Using analytics to attack addiction

Carlson School

Substance use disorder, or addiction, is a disease affecting more than 20 million Americans.
Talk to anyone who works with those who are battling addiction and they will tell you that only through understanding all of the unique circumstances someone faces can you determine which treatment methods have the best chance of leading to long-term sobriety and health.
The renowned Minnesota based Hazelden Betty Ford Foundation has taken that approach in treating its patients, but recently its leaders realized they had not applied that same thinking to the hundreds of thousands of patient records accumulated at its facilities over the years.
To do that, they turned to the Carlson School’s Analytics Lab. During a semester-long project, a team of five Masters of Business Analytics students used artificial intelligence to analyze more than 250,000 records in search of data patterns.
“It was a little challenging,” says team member Mainak Roy. “It took us about a month just to get our hands on the data and dig into it.”
The students designed an approach grounded in finding “unobvious and of interest” information. First, they determined what defined a “typical” patient at Hazelden Betty Ford—age, gender, race, profession, etc. Then they looked at how often that patient came to Hazelden Betty Ford, what those visits were like, the diagnoses they received, and the patient’s feelings about the experience.

As part of the process, the team built a dashboard to show the data in an easy-to-understand, visual way. Building that tool has provided Hazelden Betty Ford with the ability to continue analyzing its data in the future. It also allowed the team to take the next step in its work: understanding who returned to using substances after their initial treatment, a key success metric for Hazelden Betty Ford.

“It was great to be able to work on an important cause,” says team member Batool Fatima. “It was a type of project that I had never experienced before, but I quickly found that you can apply data analytics knowledge to solve all sorts of problems.”
Team members found that gender, age, type of substance used, and the number of substances used all correlated—in varying degrees—to a patient’s risk for returning to use.

Among students’ findings

  • Almost 65 percent of patients were male
  • Close to 45 percent of patients had a history of misusing 2-3 substances
  • Patients who were 18 years old experienced the highest relapse rate
  • Females returned to use less often than males, especially between ages 20 and 40
  • Those with an alcohol use disorder returned to substance use at a consistent rate across all ages, while those addicted to other drugs experienced a lower rate of relapse the older they were

One somewhat surprising finding was that the risk for relapse may be less related to the number of substances used than the type of substance.
“Our intuition was that more substances would translate to more severe addiction and higher rates of relapse,” says Roy. “But that was not the case. In fact, it was just the opposite.”
This finding will help add to the knowledge base that enables clinicians to adjust care for patients who are at higher risk. It also will be the foundation for a new Hazelden Betty Ford research publication that the nonprofit plans to share with the entire addiction-treatment field.
“There aren’t many treatment systems that use data to further research in this area,” says Quyen Ngo, executive director of Hazelden Betty Ford’s Butler Center for Research. “Thanks to the Carlson School’s cutting-edge technology and expertise, we were able to examine real-world data in a way—and in a volume—that is new to us and to our field… [and that get] to the core of how addiction, treatment, and recovery work,” says Ngo.

This story was adapted from the original version