Objective: Cosmos is a vendor-facilitated platform created in collaboration with healthcare systems using electronic health records (EHRs) from Epic Systems Corporation. Participating organizations can use the platform to advance research and apply insights at the point of care. This manuscript outlines evolutions in the platform’s infrastructure, growth of available data, and usage. Methods: Data from participating systems is sent via encrypted Health Level Seven Clinical Document Architecture to the Cosmos host at Epic Systems Corporation. Deduplicated data is stored in two relational Structured Query Language (SQL) databases, which can be accessed remotely. Aggregate counts from a limited data set (LDS) are accessed through a no-code data visualization and analysis tool. Line-level data from an expert-determined de-identified dataset is accessed in a secure virtual computing environment. Data quality and user support frameworks facilitate a feedback loop improving data quality and researcher efficiency. Platform growth was measured as participating organization counts, data elements, and proportion of patient records with data spanning over five years . Medical literature was reviewed for peer-reviewed studies using Cosmos data. Results: Over 330 organizations across Canada, Lebanon, Saudi Arabia, and the United States submit data to Cosmos. The platform includes more than 300 million deduplicated patient records, 39% of which have over five years of clinical data. Ongoing expansion of data elements within Cosmos has allowed the platform to serve as the data source for over 100 peer-reviewed studies across multiple disciplines. Discussion: With its expanding size, tooling, and breadth of data elements, Cosmos may enable a wide range of study types. Cosmos remains subject to limitations such as variation in workflows, documentation error, and bias inherent to a population sourced from healthcare organizations using a single EHR vendor. Conclusion: Cosmos demonstrates a use case of vendor-facilitated collaboration to enable a broad spectrum of research.
Martell et al. (Fri,) studied this question.