Skip to page navigation menu Skip entire header
Brown University
Brown Digital Repository
Skip 13 subheader links
  1. Top
  2. ...
  3. Warren Alpert Medical Sc…
  4. bdr:1148815

Building the data pipeline for a real-time predictive model of ED recidivism and opioid overdose

Full Metadata

Overview

Year:
2020
Contributor:
Song, Sophia (Author)
Bai, Eric (Author)
Aluthge, Dilum (Author)
Sarkar, Indra Neil (Author)
Marshall, Brandon (Author)
Beaudoin, Francesca (Author)
Ranney, Megan (Author)
Brown University. Alpert Medical School. Scholarly Concentration Program. Non-Scholarly Concentrator (research program)
Subject:
Machine learning
Emergency medicine
Opioids
DOI
https://doi.org/10.26300/kxf8-bs13

Files

Description

Abstract:
A small fraction of frequent Emergency Department (ED) users contributes disproportionately to ED visits. Frequent ED users are at higher risk of poor healthcare outcomes. However, targeted interventions for this vulnerable population improve healthcare outcomes and ED utilization. Big data, in the form of electronic health records, is a largely untapped resource for creating predictive models to identify and improve care of high-risk patients in the ED at the time of care. The objective of this study is to develop a predictive model to predict risk of 30-day ED return and 30-day opioid overdose. We extracted electronic health data of all adult patients who presented to a Lifespan healthcare system ED from June 1, 2016 to May 31, 2017, with a one-year look forward and look back period. 15 inputs were used to train and compare a panel of ten commonly-used machine learning classifiers using the Python package Sci-kit learn to predict risk of 30-day ED return and 30-day opioid overdose. We found that random forest and decision tree classifiers had high sensitivity (75.4% and 76.3%, respectively) and specificity (88.9% and 88.2%, respectively) with quick run-times (74.2s and 15.4s, respectively). We identified several factors that predict 30-day ED recidivism and opioid overdose and demonstrated that out-of-box machine learning models perform well on longitudinal EHR data. These results may have implications for the development of more accurate models based on EHR data and may assist clinicians in identifying and caring for vulnerable, high-risk patients.
Notes:
Scholarly concentration: Non-Scholarly Concentrator

Content

Access Conditions

Use and Reproduction
All rights reserved

Citation

Song, Sophia, Bai, Eric, Aluthge, Dilum, et al., "Building the data pipeline for a real-time predictive model of ED recidivism and opioid overdose" (2020). Warren Alpert Medical School Academic Symposium. Brown Digital Repository. Brown University Library. https://doi.org/10.26300/kxf8-bs13

Relations

Has Parts:

  • Child thumbnail
    Building the data pipeline for a real-time predictive model of ED recidivism and opioid overdose
    • Type: Video
    • Order:
    • View
  • Child thumbnail
    Building the data pipeline for a real-time predictive model of ED recidivism and opioid overdose
    • Type: Stream
    • Order:
    • View

Collection:

  • Warren Alpert Medical … Thumbnail

    Warren Alpert Medical School Academic Symposium

    The Warren Alpert Medical School Academic Symposium is an annual event at Warren Alpert Medical School of Brown University that provides Year II medical students a venue to present their summer research in a poster format. Participation in the Symposium …
    ...