With the United States gripped by an unprecedented opioid abuse epidemic, regulatory bodies, researchers, and companies are scrambling to find novel ways to alleviate the growing problem. As the causes of the crisis are complex and multi-faceted, so too will be the solutions.

Artificial intelligence is already being widely employed in the fight against opioid abuse—in spaces like surveillance and detection to treatment and recovery, and by stakeholders from law enforcement to healthcare players up and down the supply chain.

Prediction and prevention

Artificial intelligence has been playing an increasingly prominent role in assisting researchers, healthcare providers, and law enforcement to identify and analyze patterns of opioid-misuse behavior.

Just last month, Health Share of Oregon and healthcare analytics company Health Catalyst announced a partnership to create a tool containing machine-learning algorithms to predict individual members’ likelihood of experiencing opioid addiction. IBM has also deployed Watson in various partnerships throughout healthcare to utilize machine learning in a predictive capacity. In collaboration with MAP Health Management, IBM Watson will combine prescribers’ handwriting analyses, real-time data from smart devices, and patient-risk models to assist treatment providers, healthcare insurers, and care managers in intervening appropriately to prevent relapses in patients identified as high risk.

Digital pill technology is currently being explored as a tool used to track and predict patients’ patterns of opioid use. In a study reported in late 2017, capsules containing oxycodone alongside an ingestible sensor were administered to 15 acute fracture patients, allowing investigators to monitor medication usage. “As an investigational tool, the digital pill provides a direct measure of opioid ingestion and changes in medication-taking behavior,” said senior author Edward Boyer, MD, PhD, of Brigham and Women’s Hospital’s Department of Emergency Medicine. “This technology may also make it possible for physicians to monitor adherence, identify escalating opioid use patterns that may suggest the development of tolerance or addiction and intervene for a specific medical condition or patient population.”

Smartphone apps have likewise been showing promise in mitigating the crisis via predictive data analysis. A startup called Triggr combines data from a user’s smartphone—screen engagement, texting and email patterns, sleep history, location—as well as communications between the user and company staff to inform a series of algorithms that assess likelihood of relapse. As relapse likelihood approaches a threshold level, Triggr engages in actual intervention, engaging in outreach to the user and alerting the user’s specified relations or care providers.

Detection and enforcement

Machine learning also shows promise in detecting illegal online sales of opioids. In 2017, researchers tested a methodology that used keywords to aggregate a large number of tweets and employed unsupervised machine learning and Web forensic analyses to identify the illegal marketing of opioids. Researchers say the methodology can aid criminal investigators in identifying violators of the Ryan Haight Online Pharmacy Consumer Protection Act.

The DOJ recently formed the Opioid Fraud and Abuse Detection Unit, a focused data analytics program with the purpose of identifying and prosecuting doctors, pharmacies, and medical providers who contribute to the opioid epidemic. The unit will analyze aggregated indicators—such as which physicians prescribe opioids at rates far exceeding those of their peers, how many of a prescriber’s patients die within 60 days of an opioid prescription, the average age of a physician’s opioid recipient, pharmacies that dispense disproportionately large amounts of opioids, and regional hot spots—to aid in enforcement.

Researchers are also seeking ways to identify emerging overdose trends in order to improve prevention and response measures. For such measures to be successfully targeted, the reliability of trend data is paramount. The Event and Pattern Detection Laboratory of Carnegie Mellon University has been investigating machine-learned statistical methods that can detect clusters of opioid-related overdoses that share geographic, demographic, and/or behavioral commonalities. These models may assist local public health agencies to direct prevention and response efforts to targeted locations and subpopulations, increasing the administration of medication-assisted treatment such as methadone and suboxone and improving patients’ treatment compliance. The researchers plan to apply their work “to better handle multidimensional, correlated space-time data, and apply them to new overdose-related data sources, including data from the state of Kansas’s prescription drug monitoring program, in collaboration with public health partners.”

To alert law enforcement and health officials to opioid overdose spikes in their communities in real time, the Washington/Baltimore High Intensity Drug Trafficking Area created a web app called ODMAP, the first of its kind with a national scope. The app combines street-level data with tools from digital mapping company, Esri. Indiana and other states hit particularly hard by the opioid epidemic have also successfully prioritized complex data analytics tools in their anti-opioid abuse efforts.

What can wholesale distributors do?

With the ongoing digitization of the supply chain, including serialization as well as real-time sophisticated drug tracking software, wholesale distributors are already able to limit the flow of opioids into the wrong hands. Going forward, WDs are harnessing new technologies to improve their ability to combat the opioid crisis—investing millions in improved tracking tools and analytics to refine anti-diversion and SOM frameworks.

As Healthcare Distribution Alliance (HDA) President and CEO John M. Gray recently stated on behalf of his association, “For our part, we’re ready to move forward with practical solutions to improve communication between all entities in the supply chain and with law enforcement to mitigate abuse and misuse before it occurs.” These practical solutions must and will include artificial intelligence in its many forms.

Areas ripe for wholesale distributors to leverage artificial intelligence include blockchain technologies, machine learning, big data, and predictive analytics. Wholesale distributors need to keep abreast of emerging technological efforts across the industry as well as academic research and enforcement, and should consider partnerships and associations across such spaces.

No single effort will abate the ongoing crisis of opioid abuse; the solutions are multi-faceted and span disciplines and industries nationwide. Increased national attention, cross-industry collaborations, and artificial intelligence technologies offer hope that the worst may be behind us.