Specialty Pharmacy ContinuumExperts Call for Stronger Healthcare AI Oversight

🎓 Expert Commentary / Peer Perspective

Experts speaking at AXS26 called for stronger governance of healthcare AI technologies, particularly those influencing patient care decisions, documentation, risk assessment, and operational workflows. Shawn Griffin, MD, of URAC argued that healthcare organizations often lack adequate assurances that AI tools have been sufficiently validated for safety and accuracy. Speakers emphasized the importance of understanding how AI models are trained, tested, monitored, and reviewed before deployment. While AI may improve medication adherence programs, prior authorization workflows, and predictive analytics, concerns remain regarding bias, transparency, reliability, and potential effects on health equity. The discussion focused on responsible implementation rather than broad adoption or rejection of AI technologies.

Professional Impact

• Healthcare organizations are increasingly expected to evaluate how AI systems were trained, validated, and monitored before operational use.
• AI tools may support medication adherence initiatives by identifying patients at risk for nonadherence and facilitating targeted interventions.
• Analysis of free-text clinical records may help identify barriers to therapy, including cost concerns and medication-related challenges.
• AI-assisted prior authorization workflows may reduce administrative burden and accelerate processing for approved therapies.
• Speakers stressed that human oversight remains essential, particularly when AI-supported decisions affect treatment access or coverage determinations.
• Bias within training datasets may contribute to healthcare disparities if model performance varies across patient populations.

Action Items

• Review vendor evidence regarding model validation, accuracy, and performance in relevant patient populations.
• Evaluate whether AI outputs are independently monitored for bias, drift, or inaccurate recommendations.
• Establish governance processes defining when human review is required before acting on AI-generated recommendations.
• Monitor AI-assisted workflows for unintended effects on patient access, equity, and clinical operations.
• Document oversight practices and accountability structures for AI tools integrated into pharmacy or managed care workflows.