Novartis

Clinical trial site selection was a labor-intensive process taking weeks or months, requiring analysis of vast datasets to identify sites with the right infrastructure, patient population, and investigator expertise. Novartis also struggled to ensure diversity and inclusivity in trial recruitment, and relied on disparate data sources that limited decision-making quality.

Novartis developed an AI-powered platform drawing on data from 460,000 clinical trials, 700,000+ clinical sites, and 600,000 industry-wide principal investigators to automate feasibility analysis and site selection in minutes. The platform incorporates a 'Unified Ontology' digital twin integrating semantic and kinetic data elements, along with patient diversity data covering 1,100 diseases. Investigators are scored with representative 'HAT' profiles (e.g., fast-starter, high-performance) to optimize recruitment outcomes.

In a US pilot involving ~1,700 patients, PIs with high representative HAT scores recruited 2.7x more targeted Black or African American patients than peers, significantly improving trial diversity. PIs with combined fast-starter and high-performance HATs achieved 3.4x higher than median recruitment rates. The initiative reduced site selection timelines from weeks/months to minutes.

• AI-driven site selection can dramatically accelerate recruitment timelines while simultaneously improving demographic diversity when investigator performance profiles are incorporated into selection criteria.
• Unifying disparate internal data sources into a single ontology-based platform is a prerequisite for scalable AI-powered decision-making in clinical operations.
• Human capability-building alongside technology deployment is essential — Novartis invested in training programs to ensure employees could effectively leverage the new tools.