Skip to main content
Open this photo in gallery:

AI and machine learning are helping to battle tough opioid addiction.

With the opioid overdose crisis impacting communities right across Canada, the healthcare, social and legal network designed for helping addicts is becoming overwhelmed. The social implications are horrendous but the salt in this wound is that the costs associated with dealing with the effects of opioid addiction have become unmanageable.

This intersection between a massive social problem and a huge financial price tag is what prompted researchers at IBM to begin looking at using artificial intelligence (AI) and machine learning to help those in the trenches of the opioid crisis.

Dr. Lisa Latts, Deputy Chief Health Officer for IBM Watson Health, based in Colorado, says that part of the problem in getting a handle on managing complex issues such as opioid abuse is that “there is just so much information out there, there is just no way a human can get through all of it.” To help alleviate — and potentially help address this part of the problem — IBM researchers began to explore how AI could help frontline doctors and counsellors deal with the immediate problems they face when working with drug abusers. In essence, AI stepped in to help synthesize thousands of disparate data points and help form an individualized treatment recommendation that could help an addict stay clean.

It’s not just a matter of slicing and dicing through structured data such as spreadsheets, explains Dr. Latts. It’s also the ability for machine learning to help speed up the process of synthesizing this structured data with “unstructured” information.

What did this mean? It meant using AI and machine learning to develop a box of tools that could be employed by the patient or by frontline workers. It also meant a system of patient management that helps insurers and government healthcare agencies to reduce the costs associated with untreated addition or crisis intervention.

As Dr. Latts explains it: The opioid crisis has hit virtually all areas of society, from insurers and employers to patients, doctors and the counsellors who treat the addicts. According to Canadian federal government statistics, there were 4,000 opioid-related deaths in 2017. That figure is expected to grow in 2018, especially with the growing prevalence of the synthetic drug fentanyl, which accounted for 70% of the deaths in 2017.

In the U.S., opioid use was responsible for 50,000 deaths in 2017. By the end of 2018, the cost to treat opiate addiction will climb to more than $75 billion per year. The toll of opiate abuse is unlike anything we’ve seen before, Dr. Latts says.

As part of IBM’s efforts to help deal with the multi-faceted nature of the opioid epidemic, the company’s Watson Health division has developed a coordinated set of tools that uses AI and machine learning to fight the crisis. And so far, the results are promising. For instance, Watson Care Manager has been used in juvenile drug courts in the U.S. to help judges make critical decisions about sentencing and treatment options for opioid abusers, says Dr. Latts. Through “natural language processing,” the Watson tool can search through notes from parole officers, parents, case managers and doctors to help the judge quickly figure out the best course of action. This is key considering judges have very little time — usually only 10 minutes, at most — to make rulings that could have a huge impact on an opioid abuser’s future.

What really makes AI and machine learning so useful is its ability to synthesize a myriad of information in a fast, efficient way. Using AI, doctors’ on the front-line of the opioid epidemic could get access to doctors’ records, counselling notes, and information from other sources, to give them a more targeted and effective plan for treatment.

For instance, there’s a pilot program that uses AI at the patient level to create a “digital tether” between the counsellor and a client who has just finished rehab and is facing a big struggle to stay clean.

The pilot program is known as TryCycle and it’s the brainchild of Ottawa-based John MacBeth and his Connecticut-based partner, Ken House. Developed through a partnership with IBM Canada, TryCycle is a smartphone-based application, designed to create a digital connection between recovering patients and their counsellors. What makes it work is that the app uses the astounding analytic powers of IBM Watson, combining Artificial Intelligence and sophisticated analytics to create a "question-answer" solution.

To help determine urgency, the TryCycle app allows for real-time reporting and monitoring of a patient’s vulnerabilities, emotions and overall well-being. MacBeth says that based on answers provided by the recovering addicts to a series of specially-curated questions, patient comments at the end of the survey, as well as audio responses, the app can help gauge emotions and sentiment.

These structured and unstructured responses are then analyzed by the TryCycle data and IBM Watson Natural Language Understanding, Tone Analyzer and Watson Studio. This machine learning is able to identify the risk of a relapse in real-time. Plus, the TryCycle system can even analyze the patient’s movements — noting any changes in location, for example, that would indicate the potential for relapse. In this way, AI provides a link patient and coach so that help can come quickly and when it’s needed most.

Still, the app is in the early stages of testing. The next stage is to launch the program at a commercial level at a proposed cost of $75 per month for each patient. MacBeth thinks this is a small price to pay for unchaining the counsellor response “from the tyranny of a calendar.” He adds that “it makes absolutely no sense that a person who is in a crisis situation now has to wait two or three weeks to get an appointment with a practitioner. With something like TryCycle, response times can change and this reduces the strain on already stressed emergency and health resources; more importantly, it can help save lives.

Dr. Latts acknowledges that IBM’s foray into AI and machine learning solutions to help solve addiction problems is just a start. And while much of this technology is still in the formative stages, the hope is that these types of tools can help front line clinicians and patients make more informed decisions to help address the opioid crisis.

Advertising feature produced by Globe Content Studio. The Globe’s editorial department was not involved.

Follow related authors and topics

Authors and topics you follow will be added to your personal news feed in Following.

Interact with The Globe