Why Suicide Risk Assessment is Ineffective and What to do Instead with Dr. Matthew Large

In this conversation, Dan  chats with Dr. Matthew Large about his experiences in the field of mental health, his research on suicide prediction, and how Suicide Risk Assessment can often times be ineffective. 

Dan and Dr. Large dive into a number of hard-hitting  topics in this 7th episode: the challenges of balancing clinical work and research, the limitations of Suicide Risk Assessment, and more! Dr. Large shares personal stories on their career and what led  to a life in medicine and psychiatry. They also touch on the progress and changes in the field of mental health, including the closure of asylums and the improvement in the treatment of mental illness. The conversation explores the limitations of suicide risk assessment and the over-reliance on categorizing patients as high or low risk. Dr. Large touches on how risk assessment tools have modest predictive abilities and that suicide is a complex and unpredictable phenomenon. Dan and Dr. Large suggest that clinicians should prioritize building rapport with patients and providing individualized care rather than relying solely on risk assessment.

🎞️ Video version of the show@PsychotherapyAppliedPsychology on YouTube
🛜 Check out the website: Listen to every episode on your podcast player of choice

Connect with Dan
☏ Leave a voice message on Speakpipe
🔗 LinkedIn
🐥 @TheAPPod on twitter
📬 TheAppliedPsychologyPodcast@gmail.com
📸 Instagram

Transcript

Broadcasting from the most beautiful city in the world, I am your host, Dr. Dan Cox, a professor of counseling psychology at the University of British Columbia. Welcome to episode number 7 of Psychotherapy and Applied Psychology, where we dive deep with the world's leading researchers to uncover practical insights, pull back the curtain and have some fun along the way in an attempt to bring insights to practitioners and those training to be practitioners. If you find the show enjoyable, kindly consider sharing it.
You can conveniently tap the share button on your phone and send it via text, share on social media or any other method that you prefer. Your support in spreading the word is greatly appreciated. On today's show, for the very first time, I'm going to be welcoming-
let's do that part again. On today's show, for the very first time, I'm going to be welcoming a psychiatrist as my guest, and I could not have been more thrilled with the conversation that I had with this leader in the area of suicide risk assessment and prediction. In our conversation, we discussed suicide prediction, the limitations of risk assessment, psychiatric hospitalization, individualized care, and litigation risk management.
We begin this conversation with my guest answering my question about an early career experience that he had, where he was unsure if he could do the work that he does now. Without further ado, I'm thrilled to have one of the world's foremost experts in suicide risk assessment and prediction. He is the conjoined professor in the School of Psychiatry, at the University of New South Wales and the clinical director of mental health in the Eastern Suburbs Mental Health Service.
Please welcome Dr. Matthew Large.
Thank you.
So I started my career in mental health as a teenager. I got into med school, but I really wasn't sure whether I wanted to do medicine. So hopped on my bicycle and became a nurse at a local hospital.
And so I was a training nurse, and at that stage, there were very few male nurses, and the hospital was very unused to us. There was this sort of suggestion that we were all gay, which about half of us were. But we got shunted down to the psychiatry world.
And it's always difficult to know why you've done something much later. But I do think this experience was quite formative. My job was really just to sort of heard the patients and chat to them.
I don't know what the psychiatrist's job was. They seemed to smoke a lot of cigarettes and drink a lot of coffee and play chess. And there's a lot of psychoanalytic pronouncements and a lot of stelazine prescribed.
So our job was really just to talk to the patients. And then one morning I was coming into the ward through a back entrance into a workshop area, recreation area underneath the ward. And there was a young man hanging with a rope from the workshop, from exposed plumbing.
And my job, I was with another nurse. My job was to pick this young fellow up so that the other nurse could remove the noose from his neck. And then it was absolutely shambolic resuscitation.
And it had a bad outcome. And I really thought lots of things about that, I have to say. And I had no debriefing.
I didn't really talk about it for about 10 years. I was really quite shocked by it. But it was one of the things that sort of said, OK, like maybe I will go to medical school after all, because I just don't think I can do this.
And then, and this is in the late 70s, early 80s, the atmosphere was very much one of anti-psychiatry, that psychiatric syndromes were an invention of psychiatrists and a product of culture. And then, and I'm getting to the second part here. I wanted to do infectious diseases, and I went to southern India.
