(217) Development of Computable Operational Definitions to Maximize Comparability & Consistency Across a Multi-Data Source Global Real-World Effectiveness Program
Background: In late 2021 multiple regulatory agencies began authorizing use of AZD7442, a combination of two monoclonal antibodies (tixagevimab/cilgavimab), for pre-exposure prophylaxis (PrEP) against COVID-19 in moderately to severely immunocompromised (IC) patients. However, the rollout and essential effectiveness assessment of AZD7442 faces several challenges including definitions of IC eligibility, prioritization of patients, dosage, distribution, common clinical practice, record-keeping around administration, and availability and completeness of key data which are heterogeneous. A global real-world study is being conducted which navigates this complexity.
Objectives: We developed computable operational definitions (cODs) to contextualize results across a multi-data source global real-world effectiveness program.
Methods: To maximize consistency and comparability across studies, data source-agnostic cODs were established for each study element (e.g., eligibility, exposure, outcomes) for the umbrella protocol. However, selected data sources (e.g., Clalit, VA, UPMC, DOD) are completing the analysis independently of each other to ensure context and uniqueness of the underlying health system is accounted for. Thus, cODs were tailored to reflect differences for each data source. Development of cODs was an iterative process, including review of literature and published protocols as well as input from epidemiology, statistics, informatics, and medical teams. Where possible, clinical concepts were represented by commonly structured data types and using standard validated code lists. The computable nature was facilitated via definition components documented at a granular level in a robust technical platform.
Results: cODs for 174 study elements were developed for the umbrella study protocol: 17 eligibility criteria, 1 exposure, 68 baseline characteristics, & 88 outcomes (for 16 objectives). These cODs encompass 38 distinct data variables and 82 standards-based code lists (further delineated by code system). All data sources required some adaptation, focused primarily on coding schemas (e.g., ICD-10-CM vs. ICD-9 vs. SNOMED diagnosis codes) or definitions unique to the data source (e.g., socioeconomic status, eligible IC alignment with national recommendations, setting of AZD7442 administration). Direct comparisons across data sources were demonstrated to highlight differences in cODs and their potential impact on the interpretation of analysis results.
Conclusions: Every data source is distinct within and across countries, and the unique context is important to interpretation of results. Data source-agnostic cODs are foundational to maximize comparability of study results, and support reproducibility.
