Donate
News and events

AI predicts hospital patients most at risk of COVID-19

Last updated

28/05/24

Share

Ashleigh Myall, one of our AMR PhD students, has developed a machine learning tool to help predict patients’ risk of contracting COVID-19 after admission to hospital.

There is a need to develop predictive models to identify patients who are most at risk of contracting COVID-19 and intervene to mitigate against poor patient outcomes. We can do this alongside the usual measures to minimise outbreaks and further transmissions.
Ashleigh Myall

Developed alongside researchers at Imperial College London and the Infection Prevention and Control (IPC) unit at Imperial College Healthcare NHS Trust, the tool was able to predict patients at high risk of developing COVID-19 with 87 per cent accuracy in a Lancet Digital Health study.

Identifying patients most at risk can help with reducing onward transmission to other patients and staff. The framework may also be used to identify the risk of other infectious diseases in the future, not just COVID-19.

Close up of hospital equipment with blurred patient in distance

Contracting COVID-19 whilst in hospital

Throughout the COVID-19 pandemic, some patients developed the infection during their hospital stay.

COVID-19 infections that have developed after admission to hospital account for 12-15% of all COVID-19 cases in healthcare settings, and reached 16.2% at the peaks of the pandemic.

Sign up to our newsletter

Keep up to date with all our latest news and events, as well as ways you can get involved in our fundraising activities.

Sign up now

12-15%

of all COVID-19 cases in healthcare settings have developed after admission to hospital

Meet the researchers

Ashleigh, lead author of the study, is an applied mathematics PhD student on our National PhD Training Programme in AMR Research. He is part of the ASPIRES consortium, which is aiming to improve the use of antibiotics in surgical settings.

Specifically, Ashleigh focuses on how pathogens (disease-causing bugs) spread in nosocomial infection (these are infections which are picked up by patients when they are in the hospital).

The new COVID-19 study was co-authored by Professor Mauricio Barahona and Professor Alison Holmes.

PhD student Ashleigh Myall
This framework could be used as part of a range of surveillance tools to enhance infection, prevention and control strategies, especially during the winter months when COVID-19 infections spread more easily.
Ash Myall

A machine learning framework

Traditionally, predictions of infections in healthcare settings have focused on age, gender, comorbidities and patients’ lengths of stay, but have not taken into account patient contacts, locations, or patient flow through the hospital.

The researchers used routine hospital and patient data from the first two COVID-19 waves. They were able to train the tool to identify risk factors such as contact with infectious patients, where beds were situated, and how patients were moved, alongside the traditional categories of age and gender.

The tool analyses these risk factors and then gives a predicted risk score between 0 and 1.

87%

of hospital COVID-19 cases studied were predicted using this tool

The team will be carrying out further work to extend the framework to include the omicron strain, other infectious diseases and understand how it could be integrated into healthcare settings.

Sid Mookerjee, co-author of the study and operational lead for antimicrobial stewardship, surveillance and epidemiology at Imperial College Healthcare NHS Trust, said the tool could "help hospital managers and clinicians design safe and effective patient care pathways and bed management, helping deliver world class care.”

The study is published in Lancet Digital Health.

Support research like this

Many of the diseases and conditions that affect us as humans have been overcome as a result of medical research. Support us today.

Support us