Fighting Addiction Before It Begins

| February 9, 2018

USC Viterbi’s Yan Liu uses machine learning to fight the opioid addiction crisis sweeping the nation.

USC Viterbi’s Yan Liu, Philip and Cayley MacDonald Endowed Early Career Chair and Associate Professor of Computer Science and Electrical Engineering-Systems (Photo/Will Taylor)

“No part of our society – not young or poor, rich or old, urban or rural – has been spared this plague of drug addiction and this horrible, horrible situation that’s taking place with opioids. This epidemic is a national health emergency.”

President Trump made that statement on Oct. 26, 2017 at the White House in a speech about America’s opioid crisis. According to The New York Times, deaths involving synthetic opioids have risen from approximately 3,000 per year to 20,000 per year in the past three years alone.

At the ceremony, the president also announced his plan to combat opioid addiction on the national level, with one of his goals being “to teach young people not to take drugs,” to stop addiction before it can occur.

And now, researchers at USC Viterbi School of Engineering are working to combat addiction before it can begin.

One of the leaders in this field, Yan Liu, USC Viterbi associate professor in computer science, heads the USC Machine Learning Center, or MASCLE. This interdisciplinary center – with representatives from USC Viterbi, the USC Marshall School of Business and the USC Dornsife College of Letters, Arts and Sciences – is dedicated to uniting machine learning researchers from different educational sectors to harness the potential of machine learning to aid society.

In one of MASCLE’s recent groundbreaking studies, which came out in 2017, deep learning and computer mapping were used to study opioid addiction and adverse effect analysis. In simpler terms, the research team used computer learning models – similar to those used in self-driving cars, or those in the iPhone X’s new facial recognition software – and programmed them to sift through a patient’s medical history, identifying trends that may indicate opioid dependency and addiction; in essence, learning how to predict opioid dependency.

Identifying these patterns is extremely important. One of Liu’s ultimate goals with her research is to successfully predict, based on medical history, whether a patient is susceptible to developing an addiction to opioid-based pain medication before physicians ever prescribe this medication. If key warning variables can be identified, physicians can then take steps to combat this addiction before it can begin, prescribing the patient alternative forms of pain medication to avoid their exposure altogether.

“We conducted this research because of its importance and urgency,” said Liu, the Philip and Cayley MacDonald Endowed Early Career Chair, who joined USC Viterbi’s faculty in 2010.

The study itself achieved its goal, according to Liu: it demonstrated that deep learning models have the potential to be extremely useful tools when combating opioid addiction. The models used in the study looked at each patient’s medical history, specifically focusing on several key factors. These included the patient’s prescription history, as well as whether or not they had been diagnosed with any other relevant disorders that could indicate an increased risk of developing an addiction to opioid pain medication: alternative substance abuse, recreational drugs, alcohol, or anxiety disorders.

Based on these traits, the models classified all relevant patients – those that had previously been prescribed opioid medication – into three groups: short-term users (ST), long-term users (LT), and opioid-dependent users (OD). This research was done in collaboration with the Mayo Health Clinic using electronic health records provided by the Rochester Epidemiology Project (REP). Researchers used a sample size of 102,166 patients (narrowed down from an initial 142,377), one of the biggest datasets ever used for this research according to Liu, to achieve the most accurate data possible. The results are as follows: 79% were short-term users and 21% were long-term users. According to the algorithm, 3.47% were also opioid-dependent– a significant increase from the 0.7% that had been diagnosed with opioid dependence prior to the study by their doctor.

As this research was conducted using data taken from medical records, Liu’s team was careful to be attentive to the privacy rights of their research subjects.

“Privacy is our top concern when we conduct human subject research,” Liu said. “We only used the data of patients who agreed to share their medical records for study usage, and we followed the strict data use agreement in this study– we did not utilize or disclose any direct identifiers of the individual.”

This is the first preliminary study done using these types of deep learning models. Given the project’s success, Liu anticipates future research to refine results and further aid physicians in combating the opioid crisis.

“This is definitely an ongoing project,” said Zhengping Che, a Ph.D. student advised by Liu who has worked on the project. “Our next step in this research is to use the results achieved with this model to help us build a more powerful model, one that is better able to predict opioid dependency risk.”

With opioid abuse now a declared national emergency, this particular field of research has never been more important. And USC Viterbi’s MASCLE seems well on its way to help combat this crisis from the ground up.

“The understanding of this severe problem is limited,” Che said. “Given the huge amount of electronic health record data and successful applications of deep learning models in other areas, we believe deep learning can be an effective solution to this important problem.”

Published on February 9th, 2018

Last updated on May 16th, 2024

Share this Story