COVID-19 AI Toll Determines Patients Most At Risk of Dying from the Virus

A tool that uses artificial intelligence has recently been developed by a team of scientists in response to the pandemic that has been hitting the world hard for more than two years now.

As indicated in a EurekAlert! report the researchers developed and validated an algorithm to help medical professionals determine who is most at risk of dying from COVID-19 when hospitalized.

The tool, which employs AI, could help doctors direct critical care resources to those who need them most and be particularly valuable to resource-limited countries.

According to David Gomez-Varela, the senior author and leader of this international project, the appearance of new SARS-CoV-2 strains, waning immune protection, and relaxation of mitigation measures mean surges of infection hospitalizations are likely to continue to be seen.

A Need for AI Tools

Gomez-Varela, also the former Max Planck Group Leader and the current Senior Scientist at the Division of Pharmacology and Toxicology at the University of Vienna in Austria, added that there is a need "for clinically valuable and generalizable triage tools" to assist the hospital resources' allocation for COVID-19, specifically in areas where resources are scarce.

However, such tools need to cope with the constantly changing scenario of the global pandemic, not to mention need to be easy to implement.

To develop such a tool, the researchers used biochemical data from routine blood draws carried out on almost 30,000 patients hospitalized in more than 150 hospitals in Spain, the United States, Bolivia, Argentina, and Honduras between March 2020 and February 2022.

This means the team was able to collect data from people who have different immune statuses like vaccinated and unvaccinated, those who have natural immunity, and those infected with each SARS-CoV-2 variant from the disease that occurred in Wuhan, China, to the most recent Omicron strain.

AI-Based Prediction Models

Lead author Riku Klén, a Finland-based University of Turku associate professor, explained that the intrinsic variability in such a "diverse dataset is a great challenge for AI-based prediction models.

Called the COVID-19 Disease Outcome Predictor or CODOP, the resulting algorithm uses measurements of 12 blood molecules that are typically collected during admission. This means that the predictive tool can be easily incorporated into any hospital's clinical care.

CODOP was developed with several processes; first, using data from patients hospitalized in over 120 hospitals in Spain to "train" the AI system to predict a poor prognosis' hallmarks.

The next step was to guarantee that the tool worked regardless of the patients' immune status or COVID-19 variant, so they tested the algorithm in various subgroups of geographically dispersed patients.

Furthermore, the tool still performed efficiently when predicting the risk of in-hospital death during this inconsistent or unstable pandemic scenario, suggesting that the measurement the tool is based on are meaningful biomarkers of whether a COVID-19 patient is possibly deteriorating.

Algorithm That Can Predict Death or Survival

To test if the time for taking blood tests affects the performance of the tool, the team compared data from different time points of blood drawn before patients either died or recovered.

The researchers of the study published in eLife discovered that the algorithm can predict the death or survival or death of hospitalized patients who have high accuracy until nine days before any of the outcomes occur.

Gomez-Varela explained that the performance of CODOP "in diverse and geographically dispersed patient groups and the ease of use suggest it could be a valuable mechanism in the clinic, specifically in reduce-limited nations.

Related information about fighting COVID-19 with AI is shown on Mount Sinai Health System's YouTube video below:

Check out more news and information on COVID-19 and Artificial Intelligence in Science Times.