Reconsidering Addiction and Opioid Abuse: Is There a Causal Chain between Opioid Addiction, Morbidity and Mortality? (2016-2017)

Background

In recent years, medical and public health researchers have noted an increase in morbidity and mortality related to opioid use. These researchers have determined that there has been a precipitous rise of both medical and nonmedical opioid abuse. Of concern for both medical professionals and policymakers is the suggestion that opioid abuse, whether medically-sourced or not, has become the second most common illicit drug, behind only marijuana.

Opioid abuse is often assumed to be the result of opioid addiction. It is easy to see why. Prescription drugs have been exhaustively researched in this area and have very distinct addiction patterns, including cravings and withdrawal symptoms, which have been tested previously in medical laboratories. As a result, the current research on this topic inherently assumes that an increase in opioid-related morbidity or mortality is causally linked in a linearly scaling relationship to similar patterns in addiction. This is not necessarily the case, however, and belies a problem of a different scope entirely. Indeed, addiction has not increased in step with opioid use and abuse. Consequently, the causal chain between drug use, addiction, hospitalization, illness and death needs to be reconsidered.

In short, to say that someone died from a prescription drug overdose is one thing. To say that an addiction to prescription drugs caused that overdose is another thing entirely, and not necessarily the case.

Project Description

This project seeks to address this problem by linking together large datasets that have not previously been integrated in the same study. Death records, arrest and conviction records, credit reports and other public records will help us gain a more accurate understanding of the lives of individuals, and thus better estimate the probability that a decedent was addicted to opioids, heroin or some other pain reliever. We will use the classification and regression tree techniques of statistical/machine learning on combined datasets to make predictions about an individual’s choices while alive.

Using recent innovations in statistical learning and other computational techniques, we can build catalogues of major life events related to prescription (and nonprescription) drug use and discover patterns of addiction or abuse. Team members will use topic discovery/natural language processing algorithms to “read” these records and better classify whether or not decedents had multiple deadly illnesses, addictions and so forth. From these measures, we will apply statistical classifiers to estimate probabilities of cause of death, addiction and morbidity that can better help us understand and predict patterns of addiction and help us build a more complete picture of the chain between drug use, addiction and death. This may lead to algorithms that enable physicians to better supervise their patient’s opioid use.

Team members will connect with the repositories of these records to ascertain how to stitch together existing data sets and to estimate the costs of a large study of this kind. Our pilot study will allow us to test our models to ascertain our ability to assign probabilities over the cause(s) of death and the path of illness, addiction and opioid abuse for each decedent.

Anticipated Outcomes

TBD

Timing

Summer 2016 (May 16 – July 1, 2016) – Spring 2017

Crediting

Independent study credit available for fall and spring semesters; summer stipend

Faculty/Staff Team Members

Thomas Buchheit, School of Medicine - Anesthesiology*
Christopher Edwards, School of Medicine - Psychiatry & Behavioral Sciences - Behavioral Medicine*
Mathew McCubbins, Trinity - Political Science*
Steven Prakken, School of Medicine - Anesthesiology*

Undergraduate Team Members

Noah Davis, Biomedical Engineering
Khuong Do, Statistical Science (AB)
Juliana Zhang, Neuroscience (AB)

* denotes team leader

Status

Active