Brief description of the Sentinel Initiative and data
The US Food and Drug Administration (FDA) leads the Sentinel Initiative,25 which aims to develop new ways to assess the safety of approved medical products including drugs, vaccines, and medical devices. The initiative was created to meet a mandate by the US Congress in the FDA Amendments Act of 2007. Launched in 2009, it has developed into “the largest multisite distributed database in the world dedicated to medical product safety.”26 The Biologics Effectiveness and Safety (BEST) System is also part of the Sentinel Initiative, launched in October 2017 to expand and enhance the Center for Biologics Evaluation and Research (CBER) “access to new and better data sources, methods, tools, expertise, and infrastructure to conduct surveillance and epidemiologic studies.”
The Sentinel System data consists primarily of administrative claims and some EHRs from US national and regional health plans, integrated delivery systems, and Medicare fee‐for-service data from Centers for Medicare and Medicaid Services (CMS). These partners transform their data into the Sentinel Common Data Model, but maintain possession and operational control over their data. During the data transformation, identifiable patient information is removed or masked. The standardized format enables the use of routine analytic programs, enabling rapid analyses without the need for customized code.27 In all, the Sentinel Distributed Database includes claims data on approximately 70 million individuals per year from all 50 US states, covering 800 million person‐years.
Other Sentinel partnerships25 include EHR‐based organizations and networks—HCA Healthcare, a hospital system with over 140 facilities, the National Patient‐Centered Clinical Research Network (PCORnet), and EHR data aggregators (TriNetX, Veradigm, IBM Explorys).
The COVID-19 pandemic prompted a rapid expansion and enhancement of Sentinel. In response, Sentinel created the Rapid Distributed Database with a subset of the national health plans and integrated delivery systems. While data in the typical Sentinel Distributed Database is generally 6 months old, allowing for claims to “settle,” the Rapid Distributed Database is refreshed approximately every 2 months, enabling the FDA to have access to more timely data. The Rapid Distributed Database also maintains the capability to allow for the conduct of targeted chart reviews.
To ensure completeness of information, the inpatient EHR data from HCA Healthcare includes discharged patients only. Importantly, in addition to standardized data on diagnoses, procedures, and medications, nursing documentation notes have been included to supplement diagnosis and procedure codes for identification of oxygen related therapy and non-invasive mechanical ventilation.
The ability of the FDA to quickly use Sentinel claims data and HCA Healthcare data to study COVID-19 was facilitated by past work on influenza in collaboration with the FDA’s Office of Counterterrorism and Emerging Threats28 and an early validation study to assess the performance of an International Classification of Diseases, Tenth Revision (ICD-10)-based algorithm to identify inpatient COVID-19 patients.
As with all studies using RWD, a robust assessment of whether a study is fit-for-purpose29 is required for every proposed study to determine the best of the available sources for COVID-19 work in the Sentinel System. With the Rapid Distributed Database, the “freshness” of the data comes with challenges, including the need to understand completeness of the data by care setting (e.g., inpatient data lag behind ambulatory data) and claim type (e.g., encounter-based diagnoses lag behind outpatient pharmacy dispensations). Further, investigators now and in the future will have to understand and possibly adjust for changes in general health care utilization during the pandemic, or even issues such as the use of certain specialized wards for isolation of severe COVID-19 cases. As always within Sentinel, data quality review is a high priority and robust methods are in place for the various data sources.
