AI-enabled Pelago expands from SUD care to mental health, behavioral addictions

Pelago, a virtual provider of substance use disorder (SUD) treatment, has recently expanded its services to include mental health and behavioral addictions. This new offering, known as the Pelago Behavioral Health Platform, allows for the assessment and treatment of co-occurring conditions within a single plan. The platform includes programming for behavioral addictions such as gambling, gaming, and binge eating.

The expanded services are available to employers and payers, with member access beginning in November. The announcement of this expansion occurred at the Behavioral Health Tech (BHT) conference in Nashville.

Since its inception, Pelago has focused on treating SUD patients affordably, achieving an annual savings of nearly $12,000 per patient compared to traditional residential treatment or intensive outpatient programs. This cost-effective approach is made possible by identifying individuals with untreated SUD, engaging them in treatment, and doing so at a lower cost than the more expensive options.

Pelago's employer clients face rising healthcare costs and a sense of misaligned incentives within the traditional behavioral health system, which tends to prioritize session-based care. Pelago Co-founder and CEO Yusuf Sherwani, M.D., noted that these clients have requested that Pelago expand its services using an acuity-based clinical model.

When a new Pelago member first joins, they undergo an intake and triage process using Sona, the company's clinical AI agent. Available via both voice and text, Sona assesses the member's acuity level and routes them accordingly. Higher-acuity members are immediately connected with a clinician, while lower-acuity members may work primarily with Sona. However, the company sees the best clinical outcomes when members utilize both Sona and a human clinician, as needed.

Sona's model is based on an extensive repository of over 800,000 conversations and takes into account each individual user's chat history. If a patient consents to sharing data with Pelago, Sona can also incorporate summaries of therapy sessions. Preliminary studies involving more than 9,000 Sona conversations have shown promising results.

Members who began with clinical-range depression or anxiety scores and completed repeat assessments experienced improvement in their PHQ9 or GAD7 assessment scores by at least five points within about a month. Additionally, participants reported 20-25% fewer poor mental health days each month.

Safety is a primary consideration for Pelago when developing Sona. The AI agent is equipped with conversational guardrails to detect and escalate potential suicide risks. In fact, Sona identifies suicidal ideation at a higher rate than the PHQ9, with about 20% of patients reporting such ideation.

Beyond merely identifying and managing risks, Sona tailors its recommendations based on each individual's specific plan benefits. It can generate cognitive-behavioral therapy (CBT) lessons and assist patients in practicing these techniques between sessions, reporting progress back to the clinician. Sherwani explained that Sona feels like an extension to the clinician, providing an additional layer of support for patients.

Pelago's voice-based Sona agent has shown better outcomes than its text-based counterpart. Patients tend to engage in longer conversations with the voice agent, as it is easier and more akin to speaking with a human therapist. Moreover, the voice agent can pick up on important intonations that may be lost in a text-based chat. Sherwani believes that payers today are grappling with the question of whether covering therapy for all individuals, regardless of their clinical acuity, is a viable option.

He believes that there are only two potential solutions: return to more aggressive utilization management, which could harm access, or utilize technology to match individuals to the appropriate level of care.