PATHFinder Agent: AI-Powered Tailored Prenatal Care
The PATHFinder Agent, an AI system designed to enhance prenatal care, collects patient health information and social context through structured conversations. This initiative aligns with the guidelines set forth by the American College of Obstetricians and Gynecologists (ACOG) under the PATH (Plan for Tailored Healthcare) framework. It creates personalized care plans and highlights community resources available through Michigan 211. The system operates in four stages: patient intake, dynamic interaction, plan synthesis, and clinician oversight. An evaluation of advanced large language models (LLMs) using expert-curated criteria across five clinical dimensions revealed that GPT-5.2 attained the highest average score of 77.6%, while also pinpointing significant gaps in antenatal testing recommendations.
Key facts
- PATHFinder Agent is an end-to-end conversational agentic system for tailored prenatal care.
- It is based on ACOG's PATH (Plan for Tailored Healthcare) guidelines.
- The system gathers patient health and social context through structured dialogue.
- It curates individualized prenatal care plans and surfaces community resources from Michigan 211.
- The workflow has four stages: patient intake, dynamic interaction, plan synthesis, and clinician oversight.
- GPT-5.2 achieved the highest average score (77.6%) on expert-curated rubrics.
- The evaluation covered five clinical dimensions.
- Key gaps in antenatal testing recommendations were identified.
Entities
Institutions
- American College of Obstetricians and Gynecologists (ACOG)
- Michigan 211
Locations
- Michigan