And there in the Jwandi highlands, saw a whole group of patients in a very remote area who had very much the same presentations as people in Sydney, in suburban Sydney, OCD, anorexia, paranoia, mania. And it dawned on me that these people actually, just really basically weren't well. And I suppose, maybe I didn't swallow a whole biomedical model of psychiatry, but I certainly loosened my view that it was a social problem.
And I've had a strong interest in suicide, really, after this event, as I think I was 19.
So there too, why I became a psychiatrist, I think, and where my interest in suicide comes from, I think it's always hard to reflect on your motivation.
You said that, just listening to your story, I was sort of like, man, that's a very impactful story, very impactful experience. You said that when that happened, you said, I don't know, something like, I don't know if I can do this or something like that, and then that was what led to your decision to go to medical school.
Yeah, yeah. I thought being a nurse was an extraordinary experience, actually. You're at the bottom of the social hierarchy.
People didn't listen to you at all. It didn't really matter to them that you'd had this kind of dreadful experience, and as a teenager, and no one spoke to me about it afterwards, you just expected to carry on the next day. And of course, I didn't tell the people I was working with that I could have gone to medical school.
So it was probably one of the experiences that made me think. And I went to medical school with no view of doing psychiatry. At that point, I was quite angry about how psychiatrists treated patients, and their failure, I suppose, to save this young man who was only a year or two older than me, and their failure to protect me from having this experience.
That makes any sense.
Yeah. In Australia, do you feel like, I would imagine that everything is, of course, sort of culturally based, things are different, but I would imagine there's a lot of similarity. How have, what's been your experience of how things have changed?
Look, things have changed. There's, I suppose, much less stigma about mental illness, and the asylum system has been taken apart, and patients are treated more respectfully and they have more rights. I mean, a very classic example is how we used to treat, how I saw people with anorexia nervosa being treated.
So, if you were a 16-year-old girl presenting with a low body mass index in the late 1970s, a kindly psychologist would talk to you about all the things you liked and make an inventory of all the things that you liked without telling you that all of those will be stripped from you and returned contingent on weight gain. And in the meantime, you would have to lie in bed with no visitors, no contact with the outside world. And it was really an extraordinarily cruel form of treatment.
And, I mean, obviously, we don't do that anymore. But I think, you know, psychiatry or the treatment of mental illness in hospital has progressed, perhaps nowhere near as far as I would like it to. But it certainly has progressed.
Do you think that there are one or two specific events that really helped to push the field forward? Or do you think it's been a slow progression? I mean, it's been slow, but you know what I mean.
So I guess one of them intersects very much with our topic. In that, you know, obviously, in 1975, O'Connor versus Donaldson, the US Supreme Court ruled that it was unconstitutional to detain a person in hospital merely because they were mentally ill and that there must be something more. And that something more was, you know, widely interpreted as, in the US, the term is dangerousness.
Outside the US, the term is need for protection from serious harm. And I think that, you know, in the 1970s and early 80s, you know, coincided with the sort of, you know, fiscal conservatism. So the human rights angle and fiscal conservatism collided, you know, allowing the closure of hospitals.
Now, what's happened later is not universally good by any means. But I think that's been a big change from, you know, my training in the 80s, training and experience in the 80s and my practice these days.
So what got you into the study of suicide prediction?
So I was a junior consultant. And again, with the caveat that I think it's hard to know all your motivations, really, we're very complicated creatures. I was a junior consultant and the in the wash up of the, you know, Christopher is a death in England, the Christopher death of Christopher Clunas, the death of Jonathan Zito by Christopher Clunas, the NHS introduced a requirement that there be a risk assessment at the point of entry discharge into psychiatric services.
And that was introduced in New South Wales. And that would have been in, I suppose, about 1997 or something like that by the time it washed up on our shores. And I had had some years earlier, not when it was released, I've read Alex Bocconi's paper about the prediction of suicide.
