Your Phone Can Detect Hidden Mental Health Risks

Smartphones are commonly used to track physical health indicators like sleep, physical activity, and heart rate—but new research suggests they can also uncover signs related to mental health. In a study published in JAMA Network Open, scientists from the University of Michigan, University of Minnesota, and University of Pittsburgh used smartphone sensors to unobtrusively monitor daily behavior. These sensors captured routine activities—such as movement, sleep habits, and phone use—and revealed unexpected patterns that reflect aspects of psychological well-being.

Funded by the National Institutes of Health, the study found that a variety of mental health conditions often involve overlapping behavioral trends, including spending more time at home, waking up later, and infrequent phone charging. These behaviors may indicate a person’s standing on the “p-factor,” a general dimension that connects multiple forms of mental illness.

Aidan Wright, professor of psychology and the Phil F. Jenkins Research Professor of Depression at the University of Michigan’s Eisenberg Family Depression Center, explained that the research team found certain behaviors—such as reduced phone calls or decreased physical activity—corresponded with particular issues, including social withdrawal or physical illness.

“These findings suggest that major forms of mental illness are detectable from smartphone sensors, indicating that this technology could potentially be used for symptom monitoring and research on wide-ranging psychiatric problems,” said Wright, the study’s senior author.

The study involved data from smartphone sensors used by 557 adults over 15 days in 2023, making it one of the largest of its kind. Despite widespread enthusiasm in using phone sensors and wearables to diagnose and track mental illness, progress has been modest, Wright said. “This is, in part, because most digital psychiatry work has not used what we know about how mental illness is organized within people when selecting targets to predict and monitor,” he said.

Digital psychiatry has relied on diagnoses from the Diagnostic and Statistical Manual of Mental Disorders, or DSM-5, which are poor targets for detection and monitoring because they are heterogeneous. This means that they are combinations of different types of symptoms that might have different behavioral signatures, and at the same time, often share symptoms with other diagnoses, Wright said.

Adding to the problem is that in clinical settings, most individuals have more than one diagnosis, making it unclear which might be responsible for their behavior, he said.

“In other words, these diagnoses do a poor job of parsing mental illness,” he said.

Whitney Ringwald, assistant professor of psychology at the University of Minnesota and the study’s lead author, said the findings allow for a better understanding of why different forms of psychopathology might impair afflicted people’s functioning in daily life.

Mental illness often comes on insidiously and is best treated early before it becomes severe and debilitating. Wright said monitoring it, however, is hard and “what we have in place is far too little and not nearly up to the task.”

“The ability to use passive sensing to connect someone with help before things get really bad would have huge benefits, including better outcomes, reduced costs, and lower stigma,” Wright said.

4155/v