Do you have a connected bracelet, such as Xiaomi’s latest Smart Band 7? These activity trackers really don’t cost much anymore. And according to researchers, it would be enough to dope them with a little AI so that they can have a leading role in health – in particular in terms of screening for the coronavirus (COVID-19).
In a study published in BMJ Open, researchers from Dr. Rish Mediacal Laboratory (Lichtenstein) indeed believe that “Wearable sensor technologies can help detect Covid-19 in the period before the onset of symptoms”.
Researchers bet on connected bracelets to better detect COVID-19, earlier
Indeed, machine learning models are able, under certain conditions, to establish complex links between data from biometric markers and a patient’s state of health. We remember, for example, screening experiments capable of detecting people with COVID-19 by analyzing the sound of their cough.
With the added bonus of fairly high reliability. The idea is to adapt the same logic based on the data that current smart bands can already collect – without adding an additional sensor. By aggregating this data in a machine learning model dedicated to COVID-19 screening, your watch could thus display an alert in the event of a suspected infection.
This can thus allow the wearer to immediately carry out a laboratory test and to isolate himself earlier, making it possible to drastically reduce the spread of the virus. To reach this conclusion, the researchers tested the capabilities of the “Ava” bracelet: a tracker normally intended for women wishing to know the ideal fertility window to get pregnant.
The device collects data on respiratory rate, heart rate, heart rate variability, skin temperature at the wrist, and blood flow. The researchers thus followed 1,163 patients under the age of 51 from the start of the pandemic until April 2021.
Everyone had to wear the bracelet at night – and the bracelet was connected to an application in which the user had to mention anything that could falsify the measurement (consumption of alcohol, drugs, medication, etc.). They also had to report any symptoms suggestive of Covid-19. In parallel, participants were required to perform regular antigen tests (and PCR for sick people).
The researchers thus demonstrated the presence of significant physiological changes in the period preceding the onset of symptoms. However, thanks to an algorithm, 68% of participants who were ultimately infected with COVID-19 were able to be alerted two days before the appearance of the first symptoms.
The researchers are currently continuing their study, in order to confirm the results and refine their algorithm. The new study analyzes physiological data from 20,000 patients residing in the Netherlands. The results should be available by the end of the year.