He did this really wonderful and very advanced study for the early 80s, where he collected an awful lot of information about 480 consecutively admitted veterans, psychiatrically admitted veterans, and then followed them up for five years and had lots and lots of data points, hundreds and hundreds of data points, including all the things that you would conventionally think might matter. And metrically, the study showed that his high-risk categories, however he defined them, picked up fewer than or around 50% of the future suicides, but also captured a very large number of people who were not going to suicide. And so it had a low sensitivity and an even low positive predictive value.
And the strange thing about that paper, and I remember reading it at the time, he said, well, these are the results, but risk assessment seems to work, so we should continue to do it anyway. And I always found that rather jarring. So if you view yourself as a scientist, and I view myself as a social scientist to some extent, you have to follow the data.
And that niggled me. And then a few years later, I did some studies of risk categorization in psychiatric hospitals. I mostly work in psychiatric hospitals and found very similar result to pecornies.
And then I realized that no one had actually done a meta analysis of longitudinal perspective studies of risk categorization. I was very surprised that no one had done it. And as it happened, there were two other groups working on it at the same time.
My paper came out first by a few weeks. So I was very pleased with myself. And it had results that were so similar to peonies.
We found 53 studies. The odds ratio of suicide of higher risk patients compared to lower risk patients was 4.84, which is quite a modest odds ratio. Not that much greater from greater than the association between male gender and suicide, which in Australia and the US is about 3.5.
The positive predictive value was very low. I think it was about 5% over five years, so about 1% per year.
And the sensitivity was only 50%. So 50% of the suicides occurred in the low risk group. And that really also got me thinking about my earlier, one particular thing earlier in my medical career.
So when I finally finished medicine and was working as an intern in a major hospital, it was right at the peak of the AIDS epidemic. And Sydney was like San Francisco. It's a very gay city.
And we had lots and lots of patients. And we were told at the beginning of the year that we should wear gloves if we were drawing blood from, you know, Haitians, heroin addicts, homosexuals and hemophiliacs. And that was a risk that we...
They were the risk groups. And everybody else, we would draw blood without wearing gloves. And about halfway through the year, I can't remember who it was, but it wasn't me.
He said, this is ridiculous. Anyone can get AIDS. Let's just adopt universal precautions.
And so I think that we need to... You know, if you've got a test that only picks up 50% of future suicides and we want to prevent suicide, we can't be satisfied that low risk patients are not at risk of suicide. And that's particularly true among kind of hospital-treated patients who are classifiers, those presenting to emergency departments with suicidal thoughts and behaviours, those who are admitted and those who are discharged.
They have extraordinarily high rates of suicide. And even patients who are at low risk of suicide, or at lower risk of suicide, according to our instruments, still have very elevated suicide rates. And I just think it's bad medicine to be allocating resources and only worrying about people who we think are high risk.
So I want to sort of... You said a lot right there, and I want to delve a little bit deeper into it. And particularly when we're talking and talking about odds ratios and positive predictive values and things to try to help, particularly the clinician who might be listening, make sense of what that means in the real world.
But let me sort of set you up with my experience, my clinical experience and my clinical training in terms of what I was supposed to do when I suspected a client, when I suspected a patient was suicidal. So what I was trained to do is when you suspect a client suicidal after you sort of ask an initial question, are you thinking of potentially ending your life, are you thinking of potentially killing yourself, and they indicate yes or sort of something like that, then what I was taught to do is get out some sort of a form, and depending on where I was working, the form varied, and it had a handful of questions. Well, I shouldn't say a handful, anywhere from a half a dozen to 20 probably.
A lot of variability to ask the client to assess their suicide risk. And based on how they answered the question, and it was never quite clear what the exact algorithm was, that if they were categorized as high risk, then that meant potentially that I should talk to them about hospitalization, the possibility of hospitalization because they might die by suicide, or they might, they're at higher risk for a suicide attempt.
And I think that that training experience for people in my field is very common. So what is, what's wrong with that approach?
So I think that is a very common approach still. And I mean, there are literally hundreds of these different forms that you're talking about, and no form has ever, you know, gathered momentum as the way to do this, because they're all equally modest in their abilities. And if you look at NHS Trust, so that, and also Health Services in Australia, there are generically, there are locally derived non-empirical form that is used.
And so what's wrong with that? Well, there might be some right things about it. You can gather a bit of information, make sure your history is a little bit more complete.
