Contributor: How Health Systems Can Rebalance Behavioral Health Access

Walk into your local grocery chain’s cereal aisle and you’ll experience something akin to getting a specialty care referral today: too much choice between too many options, all of which are promising roughly the same thing.

Choosing 1 cereal should be simple. Instead, it’s not uncommon to see people standing there comparing boxes, reading labels, and second-guessing themselves before settling on something familiar. The stakes are obviously much higher in health care, and even more so in behavioral health.

From the onset of the pandemic through roughly 2024, health care’s primary mission was expanding access. Health plans broadened their networks, virtual care options exploded, and digital mental health tools multiplied. In many respects, these missions succeeded.

In 2026, the problem looks somewhat different. There’s now an abundance of access, but that comes with its own set of problems. One behavioral health referral from an emergency department can produce dozens of potential providers. For most care managers, working that list means tabbing between a directory, an insurance portal, and a spreadsheet, trying to figure out who is actually accepting new patients, which insurances they take, and whether their listed hours are still accurate. If the first few options didn’t pan out, the search hits a dead end. However, if the patient gets a list and is told to call around, this is where they are likely to disengage and where they are most likely to end up back in the emergency department.

The second failure point comes after the first appointment, if patients even get there, when they determine a provider isn’t a fit. Among patients who drop out of behavioral health treatment prematurely, 70% do so after the first or second visit.¹ A significant share of these exits come down to a match that wasn’t right to begin with.

When patients don’t get the behavioral health care that they need, their physical health worsens, too, especially for patients with chronic conditions. For example, a patient with heart failure and depression has a 223% higher mortality rate compared with a patient with heart failure who doesn’t have depression.² This is because patients with untreated behavioral health issues are less likely to adhere to treatment, leading to worse outcomes overall. This should matter to health systems for the poor clinical outcomes alone, but given increasingly tight margins, the reality is health system leaders need to make a financial case for behavioral health investment. Lack of treatment adherence due to untreated behavioral health represents millions in fee-for-service revenue lost across both physical and behavioral health care. It’s estimated that for an 800,000-patient health system, that loss is over $150 million annually.

Health systems already have the data to do better. They know which providers have above-average no-show and cancellation rates. They track satisfaction scores. They can see what happens after a care episode: whether a patient refilled their prescription, kept their next appointment, or stopped showing up. This is the same signal set streaming services use to figure out what to recommend next.

Netflix doesn’t just know what you watched. It knows who started something and stopped, who finished it in 1 sitting, who came back 3 days later, and how they rated it against everything they watched before. Health systems have equivalent data on providers and patients. The gap is not the data. It’s that nobody built the infrastructure to act on it at the moment a referral is being made.

Today, the introduction of agentic artificial intelligence (AI) has made this gap bridgeable.

A well-designed agent pulls from provider websites, state registries, and claims data simultaneously; reconciles what conflicts; and surfaces a picture of who a provider is today rather than when they first credentialed. There’s a meaningful difference between a therapist who does cognitive behavioral therapy (CBT) and one whose practice is built around CBT for pediatric anxiety. There’s a meaningful difference between a provider trained in exposure and response prevention and one who applies it specifically to obsessive-compulsive disorder or posttraumatic stress disorder.

That same agent changes what the matching workflow looks like. Instead of working through a spreadsheet until the search runs out of options, a care manager works through a single conversation: narrow by proximity, insurance, subspecialty, patient preference, and the specific clinical approach the patient actually needs. If the closest match doesn’t exist within the initial parameters, the agent widens the radius and returns the next best option rather than stopping. The care manager stays in the search instead of abandoning it.

After the match is made, care managers don’t stop working. To ensure patients actually show up, they make the confirmation call the day before the appointment. They send the text reminder because they know from experience that a patient who hears nothing between the referral and the appointment is unlikely to show. This follow-through is what moves the needle on the 70% dropout number. Application programming interface integrations that enable automated appointment confirmation and patient outreach are routine configurations at this point. The barrier is not technical; it’s that health systems haven’t connected these capabilities into a workflow that runs without a care manager initiating every step.

This is what agentic AI makes possible at scale: data aggregation across sources that no individual can monitor continuously, matching logic that improves as encounter patterns accumulate, and a confirmation and reminder loop that runs in the background without consuming the hours care managers don’t have. What’s left for the care manager is the work that actually requires a human: the first call with a patient who is apprehensive or reticent about treatment. It’s a conversation that helps someone in a vulnerable state understand what they’re being asked to do and why it’s worth doing.

References

1. Olfson M, Mojtabai R, Sampson NA, Druss B, Wang PS, Wells KB, Pincus HA, Kessler RC. Dropout from outpatient mental health care in the United States. Psychiatric Services. 2009;60(7):898-907. doi:10.1176/ps.2009.60.7.898
2. Jiang W, Alexander J, Christopher E, et al. Relationship of depression to increased risk of mortality and rehospitalization in patients with congestive heart failure. Arch Intern Med. 2001;161(15):1849–1856. doi:10.1001/archinte.161.15.1849