At present, diagnostics and line of treatment for COVID-19 is the focus of research communities worldwide. Among several research initiatives underway, collaborative efforts of researchers at King’s College London, health Science Company ZOE, and Massachusetts General Hospital resulting in the development of artificial intelligence diagnostic is an important one.
The AI-based diagnostic tool predicts if an individual is likely to develop COVID-19 based on their symptoms. The findings of the research is available for access in the May 11, 2020 issue of Nature Medicine.
AI Diagnostic combines Data of COVID Symptom Study App
To predict COVID-19 infection, the AI – based diagnostic tool uses data from the COVID Symptom Study app, for comparing people’s symptoms and results of traditional COVID tests. The diagnostic tool is expected to provide help for populations where access to testing is limited. To establish the accuracy of the tool, two clinical trials, one each in the US and the UK, are due to start shortly.
Meanwhile, more than 3.3 million individuals worldwide have downloaded the app and are using it for daily status report of their health. The reporting on the app is about state of feeling well or any visible symptoms such as persistent cough, fatigue, fever, and loss of taste or smell.
To establish the diagnostic tool, researchers analyzed data from just under 2.5 million individuals in the US and the UK who have been regularly logging into the app to report their health status. Of which, one-third of the individuals had logged symptoms associated with COVID-19. Of this population, 18, 374 reported to have taken test for COVID-19, with 7, 178 having tested positive.
Researchers propose combining AI prediction with widespread use of the app could help to identify individuals who are likely to develop the infection as soon as the first symptoms appear. This, thus would help to focus on tracking and testing efforts where most need.