Machine learning model uses blood plasma proteins to predict survival of COVID-19 patients


A single blood sample from a critically ill COVID-19 patient can be analyzed by a machine learning model that uses blood plasma proteins to predict survival, weeks ahead of the outcome, according to a new study published this week in open access journal Digital Health PLOS by Florian Kurth and Markus Ralser of Charité – Universitätsmedizin Berlin, Germany, and their colleagues.

Healthcare systems around the world are struggling to accommodate high numbers of critically ill COVID-19 patients who need special medical care, especially if they are identified as high risk. Clinically established risk assessments in critical care medicine, such as SOFA or APACHE II, show only limited reliability in predicting future disease outcomes for COVID-19.

In the new study, researchers investigated the levels of 321 proteins in blood samples taken at 349 time points from 50 critically ill COVID-19 patients treated at two independent healthcare centers in Germany and Austria. A machine learning approach was used to find associations between measured proteins and patient survival.

15 of the patients in the cohort died; the average time between admission and death was 28 days. For patients who survived, the median hospital stay was 63 days. The researchers identified 14 proteins that, over time, changed in opposite directions for patients who survived versus patients who did not survive intensive care. The team then developed a machine learning model to predict survival based on a single point measurement of relevant proteins and tested the model on an independent validation cohort of 24 critically ill COVID-10 patients. The model demonstrated high predictive power on this cohort, correctly predicting outcome for 18 of 19 patients who survived and 5 of 5 patients who died (AUROC=1.0, P=0.000047).

The researchers conclude that blood protein tests, if validated in larger cohorts, may be useful both in identifying patients at highest risk of mortality, as well as in testing whether a given treatment alters the projected trajectory of an individual patient.


Journal reference:

Demichev, V. et al. (2022) A proteomic survival predictor for COVID-19 patients in intensive care. Digital Health PLOS.


Comments are closed.