In the decade since the launch of Mini-Sentinel, the biomedical research community has demonstrated the benefits of marshalling unstructured clinical data to better understand disease etiology and improve care delivery. Often referred to as precision phenotyping, advanced analytic methods are being applied to structured and unstructured EHR data to characterize the timing, severity, and complexity of patients’ health conditions. This prior work to characterize patients’ health conditions suggests that harnessing ubiquitous, unstructured EHR data is beneficial for generating information on populations, exposures, health outcomes of interest, and covariates, which can help achieve Sentinel’s medical product safety surveillance objectives, in general, and surveillance efforts in response to the COVID-19 pandemic, in particular.
In response to the pandemic, the FDA Sentinel System has focused on addressing the following questions:
- How can the Sentinel System be optimally utilized to conduct post-market surveillance on the safety and effectiveness of COVID-19 treatments?
- Which health outcomes of interest of COVID-19 disease, and COVID-19 treatments, merit close monitoring using the Sentinel system?
- What outcomes must be measured to assess treatment effectiveness?
- Which covariates and comorbid conditions must be measured to enhance inference based on observational studies?
- Which of these outcomes and covariates can be better, or solely, captured from unstructured EHR data?
- How can unstructured EHR data be leveraged to answer important questions about the pandemic?
- How can FDA utilize the Sentinel System to prepare for the next pandemic?
Methods that are developed through the demonstration project will help advance medical product surveillance through the following objectives:
