Date: August 5, 2026

We’re thrilled to share that Evra Health will be presenting our work at Psych Congress 2026 this September in Orlando, Florida.

Our poster, “The Evra Behavioral Health Model of Care: A Human–AI Framework for Continuous Monitoring, Lifestyle Support, and Clinical Escalation,” explores a question that is becoming increasingly urgent as artificial intelligence enters behavioral healthcare:

How can we use AI to expand access and continuity of behavioral health support—without attempting to replace the clinicians, judgment, and therapeutic relationships at the center of good care?

Our answer is a harmonized human–AI model of care.

The gap between visits

Behavioral health disorders affect millions of people, yet timely access to evidence-based care remains constrained by clinician shortages, fragmented delivery, and limited support between appointments.

And much of what influences our mental well-being happens during those hours and days between clinical encounters: how we sleep, move, eat, manage stress, connect with others, adhere to treatment, and respond to changes in our lives.

This creates an important opportunity for AI—but also an important boundary.

AI alone is insufficient for diagnosis, prescribing, psychotherapy, or crisis management. At the same time, clinician-only care is difficult to scale continuously between visits.

Our model is designed around the complementary strengths of both.

What should AI do, and what should remain human?

In the framework we’re presenting at Psych Congress, Evra functions as a continuous support layer around established behavioral healthcare.

The model begins with a comprehensive understanding of the individual: medical history, medications, goals and preferences, barriers, lifestyle data, wearable signals, and validated measures such as the PHQ-9 and GAD-7.

From there, the AI layer can help with continuous monitoring of mood, energy, symptoms, and adherence; psychoeducation; reminders and nudges; lifestyle support; care navigation; and identification of meaningful behavioral trends. Evra can synthesize these signals into longitudinal summaries that help make the time between clinical visits more visible and actionable.

But there is a clear line around what AI should not do.

Human clinicians retain responsibility for diagnosis, treatment planning, psychotherapy, medication management, risk assessment, crisis care, and complex clinical decision-making.

This distinction is fundamental to how we think about responsible AI in healthcare.

The goal isn’t to build an AI therapist. It is to build an intelligent infrastructure that helps patients and clinicians stay better connected between visits.

From monitoring to meaningful escalation

Continuous monitoring is only valuable if the system knows what to do when something changes.

That’s why another core component of our model is structured clinical escalation.

Routine, lower-acuity needs may be supported through lifestyle coaching, education, reminders, and self-management tools. Signals of clinical deterioration—such as worsening PHQ-9 or GAD-7 scores, functional decline, or persistent low mood or anxiety—can trigger timely follow-up with a therapist, psychiatrist, or primary care clinician.

And when crisis indicators emerge, self-guided pathways should no longer be the primary response. The framework incorporates deterministic escalation to appropriate human and emergency resources.
This lower-acuity → clinical deterioration → crisis architecture is central to the model.

AI can monitor. AI can educate. AI can identify trends. AI can help people navigate care.

But humans remain responsible for care that requires human clinical judgment.

Bringing behavioral and physical health back together

Our Psych Congress work also reflects something we believe deeply at Evra: mental and physical health cannot be treated as separate systems.

The framework draws from collaborative care, measurement-based care, lifestyle psychiatry, and digital monitoring. It considers not only self-reported symptoms, but also lifestyle behaviors and passive signals such as sleep, activity, heart rate, and heart-rate variability.

That matters because sleep, movement, nutrition, stress, social connection, cardiometabolic health, and mental well-being continuously interact.

This is also why our work at Psych Congress is a natural extension of the research we have been presenting across preventive cardiology and lifestyle medicine.

The common thread is a shift away from episodic snapshots of health toward a more longitudinal model: one capable of understanding what happens between visits and helping translate those signals into earlier, more personalized support.

A new model for clinicians, too

The opportunity isn’t only patient-facing.

In our framework, a provider dashboard can bring together PHQ-9 and GAD-7 scores, mood trends, sleep, activity, adherence, wearable-derived signals, and longitudinal summaries for clinician review.

The aim is not to give clinicians more data.

It is to help surface the right information at the right time.

That creates an important hypothesis we intend to test: whether AI can support appropriate lower-acuity, continuous activities under clinician-defined safety protocols, while allowing scarce clinician time to be focused on the work that most requires their expertise—diagnosis, psychotherapy, medication management, crisis intervention, and complex care.

If successful, this kind of task-sharing could help address one of behavioral healthcare’s fundamental constraints: there simply aren’t enough clinicians to provide continuous support to everyone who could benefit from it.

Building the evidence alongside the technology

There is still much to learn.

Our current framework is a conceptual model, and we believe it is important to distinguish what is promising from what has been prospectively demonstrated.

At Psych Congress, we outline a research agenda evaluating patient engagement, symptom outcomes, healthcare utilization, clinician workload, equitable access, safety, and cost-effectiveness.

We are also exploring the economic hypothesis that better continuity, earlier identification of deterioration, and more efficient use of clinician resources could create value for health systems. Importantly, the economic estimates in our current model are derived from comparable published interventions and are hypothesis-generating—not demonstrated Evra savings. Prospective validation will be essential.

That distinction matters to us.

Responsible health AI shouldn’t simply move fast. It should be designed with clear boundaries, studied rigorously, and integrated thoughtfully into the clinical systems and human relationships that already work.

Human + AI, not human vs. AI

As the conversation around generative AI and mental health accelerates, it’s tempting to frame the future as a competition between clinicians and machines.

We think that’s the wrong question.

The more interesting opportunity is determining which tasks technology can safely and effectively support, which must remain human, and how we design the connection between the two.

Our hypothesis is that thoughtfully designed AI can help make behavioral healthcare more continuous, personalized, and accessible while preserving—and potentially strengthening—the role of human clinical expertise.

As our poster concludes, rather than positioning AI as a substitute for behavioral health professionals, a harmonized human–AI model may provide a practical pathway for expanding support while preserving the central role of clinical expertise.

We’re excited to bring this work to Psych Congress 2026 in Orlando this September, learn from the behavioral health community, and continue building the evidence for what safe, effective human–AI collaboration in healthcare can look like.

See you in Orlando.

-Amitha