OBJECTIVES: Retrospective medical claims data are a cornerstone for health economics and outcomes research (HEOR) studies, but they lack clinical specificity and offer limited level of detail in specific care settings. This study aimed to identify constraints that hinder answering specific research questions within retrospective real-world studies reliant solely on claims data and systematically compare these limitations against alternative real-world evidence (RWE) sources, such as hospital service-level data, electronic health records (EHR), laboratory data, and physician affiliations data. The research sought to delineate potential inadequacies of claims data and propose strategies for leveraging diverse real-world sources to address these limitations and enhance the depth and reliability of retrospective real-world studies.

METHODS: A systematic assessment of different RWE sources, including claims, hospital service-level data, EHR, and laboratory data, was conducted to assess factors such as data granularity, data completeness, and capture of clinical and economic metrics for specific HEOR use cases. The assessment framework was paired with a literature review of published material on data sources, product catalogs, and data dictionaries based in the United States.

RESULTS: For each HEOR use case, 6-10 data products were evaluated. Over half lacked at least two or more critical components, such as clinical specificity, cost information, longitudinal capture, linkability, demographic or geographic representativeness or level of reporting detail. In use cases requiring visibility into inpatient hospital or intensive care settings, claims datasets provided little to no insight into treatments, physician encounters, or service details. Hospital billing or electronic health records (EHR) datasets provided treatment specificity but low representativeness. Linked sources (e.g., claims + EHR) often excluded critical metrics (e.g., mortality) to comply with HIPAA regulations.

CONCLUSIONS: Novel analytical design techniques were documented that allowed for capture of additional clinical specificity in several use-cases through the combination of two or more unlinked data sources while enabling population-level insight generation.