But going back to those metrics, the sensitivity of this sort of form is about 50%. So the sensitivity is the proportion of future cases that you'll capture in your higher risk group. So they miss out on about half the patients.
And the positive...
So sorry, so of the people who do end up dying by suicide, about half of those folks are in the high risk group, and about half of those folks are in the low risk group.
That's right. That's correct.
So really then... Right. So that calls into question these categories overall.
And if you look at the other metrics, so for a high risk patient, very few high risk patients will die by suicide. That's the positive predictive value. The proportion of high risk patients who will die by suicide, it's a different measure, but it's of the order of 1% per year and commensurately lower over shorter periods of time.
And so that means if you were, for example, to hospitalise all your high risk patients, you would be hospitalising 99 patients, many of them may not want to be hospitalised or may not benefit from being hospitalised in order to potentially capture one person who may go on to suicide. And the other statistic that I mentioned is the odds ratio, which roughly translates into a multiplied probability of the likelihood of high risk patients versus lower risk patients suiciding. So the meta-analytic figure appears to be around five.
So higher risk patients are about five times more likely to suicide than lower risk patients. So suicide risk assessment does give you a little bit of information, and I think that is generalizable to the patient. It does tell you something about the patient.
It just isn't enough to make a clinical decision. And I mean, if you spoke about asking people about suicidal ideation, well, I've done some studies of suicidal ideation as well, some meta-analyses and some primary research. And in community samples, only about 20% of future suicidal people who pass away by suicide have disclosed suicidal ideation at a specific earlier time.
It's a very insensitive test. And the odds of suicide among ideators, suicidal ideators isn't much higher than the odds of suicide among non-ideators, particularly if you have depression, doesn't seem to make any difference at all. And if you were to play Desert Island Risk Assessment, where you were only allowed one risk factor, and you could choose between suicidal ideation and gender, I would argue that you should choose gender, because they're equally statistically related with future suicide.
And gender is a lot easier to ascertain than suicidal ideation. I mean, I partly, it's an exaggerated case, but I do think there is a valid point about the weakness of the association between suicidal ideation and suicide. And that's not to say that you shouldn't ask patients about suicidal ideation.
I mean, obviously, someone who's telling you that they're feeling suicidal is, you don't need, there's no element of prediction in that. They're telling you that they're feeling so bad that they're thinking of ending their life. And obviously, those people need care and treatment and attention.
You need to alleviate that somehow. But there are lots of very distressed people, equally distressed people who also need care and attention but don't disclose suicidal ideation. And I'm just suggesting that our kind of rather naïve, positivistic view of who will die by suicide shouldn't really be the guide to how we interact with people.
I think we need to interact with people in a way that is much more like universal precautions. We need to be careful in all our interactions with every patient, and no patient should go home empty-handed. And I mean, my own personal view and practice is that suicidal ideation, even quite severe suicidal ideation, isn't sufficient remit to compel a person into hospital care.
There does need to be something more than that, and most people are happier and better treated in their homes.
I want to come back to the hospitalization thing. I want to come back to a couple of things you said, but I'm going to go before we leave some of the more data aspects of this. So you talked about for folks in the high risk group that about 1% of them per year will die by suicide.
But for most clinicians, outpatient clinicians, I should just talk about psychologists because that's what I'm really familiar with. That's not what they're interested in. And what I was interested in is what's the probability that this person is going to attempt in the next, really, honestly, 24 hours, maybe up to 72 hours.
Really, I'm thinking about if I let this person leave my office, will he attempt suicide tonight? What do we know, so that year figure for most of the psychologists in practice, the therapists in practice, it's actually not what they're concerned about when they're doing the assessment. So what do we know about that 24, 48, 72, whatever, that really short time span in terms of this predictive ability?
Okay, so firstly, that figure of 1% comes from hospital-treated patients. So they're a little bit different to patients who you might see as an outpatient, be much lower than that. I mean, the suicide rate in the general community is one in 10,000, not one in 1%, and presumably patients who we know that having contact with the hospital is arguably the risk factor for suicide associated with odds of suicide of 20 to 50, depending on what study you look at.
So it would be lower than 1%. Remarkably, very little is known about the term immediate suicide risk is used very widely, but there is no evidence that shorter timeframes in assessments work any better than longer timeframes. I was just recently reading a study at one of the few studies that uses the Columbia Suicide Severity Rating Scale over time.
