Skip to main content

Extending Machine Learning Methods Development in Sentinel: Follow-up Analyses for Anaphylaxis Algorithm and Formalization of a General Phenotyping Framework (Phase 3)

Basic Details
Date Posted
Status
In progress
Description

This pilot activity is one in a series being launched to support FDA’s use of the post-market active risk identification and analysis (ARIA) system to assess the safety of regulated medical products. The purpose of this pilot series is to develop a framework to use machine learning and natural language processing (NLP) techniques to improve health outcome of interest (HOI) identification algorithms that may later be used in the larger Sentinel Distributed Database to assess drug safety questions.  The current task builds on the following two ongoing activities:

  1. Machine Learning Algorithm Development for Anaphylaxis (Phase 1: Jul 2018 – Aug 2020)
  2. Machine Learning Algorithm Development for Acute Pancreatitis -and- Multi-site adaptation for Anaphylaxis Algorithm (Phase 2: Apr 2019 – Sep 2021)

The purpose of this activity is to continue to learn how to improve the accuracy with which we identify HOIs in Sentinel using electronic health record (EHR) data. The current activity involves the following aims: 

  • Aim 1: Expand on the Phase 1-2 anaphylaxis analysis plan and conduct additional secondary analyses
  • Aim 2: Develop and conduct a more scalable automated NLP feature engineering process (i.e., compare a PheNorm-like automated model to the current model based on manual feature curation)
  • Aim 3: Further develop the high-level general framework for HOI identification into a more formalized and comprehensive guidance document for Sentinel (i.e., a publishable manuscript)

 

Related Links
Related Link(s)
Sentinel Methods Project: Validation of Anaphylaxis Using Machine Learning
Sentinel Methods Project: Validation of Acute Pancreatitis Using Machine Learning and Multi-Site Adaptation for Anaphylaxis
Presentation: Scalable Phenotyping for Safety Outcomes Using Electronic Health Record Data
Finding Uncoded Anaphylaxis in Electronic Health Records to Estimate the Sensitivity of ICD10 Codes
Workgroup Leader(s)

Jennifer Nelson, PhD; David Carrell, PhD; Kaiser Permanente Washington Health Research Institute, Seattle, WA

Robert Ball, MD, MPH, ScM; Danijela Stojanovic, PharmD, PhD; Office of Surveillance and Epidemiology, Center for Drug Evaluation and Research, US Food and Drug Administration, Silver Spring, MD

Workgroup Member(s)

Adebola Ajao, PhD; Sara Karami, PhD, MPH; Mingfeng Zhang, MD, PhD; Office of Surveillance and Epidemiology, Center for Drug Evaluation and Research, US Food and Drug Administration, Silver Spring, MD

Yong Ma, PhD; Yueqin Zhao, PhD, Office of Biostatistics, Center for Drug Evaluation and Research, US Food and Drug Administration, Silver Spring, MD

David Cronkite, MS; Ron Johnson, MA; Vina F. Graham; Charissa Tomlinson; Jing Zhou, PhD; Kaiser Permanente Washington Health Research Institute, Seattle, WA

Mara Bann, MD; James Floyd, MD, MS; Patrick Heagerty, PhD, MS; University of Washington, Seattle, WA

Adi Bejan, PhD; Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN

Brian Hazlehurst, PhD; Daniel Sapp; Matthew Slaughter, MS; Kaiser Permanente Center for Health Research, Kaiser Permanente Northwest, Portland, OR

Adee Kennedy, MS, MPH; Mayura Shinde, PhD, MPH; Department of Population Medicine, Harvard Pilgrim Health Care Institute and Harvard Medical School, Boston, MA

Susan Gruber, PhD, MPH; Putnam Data Sciences, LLC, Cambridge, MA

Xu Shi, PhD; Department of Biostatistics, University of Michigan, Ann Arbor, MI