Evaluating ontology coverage and internal structure to better align patients with clinical trials

Abstract

ABSTRACTRecruiting patients for a clinical trial is a time‐consuming activity that can delay recruiting and the reporting of clinical trial results. Ontologies can be used to align the inclusion criteria of a clinical trial with candidate patients but such an approach requires that (a) the concepts within an ontology adequately cover patient characteristics that are likely to be used in a clinical trial; (b) the quantitative similarity scores decline gracefully patient characteristics become less suited to the trial protocol; and (c) the alignment must be robust despite surface level differences in how eligibility features are described. The concepts are extracted from grant proposals and eight archetype breast cancer patients are mapped as matching a breast cancer clinical trial protocol does. The Medical Subject Heading (MeSH) and National Cancer Institute Thesaurus (NCIt) are evaluated with respect to coverage, connectivity, standard similarity properties and the granularity of rankings.