(142) A Descriptive Study Comparing Artificial Intelligence (A.I.)-Driven Preferred Specialty Drug Network Versus Traditional Specialty Drug Network on Drug Costs in a Multi-State Regional Health Plan

Background: : Amid escalating healthcare expenditures, pharmacy benefit managers (PBMs) are under increasing pressure to deploy data-driven, innovative strategies to manage drug costs effectively. The rapid adoption of artificial intelligence (AI) across the healthcare sector — particularly in applications such as predictive analytics, utilization management, and network optimization — has positioned it as a transformative tool for PBMs.1,2,3,4 While early adoption has demonstrated operational promise, real-world evidence evaluating AI’s impact on drug cost containment remains limited.4

PBMs contract with specialty pharmacies at varying reimbursement rates; however, patient self-selection may lead to utilization of higher-cost, non-preferred pharmacies. RealRx leverages artificial intelligence (AI) within the prior authorization process to direct fulfillment to the lowest-net-cost pharmacy for each drug. This evaluation addresses a key evidence gap in AI-driven network optimization strategies in managed care.

Objective: : To evaluate the cost impact of implementing an AI-optimized preferred specialty pharmacy (PSP) network compared to a traditional specialty pharmacy (TSP) network within a multi-state regional health plan.

Methods: : This retrospective, descriptive analysis used pharmacy administrative claims data and contracted specialty pharmacy network rates from a regional health plan between April 1, 2024, and March 31, 2025. The analysis focused on the 10 highest-expenditure specialty drugs by Average Wholesale Price (AWP). Per-drug costs were calculated by applying the best network rate (PSP rate) and the third-best network rate (TSP rate) to each claim’s AWP. These calculated costs were then compared for each individual drug and aggregated.

Results: : Annual costs across all 10 drugs were lower under the PSP model, resulting in a total cost differential of $1.726 million. The savings observed for all 10 drugs were as follows: Trikafta ($609k), Humira ($231k), Stelara ($205k), Rinvoq ($163k), Enbrel ($155k), Dupixent ($106k), Skyrizi ($84k), Taltz ($78k), Verzenio ($55k), and Cimzia ($41k).

Conclusions: : AI-driven specialty network optimization yielded substantial cost savings compared to traditional specialty pharmacy selection methods. These findings suggest that integrating AI into PBM core operations can enhance financial performance by aligning high-cost therapies with the most cost-effective dispensing channels. Nevertheless, further research is warranted to explore AI’s broader implications on clinical outcomes, operational efficiency, and effective drug utilization in managed care.

Sponsorship:: None