The Sentinel System and COVID-19 observational research
Sentinel COVID-19 related goals include describing the course and outcomes of COVID-19 illness (including changes over time) in various demographic groups determining prognostic factors for COVID-19 based on data early in the course of illness. These data may allow the examination of determinants of COVID-19-related complications, and provide a benchmark—serving as an external control—for single-arm trials of COVID-19 treatments in relevant patient populations. Sentinel is also used to study the COVID-19 pandemic in real time.32
Sentinel developed a COVID-19 Natural History master protocol to serve as a guide for future COVID-19 studies. The master protocol includes translation of inclusion and exclusion criteria and clinical endpoints and definitions that reflect clinical concepts (such as COVID-19 severity categories and complications). Code lists are provided to help define cohorts (including special populations) and pre-existing conditions. This harmonization process allows for the assessment of homogeneity of findings across data sources, regions, and countries, and promotes collaboration and communication across regulatory agencies. This Master Protocol27 is available on the Sentinel website.33
Sentinel Initiative COVID-19 case studies
Sentinel developed a protocol to estimate the 90-day incidence and risk of death from arterial and venous thrombotic events among patients with COVID-19, and compare their risk of these events to that of patients with seasonal influenza, using propensity score-based adjustment. Within Sentinel, the protocol is being implemented with data from integrated delivery systems and national claims partners. This protocol has been posted on the Sentinel website and was also disseminated through the COVID-19-Evidence Accelerator sponsored by the Reagan-Udall Foundation for the FDA. The European Medicines Agency (EMA) will conduct a parallel study among patients with COVID-19, and a meta-analysis may ultimately be conducted with the EMA- and FDA-sponsored results.34,35
Sentinel's pandemic preparation activities mentioned above were leveraged to develop ongoing surveillance reports of COVID-19 patient census and clinical characteristics in HCA Healthcare, generating bi-weekly reports that examine patient demographic and medical characteristics, treatment and care received, and complications from COVID-19 using inpatient electronic medical record (EMR) data. The work defines markers of illness severity similar to those used in clinical trials. Importantly, the project has examined the capture of oxygen-related therapy. Whereas a code-based approach identified 28% of patients diagnosed with COVID-19 as having received oxygen therapy, adjudicated standardized nursing documentation data revealed that 79% received ventilation or supplemental oxygen (capture of invasive mechanical ventilation did not significantly change when using nursing documentation). These inpatient EHR data are updated approximately every 2 weeks, and analyses are restricted to discharged patients with complete billing.36
Little information is available to support understanding the natural history of COVID‐19 disease in pregnant women, or the impact of COVID‐19 treatment on pregnant women or the developing fetus. In order to study the impact of COVID-19 on pregnancy, the FDA has partnered with the EMA on their CONSIGN (COVID-19 infectiOn aNd medicineS In preGNancy) project, which will study COVID-disease in pregnant women across 8 European countries. Sentinel is implementing a parallel study using validated analytic tools to estimate the prevalence of select medicines in pregnant women with and without COVID-19, and in nonpregnant women with COVID-19. The study will also describe disease severity and clinical outcomes of pregnant women with COVID-19, according to treatments received during pregnancy, and compare these data with those of nonpregnant women of reproductive age with COVID‐19.37
Sentinel COVID-19 impact and perspective
The success of the Sentinel System goes beyond the ability to have rapid access to COVID-19 related data. The work has successfully demonstrated the ability of the FDA to take on more ambitious protocols (such as the pregnancy study cited above), conduct drug monitoring projects (https://www.sentinelinitiative.org/methods-data-tools/methods/near-real-time-monitoring-critical-drugs-care-patients-covid-19)38 on a wide scale, and conduct collaborations with other regulatory agencies. The Rapid Distributed Database will continue into 2022 and will be reassessed to determine whether the initiative will continue.
In May 2021, Acting FDA Commissioner Janet Woodcock, MD, summarized Sentinel and related initiatives as ways that “during this global pandemic, the FDA worked to protect the American public by using every tool at its disposal to quickly review and act on new therapies for COVID-19.” She concluded that in addition to reducing administrative costs and time, master protocols “can also increase data quality and efficiency through shared and reusable infrastructure . . . advantages [that] are of particular importance during a public health emergency such as the current SARS-CoV-2 pandemic, where there is a critical need for efficient drug development. The FDA expects master protocols to continue to play an important role in addressing the public health needs created by the pandemic and in generating clinical evidence in general.”39