Machine learning model incorporates data from 14 proteins derived from a single blood sample to predict survival in patients with COVID-19
WEDNESDAY, Jan. 19, 2022 (HealthDay News) â Using proteomics data, a machine learning model may predict survival in critically ill COVID-19 patients, according to a study published online Jan. 18 in PLOS Digital Health.
Vadim Demichev, Ph.D., from Charité-Universitätsmedizin Berlin, and colleagues studied proteomes for two independent cohorts of patients with severe COVID-19 who required intensive care and invasive mechanical ventilation. The authors quantified 321 plasma protein groups at 349 time points in an exploratory cohort of 50 critically ill patients receiving invasive mechanical ventilation.
The researchers found that the SOFA (Sequential Organ Failure Assessment) score, Charlson comorbidity index, and APACHE II (Acute Physiology and Chronic Health Evaluation) score showed limited performance in predicting COVID-19 outcomes. However, a proteomic analysis identified 14 proteins that showed different trajectories between survivors and nonsurvivors. A machine learning model that used proteomic measurements could accurately predict survival using data from the first time point at maximum treatment level, which was weeks before the outcome (area under the curve of the receiver operating characteristic, 0.81). In an independent validation cohort, the model correctly predicted survival for 18 of 19 patients and nonsurvival in all five patients who died (area under the curve of the receiver operating characteristic, 1.0).
“We were able to accurately predict survival in critically ill patients with COVID-19 from single blood samples, weeks before the outcome, substantially outperforming established risk predictors,” the authors write.
Copyright © 2021 HealthDay. All rights reserved.