And actually, their scale is more accurate over a longer period of time. And the positive predictive value of a high risk category in the first few days associated with essentially suicidal ideational plans on the Columbia scale is 1 in 1600. That's the figure.
The PPV was 0.006, which translates to 1 in 1600. And so 1600 people is a lot of people to do something restrictive for. It's a lot of people to do anything for actually to try and capture or try to treat one person.
And this really comes down to what I think is the real problem with risk assessment is not really the metrics. So the question is, the metrics obviously impact on it. What would you do for a high risk patient that you wouldn't do for a lower risk patient?
And you have to think of something that can be morally and ethically and practically delivered to a high risk patient that should be morally, ethically and practically denied to a lower risk patient. And so whether you do a risk, classification is a human enterprise. So if you're making violins, I imagine that you collect wood very differently to if you're making a bonfire or a barbecue.
You know, we classify things for human purposes.
And so how we classify things depends on what range of tools we have available to us. And I sometimes like to think about a hypothetically excellent suicide prevention pill. And so you had a pill that you could give a patient that would stop them from taking their life for the next, I don't know, two months or so.
And the only justification I can think of for only giving that to high risk patients would be cost. So there might be some instances where, you know, if you had a limited supply of these high risk suicide pills, that you could deliver them on the basis of a risk assessment. But I actually just don't think we've got those instruments to us, those instruments available to us.
I mean, there's a very thorough meta analysis by Fox published a few years ago, looking at the, you know, randomized trials of suicide prevention that have been published in the last 50 years, a very sobering study. And it shows that, you know, our ability, psychotherapeutically and pharmacologically, to prevent suicide is actually quite modest, you know, disturbingly modest.
So why do we keep doing this?
That's a great question.
And of course, by this, I mean, just for the listener, by this, I mean, why do we keep, why do we keep using these risk assessment tools that are terrible at, well, just to be clear, these tools actually do increase our ability to predict. However, they don't do it in a clinically helpful way, because as you just articulated, I think it was the one in 1600, or I think that was right, that still, when you're dealing with this patient who's sitting across me right now, they're in the high-risk group, but the only one in 1600 will go on to, I don't know if that study was attempt or die by suicide. That, yeah, that's closer than one in 10,000 or whatever the low-risk group might be, but it's still really not clinically useful.
But we continue to collect these data and make very serious, consequential decisions based on these data. Why do we keep doing that?
So I think it's a central question. Firstly, I do think it takes medicine and psychology, I think it takes a long time to change. There's some evidence that there's a paper that says it takes 20 years for reforms to happen in medicine.
And I remember when I was a nurse, if you had a big heart attack, we'd force you into bed rest for 10 days. And we literally pushed people into bed, stopped them from getting up and walking around. And if you had a smaller heart attack, it was a week.
But we now know that that actually kills people. And that practice, you know, went on in the absence of evidence for years. So...
But I think risk assessment is a little bit different in that we... I think we have our own, you know, risk assessment mechanisms. We worry about other people.
So we've got our own internal, you know, risk instruments. And we're, I think, a little bit aware that those instruments are imperfect, and we would like some external help to help us with that. I think it's also...
It's a beguiling idea that we can see into the future. Suicide, I think, is a very, very complicated and disturbing problem. I don't think it's any...
I don't think it's really whole... There's much contention that to be or not to be is one of the most well-known phrases. It's in the English language.
But, you know, it's a very complicated problem, and having a simple solution to it is attractive and beguiling and, of course, wrong. And then there's this... your actual feedback that you get.
So I've mostly worked in emergency departments and inpatient units, but a lot of time in emergency departments. So you see a patient in the emergency department, and you think they're high risk, and you admit them to hospital. And, of course, hardly any of them suicide, so you think you've done the right thing.
And then you send another group home, and hardly any of them suicide, so you think you've done the right thing. So you don't actually get any real feedback about your decision, and you can be just very satisfied with yourself until such time as you actually do have a suicide.
