Skip to navigation Skip to main content Skip to footer

Approved Research

Use of machine learning and artificial intelligence to predict the outcomes of SARS-CoV-2 positive patients from pseudo-anonymised human data and to use as a prognostic tool of severity and fatality.

Principal Investigator: Dr Louise Mackenzie
Approved Research ID: 65571
Approval date: March 7th 2022

Lay summary

Aims: To help hospitals identify patients most at risk of dying from COVID-19 when they arrive at hospital.

Scientific Rationale: 

Identifying people who are at high risk of not recovering from the COVID-19 virus is crucial to help increasing the number of people who survive. Some groups of people are more likely to suffer from COVID-19 than others. These people need to be identified so that they can receive better treatment sooner and have a better chance of recovering.

We have previously used data from patients who had their blood taken for testing by the hospital and were well enough to remain in the community. Our methods were able to predict individuals who were infected COVID-19 virus, and those who were not infected using routine blood tests taken in hospital, without knowing other symptoms or details about these individuals. By comparing to patients from a time before the disease existed, we can find key patterns in the biomarkers that are specific to COVID-19. We wish to further these findings to identify early stage or asymptomatic cases to help prevent further waves of the pandemic.

We have developed a new, free and simple calculator for doctors that can help them identify quickly and easily who needs most support early on in the infection. We need to check that this calculator is accurate and user friendly, and so need the data to check this. In addition, we will go onto further develop our calculator to help doctors better understand how the disease is progressing.

 Our research published in International Immunopharmacology Journal (Elsevier) https://doi.org/10.1016/j.intimp.2020.106705 shows how machine learning and artificial intelligence models, are used on full blood counts to a high level of accuracy. We have two more papers ready to get published as well as a web-based App that can be used already.