Photo Credit: peterschreiber.media

The following is a summary of “Identification of patient demographic, clinical, and SARS-CoV-2 genomic factors associated with severe COVID-19 using supervised machine learning: a retrospective multicenter study,” published in the January 2025 issue of Infectious Disease by Nirmalarajah et al. 

The severity of COVID-19 impacted by a range of viral and host-related factors, including demographics, pre-existing conditions, and genetics, complicating the prediction of clinical outcomes across different severe acute respiratory syndrome coronavirus (SARS-CoV-2) variants.  

Researchers conducted a retrospective study to establish a dataset COVID-19 patient with linked clinical and viral genomic data to examine associations between SARS-CoV-2 genomic signatures and clinical disease phenotypes.  

They recruited adult patients with laboratory-confirmed SARS-CoV-2 from 11 healthcare institutions in the Greater Toronto Area (GTA) between March 2020 and April 2022. Supervised machine learning (ML) models were developed to predict hospitalization using SARS-CoV-2 lineage-specific genomic signatures, patient demographics, symptoms, and pre-existing comorbidities. The relative importance of these factors was evaluated.  

The results showed complete clinical and viral genome data were obtained from 617 patients, with 50.4% hospitalized. Hospitalized patients were older, with a mean age of 66.67 years (SD ± 17.64 years), compared to 44.89 years (SD ± 16.00 years) for outpatients. SHapley Additive exPlanations (SHAP) analyses identified underlying vascular disease, pulmonary disease, and fever as key factors linked to hospitalization. Models based on amino acid sequences of spike, nucleocapsid, ORF3a, and ORF8 proteins revealed that variants preceding variants of concern (VOCs) were associated with hospitalization.  

Investigators concluded that both clinical and viral genomic data provided valuable insights into factors influencing COVID-19 severity; clinical features demonstrated stronger predictive power for hospitalization compared to viral genomic characteristics across the diverse SARS-CoV-2 landscape. 

Source: bmcinfectdis.biomedcentral.com/articles/10.1186/s12879-025-10450-3