Introduction
In March 2020, we launched an international research consortium called 4CE (Consortium for Clinical Characterization of COVID-19 by EHR) to use data in hospital electronic health record (EHR) systems to gain early insights into the COVID-19 pandemic. At its peak, 4CE had 263 members representing 342 hospitals across 8 countries and 4 continents. Despite it being a volunteer network, without any funding for participating sites, its members were highly engaged. This presentation describes several Team Science aspects of 4CE.

Methods
(1) Multidisciplinary “Triads”. In establishing 4CE, we brought together three types of experts: physicians, who could present research questions relevant to the patients they were treating and provide important clinical insights; biomedical “informaticians”, who understood the EHR data and its limitations; and statisticians/epidemiologists, who could design the analytical methods. (2) Federated Architecture. For both patient privacy concerns and regulatory reasons, it is difficult for hospitals to combine patient-level data into a central database, especially in an international project. We therefore used a organizational approach where we collectively developed statistical analysis code to answer a research question, manually distributed this to sites, had a local “triad” team at each site run the experiments, and then shared results in the form of only aggregate counts and statistical summaries. (2) Changing Motivations. Many 4CE members were first motivated by personal reasons, such as concerns about their patients or family. By 2021, vaccines and better treatments were available, but there was academic interest in continuing to be part of a large consortium that was in a unique position to study the many remaining unanswered questions about COVID-19. By 2022, it was difficult for sites to prioritize the effort required for consortium-wide studies and most activity in 4CE moved to smaller working groups, each with only a few hospitals that worked on specific topics.

Results
4CE released its first preprint just four weeks after our initial meeting and 4CE has since published dozens of papers in a wide range of topics, including risk factors for severe COVID-19 disease, neurological effects from COVID-19, conditions seen in children, and “long-COVID” symptoms. Simple collaboration tools like Zoom, Slack, and GitHub for code sharing helped us onboard sites rapidly, lowered barriers to participation, and enabled us to track the level of participation over time.

Conclusion
From the beginning, a key focus of 4CE was trusting the data from the hospital EHRs. Back in March 2020, there were still very few patients at any individual hospital with COVID-19 and standardized diagnosis codes for the disease had not yet been developed. Over time we faced new challenges such as harmonizing the results from the different types of laboratory tests used in different countries. The approach we have used in 4CE has helped us build that trust by having both a broad global perspective on the various problems as well as local multidisciplinary teams within each site who understand the complexities of their hospital’s specific patient population, EHR databases, and clinical practices. For more information visit our website at CovidClinical.net.