDH levels. Cluster 2 (N=370) was characterized by lower chest x-ray score and higher serum albumin levels. Cluster 3 (N=141) was characterized by older age, history of diabetes melitus, higher chest x-ray score, more severe vital signs, higher creatinine levels, lower hemoglobin levels, lower lymphocyte counts, higher CRP, higher D-dimer, and higher LDH levels. When compared to cluster 2, cluster 3 was associated with increased risk of 90-day mortality (HR, 6.24; 95% CI, 2.42-16.09) and 90-day intubation (HR, 5.26; 95% CI 2.37-11.72). While survival in cluster 1 was 100% and intubation rate was non-significant increase when compared to cluster 2 (HR, 1.40, 95% CI, 0.18-11.04)

Conclusions: Our study characterized hospitalized patients with COVID-19 pneumonia to 3 clinically distinct phenotypes with different risk for respiratory failure and mortality.