Psychiatry Measurement & AI: Improving Visibility & Outcomes in Behavioral Health
The conversation around artificial intelligence in healthcare often centers on potential and disruption. But a recent discussion with Taft Parsons III, MD, chief psychiatric officer at CVS Health/Aetna, highlighted a more fundamental challenge: visibility. The current system, reliant on claims data, often obscures the nuances of care, hindering both quality improvement and trust between payers and providers.
Dr. Parsons emphasized that what payers currently receive is simply a claim – a coded record of a service provided. This claim lacks crucial details about the care itself, such as whether measurement-informed care was utilized, if treatment aligned with evidence-based guidelines, or even if the patient is demonstrably improving. This flattening of complex clinical encounters into simple claims is a core issue, and one that AI could exacerbate or alleviate depending on how it’s governed.
The Limits of Current Data
The problem isn’t necessarily payer indifference, Dr. Parsons explained, but rather the constraints of the existing system. Payers are limited to the information contained within claims. Access to detailed patient charts is often limited, and standardized outcomes data is lacking. This creates an architecture of mistrust, forcing both parties to operate with incomplete information.
This lack of visibility extends to the crucial area of measurement-informed care. While the importance of consistent measurement is widely acknowledged, the distinction between “performative” and “functional” measurement is often overlooked. Performative measurement involves administering a tool simply because it’s required, without altering clinical decisions based on the results. Functional measurement, actively informs treatment adjustments and deeper assessments.
Dr. Parsons highlighted that consistent measurement, even with imperfect instruments like the PHQ-9, is valuable because it forces clinicians to reassess patient outcomes when progress stalls. His experience at Kaiser Permanente demonstrated that consistent measurement correlated with roughly double the positive outcomes, not because of the instrument itself, but because it prompted reassessment and adjustments to care plans.
Acknowledging Variation in Quality
Perhaps the most candid point Dr. Parsons made was the acknowledgement of significant variation in the quality of behavioral healthcare. He noted that chart audits often reveal a gap between clinicians’ self-assessment of their care and the actual quality of care delivered. This isn’t an accusation, but a recognition of a reality that must be addressed to drive improvement.
He further pointed out that many innovative care models don’t statistically outperform usual care, not necessarily due to ineffectiveness, but because the current measurement infrastructure is unable to capture the nuances of improved care. This underscores the need for better data collection methods and a more sophisticated understanding of what constitutes quality in behavioral health.
The Infrastructure Gap and the Role of Technology
Even with a commitment to measurement, many clinicians lack the necessary infrastructure to effectively implement it. Many providers still document outside of interoperable electronic health records, sometimes even on paper, making outcomes data invisible to payers. The Wyanoke Group, parent company of Healio, is working to improve this through platforms like Healio, which provides news, information, and education for healthcare professionals.
Aggregator platforms like Headway, Alma, Grow, and Rula are beginning to address this gap by enabling measurement transmission at scale. Although, smaller, single-shingle practices often remain underserved by modern tools and value-based models. Dr. Parsons emphasized that if professional associations advocate for specific practices, they must also ensure that the necessary infrastructure is in place to support them.
Aetna’s Stance on AI Governance
Concerns about AI being used to automate denials are prevalent among clinicians. Dr. Parsons addressed this directly, stating that Aetna has a firm policy against using AI to make adverse decisions on cases. The company’s focus is on leveraging AI to reduce administrative burden and free up clinicians to focus on patient care, specifically through ambient documentation – AI-generated drafts of clinical notes.
However, he cautioned that governance is an ongoing discipline and that maintaining this standard will require vigilance as AI capabilities evolve and cost pressures mount. The goal is to create a “human-in-the-loop” system where AI assists clinicians without compromising their judgment or patient care.
Beyond the Visit: The Importance of Continuous Care
The future of behavioral health, Dr. Parsons suggested, lies in recognizing the importance of the work that happens *between* visits. This includes ongoing monitoring, support, and adjustments to treatment plans. The challenge is determining how to appropriately reimburse for this continuous care and ensure it’s not an invisible burden placed on clinicians.
He acknowledged that the payment models are still evolving and that the tools will help clarify the appropriate structures in the coming years. Platforms like Zenara Health are exploring ways to integrate continuous care into value-based models, utilizing chronic care management codes to support ongoing patient engagement.
The Need for Shared Understanding
Dr. Parsons emphasized the importance of recognizing that payers and clinicians have different, but not necessarily conflicting, objectives. Payers manage populations, while clinicians focus on individual patients. Acknowledging this distinction is crucial for designing systems that honor both perspectives.
He also stressed the need to address fraud and discrepancies between billed and delivered care, as these undermine trust for everyone involved. Aetna actively investigates such cases to ensure that legitimate care is appropriately reimbursed.
The key takeaway is the need for visibility. If quality is not visible in the data, it remains invisible to those who fund care. By improving data collection, embracing measurement-informed care, and fostering a shared understanding of quality, the healthcare system can move towards a more effective and equitable future for behavioral health. SLACK Incorporated’s 2026 Journal Subscription Rates provide further insight into the evolving landscape of medical publishing and data dissemination.
Ravi Hariprasad, MD, MPH, is a board-certified psychiatrist, public health physician, and engineer trained in operations research at Cornell University. He previously architected collaborative care at Kaiser Permanente Oakland and Headspace Health. He is the founder and CEO of Zenara Health, which designs human-in-the-loop AI platforms, clinical workflows, and safety governance for health systems and digital health companies. He can be reached at [email protected].