How LLMs can Improve Care Quality in Behavioral Health, Expand Access, and Accelerate the Shift to Value-Based Care
The next frontier in mental health care delivery 📱
After last weeks’s drop on Measurement Based Care (MBC), we continue the theme and turn to continuous care and the opportunity ahead of us thanks to LLMs (Large Language Models). To keep things fresh, we’re trying out a more concise format - we welcome your feedback!
While we are both clinician technologists at heart, neither of us would dare pitch ourselves as AI or LLM experts. As you’ve come to expect from Two Docs and a Stack, we’ve teamed up with one of the foremost healthcare LLM experts out there to bring you this piece. Please join us in welcoming Luis Voloch - an MIT-trained computer scientist, experienced healthcare founder (previously at Immunai and now at Jimini Health), and lecturer at Stanford’s GSB.
LLMs in Behavioral Health
Imagine a world in which everyone seeking mental healthcare has a real-time, accurate pulse on their well-being and knows precisely how they are responding to treatment. In healthcare delivery, AI can revolutionize how we deliver psychotherapy and enable continuous care models.
Through the adoption of Large Language Models (LLMs), we will soon be able to create continuous care models that improve both care quality and our ability to assess patient well-being. Continuous care means providing ongoing support and treatment to patients, rather than relying solely on episodic sessions. We call this moving treatment beyond the "four walls" of the office (or Zoom session). This combination of continuous care and high-fidelity outcome data will, in turn, accelerate behavioral health's shift into its value-based era.
LLMs can transform psychotherapy by facilitating always-available communications (primarily via text, but also audio and more) between patients, therapists, and AI coaches. This real-time, text-based approach would be unthinkable without the help of LLMs, as the amount of therapist time required would be staggering. But through AI, we can give individuals access to timely support whenever they need it, making therapy more effective, more accessible to a broad and diverse population, and more suited to implementing value-based care.
What is continuous care?
Continuous care powered by AI brings revolutionary features: much more granular and detailed data about patients and the ability to use AI to mine that data for insights. This data creates a remarkable potential to advance our understanding of patient well-being.
Traditionally, therapy sessions occur at fixed intervals, leaving massive gaps in our knowledge of patients' emotional states between appointments. In a typical weekly treatment cadence, a therapist is with the patient for just 1 of the 168 hours in a week. With continuous care, we can bridge these gaps by providing a channel for patients to express their thoughts and emotions through their phones and computers any time they want, giving us far more granular data about how they are doing.
AI can mine these interactions, identifying patterns, changes in mood, and even subtle nuances in language that may signal shifts in mental and physical health. This large, continuous data stream offers clinicians a real-time, deeper window into their patients' well-being, allowing clinicians to detect issues earlier, intervene when necessary, and provide more personalized and timely support.
Several groups have already demonstrated that AI can assess aspects of patient mental health using standardized scales such as the PHQ-9 and GAD-7 with clinician-level accuracy. Moreover, the Machine Learning literature around the accurate assessment of medical conditions is now robust and has shown that LLMs can successfully assess various health conditions.
By harnessing AI's data-driven insights, therapists can respond to patients more quickly and effectively, track progress more closely, and ultimately contribute to better outcomes and a more holistic understanding of patient well-being throughout their therapeutic journey. This combination of continuous care and AI-based insights promises a step-function advancement in psychotherapy, enhancing the quality of care and ultimately improving the lives of those seeking help.
Continuous care powers outcome measurement…
The enhanced understanding of patient well-being facilitated by continuous care and AI-based insights represents a gold mine for measuring outcomes in the context of value-based care. Traditional methods of measuring therapeutic outcomes rely on periodic assessments or self-reported data, which can be subjective and reflect only a snapshot in time. However, continuous care will give us a comprehensive and dynamic picture of a patient's progress over time. This wealth of data not only allows for more accurate and objective measurement of treatment effectiveness but also opens the door to the development of value-based reimbursement models for therapy.
By tracking tangible improvements in mental health over time, clinicians and payers can establish clearer metrics for determining the value of psychotherapy. These metrics, in turn, pave the way for payment models that reward therapists and healthcare organizations for real outcomes rather than the quantity of services.
Enabling direct feedback to clinicians and patients
The wealth of data generated by AI-enabled continuous care also serves as a valuable resource for providing feedback to clinicians and driving professional development for therapists. In the traditional therapy model, feedback is often based on limited and intermittent patient interactions (at most one hour per week). In contrast, in continuous care models, data will be available many times per week, providing more visibility. With a constant stream of data, therapists can receive real-time information about their patients' progress, enabling them to adjust treatment plans and interventions. This feedback loop empowers therapists to fine-tune their approaches, optimize their strategies, and provide better care.
Moreover, the data can be anonymized and aggregated to identify broader trends and best practices within the field of psychotherapy and used to develop training programs, establish benchmarks, and drive ongoing professional development. Through data, we can foster a culture of continuous improvement and provide therapists with the knowledge and techniques to deliver the best possible care to their patients.
Our take on how AI-assisted care will give mental health clinicians (and the industry) superpowers
Democratizing mental health care access and improving equity
This new paradigm promises to significantly advance health equity and improve care quality across the board. By leveraging AI and continuous care, we break down the traditional barriers to access that have long hindered marginalized communities' abilities to receive timely and effective care.
Patients can reach out for support whenever they need it, bypassing the constraints of appointment schedules and geographical limitations. Moreover, AI's ability to analyze data and detect subtle shifts in language and mood allows for heightened cultural sensitivity and personalization in therapy. It can help ensure that care is not only more accessible but also more responsive to the diverse needs of different individuals and communities.
As we advance toward value-based reimbursement models, this approach aligns incentives to prioritize patient outcomes over quantity of care, further promoting equitable care. In this way, integrating AI, continuous care, and data-driven insights can level the playing field, make quality psychotherapy accessible to all, and foster a more inclusive and culturally sensitive mental healthcare system.
We’re all excited to see where this leads for the behavioral health industry as a whole, and will be eagerly watching as smart, passionate, and motivated builders and operators bring their talents to the (in our opinion) best sub-sector in healthcare.
✌🏽 A + A
To read more about our vision for the Stack, check out our intro post here.
 |
|
As always, enjoyed y'all's thoughts!
Do you see LLM based check-ins (for lack of a better term) as completely unsupervised? Similar to Dr. Attia's question below, would the future be via a medical device approval pathway and/or would providers be able to build their own agents and, perhaps, have to attest that they will supervise/keep a human in the loop as the agents interact with their patient panel? Really hard questions that we'll have to answer in the coming months and years.
Love this, thanks for a very relevant article!
I am unabashedly bullish on the necessary (and inevitable?) adoption of ML-powered tools in healthcare, and not just for front office functions like automated charting. These tools should start to touch care relationships directly, in my view. Two follow-ups come to mind:
First, what are some other data streams you all would like to see feeding into such ML-powered tools? Continuous engagement with an LLM itself would provide useful data, but I wonder if there's an opportunity to layer on additional insights from, say, the patient's smartphone, at-home wearables, etc.
Second, when, if at all, do you think we'll see an LLM approved as a SaMD to provide CBT for certain conditions, like mild/moderate depression? To my knowledge, healthcare still hasn't crossed this rubicon, but I wonder what your thoughts are on pros/cons.