Predicting Opioid Dependence with Marijuana Use and Multi-Layer Perceptron Classifiers

15 Pages Posted: 23 Jan 2020

See all articles by Longchao Liu

Longchao Liu

Phillips Exeter Academy

Govind Chada

affiliation not provided to SSRN

Odessa Thompson

affiliation not provided to SSRN

Elaine Chu

affiliation not provided to SSRN

Date Written: December 30, 2019

Abstract

Objective. To study the correlation between opioid dependence and marijuana use and identify actionable improvements to opioid risk assessment.

Data & Methods. Labeled data sets were created from the 2016 National Survey on Drug Use and Health (NSDUH) and the eICU Collaborative Research Database (eICU). Multilayer perceptron (MLP) networks were trained and tested with the two data sets to compare the merit of the survey and medical data.

Results. While the prevalence of opioid dependence is only roughly 1%, the classifiers trained with survey data can accurately predict the subjects at risk for opioid dependence (sensitivity = .71, specificity = .80, AUC = .81) with marijuana use history.

Conclusion. The results indicate that supervised machine learning can be used to predict opioid dependence through demographic, socioeconomic, and behavioral features. The output indicates that having ever used marijuana (ever-use), a variable absent from eICU data and the Opioid Risk Tool, is important for predicting opioid dependence, highlighting gaps in medical data currently collected. The researchers propose several courses of action for the improvement of opioid risk assessment.

Keywords: Opioid Dependence, Marijuana Ever-Use, Multi-Layer Perceptron Classifiers

JEL Classification: I18, C80

Suggested Citation

Liu, Longchao and Chada, Govind and Thompson, Odessa and Chu, Elaine, Predicting Opioid Dependence with Marijuana Use and Multi-Layer Perceptron Classifiers (December 30, 2019). Available at SSRN: https://ssrn.com/abstract=3511743 or http://dx.doi.org/10.2139/ssrn.3511743

Longchao Liu (Contact Author)

Phillips Exeter Academy ( email )

United States

Govind Chada

affiliation not provided to SSRN

Odessa Thompson

affiliation not provided to SSRN

Elaine Chu

affiliation not provided to SSRN

Here is the Coronavirus
related research on SSRN

Paper statistics

Downloads
46
Abstract Views
212
PlumX Metrics