Artificial Intelligence in Behavioral Health: Navigating the Tapestry of Triage Decisions
This is an exclusive BHB+ story.
Artificial intelligence (AI) has been hailed as a game-changer in behavioral health, promising improvements in efficiency, pattern detection, and cost savings. However, when it comes to triage decisions, AI's potential pitfalls outweigh its promises. Despite this, AI adoption in behavioral health is accelerating, with both providers and payers embracing the technology.
Guardrails for AI-driven Triage
Executives in the behavioral health industry have consistently emphasized that AI should not replace clinical decision-making. Yet, the rollout of AI for patient triage is gaining momentum. Aetna, for instance, plans to introduce an intelligent triage tool in Q2 2027, featuring a brief PHQ-4 assessment for members seeking behavioral health services. This tool will direct patients to appropriate care options, including employee assistance programs (EAPs), outpatient resources, on-demand support, or higher-acuity settings, according to Miriam Ferreira, head of Aetna’s mental wellbeing division.
Ferreira highlights the importance of "care optionality" and tailoring interventions to individual needs. However, Aetna is cognizant of the industry's concerns regarding AI potentially leading to a cookie-cutter approach to care, which might result in inappropriate recommendations for cost savings.
Considerations and Safeguards
Before implementing AI for triage, Isamu Pant, an independent behavioral health consultant, emphasizes several crucial considerations. He warns against proprietary triage processes and stresses the need for clinically accepted methods and measurement tools. Pant advocates for recommendation-based AI, where the technology raises concerns to clinicians rather than making definitive decisions.
Pant, who previously worked on AI at Amazon and conducted workshops for the National Association of Addiction Treatment Professionals (NAATP), highlights the importance of understanding AI biases and inherent flaws. He suggests balancing AI insights with human judgment, incorporating social determinants of health, and ensuring transparency.
Legal Aspects and Safeguards
Victoria Cvitanovic, an AI, cyber, and behavioral health law attorney at Rudick Law Group, points out the legal risks associated with AI triage. While AI can enhance efficiency, its use raises distinct concerns regarding accountability and human oversight. Cvitanovic questions the informality of AI recommendations, potential manipulation by external factors like advertising payments, and the need for transparent audit trails to address bias.
She emphasizes that executives should prioritize accountability and human oversight over efficiency, ensuring that AI in triage is ethically sound and legally compliant.