And one thing that I'm always concerned about in my role is the welfare of people who are junior to me in the system. And time and time again, I've seen young doctors and young psychologists and social workers see a patient fill in these forms that you describe in perfect handwriting, make a totally correct categorization as to low, medium or risk. And then the patient passes away from suicide, and it happens in a big hospital, it happens once or twice a year to either an outpatient or an inpatient.
And the vast majority of those young people will have, and they're very nice people that want to work in mental health, they're caring people, and most of them will have some sort of mood disorder, maybe only an adjustment disorder, but I've certainly seen more severe disorders follow. And there's always a presumption that the risk assessment was incorrect, just by virtue of the fact of the suicide. I mean, it's crazy thinking.
I mean, at one level, I mean, a winning or a losing lottery ticket is no more likely to win before the, you know, balls pop out of the little thing. But that is the way people think about it. And so I think I've got off topic a little bit.
But I've been very concerned about the effects of investing too much in these sorts of forms for young people. And the end, the beginning of every year, when we have a group of new people start, I give a talk about suicide prediction. I tell them that probably one of you in this group, we run a big hospital, sees thousands and thousands of patients a year, will probably have a suicide, and you will have done a risk assessment, and the patient will have passed away.
And when that happens, it's very useful to be able to go back and say, look, unfortunately, this is very unlucky that this has happened to you. And of course, unlucky that it's happened to your patient and their family as well. But yeah, I sometimes worry that I've got inordinate concern about the welfare of my colleagues.
But we do have to march on and treat people the next day, often that afternoon even.
If I'm seeing patients in an outpatient context and I'm concerned about their suicide risk, often the expectation is to hospitalize them. What do you think about, what are your thoughts about hospitalization in terms of treatment for suicidality?
Well, some years ago, I did a meta-analysis of rates of suicide in psychiatric hospitals. And so I looked at all the available data, I think up until about 2011. And it showed a few things.
And one of them is that rates of suicide in psychiatric hospitals have actually gone up. And they're quite high. In more recent studies, 600,000 per annum, about 60 times that of global suicide rates.
So if nothing else, our ability to protect people is imperfect. There were some other things that were very interesting in that study, and there was this enormous heterogeneity, statistical heterogeneity, between the highest hospitals with the highest suicide rate and the lowest suicide rates. And there were some clues to why that might be.
The hospital with the very highest suicide rate ever recorded was a hospital in Australia, in a kind of top floor of a brutalist building. It was a ward that was run by a university and not a health department. The authors of the paper wrote saying that they thought that it was due to low nursing numbers, the absence of psychologists and social workers on the ward really did sound like a terrible ward, and it was promptly closed.
So I think wards can be better or worse, but the rate of suicide in hospitals is high, and I've got three sort of things that I think about that. So we know in the general community that stigma, trauma and dissociation from your normal social role are things that have some association with suicide. And a hospital does all of those things.
So something like, globally, something like, I think we found in another study, 17% of psychiatrically admitted people are violent to another person, and that's almost always other patients. It's not known what proportion of patients are actually traumatized by an assault. But psychiatric wards are horrible, frightening places.
And then there's the stigma of being a psychiatric patient, often being an involuntary psychiatric patient. And people come in and they look around and they see the other patients, some of whom are extremely unwell. They think this is their future.
So there's self stigma and stigma. And then being admitted to a psychiatric hospital for the first time often will result in losing your place at university or losing your job or losing some standing even with your friends and family. So I don't think we've concentrated enough on the adverse aspects of psychiatric hospitalization.
And then, of course, there's precious little data that hospitalization prevents suicide. In the Fox meta analysis, they said there was no data. Subsequently, there's one paper that suggests a very modest reduction in long-term suicide if you are admitted for suicidality that has emerged within the immediate period.
But it's a very modest reduction, and suicidality in general is not reduced by psychiatric hospitalization. This is a complex machine learning paper, and I've got some doubts as to whether it will be replicated. But that is the only data.
There's really no data about hospitalization. It's a contingent accident of history that we put people in hospital, and it's not an evidence-based treatment. Having said that, it's not all that long ago.
I remember admitting a patient involuntarily with essentially no suicide risk factors. And so I perhaps won't talk about the details of this, but she had a catastrophic event in her social environment and just absolutely couldn't think, and you couldn't take a history from her. And although she didn't have a mental illness, she was highly educated, had children, was well off.
