Rishi Desai Seminar, January 28, 2026

Rishi Desai Seminar, January 28, 2026

by Rishi Desai

Applied Sciences & Engineering Medicine & Health MBGMedical equipment, devices and techniquesMBGRMedical research
🎙 Rishi Desai 👥 2K 📅 February 10, 2026 ⏱ 50 min 👁 124 📄 expert opinion 🧭 2026-08-15
Available in: English (current) Français

Keywords

SentinelFDAreal-world evidenceelectronic health recordsdrug safety

Summary

In this seminar, Dr. Rishi Desai, an associate professor at Harvard Medical School and epidemiologist at Brigham and Women’s Hospital, presents an overview of the FDA’s Sentinel system and recent efforts to enhance its data infrastructure and methodological capabilities. He begins by explaining the origins of Sentinel, which was established in 2007 following the Vioxx controversy to enable the FDA to conduct its own post-marketing drug safety studies. The system relies on distributed databases from insurance claims, covering over 100 million people. However, a root cause analysis revealed that about 60% of FDA safety questions could not be addressed due to limitations of claims data, such as missing laboratory results and clinical details. To address this, the FDA initiated the Real-World Evidence Data Enterprise (RWE-DE), integrating electronic health records (EHR) from both commercial networks and academic medical centers. The development network, including institutions like Duke and Mass General Brigham, provides rich granular data for algorithm development, while the commercial network offers large-scale data for analysis. Dr. Desai highlights the use of natural language processing (NLP) to extract information from unstructured clinical notes, significantly improving the identification of conditions like obesity and smoking. He then presents a demonstration project on the risk of acute pancreatitis with SGLT2 inhibitors, which was previously deemed infeasible due to outcome misclassification. By developing a computable phenotype that incorporates lab results and imaging reports, they achieved a positive predictive value of 90% and successfully conducted the study, finding no increased risk. The seminar concludes with a discussion of the harmonized process guide for inferential studies and the system’s transition to production mode, including an ongoing study on NAION with GLP-1 receptor agonists. Overall, the presentation underscores the value of integrating EHR data into regulatory drug safety surveillance.

291 words

Critical Evaluation

Value of the Information & Strength of the Argument

The seminar provides valuable insights into the evolution of the FDA’s Sentinel system, highlighting concrete improvements in data infrastructure and methods. The argumentation is solid, grounded in published research and official FDA documents. Dr. Desai effectively demonstrates the limitations of claims data and the potential of EHR integration, using a real-world example to illustrate the practical application. The presentation is well-structured, with clear explanations of complex concepts, making it accessible to a technical audience. The value lies in its detailed description of the RWE-DE and the successful demonstration project, which provides evidence of the system’s enhanced capabilities.

Scientific Rigor, Source Quality, Title Accuracy

The seminar demonstrates high scientific rigor, with references to peer-reviewed publications and FDA reports. The speaker cites specific studies, such as the 2023 analysis of Sentinel investigations and the Floyd paper on pancreatitis validation. The title accurately reflects the content, as it is a seminar presentation by Dr. Desai. The sources are credible and relevant, and the speaker clearly distinguishes between established findings and ongoing work. The presentation adheres to scientific standards, with transparent methodology and acknowledgment of limitations.

191 words

Title / Content Match

The title accurately reflects the content: a seminar presentation by Dr. Rishi Desai on the topic of improving the FDA's Sentinel system.

Quality & Reliability

8/10

The seminar is delivered by a recognized expert in pharmacoepidemiology, presenting a detailed overview of FDA's Sentinel system enhancements. The content is based on published research and official FDA initiatives, with references to specific studies and reports. The presentation is technical and well-structured, but it is a single expert's perspective without external validation or critical discussion.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

  • FDA Sentinel Initiative — Official FDA page describing the Sentinel system, consistent with the seminar's description.

Contribution & Novelties

This seminar provides a comprehensive overview of the FDA’s Sentinel system enhancements, specifically the integration of electronic health records (EHR) to overcome limitations of claims data. The presenter details the development of the Real-World Evidence Data Enterprise (RWE-DE), including the use of natural language processing (NLP) to extract information from unstructured clinical notes, and demonstrates its application in a previously infeasible drug safety study. The presentation highlights the importance of computable phenotypes and harmonized processes for regulatory decision-making.

Pour aller plus loin :

  • FDA Sentinel Initiative — Official FDA page describing the Sentinel system and its goals.
  • Real-World Evidence — FDA’s page on real-world evidence, relevant to the seminar’s focus.
  • Natural Language Processing in Healthcare — Overview of NLP applications in healthcare, including clinical text mining.
  • Computable Phenotypes — Article discussing computable phenotypes and their use in research.
  • SGLT2 Inhibitors and Pancreatitis — Example of a study on SGLT2 inhibitors and pancreatitis (hypothetical reference; verify).

155 words

Radar Profile

The radar profile shows high scores across all dimensions, indicating a well-rounded and reliable presentation. The highest scores are in information quantity and technical level, reflecting the detailed and specialized content. The slightly lower score in global reliability suggests that while the information is credible, it is based on the presenter's expertise and may not be independently verified.

Reliability 8/10

💬 No comments were provided for analysis.