AINEWS
The federally funded project splits them across three cohorts, including one with active suicidal thoughts, and asks what keeps the well students well, not only what breaks the others.
Every year, more than 47,000 Americans die by suicide and more than 10 million seriously consider it, according to the CDC figures cited in the USC announcement. A new project at the university wants to know whether the warning signs of those crises are already in the everyday noise of a college student's life: in their sleep, the way their eyes move, the sweat on their skin, the activity in their brain, and the way they use their phone, long before a clinician would have any reason to ask.
The project, announced this week by USC's Keck School of Medicine, is called SENTINEL (Scalable Evaluation of Neurobehavioral Trajectories in Everyday Life). It plans to follow about 210 USC students for up to 36 months, evenly split across three cohorts: students with no mental-health concerns, students with depression, and students living with depression and active suicidal thoughts or behaviors. That three-cohort design is the spine of the study. By tracking all three groups over the same window, the team is betting it can learn not only who trends toward crisis but also what keeps the well students well, which is the part of the project the press release barely mentions.
SENTINEL is funded by an ARPA-H contract award of up to $4 million. The "up to" matters: a contract ceiling is a maximum the government can be billed against, not a confirmed spend, and the actual outlay will depend on milestones the team has not yet published.
The signals the platform will read are mostly passive and continuous. Lab tests, wearable sensors, and smartphone data will be combined, per the university's release, and the team points to prior work that reportedly identified distinct eye-movement, brain-activity, and sweat biomarkers in adults with suicidal thoughts. The next question is whether those signals hold up in the messier signal environment of daily life, and whether the same model that flags a future crisis can also flag its absence. The press release gestures at the earlier work but does not link to the underlying peer-reviewed publications, so the prior claims sit at the level of "the lab says," not "the literature shows."
The principal investigator is Shrikanth (Shri) Narayanan, a University Professor and vice president for presidential initiatives at USC who holds the Niki and Max Nikias Chair in Engineering and appointments across electrical and computer engineering, computer science, linguistics, pediatrics, otolaryngology, and music, per the university's release. The SENTINEL bet is that a multi-year, three-cohort design at a single site can do something the field has not yet pulled off in this population: read both risk and resilience from the same data stream.
That continuous monitoring, on a population that includes students enrolled specifically because they are living with suicidal thoughts, also forces onto the table a set of questions the press release does not address. What happens when a wearable flags a participant in real time? Who is on call, and on what clock? What does passive monitoring mean for students whose parents don't know they're enrolled, or who don't want their crisis trajectory stored in a university research database? The release mentions IRB approval but does not describe the duty-of-care protocol, the data-retention window, or the conditions under which a flagged signal becomes a clinical intervention rather than a research observation.
There are no findings yet. SENTINEL is a research launch, not a deployed product, and the 36-month horizon means the first peer-reviewed outputs are unlikely before 2027. The honest read is that a credentialed team is testing a specific bet: that the everyday signals of a college student's life contain both the seeds of a future crisis and, just as usefully, the pattern of a life that stays well, and that the bet is now funded long enough to find out which it is.