Resumen
Obtaining and structuring information about the characteristics, treatments, and outcomes of people living with cancer for research purposes is difficult and resource-intensive. Oftentimes, this information can only be found in electronic health records (EHRs). In response, researchers use natural language processing with machine learning (ML extraction) techniques to extract information at scale. This study evaluated the quality and fitness-for-use of EHR-derived oncology data curated using ML extraction, relative to the standard approach, abstraction by trained experts. Using patients with lung cancer from a real-world database, we performed replication analyses demonstrating common analyses conducted in observational research. Eligible patients were selected into biomarker- and treatment-defined cohorts, first with expert-abstracted then with ML-extracted data. The study?s results and conclusions were similar regardless of the data curation method used. These results demonstrate that high-performance ML-extracted variables trained on expert-abstracted data can achieve similar results as when using abstracted data, unlocking the ability to perform oncology research at scale.