Driving Collaborative Research: Duke’s Foundational Involvement in the National COVID Cohort Collaborative (N3C) Data Enclave

Project in a Nutshell

The COVID-19 pandemic demanded a rapid, coordinated response to understand the disease, but healthcare data are often siloed within individual institutions. N3C was developed to build and use a national data repository of real-world data (Electronic Health Record data) for understanding the prevalence, progression, and impact of COVID-19 across the CTSA consortium. 

Significance of the Project

The COVID-19 pandemic demanded a rapid, coordinated response to understand the disease’s symptoms, clinical presentation, and outcomes. However, healthcare data are often siloed within individual institutions, limiting researchers’ ability to study patient data across systems. This fragmentation not only limited researchers’ capacity to study COVID-19 at scale but also impeded progress in understanding and treating other medical conditions. 

Overarching Goals and Approach

The National COVID Cohort Collaborative (N3C) began during the COVID-19 pandemic to enable rapid, large-scale, multi-institutional research using electronic health records (EHRs) from Clinical and Translational Science Award (CTSA) institutions. It aimed to understand the presentation, risk factors, symptoms, and health outcomes of COVID-19 patients and it allowed comparison with similar patients who did not have COVID-19. Initially driven by the Center for Data to Health (CD2H) and coordinated by the National Center for Advancing Translational Science (NCATS), this effort established a unified protocol, data use agreements (DUAs), and memoranda of understanding (MOUs) to streamline participation and reduce regulatory complexity. To standardize submitted data, N3C used the Observational Medical Outcomes Partnership (OMOP) common data model. Contributing sites could submit data either using the OMOP or Patient-Centered Outcomes Research Network (PCORnet) format, as N3C developed a converter from PCORnet data into OMOP. In addition, federally managed infrastructure provided a secure environment for storage and analysis. N3C emphasized collaboration, in that researchers were encouraged to work collectively when using N3C data; this reflects the consortium’s core value of shared, transparent science. To support this collaborative model, N3C utilized a cloud-based analytic platform that allowed researchers from different institutions to access, analyze, and share findings within a secure, centralized workspace.  Phase 1 of the N3C initiative established the infrastructure for secure, standardized data collection, while Phase 2 enabled researchers to use the data for public health and clinical analyses. N3C’s design intentionally supported scalability and adaptability, allowing it to be extended for research on other diseases beyond COVID-19.

Duke’s role in the N3C startup centered on contributing data through its existing PCORnet infrastructure, including the servers, data storage, and extraction processes used to convert Duke’s EHR data into the PCORnet Common Data Model.  As one of the first 10 institutions to join the initiative, Duke drew on this infrastructure to identify and extract EHR data for COVID-positive patients along with matched COVID-negative controls and submitted these data in accordance with the N3C data submission protocol for secure and standardized data submission.  Duke also volunteered to serve as a pilot site for several innovations, including extending the data model by including viral variant information, social determinants of health (e.g., data addressing housing and food security), respirator settings for patients in the ICU, and other critical COVID-19 research data points, as well as participating in Privacy Preserving Record Linkage to securely connect records across different institutions while maintaining patient privacy. 

As of October 2025, N3C aggregates data from over 90 institutions across the United States and 603 research project data requests have been submitted and approved by the Data Access Committee, a federal review panel composed of experts in scientific research, bioethics, information technology, and human subject protects. Of these projects, 13 were submitted by Duke investigators. While N3C originally focused on COVID-19, the platform has since expanded to support a wide range of clinical and translational research topics and now contains data on approximately 22.8 million individuals. Examples of N3C-enabled research include the development of a classification model that identifies age-specific risk factors for patients at risk of Long COVID and an analysis showing that adults with preoperative swallowing difficulties face higher risks of complications after surgery, highlighting the breadth and potential impact of N3C. 

