(ASPOT03) The Value of Large Population-based Electronic Health Record and Insurance Claims Databases for Real-World Evidence Generation in Rare Diseases: a Systematic Literature Review

Background: Globally, ~10,000 rare diseases have been documented, affecting ~6% of the world's population. Due to the low individual disease prevalence, there are significant challenges in identifying existing disease-specific databases for clinical research or enrolling patients into new studies. Disease-agnostic databases covering extensive and representative parts of the society can help overcome this bottleneck. While there are many North American [NA] population-based data assets available for clinical research, the landscape of accessible non-NA databases and their feasibility for rare disease research are largely unexplored.

Objectives: To characterize the landscape of non-NA population-based databases for real-world evidence [RWE] generation in rare diseases.

Methods: We conducted a Systematic Literature Review [SLR] of 3 medical databanks: PubMed, Embase and Medline. To be included, articles had to be published after 2015 in English, German or French. Articles had to describe a rare disease study conducted in a non-NA population-based data asset and provide details on how rare disease patients were identified within the database. Our specific focus was on Electronic Health Record [EHR] and health insurance claims databases. Information on database type, size, region, the rare disease(s) studied, and the target population were extracted.

Results: Overall, 113 publications met our inclusion criteria. About 64% of the studies (73/113) utilized insurance claims databases. Among these studies, ~20% (14/73) featured database linkage between the claims data and a clinical registry database. Overall, 49 unique databases across 14 countries, including 27 insurance claims, 20 EHR, and 2 registry databases linked with claims, were identified. About 65% (32/49) of the databases were from the Europe – Middle East – Africa [EMEA] region, and ~35% (17/49) from the Asia – Pacific [APAC] region. Our search did not identify a single rare disease study performed in Latin American insurance claims or EHR databases. Claims databases covered a median population of 25 million [mio] patients (range = 0.6 – 125 mio), while EHRs covered a median population of 13 mio patients (range = 0.02 – 65 mio). Autoimmune diseases, predominantly myasthenia gravis, were the most frequently studied rare diseases in the databases identified.

Conclusions: Our search detected 49 unique databases across 14 countries in the EMEA and APAC regions, each having the potential to significantly advance RWE generation to ultimately improve health outcomes in patients suffering from rare diseases. Additional research is required to validate the algorithms used for rare disease patient identification within these databases and to harness artificial intelligence tools to enhance the classical SLR.