Assistant professor Hai Phan, of the Department of Informatics at NJIT’s Ying Wu College of Computing, said the project seeks to harness the power of machine-learning — in which computers are taught to learn on their own, without additional programming — and the wealth of online data to diminish the impact of drug abuse, which has serious social, economic and public health consequences. NJIT is collaborating on the project with Prevention Links, a Roselle-based nonprofit that works with communities to strengthen their resilience to addiction.
Referring to the crisis of drug abuse nationwide and the devastation it’s causing in communities, Phan said, “Unfortunately, local communities and organizations still lack tools and the means to address the problem more effectively in this new reality, in which online and offline worlds blend seamlessly together.”
Plans have yet to be developed for distributing the product, which Phan said will be available to community groups, nonprofits and individuals who have positive social intentions.
Phan noted a 2018 federal survey which stated that one in 10 U.S. residents over age 12 abuse illicit drugs. In New Jersey, tens of thousands of residents seek treatment for addictions annually — with many others unable, or unwilling, to get help. More than 3,000 people died of drug-related issues last year.
Using data science to tackle the problem“The number of deaths from drug abuse is higher than murder. It’s a big problem,” Phan said, noting that he was attracted to the project by the opportunity to use data science to help tackle such a major problem. Phan will present a paper on the concept at the 2019 MedInfo conference, to be held the last week of August in Lyon, France.
A plethora of apps and online tools have already been developed to help individuals with addictive disorders manage their disease. These include programs that track sobriety, connect them with others in recovery, or help them locate a support group or peer counselor nearby. (Some states are also using big data to tackle other healthcare problems, like improving outcomes for Medicaid members.)
Hai Phan, assistant professor in informatics department of NJIT’s Ying Wu College of ComputingPhan said technology has also been developed to help law enforcement and other organizations track illicit drug activity. However, these tend to focus on just one or a handful of substances, or a limited geographic region, he said, and often involve outdated data and terminology.
DrugTracker is being built to sift through the slang and other terms on platforms like Twitter and Reddit to flag illegal drug use, sales and associated behaviors — all in real time, Phan explained. Treatment and recovery organizations, and families impacted by drug use, can use these results to better understand new terms, trends and patterns in substance abuse, including the branding used by dealers to attract customers.
The online tool will combine these findings with mapping capabilities so that organizations using the system can deploy resources — like outreach workers or public health teams — to the latest hot spots for drug activity. These results could also help groups identify the days and times when their services are needed most.
“We hope to identify from social media information who is being impacted, what drugs are being misused and use the data to educate and better utilize our resource distribution,” said Morgan Thompson, Prevention Links chief executive officer.
Learning slang termsThompson said Prevention Links’ role has been to help NJIT identify key search terms and refine the platform’s search process, so that it weeds out false positives; for example, the tool must be able to differentiate between a place name, like “Molly’s Bar,” and other uses of the word “Molly,” a slang term for the party drug MDNA, or Ecstasy.
The system is also being designed to recognize images, Thompson said, something that is critical when tracking heroin use and sales. Dealers don’t often use common terms like “heroin” online, she said, but they will post a picture — or a written description of the “stamp,” the telltale image on the glassine envelope used to package the drug. These stamps function as a calling card for the dealer and become familiar to drug users, who seek out certain brands known for their strength.
“The branding is big — especially right now. The stamps are what people are looking for,” Thompson said, noting that a particularly popular option right now is “Donald Trump,” printed with the name or an image of the president.
Phan said NJIT is working on final details so that DrugTracker can go live within the next six to nine months; he said it will evolve after its launch, with new features designed to keep pace with the changing nature of the addiction epidemic. The system is devised to work in multiple ways, he said, by helping organizations plan, operate and share information and also allowing individuals to connect with addiction-related services.
Eventually the system will be available free of charge to communities, groups or individuals with positive social-service mission, Phan said; interested organizations are encouraged to contact his office for more information.
“Our hope and purpose are to help families, local communities, and local organizations in order to reduce the tremendous economic, societal, and public health challenges” associated with addiction, Phan said.