I just couldn't release her from hospital because I didn't know what she was thinking or what would happen next. And there, I thought epistemically, she was a person who I regarded as something, as a kind of suicidal person in a way. Not that I thought that it was particularly likely that she would suicide, but there was a judgment that I made about her at that moment.
And of course, when she settled down, it was very easy to release her from hospital, and she was very grateful. And so that touches on this issue of what is the uncertainty about suicide and John Maynard Keynes said that there were two sorts of uncertainty. There was some uncertainty stemming from lack of knowledge or epistemic uncertainty, and then there was uncertainty due to chance factors.
And so if you see a 16-year-old smoker, they may or may not develop lung cancer, and further information isn't really going to tell you whether they will. But two decades later, if they present coughing up blood, then they may or may not have lung cancer, but you can work that out by a chest x-ray, CT scan, and ultimately by biopsy. And so I used those sorts of concepts to try to think about suicide prediction.
So I've done a few studies of this now. So in three studies, I've looked at the number of risk factors incorporated in the suicide, used in the high-risk model, and the strength of the suicide prediction. And collecting more information, suicide models that have more information don't work any better than suicide models that have, rely on less information.
And in my 2016 study, the odds of suicide was, which I mentioned earlier, was 4.84, which is not really any higher than any individual risk factors. So collecting more information isn't necessarily going to help you know what the patient's going to do next. And that has a lot of clinical implications.
The big one it has for me is if you've got a patient who's distressed and who you're having difficulty establishing rapport with, and you find you've got a choice between taking a complete history that will satisfy other people when they read it, or getting along with the patient. You should definitely choose getting along with the patient. You shouldn't choose interrogation, because that information isn't going to help you about the most significant outcome that might occur.
Just to highlight that point, and there is some writing on this that when you're doing this risk assessment, how I sort of communicated to folks is like, if I'm doing a risk assessment, I'm paying attention to this assessment, making sure I collect the information, but I'm not like, I think almost no clinician is there therapeutically best. It sort of takes you out of that moment and sort of some of those skills that we have and that are so therapeutic sort of evaporate as we focus on collecting these data.
I couldn't agree with you more, and I've got a specific way of thinking about this. And so I think, you know, we're risk assessing the patients and the patients are risk assessing us. They're risk assessing our likelihood of putting them in hospital.
And it's almost like a prisoner's dilemma. And this is obviously for mentally competent patients, for patients who are hallucinated or severely depressed. They may not think too much about what you're thinking.
But a lot of patients do think about what you're thinking and we think about what they're thinking. And I always try to find some kind of humorous way, if I can, of getting past that. And one way is to reassure the patient that, probably earlier than most people, that my intention is definitely not to put you in hospital, but to help you.
And that usually cuts past it. And we spend a lot of time doing suicide risk assessments, and mental health resources are very scarce. I personally think, feel much more freed up as a clinician.
I don't fill in any forms anymore. I've obviously given it up a long time ago. I don't criticize other people who do, because people do it for their own reasons.
Interestingly, I've never been a sort of issue entrepreneur saying to health service, look, you must remove all these forms. But actually, almost not doing that has been more powerful than doing it. So in my own service, people have given up the forms, although I never asked them to.
But they're very aware that I've given lots of talks on this. So in New South Wales, there are some little things about high, medium and low done by the nurses, but the doctors are not required to make a categorization about that. But we are required to take a full history from everyone and to talk with their family, to give information about suicide risk if they want it, to provide something for everybody.
And if someone wants to come into hospital, it's a completely different matter. I mean, we may not always be able to accommodate them, but that's a very different matter to compelling someone into hospital.
So what do you do when clinicians say to you, but I'm required to do a suicide risk assessment? How do you respond to that?
So I don't ever think I've seen a low risk patient in my whole career. I really don't think I have. I've seen severely mentally ill people in hospital.
So that kind of strikes out low risk. And we know that the positive predictive value of high risk is quite small. So I'm very sparing about low risk.
So if you have to do it, medium, you can't be more than one wrong. So if you have to do it, and I'm not being facetious here, actually, I would just choose medium, go on and try to be helpful to the patient, not spend too much time agonizing over it. But just try to do something that will be helpful.
