[Summary] Digital Phenotyping: Technology for a New Science of Behavior (Insel 2017)

TLDR; In a field that hardly uses measurement; and where clinicians are too busy and patients are unmotivated; we need an objective, passive, ubiquitous device to capture actionable data.

https://twin.sci-hub.do/6540/08dec3140109259076efa0ea6dc7f5e1/insel2017.pdf

TLDR; In a field that hardly uses tools for measurement; and where clinicians are too busy and patients are unmotivated; we need an objective, passive, ubiquitous device to capture actionable information. Digital phenotyping (prediction and diagnosis via smartphone usage) is already showing promising results, but we need additional clinical and real world research and validation.

“What the field needs is an objective, passive, ubiquitous device to capture behavioral and cognitive information continuously. Ideally, this device would transmit actionable formation to the patient and the clinician, improving the precision of diagnosis and enabling measurement-based care at scale.”

In practice, clinicians use few standard, validated tools for measurement.

• “Over the past 4 decades, behavioral expertise, once the strength of psychiatry, has diminished in importance as psychiatric research focused on pharmacology, genomics, and neuroscience, and much of psychiatric practice has become a series of brief clinical interactions focused on medication management. In research settings, assigning a diagnosis from the Diagnostic and Statistical Manual of Mental Disorders has become a surrogate for behavioral observation. In practice, few clinicians measure emotion, cognition, or behavior with any standard, validated tools.”

NIMH is attempting to change this with the RDoC

• “Some recent changes in both research and practice are promising. The National Institute of Mental Health has led an effort to create a new diagnostic approach for researchers that is intended to combine biological, behavioral, and social factors to create “precision medicine for psychiatry.”2”

It’s impractical to expect more forms to be completed by busier clinicians and patients with mental illness; therefore we need a device that constantly captures and transmits actionable information to patients and clinicians for precision diagnosis and measurement.

• “Expecting mental health clinicians to complete rating forms may be a challenge for those with less than 15 minutes per patient encounter. Asking patients to complete rating forms might seem like an efficient alternative but it is unclear if patients with serious mental illness who are often nonadherent to their medication would be more compliant with self-ratings. What the field needs is an objective, passive, ubiquitous device to capture behavioral and cognitive information continuously. Ideally, this device would transmit actionable formation to the patient and the clinician, improving the precision of diagnosis and enabling measurement-based care at scale.”

“Digital phenotyping” is measuring behavior using smartphone sensors and interactions, and is already proving promising in predicting some clinical ratings.

• “Could anyone have foreseen the revolution in natural language processing and artificial intelligence that is allowing voice and speech, collected on a smartphone, to become a possible early warning sign of serious mental illness?”

• “Digital phenotyping is the term now used for describing this new approach to measuring behavior from smartphone sensors, keyboard interaction, and various features of voice and speech. Already digital phenotyping is revealing new aspects of behavior that appear clinically relevant. In one study of 48 individuals, Saeb et al6 described behavioral entropy based on the variation in several sensor measures as a correlate of mood ratings. The study by Bedi et al7 proposed measures of semantic coherence from speech samples as a predictor of psychosis. Although most of the early studies, seeking validation, have measured the correlation of digital phenotyping features with standard clinical ratings, it is not clear whether smartphone measures collected continuously in a patient’s ecosystem will prove better at predicting clinical outcomes than episodic rating scales collected in a clinical setting. Even if digital phenotyping is more successful at predicting outcomes, the real question is whether this information can be used to monitor and improve patient outcomes.”

We need more studies on whether digital phenotyping has clinical effectiveness and create better real world outcomes.

• “Two major hurdles must be overcome if digital phenotyping is to move from the current stage of hype to the hoped-for stage of public health effect. First, the technology must demonstrate its value, not in economic terms but in terms of clinical effectiveness. Will digital phenotyping confer better outcomes in the real world of clinical practice? Will better measurement reduce morbidity and mortality? Currently, there are no studies with which to answer these questions, but there are few areas of medicine in which better measurement alone confers better outcomes. Data that improve decisions or improve efficiency will be helpful, but bridging the “last mile” from better data to better care is a major challenge. “

The other important challenge is trust and privacy in constantly collecting data. 

• “As important as demonstrating value is the issue of ensuring trust. Digital phenotyping research has been limited to consenting research participants who are essentially collaborators. Questions of privacy and agency need to be addressed in the research environment, but these questions become even more acute in clinical practice or when digital data are part of population surveillance for disease risk. Proponents may point out that digital data of each person are being collected all the time without that person’s awareness, and that everyone must adjust to a world without privacy protection for actions and information on phones. But will people feel the same way about their behavioral data being collected for health purposes? Who will own these data? Will the data be used to empower patients and families to make better de- cisions about health issues or, like online data collected today, used to identify consumers and link to potential markets?”