In the digital era, researchers are working with tech companies to monitor how people interact with the digital world to better understand their mood, cognition, and behavior, but there are challenges, including ethical issues.
Consider this situation: A family member starts avoiding social gatherings, doesn’t respond to messages quickly, becomes isolated for long periods of time, and stays online late at night becoming sleep deprived.
You sense anxiety and depression in social media posts and you automatically wonder if your family member is undergoing some problems. That is the basis of digital phenotyping, a term coined by the Harvard T.H. Chan School of Public Health for the emerging field, which is focused on trying to assess people’s well-being based on their interactions with digital devices, such as smartphones.
Phenotypes are physical traits such as eye or hair color, height, voice, metabolism, or shoe size, that are influenced by genotypes (and sometimes by external factors such as nutrition). Phenotyping is used in clinical applications such as the discovery of disease genes and pharmacogenomics.
Want to publish your own articles on DistilINFO Publications?
Send us an email, we will get in touch with you.
In the digital era, researchers have joined hands with tech companies to monitor how people interact with the digital world to better understand their mood, cognition, and behavior. The vision is to analyze this data based on an individual’s genetic makeup for a new level of behavioral observation.
How does this work?
Smartphones linger in our hands longer than they used to. A 2017 report from App Annie states that an average smartphone user:
- Spends 2.25 hours a day using apps
- Has 60–90 apps installed and uses about 30 per month and 9 per day
- Mostly uses social networking, communication, and productivity apps (apart from common utility apps)
The data generated from extensive smartphone use can shed light on our physical and emotional health. Smartphones generate two forms of digital data: Active or user-generated data including content in texts, calls, and social media (for social engagement); and passive data including spatial location, time spent in various locations, driving speed, and phone usage patterns all collected via the phone’s sensors.
J.P. Onnela, an associate professor of biostatistics at Harvard Medical School who is currently developing an analytical model on digital phenotyping using the Beiwe platform, states that data from personal digital devices is more reliable and accurate in understanding social and behavioral phenotypes than data acquired via interviews or surveys. We can use this data to monitor patients for treatment effectiveness, to identify psychiatric disease phenotypes, to predict bipolar disorder and schizophrenia, and to support drug trials and precision medicine. The applications are endless.
Following are some applications of digital phenotyping in action today:
- Facebook recently announced an algorithm that scans posts to see if users are exhibiting signs of suicidal thoughts and alerts a Facebook review team.
- Mindstrong Health, a California-based startup, has developed a research app to continuously monitor users’ phone habits, keyboard accuracy, and speed for hints about mood and memory changes associated with depression. The company’s research has shown that specific biomarkers from human smartphone interactions that are associated with cognitive control and reward correlate highly with activity in brain areas that are implicated in those same domains. Paul Dagum, CEO of Mindstrong Health, believes that this data can be used at an aggregate level for public health, such as creating a heat map of the planet to show spots of emotional volatility or cognitive decline from stress, contagion, or toxins.
Surprisingly, this kind of digital phenotyping is not new. A few years ago, John Brownstein, a professor at Harvard Medical School and CIO at Boston Children’s Hospital, built a real-time public health surveillance system called HealthMap that mines internet data like news stories, blogs, government websites, and social data to monitor disease outbreaks. Analysis of public health using nonclinical data sets is called digital epidemiology. In 2014, HealthMap picked up on the deadly Ebola outbreak in West Africa a week before an official announcement was made. John Brownstein and Sachin H. Jain, CEO of CareMore Health, coined the term digital phenotype in 2015.
- Sharecare, an Atlanta-based digital health company, offers a wellness app with an optional feature that analyzes users’ stress levels during phone calls. The system delivers real-time feedback like “You seemed anxious” or “You seemed balanced” and characterizes users’ relationships with the people they call in terms of attitudes like dominance or affection.
With apps like these, users’ privacy is invaded. For example, although Sharecare says it doesn’t record the content of the calls it scans, it does collect phone numbers without informing the user.
The challenges aren’t limited to ethical issues. Making sense of the dense data is extremely difficult and requires effective data modeling and analytical techniques to be clinically valid. The analysis also doesn’t consider the user’s familiarity with smartphones, their typical smartphone usage, ease of use of the apps, and the clinical accuracy of the devices.
After all, technological development is futile unless it is properly implemented and utilized. Will users be comfortable if their day-to-day actions are monitored? How will information be handled and shared with the users or their families if they are at risk? These are just some of the questions and concerns related to digital phenotyping. The promise is there, but we have a lot to do before we can fulfill it.
Date: January 23, 2019