Future Directions

Lessons from COVID-19 highlighted the need for scalable, efficient EHR reporting systems. As the immediate urgency of the COVID-19 pandemic recedes, N3C is evolving to serve as a more generalizable platform for public health research and reporting. For example, in later public health events such as monkeypox, reporting processes from individual clinics and health systems to state and federal agencies remained largely manual, underscoring the opportunity to develop infrastructure capable of handling large-scale data collection with minimal friction. 

Several key initiatives are supporting N3C’s movement to a generalizable platform. Specialized enclaves (i.e. domain-specific platforms) have been established for research areas beyond COVID-19, including cancer and renal disease, demonstrating that the platform can support multi-institutional research across multiple health domains. Additionally, N3C is being integrated into the broader NIH real-world data infrastructure, positioning it as a core component of national efforts to leverage large-scale health data for research, public health monitoring, and rapid response to emerging health threats. 

Facilitators of Success

Community and Cultural Momentum
Early in its development, N3C created a sense of startup-like momentum among its partners, marked by high engagement and widespread opportunities for collaboration. This energy stemmed from the urgency and shared purpose to address the pandemic, as well as the small, agile group of initial partnering institutions. The resulting cultural momentum for N3C participants helped drive participation and investment in the network’s goals, illustrating that cultivating community within initial partners and shared purpose can be as vital as technical infrastructure in sustaining large-scale research initiatives. For Duke, joining N3C reinforced this momentum and helped elevate internal priorities around data infrastructure, particularly accelerating the development of a COVID-19–focused datamart (a focus, subject-specific data repository).

Preexisting Infrastructure 

The COVID-19 datamart was built shortly after joining N3C, despite a very tight timeline. Duke’s N3C IRB was approved in June 2020, and by the end of that month, the datamart was operational with the first dataset transferred. Leveraging the existing PCORNet Common Data Model enabled rapid deployment of the infrastructure, ensuring timely access to critical COVID-19 data.

Reduction of Regulatory and Organizational Hurdles

N3C was launched by the Center for Data to Health in direct response to the emerging COVID-19 pandemic. The urgent need for data drove participating sites to adopt standardized DUAs and MOUs, allowing institutions to share data under one consistent regulatory structure. This approach helped reduce bureaucratic delays that might have otherwise slowed multi-site collaboration and as a result, N3C aggregated multi-site patient data at unprecedented speed, reaching millions of records within its first year. 

Furthermore, Johns Hopkins University led and managed a single centralized IRB for all N3C participants, supported by a reliance agreement that was renewed annually. This arrangement allowed Duke CTSI and other participating institutions to avoid submitting new IRB approvals each time case definitions (e.g., rules for deciding which patients count as having COVID-19 in the dataset) evolved. Leveraging this framework, Duke CTSI created a single broad extract from PCORnet for N3C — a comprehensive data pull into N3C instead of repeated dataset submissions — and applied updated definitions within the local data copy at Duke. This approach reduced administrative burden, accelerated approval, and provided the team with greater flexibility to adapt as requirements changed.

Building for Long-Term Sustainability Beyond the Pandemic

While N3C originated as an emergency pandemic response, it was intentionally designed to outlast the immediate crisis. By not solely focusing on the pandemic, N3C established a framework for continuation and expansion of use. This durability made it a meaningful facilitator for Duke, as it enabled the institution to leverage data resources previously developed for participation in PCORnet and various clinical trial networks, with the added benefit of further refining and strengthening those resources through engagement with N3C. This ensured that N3C’s infrastructure, governance, and community could support ongoing research well beyond COVID-19. 

Challenges and Learnings

Fostering a Culture of Collaboration

One of the most persistent challenges throughout the project was fostering and sustaining a culture of collaboration across a growing and varied network of institutions. As more partners joined, it became increasingly important, but also more difficult, to maintain alignment around the project's core value of working collaboratively.  Many researchers were accustomed to conducting studies independently and found it difficult to adapt to the collaborative model that N3C promoted, which included the expectation to join structured workgroups (specialized subgroups formed to solve a specific problem), share ideas early, and co-develop analyses. This collaborative model was a significant cultural shift for some researchers. While this model ultimately enabled richer, more impactful research, it initially met resistance from those who preferred to work independently or were unfamiliar with such governance structures. 

