Suicide: Predicting the unpredictable

Suicide prevention is the hardest in the hard basket, but data and strategic thinking are making the impossible seem possible.

Mental health is difficult: It’s difficult to diagnose, to treat, to predict and to understand. Suicide is beyond difficult as there are so many individual factors that lead to someone intentionally taking their life which makes it a uniquely problematic phenomenon to research.

This is reflected in the national suicide rate, which has gradually risen in the past 10 years. According to the Australian Bureau of Statistics (ABS), in 2008, suicide was the leading cause of death among people aged 15-44 in Australia.

The ABS also reported that men were three times more likely to intentionally take their own lives. The median age for suicides was 44.3 years of age for males and 44.7 for females.

In writing this piece, the topic of prevention strategies for at risk cohorts was broached with Professor Andrew Page. In his response, he reframed suicidal ideations as something to be considered as a variable risk factor that’s not just applicable to certain at-risk cohorts, but rather to just one: everyone.

“What we need to be looking at is what would increase that risk within us, because thoughts of killing yourself are wide spread. We’ve looked within university cohorts and over the past year, close to half of them have thought about killing themselves at some point in time, so we know that this is not an isolated problem.”

Prof Andrew Page is the Pro Vice-Chancellor of Research at the School of Psychological Sciences at the University of Western Australia and is part of the research leadership team at the Young Lives Matter Foundation (YLM).

Suicidal ideations are like most other aspects of mental health – variable and experienced by many people at one or many points in their lives, usually innocuous enough not to raise alarms.

Anxiety, depression, obsessive thoughts, rigid thinking. Sound familiar? In small enough cases they can be useful in navigating through life, or triggers to address the causality, yet in prolonged and chronic cases, they can be debilitating.

This way of looking at whole-population suicidal variability is not just an interesting idea to reframe suicide, it’s a practical tool that can measure risk, save lives and also reduce stigma.

Part of Prof Page’s work, in collaboration with Perth Clinic and with YLM, is the development of a world first clinical tool, the Dynamic Developmental Vulnerability Index (DDVI), which utilises artificial intelligence and machine learning to predict suicidal or self-harm behaviours. The phases of suicidal ideations begin with fleeting thoughts of suicide, progressing to planning or attempting suicide, to the ultimate tragedy of the action itself. It’s the subtlety and significance of these phases that the DDVI has been designed to identify for intervention.

Prevention through prediction

YLM was formed with the singular purpose of developing new interventions that would reduce the suicide rate, such as the DDVI. The YLM team is comprised of a multidisciplinary group of mathematicians, statisticians, psychologists, and psychiatrists.

“Young Lives Matter is about looking at creative and innovative ways that we can more effectively predict self-harm and suicide, especially with youth, so that we can more effectively target preventative efforts, Prof Page told Medical Forum.” The impetus for YLM to specifically look at predication was based on what hasn’t been done, rather than focusing on what has.

“When we looked at what was happening across the nation, there was a lot of work on prevention but there was less work in terms of prediction. And if you have finite resources the question becomes, how do you allocate those finite resources in the most effective way that you can predict?

“The problem has been that the prediction of suicide has been notoriously poor. In fact, it’s been so poor that some people have suggested we should just give up trying to predict at all. As a group, our response to that was: If it’s a problem that everybody thinks is too hard, it’s one that we should be trying to solve.

“And clearly, we need to solve it in new and different ways and not just keep on doing the same things that we have been doing, expecting them to be able to solve the problem.”

The team at YLM decided to shift the research paradigm of suicide prevention, Prof Page explained.

“Up until now, most researchers have tended to think about suicide risk by trying to identify who is at risk and so they generally find older men, people with mental health problems, substance use etc. And what we’ve said is, well, if you think about suicide risk, if you talk to anybody who’s felt a risk of suicide, what they’ll say is, ‘yes, I was thinking about killing myself yesterday, but I wasn’t the day before. I’m not today, but I might tomorrow’.”

By measuring the variability or fluctuations of suicidal ideations, instead of at-risk cohorts, the team at YLM can identify when someone is most at risk and intervene before that person can act out their ideations.

“Suicide risk fluctuates. If you’re trying to predict something that fluctuates with something that’s static, all you’re going to do is be able to predict the average, you won’t be able to predict those fluctuations. And that’s what we need to do.”

According to Prof Page, by reframing the research to focus on when someone at risk rather from who is at risk this opens the possibilities of measuring variability.

“Once you’ve changed the question, you ask it differently. So, you no longer ask, ‘who are the individuals?’ But you ask of individuals, what should I be measuring and at what time period’ to be able to work out when risk is increasing and when it is when it is decreasing.”

