CohortFlow speeds up clinical trial group curation with targeted updates
CohortFlow: A Steerable Clinical Cohort Curation Agent with Localized Semantic Correction
Artificial Intelligence
Summary
Choosing the right patients for medical studies can be tricky and slow when changes need redoing everything. The authors created CohortFlow, a tool that makes small, focused updates to patient group definitions instead of restarting from scratch. This approach keeps earlier work intact when possible, making corrections faster and requiring less computing effort, while keeping accuracy high. They tested CohortFlow on clinical trial rules and real healthcare data to show it works well.
What this means in practice
- •For hospital data teams: Speed up updates to patient groups in clinical research by only revising changed criteria, reducing delays in cohort selection.
- •For healthcare software developers: Build smarter clinical trial cohort tools that cut computation and speed correction steps by tracking dependencies between criteria.
Authors
Neil Chen, Adrian Serapio, Hanxue Gu, Evan Lee, Kang Wang, Yang Yang
Abstract
Language model agents for clinical cohort curation should accommodate evolving eligibility criteria without repeatedly processing unchanged requirements. We introduce CohortFlow, a steerable agent that represents natural language corrections as localized updates to a structured cohort specification and reconciles them with a persistent execution graph. Explicitly recorded semantic dependencies enable affected computations to be selectively invalidated while preserving valid prior results. On 100 scenarios derived from clinical-trial eligibility criteria, comprising 126 correction turns, localized steering achieves 77.0% requirement micro-F1 versus 76.4% for full regeneration using the same 8B model, a small numerical advantage that is not statistically significant ($p=0.493$). It reduces mean correction latency by 64.8% after completion and 44.3% during execution, with 38.1% fewer model calls. In a separate proof-of-concept evaluation over MIMIC, all 13 cohort states execute successfully, while correction trajectories demonstrate selective preservation of unaffected computations and recomputation of changed or dependent nodes. Together, these results show that localized semantic steering can substantially reduce correction latency and model calls while maintaining strong requirement-level accuracy.