Artificial intelligence on the frontlines of suicide prevention
What if we knew who was at high risk for suicide long before they were in danger?
If it sounds like the premise of a sci-fi novel, you’re not far off – but neither is the technology.
In fact, Dr. Zachary Kaminsky says it’s here now. He has built an algorithm using artificial intelligence to identify Twitter users at high risk of suicide.
Public posts on the popular social media platform can hold clues about who might be at risk for suicide long before they reach a crisis point. The algorithm doesn’t just use word recognition or identify suicidal ideas – it analyzes speech patterns and pinpoints the psychological concepts related to suicide. It then scores and matches these patterns alongside all the public data linked to a Twitter user.
“This isn’t identifying only people saying, ‘I’m going to kill myself.’ It’s identifying a risk pattern for people who are moving towards that point,” says Dr. Kaminsky, DIFD Mach-Gaensslen chair in suicide prevention research at The Royal’s Institute of Mental Health Research (IMHR).
He says he built the algorithm using Twitter data because it’s public. It is also a social media platform used by more than a third of Canadians between age 18 and 34 – a population amongst whom suicide is the second leading cause of death.
The algorithm can effectively comb through years’ worth of tweets in minutes and, through machine learning – part of artificial intelligence – link together all of the data to identify at-risk users.
In the years since the original paper was published in Nature’s NPJ Digital Medicine, Dr. Kaminsky has collaborated with other scientists at The Royal and at Carleton University to validate the algorithm’s performance in two online studies during the pandemic – finding that the algorithm correctly identified those with the highest levels of suicidal ideation with around 80 per cent accuracy.
Importantly, suicidality in these Twitter users was determined with depression scales, meaning that the method isn’t just identifying those who are comfortable sharing their suicidality on social media.
Having a tool that identifies those at risk for developing suicidal thoughts opens up the possibility of developing novel interventions based on this data.
“A key value-add of this technology is the ability to predict who is at risk before they reach a point of crisis. This means that we can hopefully pair this with fairly non-invasive interventions, like encouraging a timely connection to the people and supports that already exist in someone’s life,” he says. “Critically, if we do this early enough before someone is in distress, we may be able to avert periods of crisis entirely.”
Dr. Kaminsky is working alongside commercial partners at Dionysus Digital Health INC – a start-up company he helped to co-found – to develop such a digitally augmented intervention and hopes to evaluate its efficacy in future years.
He is also working to develop a new algorithm that uses personalized Twitter data to suggest the best way to intervene for someone who is in suicidal crisis.
“If we build a personalized model based on someone’s data, we may be able to generate a tool that gives crisis responders or peer supporters suggestions on how to tailor their responses in a personalized way,” he says.
Based on his preliminary analysis of over 200 people, those that received high scoring responses to their suicidal mentions had a significant decrease in their suicidality over the following days and weeks compared to those that received poorly scored responses.
Intuitively, this means when people received personalized responses from people that knew them and knew what was important to them, they were more likely to get better. Dr. Kaminsky hopes that in the future, anonymous crisis response lines may be able to use this sort of method to give their responders an inside track and boost the ability of their staff and volunteers to help those in need.
In reflecting on his work, Dr. Kaminsky adds, “At the end of the day, we want to move from just identifying those at risk of suicide to developing novel tools to help – tools that will enable interventions delivered at the right time and in a personalized manner that can be scaled and allow for improved access to care.”
More personalized treatment for people with a post-traumatic stress disorder
The University of Ottawa Institute of Mental Health Research at The Royal has received more than $1M from the DND (IDEaS program) to find biomarkers that could lead to better diagnosis and more personalized treatment for people with PTSD.
The research study, called Multi-Dimensional Assessment of PTSD Subtypes (MAPS), aims to improve understanding of post-traumatic stress disorder (PTSD), in particular the dissociative subtype of PTSD, so that it can be more easily diagnosed and more effectively treated.
People experiencing the dissociative subtype of PTSD (PTSD-DS) feel detached from themselves and the world around them, making the world seem unreal or dreamlike. They may also experience emotional detachment. These dissociative symptoms can come in addition to other symptoms of PTSD.
