Data Modeling and Analytics to Advance the Secondary Use of Real-World Data in Healthcare Research

This dissertation investigates issues related to the secondary use of real-world data for healthcare research. It introduces a methodology to extract value from real-world data, enhancing efficiency, reducing errors, and uncovering insights hidden within non-compatible datasets. Applying data modeling and analytics principles to complex datasets generates meaningful analytical data for healthcare research. The central goal is to analyze real-world data stored in discrete structures and convert it into actionable insights to improve patient outcomes. Additionally, the research advocates transitioning from unstructured formats like free text files to discrete structures, enabling users to fully leverage database capabilities to enhance efficiency and analytics and reduce errors. The real-world data methodology developed for healthcare research through research experience gained by participating with clinicians across four Use Cases, three focused on cancer research. Adhering to the extended relational model, discrete structures served as the primary source of real-world data. This approach minimizes inefficiencies such as duplication and redundant workflows, often in unstructured data and external systems. The complexity of these systems can sometimes obstruct rather than optimize workflows. Prioritizing error reduction and efficiency, relying on discrete structures and the extended relational model, proves to be a trusted and effective strategy where precision is critical, particularly in healthcare. Results from the Use Cases underscore the transformative potential of this methodology: Use Case 1: Developed a patient cohort model involving 15,306 cancer decedents for a study on end-of-life care predictors. Use Case 2: Created two patient cohort selection models, minimizing the need for labor-intensive manual chart reviews. Use Case 3: Evaluated seven genomic clinical trials, highlighting the discriminatory power of disease specificity and physician interpretive indicators. Use Case 4: Analyzed historical hospital admission data from two community hospitals, demonstrating the methodology’s real-world applicability. The findings demonstrate that the real-world data methodology for healthcare research significantly improves the precision of patient screening for clinical studies, reduces the inclusion of ineligible cases, and enhances the efficiency of clinical study analysis. These contributions underscore the vital role of real-world data, advanced data modeling, and analytics in addressing the secondary applications of artificial intelligence algorithms. Ultimately, this research presents a transformative vision for the future of healthcare research.