Recognition of contributions in papers also posed a challenge; for instance, some contributors who undertook substantial work were credited only as consortia authors (authors credited collectively as a single entity rather than individually), which was not always seen as fully reflecting their efforts. Despite this, a key strength of N3C was its ability to connect researchers across institutions, enabling collaborations between groups with shared research interests that may not have otherwise intersected. 

Resource Constraints 

Participation in N3C generally did not include additional funding, creating uncertainty for some institutions about how to support and compensate personnel involved in the initiative. Duke sustained the project through relying heavily on the CTSI-funded PCORnet team. However, the work was widely regarded as important, and the urgency of the national public health emergency fostered a strong sense of purpose, motivating institutions and researchers to contribute voluntarily despite the lack of dedicated financial support.

Rapid Growth and Organizational Processes

As the network grew, the rapid increase in workgroups and domain teams (specialized subgroups focused on examining specific research domains such as diabetes and obesity) led to a high volume of meetings and required ongoing adjustments to communication and team workflows. Practices that were effective when a team was small often needed to be restructured as it grew larger. 

Navigating the Technological Learning Curve

Introducing N3C to a broad community of researchers required many to adopt cloud-based analytics for the first time, representing a significant shift from familiar tools like local analytic software or workflows. The scale and complexity of the data, combined with unfamiliar platform features such as shareable workspaces and embedded knowledge components (i.e. code, tools, or data curated from broader datasets) created a steep learning curve. Initial efforts to support adoption included group training sessions through N3C, such as “lunch and learn” workshops. Individuals who attended these sessions were expected to serve as local trainers and disseminate knowledge within their institution. 

However, platform data access policies limited these approaches. For example, N3C prohibited screenshots and recordings of training to ensure that only individuals who had signed a DUA could view the data, limiting how these trainings could be shared with other colleagues. These restrictions posed fewer challenges for institutions with an organizational DUA, such as Duke, as anyone in an institution with an organizational DUA could view the data. However, other challenges remained even for these organizations. For example, it was difficult to incorporate N3C data into coursework because students first had to apply for data access, a process that typically took three to four weeks, making it incompatible with most course timelines. As a result, many students and instructors ultimately sought alternative, openly available data, which required less time and administrative effort to access and use.

Contribution to Case Study Development:

Rishi Kamaleswaran, PhD, MS
Director of Duke CTSI Informatics
Associate Professor in Surgery and in Anesthesiology

Resources

CTSA Resources Used

References
Haendel MA, Chute CG, Bennett TD, et al; N3C Consortium. The National COVID Cohort Collaborative (N3C): Rationale, design, infrastructure, and deployment. J Am Med Inform Assoc. 2021;28(3):427‑443. doi:10.1093/jamia/ocaa196

Alnajar A, Kareff SA, Razi SS, et al. Disparities in survival due to social determinants of health and access to treatment in US patients with operable malignant pleural mesothelioma. JAMA Netw Open. 2023;6(3):e234261. doi:10.1001/jamanetworkopen.2023.4261

Wang WK, Jeong H, Hershkovich L, et al. Tree-based classification model for long-COVID infection prediction with age stratification using data from the National COVID Cohort Collaborative. JAMIA Open. 2024;7(4):ooae111. doi:10.1093/jamiaopen/ooae111

El Qadir NA, Jones HN, Leiman DA, Porter Starr KN, Cohen SM; National COVID Cohort Collaborative (N3C) Consortium. Preoperative dysphagia and adverse postoperative outcomes in middle aged and older adults. J Clin Anesth. 2025;100:111688. doi:10.1016/j.jclinane.2024.111688