COVID VOICE [COVID-19]
Research type
Research Study
Full title
Coronovirus Infectious Disease 2019: Ventilator Outcomes using artificial Intelligence, Chest radiographs and other Evidence-based co-variates (COVID VOICE)
IRAS ID
284442
Contact name
Thomas Booth
Contact email
Sponsor organisation
King's College London
Clinicaltrials.gov Identifier
N/A, N/A
Duration of Study in the UK
1 years, 0 months, 1 days
Research summary
Research Summary
COVID-19 is a novel infection with serious clinical manifestations, including death. There is the potential for the disease to produce enough severe illness to compromise health care infrastructure. Such demands may create the need to ration medical equipment and interventions. \n\nOnce a decision-support tool to triage patients for ventilator use has been validated both analytically and clinically, an algorithm may be useful in the tail of the first peak in the UK, or in subsequent second or third peaks of the pandemic. It is also noteworthy that transfer learning would also allow the algorithm to be used in future viral epidemics or pandemics.\n\nWe aim to build a decision-support tool to triage patients for ventilator use. Specifically the Primary Objective is to use chest radiographs and clinical data to determine whether a patient can survive with a ventilator. The Secondary Objective is to use chest radiographs and clinical data to determine whether a patient can survive without a ventilator.\n
Summary of Results
For the retrospective data:
Abstract: Purpose of the research: To develop a triage decision tool capable of predicting the outcome of ventilation (death or survival) using chest radiographs, key blood markers and clinical history available to the clinician at the time of decision for intensive care unit transfer.
Principal results: For retrospective data, the automated triage tool was capable of accurately predicting COVID-19 patient-ventilator survival outcome moderately well for different models (combined model, using chest radiographs and temporally-relevant clinical covariates from free text, as well as simpler models employing either imaging or clinical/demographic data alone). For prospective data the results were similar.
Conclusion: We have shown that our automated triage tool is capable of accurately predicting COVID-19 patient-ventilator survival outcome moderately well, using multi-modal data readily available to clinicians in a real-world ICU triageREC name
London - Harrow Research Ethics Committee
REC reference
20/HRA/2595
Date of REC Opinion
14 Jul 2020
REC opinion
Further Information Favourable Opinion