The rise of wearable technology, such as fitness trackers, has given unprecedented access to real-time data that could be leveraged to predict the next influenza (flu) outbreak.
In the US, approximately 7% of adults and approximately 20% of children are infected with the flu every year, according to estimates from the World Health Organization (WHO). Globally, the flu kills as many as 650,000 individuals each year. In a new study published in The Lancet, Dr Jennifer Radin and colleagues used de-identified data from Fitbit users on sleep and resting heart rate (RHR) to improve real-time prediction of influenza-like illness (ILI).
Current surveillance methods to predict flu rates are incredibly slow, as they mainly use data that is reported to the Centers for Disease Control and Prevention (CDC). This type of reporting is often delayed by one to three weeks, with data that is often revised months later. This delay makes it nearly impossible to deliver response measures to a flu outbreak, such as deploying vaccines or anti-viral medications, as outbreaks are quick to spread to new geographical regions and vulnerable populations, including children and the elderly.
Previous studies have tried leveraging social media and the power of crowdsourced data, such as Google Flu Trends and Twitter, to provide real-time ILI information. However, these approaches were variable in their success, most likely due to outside factors. For example, it is impossible to separate the behaviour of people who have the flu from those who are influenced by and reacting to media coverage during an influenza outbreak.
This study used de-identified Fitbit data from 200,000 people from the top five US states with the most fitness track users: California, Texas, New York, Illinois, and Pennsylvania. More than 13.3 million total RHR and sleep measure data points were taken. RHR tends to spike when an individual is sick and drops back down to normal after they have recovered. Sensor-based data offers the first objective and real-time measurement of illness in a population. This could be used to reduce the effect of overestimation during epidemics.
Furthermore, the ability to harness this data on such a large scale may help to provide estimates with granularity at the state and local level, giving public health responders the ability to act quickly and precisely on suspected outbreaks, corralling the outbreak before it has a chance to take hold.
With the incidence of influenza increasing due to increasing population size, and globalization making the spread of infection easier, wearable sensors that detect fluctuations in RHR and sleeping patterns could provide low-cost, practical solutions to the early diagnosis of influenza or other infectious diseases. This will allow for accurate and timely response measures at both the individual and population level.