The current iteration of the DDVI is being trialled at Perth Clinic as a self-report questionnaire administered on iPads. When patients initially present to the hospital, they are asked to complete the questionnaire and continue to report on it daily. The reports are then processed by machine learning algorithms based on dynamic risk factors for self-harm and suicidal behaviours to predict individual risk.

If a patient reports consecutive days of increasing self-harm or suicidal ideations, staff are alerted and resources can be allocated prior to a suicidal action. If the same patient reports reduced ideations, their risk is lower and the hospital’s resources can be focused on a higher risk patient.

Beyond the hospital, Prof Page envisages self-administered questionnaires or possibly a wearable device for individuals to monitor their own risk and if there is a flag, such as reduced sleep or elevated vital signs, they make the data available to a health care professional, such as their GP.

In its clinical use, the DDVI has provided encouraging results, however, Prof Page does not see it as a one size fits all, rather the dynamic risk factors can be adapted for different people in different situations such as for Aboriginal people in their communities, or school students.

Back to basics

Suicide affects everyone it touches but there is no avoiding the overwhelming majority of suicides in Australia are male. Yet, what are the factors that make men so much more susceptible to taking their own lives?

For one thing, it is a lack of communication, according to Emeritus Professor John Macdonald: “If only men would talk, then that would be a sure path to prevent suicide.”

Prof Macdonald is the former Director of the Men’s Health Information & Resource Centre and Foundation Chair in Primary Health Care at Western Sydney University and patron of the Australian Men’s Shed Association. Prof MacDonald has over 30 years’ experience in primary health care and education with a particular focus on the social-determinants of health.

“In addition to talking, there must be an effort to acknowledge that behind the suicide ideation could be unemployment, family issues, such as separations from children. We tend to quickly label anybody who approaches us in this context as being mentally ill and label them as depressed and understandably, we medicate them.

“If we look at the rate of medication over the past 15 years for depression, it’s increased phenomenally. We’ve equated suicidal ideation with depression and I think we should have the courage to question that and go behind what is causing the depression.”

“This is not to say we who look to the social determinants have the easy answers, however, it’s much easier and understandable to prescribe something the literature says is useful for these cases than to actually deal with the issues behind it.”

If we see someone as being emotionally distressed or mentally ill, it will, of course, influence the way we deal with them, prescribe for them, or seek to help them. We should focus on the issue of emotional distress rather than just quickly label someone as being mentally ill.

As part of a study investigating the social determinants of suicide, Prof Macdonald interviewed families of men who had taken their own lives and men who had made serious attempts to end their lives.

“The general conclusion was many of the cases were a result of an accumulation of terrible life events, which led people to a situation where they could no longer cope. And those things could be unemployment, separation from a spouse, loss of home.

“If someone had been sexually abused or abused as a child, that didn’t set them up well to deal with the [stresses of life] that we all have to deal with. If sad life events accumulated that was often what led people to say, ‘well, that’s enough’.”

Prof Macdonald said social support is an essential component of the social determinants of male health, citing his involvement in the Australian Men’s Shed Association, which he sees as essential for men to be in contact with others that are in the same or similar situations, particularly for older men, in reducing or preventing emotional distress.

Modelling the care

Emergency rooms can be chaotic places with endless streams of variable presentations, each requiring a multidisciplinary team of clinicians to provide care, which leads to a formidable balancing act of resource scaling to meet these demands, especially for patients experiencing emotional distress and who are at risk of self-harm or suicide.

With this in mind, it was unusual to learn that Airbnb’s newly renovated headquarters in San Francisco was an inspiration for UWA’s Professor Sean Hood to research service design for emotional distress and mental health presentations to a Perth ED.

Professor Sean Hood is the head of psychiatry in the UWA Medical School, Associate Dean of Community and Engagement in the Health and Medical Sciences Faculty and a Director of UWA Young Lives Matter Foundation (YLM).

According to Prof Hood, Airbnb hired a team of designers to follow their staff for around three months to monitor how they worked and moved through the office before reporting back with design suggestions of an office redesign.

Based on this approach, Prof Hood, formed a small team of psychiatrists and mathematicians to go into the emergency department at Sir Charles Gardner Hospital and track patients who presented with self-harm or suicide risk.

The team then noted the steps of each patient’s flow through the hospital, from the initial presentation to when they were discharged, and then the mathematicians coded these steps into a network model which they could manipulate to identify the impact on the patient of each clinician who saw them and then tested what would happen if they removed a particular clinician from the patient flow.

Such a role was the psychiatric liaison nurse, whom the team found was vital as they were the key team member in the patient flow that would capture patient data specific to self-harm and suicide.

“At the time, the hospital was reviewing this service and we could show them what would be lost if the service was removed; what vital information would not be captured anywhere else in the system, Prof Hood said”

Data driven service delivery

Now imagine combining data captured from the Dynamic Developmental Vulnerability Index, with emergency department patient flow data, and having an algorithm to predict risk of self-harm or suicide during a patient’s journey through the hospital. Targeting resources could be a whole lot easier.

