Prediction of COVID-19 mortality using machine learning strategies and a large-scale panel of plasma inflammatory proteins: A cohort study.

Abstract

Objective: To apply machine learning algorithms to generate models capable of predicting mortality in COVID-19 patients, using a large platform of plasma inflammatory mediators.

Desing: Prospective, descriptive, cohort study.

Setting: 6 intensive care units in 2 hospitals in Southern Brazil.

Patients: Patients aged > 18 years who were diagnosed with COVID-19 through reverse transcriptase reaction or rapid antigen test.

Interventions: None.

Main variables of interest: Demographic and clinical variables, 65 inflammatory biomarkers, mortality.

Results: Combinations of two or three proteins yield higher predictive value when compared to individual proteins or the full set of the 65 proteins. A proliferation-inducing ligand (APRIL) and cluster of differentiation 40 ligand (CD40L) consistently emerge among the highest-ranking combinations, suggesting a potential synergistic effect in predicting clinical outcomes. The network structure suggested a dysregulated immune response in non-survivors characterized by the failure of regulatory cytokines to temper an overwhelming inflammatory reaction.

Conclusion: Our results highlight the value of feature selection and careful consideration of biomarker combinations to improve prediction accuracy in COVID-19 patients.