Could AI spark the opioid epidemic’s end?

The opioid epidemic has complicated researchers for over two decades as they try to understand the changing social and systemic factors that lead to opioid abuse and identify overdose hot spots.

Clinicians provide safe, effective treatment and other resources to addicts, which is tedious and often flawed.

AI may be the moonshot that ends the opioid epidemic, as researchers and clinicians examine its widespread impact.

Healthcare is notoriously slow to adopt new technology. And this tendency has consequences. According to one report, the industry loses $8.3 billion a year by not adopting advanced electronic health records.

But the opioid epidemic kills more than the ledgers show. Drug overdoses have killed over 1 million people since 1999. In 2021, America had 106,699 drug overdose deaths, a record high per capita. Prescription painkillers like Vicodin and Percocet and “street” drugs like heroin caused 75% of these overdoses.

The epidemic has persisted despite billions of dollars spent on outreach, education, and prescription monitoring by the CDC and NIH.

I have studied the opioid epidemic in rural and urban America, including New York City and rural southern Illinois, for a decade.

Most in my field agree, reluctantly, that identifying drug users’ complex risks requires a lot of guesswork. Which drugs will they get? Will they smoke, snort, or inject? Who will they call for help if they overdose?

Not that one. Practitioners also battle inconsistent federal and state opioid use disorder treatment guidelines like suboxone. They also have to catch up with unpredictable drug supplies contaminated with cheap, synthetic opioids like fentanyl, which is causing opioid-related overdose deaths to rise.

The public has been fascinated by AI developments like ChatGPT, but public health researchers and biomedical engineers have been quietly creating an AI-fused revolution in medicine, with addiction prevention and treatment the latest beneficiaries.

This innovation uses machine learning to identify people at risk of opioid use disorder, treatment disengagement, and relapse. Researchers from the Georgia Institute of Technology developed machine-learning techniques to identify Reddit users at risk of fentanyl misuse, and others developed a tool to find misinformation about opioid use disorder treatments, which could help peers and advocates educate.

Sobergrid and other AI-powered programs are developing the ability to identify at-risk relapsers based on their proximity to bars and connect them to recovery counselors.

The most significant changes reduce overdoses, often caused by drug mixing. Purdue University researchers have developed and tested a wearable device that can detect overdose and automatically inject naloxone. Tools to detect hazardous contaminants in drug supplies could greatly reduce fentanyl-fueled overdoses.

Despite its potential, could facial recognition technology be used to find high people, causing discrimination and abuse? Uber tried to patent a drunk passenger detection technology in 2008.

Are chatbots already plagued by dis/misinformation? Should malicious parties insert false information into chatbots to mislead drug users about risks?

Since Fritz Lang’s 1927 silent film “Metropolis,” the public has been fascinated by the idea of humanlike technology making lives easier and richer. From Stanley Kubrick’s 1968 film “2001: A Space Odyssey” to “I, Robot” and “Minority Report” in the early 2000s, these wistful visions have become existential dread.

Researchers, clinicians, patients, and the public must keep AI honest and prevent it from making humanity’s biggest problems, like the opioid epidemic, insurmountable.