Digital phenotyping in psychiatry: When mental health goes binary

While the scientific explorations have churned out the role of micro/nano molecules and omics in psychiatry, the art of observation of an individual in its entirety, in an interaction, has gone underemphasized. The microscopic scrutiny of fractions somehow has overshadowed macroscopic study of interactions. While our classificatory systems are concerned with rise of gaming disorders and other Internet-related mental health problems, industries have been closely observing online behaviors of these users, making sense of their need/motivation, likes/dislikes, and utilizing these to promote sale and reap benefits. This concept of observing online proxies of behavior and emotions to understand human psyche has ramified into dimensions of mental health care and is called digital phenotyping.

Digital phenotyping is defined as moment-by-moment quantification of individual-level human phenotype in situ, using data from personal digital devices.[1] Given that, of more than 3 billion people with Internet access around the world, one-third resides in India or China; this concept has enough catchment population to study human behavior, or aberrations thereof.[2] The study has shown that an average smartphone user has around 60–90 apps installed in his phone, uses 30 of these in a month or 9 a day, and spends around 2.25 h on these apps per day.[3] Digital phenotyping, also called personal sensing, can render live proxy of human behavior and emotion, and may change overall realm of psychiatry.[1,4]

Digital phenotyping considers data from smartphones and other digital wearable. These data may be active (real inputs from user) or passive (input from sensors). Various digital biomarkers which are being utilized in digital phenotyping are geolocation, calls (outgoing/incoming/not answered, duration/timing), messages (SMS/WhatsApp, length/timing), finger taps (speed, number), status of phone (Wi-Fi, Bluetooth, and power on/off state), ambient/preferred light, data on accelerometer, gyroscope, magnetometer or pedometer, sleep, heart rate/variability, screen behavior such as scrolling/clicking/tapping, speech, and voice modulation, frequency of battery charging, call log, navigation path on net, app visit, and update frequency.[5,6,7]

Effective assimilation of these data to understand and individualize human behavior may lead us to digital personalized psychiatry, by improving diagnostic process, designing individualized treatment plans, facilitating monitoring of behavior or treatment effects/adverse effects, predicting onset or relapse of illness, preventing mental morbidity, and fostering positive mental health or by risk reduction in psychiatry.[8] Given the advancement in machine learning, the data are going to be better and more real in future and possibility of evaluation and management, more precise.[7]

Use of digital phenotyping is being explored in many behavioral problems and illnesses, such as addiction,[9] autism spectrum disorders,[10] posttraumatic stress disorder,[11] schizophrenia,[12] perinatal psychiatry,[13] mood disorders,[14] sleep disorders,[15] child and adolescent psychiatry,[16] and suicide prevention[17]. Shift of pronoun to first-person singular has been seen in depression.[18] Semantic/phonemic fluency has been found impaired in first-episode depression.[19] Latent period between space and character or intervening time between scrolling and clicking has shown to be a fair surrogate of cognitive/affective trait or state.[20]

However, there are concerns with digital phenotyping. Currently, data are mostly based on convenience sampling, and thus, the characteristics may vary with the target group. Generalization of results and enhanced precision will require well-validated research in this direction and replication of findings. Evidence would be further required for its clinical utility and scalability.[8] Collection of data is riddled with ethical issue of informed consent, privacy, transparency, and accountability.[21] These digital markers currently are being used extensively by third parties for the purpose of digital analysis and marketing. Commodification of health-care data may impinge on privacy/freedom of the users.[22] Standard guidelines of use by professional bodies, robust protocol for data collection, and statutory body for ethical scrutiny may mitigate these problems.

To summarize, though the science of digital phenotyping is still in infancy stage for its use in psychiatry, the conceptual merit and available literature holds adequate shine and promises for its robust application in future.