Signal New Pipeline Targets EHR Feature-Engineering Bottleneck for Heart Failure Research

Summary

A preprint by Soorya Ram Shimgekar, Michelle Hu, Dorisa Shehi, Daniel Kang, and colleagues describes electronic health record feature engineering as a major bottleneck in clinical research and AI, one that consumes an estimated 39 to 45 percent of data scientists' working time. This bottleneck is especially acute for heart failure, a condition affecting an estimated 6.7 million US adults, because it requires integrating fragmented EHR data with disease-specific, guideline-based clinical reasoning. The authors argue that both existing rule-based methods and large-language-model approaches fall short of handling that integration challenge. They respond with an 'evidence-linked' pipeline designed to trace clinical features back to the supporting evidence found within patient records. The stated goal is to make AI-assisted heart-failure research more reliable and auditable.