MAPS takes a comprehensive approach to exploring PTSD-DS, using various techniques and technology to look at PTSD-DS from a whole body perspective. The research team behind the study includes scientists with varying areas of expertise from the Institute of Mental Health Research (IMHR), in collaboration with The Royal Operational Stress Injury Clinic to reach out to veterans willing to share their experiences.
“This is really leveraging the strength of the IMHR when it comes to taking a bunch of experts, having them each do their thing and then putting it all together,” says Dr. Zachary Kaminsky, the DIFD-Mach-Gaensslen chair in suicide prevention research and the principal investigator for MAPS. “It’s everyone’s expertise coming together that makes this such an impactful project.”
The results so far are promising. The team initially studied information from 32 former Canadian military personnel, looking at a multitude of factors, including brain function and activity, heart rate, sleep patterns, genetics and inflammation. The team then generated an algorithm that can predict PTSD-DS with 91 per cent accuracy.
Over the past two years, the team also followed approximately 60 Canadian veterans over a more extended period of time to replicate their initial model and to understand what biological factors change with improving or worsening symptoms.
The first major result was that the team’s initial model, which predicts PTSD-DS, worked in the newly recruited participants with 85 per cent accuracy.
“This means the biomarkers we identified in the first part of the project truly represent some of the underlying biology distinguishing the dissociative subtype of PTSD,” explains Dr. Kaminsky. “Importantly, the results only really gel when multiple biomarkers are considered at once, demonstrating the power and promise of this whole body approach.”
The second major finding was that epigenetic data and brain imaging data implicated a possible dysregulation of a system responsible for making one of the major neurotransmitters in the brain, glutamate. Interestingly, this brain signalling chemical has been involved in PTSD and other stress-related conditions, suggesting the change identified may be important for understanding the disease.
“These sorts of results act as signposts, pointing us in the direction of what to investigate further,” says Dr. Kaminsky.
The team notes that while there is more work and analysis to be done, they are on their way to achieving their two goals. First, to better understand the biological underpinnings of PTSD and its dissociative subtype, and second, to create predictive diagnostics and enable personalized medicine to improve the chances of recovery for people with PTSD-DS.
Life-changing science beyond the lab: Providing real-time treatment alternatives through depression research
The Royal believes in a future where everyone living with a mental illness or substance use disorder can get the personalized help they need, where, when, and how they need it.
The brain is more complex than any other known structure in the universe. It defines our humanity. Yet we still know so little about it.
In the nation’s capital, more than 5,000 individuals suffer from difficult-to-treat depression, meaning it will not improve despite multiple rounds of treatment.
Medication can be a blunt instrument. With few scientific advances in treating mental illness over the past fifty years, we need medications that work faster, with more rapid-acting antidepressant effects as traditional medications to treat depression can take 4-6 weeks to begin to work. Second, we need medications that are effective for more people, medications with improved treatment response and remission rates. Ketamine, a novel treatment strategy for depression, appears poised to meet both these needs.
Recent research has shown that ketamine, a medication in use for over 60 years to provide general anaesthesia and pain relief, has rapid antidepressant effects. When administered in small doses by nasal spray (esketamine), patients show a rapid decrease in depression and suicidal ideation, often within a few hours to a few days. This discovery has proven to be a life-changing – even life-saving – treatment and is considered the biggest breakthrough in the field of depression over the past 50 years.
Support to The Royal will help advance innovative treatments such as this. Here, research and clinical excellence co-exist seamlessly, together with our patients, as the research bench moves, quite literally, closer to the bedside.
The Royal has been a national leader in research on the antidepressant effects of ketamine and is ideally positioned to not only implement delivery of this innovative new treatment, but also to translate new knowledge into clinical practice throughout Canada. The first clinical trial of ketamine for depression in Canada was conducted at The Royal’s IMHR by a team led by Dr. Pierre Blier. The resulting paper describing the team’s findings was published by lead author Dr. Jennifer Phillips in the American Journal of Psychiatry, the most widely read psychiatry journal in the world.
While The Royal is a leader in this novel treatment, demand far outpaces supply. With new investments, we can scale this service nationally breaking down barriers to access this life-saving treatment.