“One of the key issues with mental health is we can look back and say ‘I see significant risk factors in that patient with poor outcomes but the same factors were present in almost everyone else we saw.’ But if we can better predict a person has a cluster of symptoms, which we might not clinically consider, and track that mathematically, we can flag that person as more at risk and provide more resources at the point in which they need them,” Prof Hood said.

“Ultimately, we are hoping this has an impact on suicide but in the short-term we can better utilise the resources that we have.

“Currently, most of the decisions are made by clinical wisdom with a clinician saying, ‘I just feel this is risky, I don’t feel happy about sending them home’ without external diagnostic guidance.”

Traditional approach

Although taking the non-cohort approach to suicide risk modelling has its merits, there are groups in Australia that have far greater risk than others, such as Aboriginal and Torres Strait Islander people.

Another project that Prof Hood and YLM is a is focused on is the use of mathematical techniques to map the utility of traditional Aboriginal healers in treating mental health.

“This is an area that’s not well understood and obviously very culturally sensitive. An aim is to know, as a psychiatrist, when it would be most useful for me to refer a person to a traditional healer. And ultimately can we get Medicare rebates or funding to support this service, which for many patients is going to be better than the alternatives, such as uprooting them from their Country and taking them to a hospital.”

Community based care

Similar to the role of men’s sheds for older men, community-controlled services in Indigenous communities can provide pivotal peer support because of the shared trauma these people have experienced, Associate Professor John Allan explained to Medical Forum.

“Throughout Aboriginal culture, the issues of dispossession and displacement from country and intergenerational trauma have brought considerable loss that is still really prominent today. This is part of the alienation many people feel.”

A/Prof Allan is the President of the Royal Australian and New Zealand College of Psychiatrists (RANZCP), the Executive Director of Mental Health Alcohol and other Drugs Branch in Queensland Health, and was the Chief Psychiatrist in both Queensland and New South Wales.

A/Prof Allan has extensive experience in the mental health and wellbeing of Aboriginal and Torres Strait Islander people, particularly in North Queensland where he spent 20 years working with communities and developing mental health services.

“Often, the care and treatment services provided, however good they are, might not be the thing that person needs. They might need Indigenous healing or peer support.”

“The Indigenous communities I have worked with that have been able to alter [the care] through community control services have fostered a sense of purpose and pride in what people are doing,” he said.

“Community development and peer support is often being controlled by councils or elders which gives communities some control over the situation rather than feel they are at the mercy of the police or health workers. That’s where there has been success in altering outcomes.

“Remembering that not all Aboriginal people are the same or have the same social circumstances, it’s important not to label but to look at the strengths of each person and community.”

Real time data collection

It would be remis not to discuss with him the grim forecasts of the University of Sydney’s Brain and Mind Centre for the potential increased risk of suicide due to the economic fallout from the COVID-19 pandemic. This is why, he said, the RANZCP has called for real-time data collection and monitoring of people at risk of suicide.

“The stresses of the pandemic have highlighted the many gaps we have. So, the big concern is the changing economic conditions. We know that unemployment is a trigger for suicide,” A/Prof Allan said.

“We need to understand the patterns of change, yet we don’t have real-time data collection or monitoring of what’s actually happening. We need to be able to identify hotspots, trends and groups of people most at risk.”

He doesn’t want real time data limited to suicides and suicide attempts but for suicidal feelings and behaviours to inform a model that can predict these patterns of behaviour. If this sounds ambitious, it is, in light of what is currently available.

A/Prof Allan said that the current predictive modelling of suicide was not robust when compared with, for example, the COVID modelling, which comes down to the difficulty in identifying and extrapolating the risk factors associated with suicide risk.

“Suicide is actually a rare event. It’s a terrible event but it’s a rare event in terms of the numbers. So, it’s actually quite hard on an individual level to predict who might do what.”

Mental health first aid

Attitudinally, there is a dissonance in how individuals feel they can help with mental health compared to physical health, A/Prof Allan said.

“If someone falls over and breaks their arm, most people know enough first aid to help, to immobilise the arm and call and ambulance. But if a person was feeling low, expressing suicidal thoughts, not everyone feels confident to talk to that person and offer assistance.

“Having a general knowledge of mental health is especially important now with hundreds of thousands of people unemployed and millions in insecure employment.

“People can be trained in mental health first aid, in just the same way they are trained to deal with a broken arm, and have confidence to get a person help and what the pathways to care are.”

Reducing the national suicide rate is beyond difficult but with an understanding of risk variability, real time data monitoring, predictive algorithms, community-based interventions, and the simple act of having a conversation, there is hope.

For more information about the Young Lives Matter Foundation: