611 Advancing data harmonization in a federated model/network for assessing patient characteristics in systemic lupus erythematosus

Background/Purpose Systemic lupus erythematosus (SLE) is an autoimmune disease with heterogeneous manifestations across patient populations. Although real-word evidence data on SLE exists via SLE registries worldwide, the data collected vary by registry due to variations in genetics, practice, and availability of therapeutic options. The Lupus Federated Data Network (LupusNet) is an interdisciplinary initiative aiming to combine and harmonize data from existing local SLE registries to create a federated, unified, global SLE database with increased patient counts and data consistency. Here, we present the harmonization methodology, as well as opportunities and challenges related to the harmonization process to inform rheumatology-related research in a federated model. Methods LupusNet includes patient data as collected in various SLE registries in the North America, South America, Europe, and Asia Pacific regions. Multiple demographic and clinical variables are collected in these registries (e.g., patient demographics, disease activity/severity, clinical events, biopsies/histology, SLE biomarkers, treatment history, comorbidities, comedication, and patient-reported outcomes). Each dataset was harmonized to the Observational Medical Outcomes Partnership (OMOP) Common Data Model v5.4. The semantic mappings involved OMOP standard terminologies (i.e., SNOMED CT for diagnosis, symptoms, severity scales; LOINC for lab tests; RxNORM and ATC classification for medication) and additional terminologies for concepts not present in OMOP standard terminologies (e.g., patient-reported outcomes, severity scales). Applied mapping conventions can be consulted at https://github.com/OHDSI/ETL--LupusRegistries/. With this federated approach, researchers can define an analysis without access to each dataset. A script can be shared and executed against each local harmonized dataset upon registry approval. Results Implemented conventions allow an accurate representation of source data with an increased sample size. The harmonization and mapping process of variables in each registry allows for improvements in data quality as well as a more efficient and informative analytic dataset that can be utilized for sub-population analyses. To date, 1833 concepts have been mapped to the Common Data Model from 3 registries (1674 in 1 registry; 137 in 2 registries; 17 in 3 registries). Across registries, data collection occurs at different time intervals, employing different disease measures or severity scales (e.g., SLEDAI, SLEDAI-2K, BILAG, LLDAS, SLAM, ECLAM), and documentation of medications may vary, including discrepancies in levels of granularity, differences in reporting medication use duration, and lack of recording comedications or occasional exposure to other drugs. These potential variations will be taken into consideration when analyzing results. Assumptions and algorithms will be created to harmonize these data (tables 1 and 2). Conclusion Mapping patient characteristics from LupusNet will allow researchers to analyze a larger population of patients with SLE to better understand disease heterogeneity, patient populations, and treatment patterns with the goal of improving patient outcomes. Lay summary Systemic lupus erythematosus can affect patients differently. Here, we report methodologic challenges associated with harmonization of the data within the Lupus Federated Data Network (LupusNet). LupusNet is an interdisciplinary initiative aiming to harmonize data from existing systemic lupus erythematosus registries to create a global, unified database with increased patient counts and data consistency.