How can pooling health record data improve influenza research?

Challenge

Vaccines for COVID-19 have significantly decreased morbidity and mortality among those infected, yet more work is needed to fully understand how well they work for different groups of people, in different real-world conditions, and to learn more about the symptoms, outcomes, and complications from this disease.

To answer these questions, the Centers for Disease Control and Prevention (CDC) established the VISION Vaccine Effectiveness (VE) Network—leveraging existing virtual networks, including the VISION influenza (flu) network to swiftly integrate data from 9 medical systems across the U.S.:

• Baylor Scott & White Health (Texas)
• Columbia University Irving Medical Center (New York)
• Children’s Minnesota
• HealthPartners (Minnesota and Wisconsin)
• Intermountain Healthcare (Utah)
• Kaiser Permanente Northwest (Oregon and Washington)
• Regenstrief Institute (Indiana)
• University of Colorado
• Vanderbilt University Medical Center (Tennessee)

Needing to engage a large data management team of skilled epidemiologists, biostatisticians, and IT specialists, able to develop and use advanced data collection and analysis tools and multimode processes, CDC selected 趣赢平台 in 2020 to serve as the VISION Data Coordinating Center. The project demands collaboration with the VISION network of clinical sites and partners to harmonize a common protocol and data management system to expeditiously and continuously collect data on vaccinations, testing, and COVID-19-like illness. 趣赢平台 provides analytic support to measure the effectiveness of COVID-19 vaccines, which informs public health policies designed to protect the public from the ravages of these diseases.

Solution

To address key questions about COVID-19 vaccine effectiveness and the clinical epidemiology of COVID-19 by sociodemographic status and among high-risk groups, 趣赢平台 established collaborative relationships with 9 health systems and partner research organizations to collect data on COVID-19-related medical outcomes in outpatient, urgent care, emergency department, and hospital settings.

We developed a common protocol for harmonized data collection across sites; coordinated approval of the study protocol and data dictionary by a common institutional review board (趣赢平台’s IRB); and constructed a research data platform for site data collection.

The 9 health systems contribute test-negative design data (biweekly) and, for a subset of sites, cohort data (monthly) from electronic health records (EHRs) and vaccine registries related to SARS-CoV-2 testing, COVID-19 vaccinations, COVID-19-associated medical events, and severe outcomes. Sites collect information on demographics, clinical characteristics, discharge diagnoses, SARS-CoV-2 testing, and vaccination status of individuals with a COVID-19-like illness associated medical event.

Data are sent from the sites to 趣赢平台 using a secure data pipeline. We then conduct quality checks and combine the data into an analytic dataset with additional derived variables (such as underlying medical conditions, which are defined based on discharge diagnoses for each encounter). These datasets are sent to CDC on a biweekly basis. This biweekly timeline is vital to informing key public health policies, such as booster recommendations, and resource allocations. Our project staff work with site staff to facilitate accurate and complete reporting of testing, vaccination, and medical encounter data. 趣赢平台 staff also combine data across platforms and sites, conduct data cleaning and quality control, produce analytic datasets, and take the lead on analyses, including vaccine effectiveness (VE) of COVID in immunocompromised populations and pregnant people, waning COVID VE, VE during different SARS-CoV-2 variant eras, and, more recently, influenza (flu) VE.

趣赢平台 uses machine learning-based methods and complex analytics (e.g., gradient-boosted regression trees) to help CDC learn more about vaccine-induced immunity. By employing Amazon Web Services (AWS), we are able to capture and harmonize data in a central database and have instituted procedures for quality assurance and control. Our staff have developed their own R packages to analyze the data, and manage a shared Git repository for central code distribution and reproducibility.

Results

趣赢平台’s research is broadening CDC’s knowledge about real-world effectiveness of COVID-19 and flu vaccines among different populations, including those who are immunocompromised. It is also contributing to CDC’s understanding of the signs and symptoms associated with COVID-19 illness; who is receiving each vaccine; and how different groups of people respond to it. These studies are providing CDC with the critical intelligence it needs to develop recommendations about how Americans can protect themselves from both diseases.

Importantly, should other pathogens emerge, VISION’s structure will be able to leverage the platform to quickly conduct studies of those diseases and assess the level of protection afforded by any vaccines developed to combat them.

Since 2020, 趣赢平台’s rapidly generated vaccine effectiveness estimates have informed CDC, the Advisory Committee on Immunization Practices, the Food and Drug Administration (FDA), and White House stakeholders as they consider next steps in safeguarding the public against both flu and COVID-19. The data have also been published in over a dozen high-impact publications, as well as covered in major news media.