CURIAL: An AI System to Detect COVID-19 in Emergency Department Patients Using Routine Blood Tests and Vital Signs
In July 2020, researchers at the University of Oxford, led by Dr Andrew Soltan and Professor David Clifton, announced CURIAL, a machine-learning model trained on routine blood tests and vital signs from 115,000 hospital presentations that could identify COVID-19 patients in emergency departments within one hour and with accuracy exceeding 90%.
Background
When COVID-19 began spreading through the United Kingdom in early 2020, hospitals faced a testing problem. The PCR tests that could confirm infection were slow, and at many points during the first wave, supply could not meet demand. Emergency departments were receiving patients with breathing difficulties, fever and other symptoms that might point to COVID-19 or might equally point to influenza, pneumonia or a range of other conditions. Clinicians needed to make quick decisions about isolation, bed allocation and treatment, often before any test result came back.
The tools available for rapid triage were blunt. Symptom checklists helped, but symptoms overlapped with too many other illnesses to be reliable on their own. Chest imaging could offer clues, but not every department had easy access to rapid radiology, and reading scans takes time and trained staff. What hospitals did have, for almost every patient who walked through the door, was routine blood work and vital signs: measurements already being collected within the first hour of arrival, for reasons entirely unrelated to COVID-19 screening.
The question was whether that data carried enough signal to distinguish COVID-19 patients from others, if you had a model capable of reading it.
What happened
In March 2020, Dr Andrew Soltan of the John Radcliffe Hospital in Oxford and Professor David Clifton of the Institute of Biomedical Engineering at the University of Oxford began training a model on exactly that kind of routine clinical data. They were joined by Professor David Eyre of the Oxford Big Data Institute. The team drew on records from around 115,000 hospital presentations and built CURIAL, a system that combined several modelling approaches including gradient boosting, random forests and neural networks, each of which learns to find patterns in data through different mathematical strategies.
CURIAL took as its input only the data that emergency departments already collected on admission: blood test results and vital signs such as temperature, heart rate and oxygen levels. No dedicated COVID-19 swab was needed. The model returned a result within one hour of a patient arriving, which matched the time it took for routine blood results to come back anyway. In prospective validation, meaning testing on patients who came in after the model had been trained rather than on historical records, CURIAL identified COVID-19 cases with accuracy exceeding 90%.
The peer-reviewed results were published in The Lancet Digital Health in April 2021, with an institutional announcement of early findings from the Oxford team appearing in July 2020. By the time of publication, the model had been tested across multiple sites in the Oxford University Hospitals NHS Foundation Trust, giving the results a grounding in real-world clinical conditions rather than a single controlled setting. The validation dataset was large enough to lend the accuracy figures genuine weight, though the work also made clear that a screening tool of this kind sits alongside clinical judgement rather than replacing it.
Why it mattered
CURIAL demonstrated that a diagnostic AI could be embedded into existing clinical workflows using data already collected on emergency admission, without requiring dedicated COVID-19 tests, thereby offering a rapid triage tool during a period when PCR testing capacity was constrained. Its validation on a large, real-world NHS dataset strengthened the case that machine learning could assist time-critical clinical decisions at scale. The work also contributed to a growing body of evidence that routine clinical data carries latent diagnostic signal extractable by neural methods.
People
Andrew Soltan, David Clifton, David Eyre
Organisations
University of Oxford, John Radcliffe Hospital, Oxford Big Data Institute, Institute of Biomedical Engineering Oxford
Sources
- New AI test identifies COVID-19 within one hour in emergency departments.University of Oxford.Official
- Rapid triage for COVID-19 using routine clinical data for patients attending hospital: development and prospective validation of an artificial intelligence screening test.The Lancet Digital Health.Primary source
Cite this page
AI Achievements. (2020). CURIAL: An AI System to Detect COVID-19 in Emergency Department Patients Using Routine Blood Tests and Vital Signs. Retrieved 2026-08-22, from https://achievements.ai/milestone/curialai-first-ai-system-detect-covid-19
@misc{achievements_curialai_first_ai_system_detect_covid_19,
title = {CURIAL: An AI System to Detect COVID-19 in Emergency Department Patients Using Routine Blood Tests and Vital Signs},
author = {{AI Achievements}},
year = {2020},
url = {https://achievements.ai/milestone/curialai-first-ai-system-detect-covid-19}
}