Predict Risk of Opioid Use Disorder

From REU@MU
Revision as of 21:18, 12 January 2020 by Praveen (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Title: Using Hospital Records of Patients to Predict Risk of Opioid Use Disorder (OUD), Fatal and Non-fatal Opioid Overdose, and ED Readmission.

Mentor: Dr. Praveen Madiraju

Approach: Develop machine learning models on hospital dataset to predict risk of opioid use disorder. Mentor has on-going relationships with the Comprehensive Injury Center at the Medical College of Wisconsin, who will provide de-identified hospital data. Students will gain experience with all aspects of data science – data cleaning, data wrangling, feature engineering, data visualization, implementing and evaluating machine learning models for accuracy, precision and recall. Summary: The United States has seen a meteoric rise in opioid epidemic during recent years with opioid overdose deaths increasing by 200% between 2000 and 2014. Hospital readmissions related to opioid overdose increased by 64% from 2002 to 2015 resulting in a devastating toll on patients, their families and communities. A handful of studies have considered predicting hospital readmission after hospitalization for heart failure, diabetic patient discharge and general medical hospitalization. However, very few studies have looked at the problem of predicting readmission for opioid related issues in an emergency department ( ED) setting using machine learning methods. The goal of this study is to utilize hospital records to identify key factors such as health data, social determinants and others that may be used to predict a patient’s risk of developing Opioid Use Disorder (OUD), overdose, or readmission at various time intervals. By assessing a patient’s risk, providers (physicians, social workers, advanced practice providers) can implement specific prevention strategies aimed at reducing the risk for their patient.

Student Research Activities: The REU fellows will perform the following major tasks:

  • Survey state-of-the-art in the area of opioid use disorder and hospital readmission.
  • Perform data cleaning, wrangling, feature engineering, and data visualization on the mentor supplied healthcare data using packages from Python or R.
  • Implement different machine learning models such as decision tree, random forest, and recurrent neural network.
  • Evaluate and select the best machine learning model by presenting results with the stakeholders of the project.

Student Background: Students need to have basic computing knowledge and introductory programming skills in Python or R.