Philadelphia-based NeuroFlow, a behavioral health tech and analytics company, wants to help its clients of commercial payers and health systems identify patients who could be at risk of mental illness and suicide and intervene early. Its bet: artificial intelligence.
More than 27 million US adults live with a mental illness but aren’t getting treatment, according to a 2022 estimate from the nonprofit Mental Health America. Untreated mental illness not only hurts the individual, but leads to unexpected costs for NeuroFlow’s provider and payer clients, NeuroFlow Chief Commercial Officer Robert Capobianco told Healthcare Brew.
NeuroFlow believes its technology can assist the medical supply chain, from doctors to health systems, to be more efficient in identifying patients who need help and prevent the need for expensive interventions down the line.
“We have to stay in front of the people just about to be acute, or [they] will continue to be high-cost members because you catch them at the end and not the beginning,” Capobianco said.
Predicting the future. With the consent of NeuroFlow’s clients and their patients, AI analyzes de-identified population health data to identify patterns relevant to mental health.
From these analyses, AI then helps find patients who haven’t been diagnosed with a mental health condition but have similar characteristics and behaviors to patients who do have a diagnosis.
These patients can be prioritized for proactive outreach, and providers can check in with them more frequently, Capobianco said.
“Instead of reacting to what’s known, now we can start to get ahead of that,” Capobianco said.
The company acquired Alluceo, a team-based care platform that incorporates mental and behavioral health into primary care, at the beginning of January to bring the platform into its analytics architecture. While Alluceo includes a non-AI model that helps integrate care and determine patient “risk and complexity,” the company plans to build and train the Alluceo model on about 1 million additional patients’ data, Capobianco said.
“We might look for things like…what physical health conditions have a co-occurring behavioral health condition?” Capobianco said. “If we were looking for known cost drivers, we might look at outpatient utilization. We might look at inpatient utilization around residential treatment, partial hospital programs.”
Navigate the healthcare industry
Healthcare Brew covers pharmaceutical developments, health startups, the latest tech, and how it impacts hospitals and providers to keep administrators and providers informed.
Language processing. AI also plays an important role in identifying potentially at-risk patients in ways that go beyond traditional methods, Capobianco said.
Say a patient hasn’t taken a formal depression assessment, but they have filled out an ecological momentary assessment, which asks them to write down how they currently feel.
If someone writes “I really don’t feel like being here today,” AI could flag that using natural language processing, and a provider’s care management team could choose to give that person a call, Capobianco said.
In a 2024 study published in the Journal of Technology in Behavioral Science, NeuroFlow researchers found that patients’ self-reports of sleep, stress, and mood were indicators that helped significantly predict the risk of suicidal ideation. The authors noted, however, that follow-up studies are needed to confirm these findings.
Potential obstacles. There are a number of other machine learning models that aim to predict the risk of mental health crises using patient and provider data, including some at institutions across the pond in Spain and the UK.
David Cooper, executive director of Therapists in Tech, a community that supports mental health professionals in tech roles, said this tech is great for value-based care—which pays based on patients’ results rather than the number of procedures performed—and can be helpful “long term.”
But he also noted a shortcoming. Right now, he worries that there aren’t enough mental health professionals to see everyone identified as a candidate for treatment by these algorithms.
Over half—53%—of 853 surveyed psychologists didn’t have openings for new patients in the American Psychological Association’s 2024 Practitioner Pulse Survey.
“The mental health crisis we have is not one of identification; it’s one of access,” Cooper said. “Just pouring more [patients seeking care] in the top of the funnel when it’s already stuck—I don’t know if that helps.”