Validation of Accuracy for LLM/ML-Extracted Information and Data — a comprehensive framework for evaluating the quality of clinical data extracted by large language models, advancing industry standards for trustworthy AI-powered evidence generation in oncology.

Representativeness analysis and mortality endpoint validation in EHR-derived oncology cohorts from Germany, demonstrating 87.7% sensitivity and 91.7% specificity for vital status classification. First peer-reviewed validation of German EHR oncology data.

Describes the general approach for applied NLP and ML methods used to extract data from unstructured EHR documents in oncology, developing 10 ML models outputting 20 distinct RWD variables. One of Blythe’s most cited first-author papers.

A public-facing policy argument for school reopening based on epidemiological evidence, published at the height of the pandemic debate. Demonstrates Blythe’s ability to translate complex science for a mass audience.

Modeling the epidemiological and economic case for an HIV vaccination campaign strategy in South Africa.

Presented at an ASCO plenary session; covered by NPR, CNBC, and The Cancer Letter. Published in The American Journal of Managed Care (2021).