FDA Sentinel System's Coronavirus (COVID-19) Activities

The US Food and Drug Administration's Sentinel System was established in 2009 to use routinely collected electronic health data for improving the national capability to assess post‐market medical product safety. Over more than a decade, Sentinel has become an integral part of FDA's surveillance capabilities and has been used to conduct analyses that have contributed to regulatory decisions. FDA's role in the COVID‐19 pandemic response has necessitated an expansion and enhancement of Sentinel. Here we describe how the Sentinel System has supported FDA's response to the COVID‐19 pandemic. We highlight new capabilities developed, key data generated to date, and lessons learned, particularly with respect to working with inpatient electronic health record data.

Administrative claims data, which comprise most of the data in the current Sentinel Distributed Database, are limited in their ability to study severely ill hospitalized patients, including those in the intensive care unit. For example, data on duration of mechanical ventilation, symptom severity, oxygen saturation, supplemental oxygenation requirements, ventilator settings, and lung imaging features are not readily available in claims data. 

Information-rich unstructured electronic health record (EHR) data may address these challenges and enhance medical product safety surveillance. COVID-19 is a compelling use case for using unstructured EHR data to improve medical product safety assessment in Sentinel since studying COVID-19 requires accurate information about its symptoms, severity, comorbidities, and treatment outcomes and because patients with COVID-19 disease often present in emergency department and receive care in inpatient settings. 

This project will yield algorithms and methods to better position Sentinel for addressing future pandemics and advance Sentinel’s long-term objective of enhancing medical product safety surveillance by incorporating EHR data into surveillance methods. Specifically, we will use natural language processing (NLP) and text mining methods to identify patients with COVID-19 disease and to extract clinical features from unstructured EHR data. These features will complement structured data already available in the Sentinel Common Data Model (SCDM). This project will also develop processes that would allow us to more rapidly prototype EHR-based information extraction algorithms for new health outcomes of interest (HOIs) in the case of future pandemics and other needs.
 

The Sentinel System can be used to address numerous questions of importance to FDA during the COVID‐19 pandemic. One population of significant interest for study is pregnant women. Evidence suggests that pregnant women are more likely to experience severe illness related to respiratory infections, including COVID‐19, than nonpregnant women (1,2). Little information is available to support understanding the natural history of COVID‐19 disease in pregnant women, or the impact of COVID‐19 treatment upon pregnant women or the developing fetus.

To address vulnerable populations, including pregnant women, Sentinel recently published a COVID‐19 Natural History Master Protocol, designed to identify multiple COVID‐19 cohorts to support a variety of on‐demand queries and subsequent descriptive and inferential studies. The protocol can be adapted to various data sources with clinically rich data and included pregnant women as a specific population of interest. The EMA (European Medicines Agency) has also funded a project, called CONSIGN (“COVID‐19 infectiOn aNd medicineS In pregnancy”) designed to understand the natural history of COVID‐19 disease in pregnant women, including disease severity and clinical outcomes of COVID-19, medications utilization, and impact of COVID-19 treatment on neonatal outcomes. The CONSIGN study will be implemented in varied data sources across eight European countries. 

This project will implement the following (two) aims of the CONSIGN protocol:

1. To estimate the prevalence of medicines used and compare this among pregnant women with COVID‐19, pregnant women without COVID‐19, and non‐pregnant women with COVID‐19.
2. To describe severity and clinical outcomes of COVID‐19 disease in 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.

(1) Allotey J, Stallings E, Bonet M, et al. Clinical manifestations, risk factors, and maternal and perinatal outcomes of coronavirus disease 2019 in pregnancy: living systematic review and meta-analysis. BMJ. Published online September 1, 2020:m3320. doi:10.1136/bmj.m3320

(2) Ellington S. Characteristics of Women of Reproductive Age with Laboratory-Confirmed SARS-CoV-2 Infection by Pregnancy Status — United States, January 22–June 7, 2020. MMWR Morb Mortal Wkly Rep. 2020;69. doi:10.15585/mmwr.mm6925a1

Among hospitalized non-pregnant adults with COVID-19, we aimed to describe the number of patients with hospitalized thrombotic events or death through 28 days. We stratified each outcome by d-dimer level and record of early intensive care services. We used electronic health record (EHR) data from the TriNetX Live™ USA Network to identify adults with positive COVID-19 antigen or PCR laboratory test results or documented COVID-19 diagnoses within ±3 days of a hospitalization record. We queried 65 health care organizations’ data from February 20, 2020 through October 16, 2020.

To inform future sample size calculations pertaining to thrombosis in COVID-19-associated coagulopathy, we describe occurrence of thrombotic events and death among patients aged 40-79 years not hospitalized at the time of COVID-19 identification. We used electronic health record (EHR) data from 64 health care organizations (including inpatient and outpatient data), provided through the TriNetX Live™ platform in their USA network to identify COVID-19 cases in the outpatient setting from February 20, 2020 through September 11, 2020. 

This presentation was given at the Reagan-Udall Foundation and Friends of Cancer Research COVID-19 Diagnostics Evidence Accelerator Parallel Analysis meeting by Noelle Cocoros from Harvard Pilgrim Health Care Institute on behalf of the FDA Sentinel COVID-19 Coagulopathy workgroup. This presentation reviews descriptive feasibility data from select Sentinel Data Partners. These data, focused on thromboembolic outcomes among COVID-19 patients, provide preliminary information for full study implementation. 

The COVID-19 Natural History Master Protocol effort involves the development of a protocol designed to identify multiple COVID-19 cohorts to support a variety of on-demand queries and subsequent inferential studies. The protocol will be adaptable to various data sources with clinically rich data and will include specific populations of interest, such as the pediatric population, pregnant women, and other vulnerable populations identified by FDA. The implementation of the Natural History protocol will prioritize data source capabilities, such as: a) the capacity to form stable population-based cohorts that can be refreshed quickly and reliably over time, b) deep on-site pharmacoepidemiology experience with a research record of both descriptive and inferential studies, and c) demonstrated analytic capability to support rapid on-demand queries which support identification of the natural history of COVID-19. Expert infectious disease specialists and pharmacoepidemiologists will lead development of this protocol and its implementation in various data sources in the Sentinel network.

This project supports FDA’s COVID-19 response through: 

1. Adding COVID-19 laboratory results data to the Sentinel Common Data Model (SCDM) for routine querying in the Active Risk Identification and Analysis (ARIA) system
2. Developing access to the “freshest feasible” rapid COVID-19 data from Sentinel data partners in the distributed data network