Agentic data science with the Data Agent Kit
cloud.google.com·6h ago
AI Research
arXiv CS.AI·Aug 7·5 min read
Tracing the Heart: An Evidence-Linked Pipeline for Heart-Failure Feature Engineering
TL;DR
Nimblemind Multi-Agent System (nMAS) automates heart-failure feature engineering from electronic health records, significantly improving predictive accuracy for heart failure phenotyping. It achieved an AUROC increase from 0.895 to 0.963 for HFrEF and from 0.870 to 0.910 for HFpEF.
✦ Why It Matters
Consider implementing nMAS for more efficient and accurate heart-failure feature engineering in your clinical research.
Key Takeaways
- Explore nMAS for automating EHR feature engineering.
- Utilize structured and rubric-scored features for better model performance.
- Assess the impact of feature engineering on predictive accuracy.
- Plan for external validation of automated systems in diverse settings.
How It Works
nMAS integrates fragmented EHR data with clinical guidelines to automate feature engineering. It generates structured features and scores them based on a rubric, ensuring compliance and traceability.
The system uses a restricted large language model (LLM) for auditing, enhancing the reliability of the generated features.