Fallback to Ad Hoc Phenotype Creation Given Poor/No Recommendations

[jshoughtaling]

@rkboyce et al - this is amazing work. We chatted on the phenotype workgroup call a few weeks ago, and I'd be very interested in contributing some work I'll present at OHDSI global this fall on CAPR-based iterative phenotype development using LLMs. I'm not sure if that fits within your vision for the StudyAgent, but after some initial testing with this tool I do see scenarios where it may be advantageous to develop fresh phenotypes on the fly when there are no suitable definitions available in the index/catalog. Please let me know if you'd like to setup a brief call to discuss further.

[rkboyce]

Hi @jshoughtaling - So sorry for such a long delay in replying. I would love to chat with you. That feature is definitely high priority on the road map. We might be able to integrate tools like Ariadne and the logic implemented in others like Phenelope to accomplish a robust fresh phenotype creation agentic flow. Could you please reach out to me directly at my email address (rdb20@pitt.edu) with suggested days/times? Please put in the subject line Study Agent and tag me here when you send it b/c I have to filter my email. You could alternatively try to tag me on the OHDSI teams with a similar message.