Predicting clinical outcomes of SARS-CoV-2 infection during the Omicron wave using machine learning

• DOI: 10.1371/journal.pone.0290221
• Publication Date: 2024-04-25T17:43:38Z
• Authors: Steven Cogill; Shriram Nallamshetty; Natalie Fullenkamp; Kent Heberer; Julie Lynch; Kyung Min Lee; Mihaela Aslan; Mei-Chiung Shih; Jennifer S. Lee
• Keywords: Male; Aged, 80 and over; Adult; COVID-19 Vaccines; SARS-CoV-2; Science; Q; Vaccination; R; COVID-19; Comorbidity; Middle Aged; Machine Learning; Hospitalization; Medicine; Humans; Female; Longitudinal Studies; Research Article; Aged; Retrospective Studies

The Omicron SARS-CoV-2 variant continues to strain healthcare systems. Developing tools that facilitate the identification of patients at highest risk adverse outcomes is a priority. study objectives are develop population-scale predictive models that: 1) identify predictors with surge infections, and 2) predict impact prioritized vaccination high-risk groups for said outcome. We prepared retrospective longitudinal observational national cohort 172,814 in U.S. Veteran Health Administration who tested positive from January 15 August 15, 2022. utilized sociodemographic characteristics, comorbidities, status, time testing hospitalization, escalation care (high-flow oxygen, mechanical ventilation, vasopressor use, dialysis, or extracorporeal membrane oxygenation), death within 30 days. Machine learning demonstrated advanced age, high comorbidity burden, lower body mass index, unvaccinated oral anticoagulant use were important hospitalization care. Similar factors predicted death. However, did not mortality risk. all-cause model showed discrimination (Area Under Curve (AUC) = 0.903, 95% Confidence Interval (CI): 0.895, 0.911) followed by (AUC 0.822, CI: 0.818, 0.826), then 0.793, 0.784, 0.805). Assuming vaccine efficacy range 70.8 78.7%, our simulations projected targeted prevention group may have reduced 30-day more than 2 5 patients.