Objective: To explore implementing regular expressions (RegEx) to streamline patient identification and classification for matching to clinical trials.
Materials and methods: To prepare approaches needed to match patients to relevant cancer clinical trials, we combined NCI's Clinical Trials Search API to extract high-level eligibility criteria, including cancer type, stage, receptor/biomarker status, with similar data of patients with appointments in the upcoming week. Using RegEx, we prospectively identified all patients with breast, liver, or lung cancers at treatment decision points at 2 Cancer Centers' and 2 community hospitals', classified their cancer type, stage, and receptor/biomarker status. We evaluated accuracy using RegEx against manual reviews.
Results: Algorithm accuracy to identify patients at treatment decision points revealed 92% True Negative and 53% True Positive rate. Staging accuracy varied from 67% to 95%, and receptor/biomarker status accuracy from 76% to 86%.
Discussion and conclusion: Using RegEx significantly reduced the number of patients requiring manual review, demonstrating a reduction in manual labor and potential biases, which can improve efficiency and inclusivity of clinical trial enrollment processes, especially in resource limited or data sensitive environments.
Trial registration: NCT05146297.
Keywords: algorithms; cancer clinical trials; racial disparities; regular expressions.
© The Author(s) 2024. Published by Oxford University Press on behalf of the American Medical Informatics Association.