A Logical Representation for Capturing the Context of Observations and Quantitative Information in Clinical Trial Reports
- Author(s): Tong, Maurine May-Lin
- Advisor(s): Taira, Ricky K
- et al.
Clinical trial experimental studies are the gold standard for obtaining evidence related to interventions for a given disease or chronic condition, and currently results are documented in free-text reports. Due to the current free-text representation, utilizing knowledge from these studies and interpreting results remains an ongoing challenge. This dissertation proposes a bridge representation that transforms information in clinical trial reports from a free-text format to a representation that is computer understandable and capable of assisting answering high level queries from bio-statisticians and clinicians. The objectives of this work are: (1) to specify a representation that will concisely synthesize fragments of information found in clinical trial reports, so users can readily understand the context of numerical data, follow the flow of the study, and assess the quality of the study; and (2) to support queries related to assessment of study quality and estimation of contextualized probabilities derived from various sections within the report (e.g., survival curve, p-values, etc.). The representation is based on a hybrid structure combining several modeling paradigms to create an intuitive and standardized way of describing the conditions of the experiments, the data generated, the analysis methods and the results. Query processing and navigation methods have been designed to operate on the representation to answer common questions related to clinical research, from the clinical and biostatistics side. Such queries include defining the conditions of the patient cohort and interventions, providing context to numerical frequency information, and providing a comprehensive summary of the methods used to compute statistical significance. The focus of the dissertation has been in the clinical research domain of oncology. The dissertation work offers a value-added and time-saving solution to standardizing and organizing information from clinical trial reports and synthesizing knowledge to advance clinical research.