Rett syndrome, a rare neurodevelopmental disorder, gained its first approved treatment in 2023 with trofinetide. However, tracking how patients respond in real-world settings remains difficult because it relies on labor-intensive chart review. A new preprint on arXiv describes a dual large language model (LLM) consensus pipeline intended to automate this process.
The pipeline uses two LLMs to analyze clinical notes and document treatment phenotypes, with a consensus mechanism to improve reliability. The authors position this as a scalable alternative to manual review, which could support ongoing monitoring of trofinetide's effects outside controlled trials.
The paper's abstract focuses on the objective and design; no results are reported in the available text. This is a methods-oriented contribution, and its practical utility will depend on validation against human chart review in future work.