Open help menu
Live Webinar

AI-Driven Clinical Decision Support for Veteran PTSD Treatment: A Bayesian POMDP Simulation Study

Umit Tokac, PhD
|

Pricing

Information

Date & Time

Description

This workshop presents a simulation study examining the development and evaluation of a Bayes-Adaptive Partially Observable Markov Decision Process (POMDP) algorithm designed to personalize PTSD treatment intensity recommendations for veterans. The algorithm integrates a Bayesian hierarchical framework with Markov Chain Monte Carlo (MCMC) estimation to model latent PTSD severity trajectories and generate individualized, data-driven clinical recommendations. Using synthetic data modeled on realistic VA clinical presentations, including PCL-5, PHQ-9, and Q-LES-Q-SF scores across three therapy sessions, the study demonstrates proof-of-concept that such a framework can accurately estimate hidden patient states and adaptively refine treatment decisions over time. Participants will explore how AI-driven sequential decision-making approaches address a critical gap in veteran mental health care, where only 4–9% of veterans with PTSD receive adequate evidence-based treatment and approximately two-thirds retain their diagnosis after completing care. The workshop will discuss both the promise and current limitations of this approach, including the path from simulation validation to potential clinical implementation in VA settings.

Learning Objectives

Participants will be able to:

  • Describe the core components of a Partially Observable Markov Decision Process (POMDP) and explain how it models clinical decision-making under uncertainty in PTSD treatment contexts.

  • Explain how Bayesian hierarchical modeling and MCMC estimation are used to infer latent PTSD severity states from longitudinal, multi-instrument clinical data (PCL-5, PHQ-9, Q-LES-Q-SF).

  • Evaluate the methodological advantages and limitations of simulation-based validation for AI clinical algorithms prior to real-world implementation.

  • Discuss ethical, methodological, and practical considerations for translating Bayes-Adaptive POMDP models from simulation to actual VA electronic health record data.

Educational Goal

The educational goal of this workshop is to introduce clinicians to AI-driven clinical decision support as an emerging tool for personalizing veteran PTSD treatment, moving beyond one-size-fits-all protocols toward more individualized, data-informed care. Attendees will leave with greater sophistication in understanding how Bayesian and POMDP frameworks can inform clinical reasoning, positioning them to more critically evaluate and embrace precision medicine approaches as they enter real-world practice.

Target Audience

  • Addiction Professional
  • Counselor
  • Marriage & Family Therapist
  • Psychologist
  • Social Worker

Presenters

Umit Tokac, PhD
Umit Tokac, PhD is an Associate Professor of Data Science at the University of Missouri–St. Louis College of Nursing and a data scientist at the National Heart, Lung, and Blood Institute through the NIH Data Scholar Program. He holds a PhD in Measurement and Statistics from Florida State University and brings interdisciplinary expertise spanning data science, artificial intelligence, and health research. His work focuses on machine learning, geospatial modeling, and AI-driven methods to address public health challenges, including climate-related cardiopulmonary outcomes, personalized mental health treatment, and surgical risk prediction. Dr. Tokac has led federally and internationally recognized collaborations and is committed to research that advances health equity and social good. An award-winning educator, he teaches graduate-level biostatistics and data science using inclusive, active learning approaches and mentors students across disciplines.

Financially Sponsored By

  • APA Division 18: Psychologists in Public Service