And the patients all have things, there are things they're upset about, the substance use can be assisted with, their housing. Patients usually have lots of targets for things that will be helpful for them.
So when working with a suicidal patient, how much do you think is a very broad question? I know there's not a one size fits all, but how much do you think that we should be prioritizing the suicidality in our treatment?
I don't think we should be prioritizing categorizing the patient. Obviously a patient who is feeling so awful that they've made an attempt on their life or are expressing suicidal ideas, I mean, it's a very important piece of psychological, almost philosophic information about the person. And I think we should prioritize those patients.
But actually, lots of our patients don't have suicidal ideation. And in a way, I'm more concerned about those people being excluded from care than I am about... I mean, my concern about high-risk patients is really...
is that we act far too quickly to put them in hospital against their will. My concern about low-risk patients is that we don't give them anything very much. So I think we just have to be modest.
The amount of information that we get from a suicide risk assessment is modest, and therefore we should only pay a moderate amount of attention to it, if that makes sense.
So what sort of pushback have you gotten?
Well, it's an ongoing controversy academically. There are two prominent international groups who are still very much pushing their own instrument. I'm not going to mention names particularly, but they're on both sides of the Atlantic.
But the metrics of their instruments are no better than Pecorni's study in 1983. And there has not been a big breakthrough in this field. But what I have noticed is it's much easier for me to get things published these days.
So I had an extraordinary experience actually not too long ago, where I wrote a very nice paper and I sent it to a good journal. And the editor wrote back to me without sending it to review, said, Look, Matthew, very nice paper, very convincing. But, you know, as a journal, we've moved on from risk assessment.
Nobody believes this anymore. Put it somewhere else. So that was kind of gratifying.
But a number of years before then, I had difficulty getting things published because people were sceptical. So that's an academic thing. It's become easier.
I've never had a pushback from patients or their families. I mean, the stakeholders involved in this are the patients, the families, the clinicians, the third-party providers, the courts, and ultimately, in some instances, even government. And talking through all of those, patients hate the idea that you're going to assess their idea of what they might do without really discussing exactly how you've done that.
They don't like it. Patients' families are reassured by absolute estimates and by actual knowledge. They don't really want to hear that their child is high or low risk.
Colleagues, there were a lot of people very invested in it. And actually, interestingly, the metrics of violence risk assessment are very similar to the metrics of suicide risk assessment, but violence risk assessment, no one's really taken it on. So you get some pushback from forensic psychiatrists that realize that if there's real insight into the weakness of suicide risk assessment, that that will call into question one of the central tenets of their profession.
So I do have some pushback from other psychiatrists. The third-party providers are, when you explain it to them, are remarkably fine. And I mean, I've been giving evidence in coroner's cases across the country for two decades now.
And I've given probably rather too many statistic lessons on the stand. But the courts, I mean, we have a duty as mental health professionals, an evidentiary duty to explain things to the courts. And the standard of medical practice everywhere is reasonableness and scientific probity and results are central to reasonableness.
So I've seen some horrible cross-examinations of people. But ultimately, the court decisions, at least in Australia, have been all quite reasonable.
Do you have any advice? One of the anxieties I hear from clinicians when I talk to them about this is, if I don't do this and document it, that I could be held liable.
Yeah, I think I heard that at the APA. I just don't buy it. I mean, I suppose you might be if the wrong experts gave evidence.
But doing a risk assessment, categorizing the patient as high risk and then sending them home. I mean, we do that all the time. We can't do anything about it.
Then that actually causes a liability. Doctor, why did you classify this person as high risk and then send them home? If you classify them as low risk and then something happens, then there's the liability of you potentially having gotten the assessment wrong.
So if you have a suicide, there's always a potential for some sort of litigation. But I'm in my mid-60s. I've been a psychiatrist for a really long time.
I've never had to give evidence to a court about a death of a patient under my own care. And I think part of the reason for that is I've been very careful to talk with families. It's not that I haven't had any suicides.
Unfortunately, I've had a handful over the years. But communicating with the family, communicating with the patient. I mean, I would like to think that a risk assessment could be replaced by a kind of risk communication, really, in which we can give actual real knowledge.
And we need to emphasize that it is a very unpredictable event with a lot of chance in it. Knowing more about the patient isn't necessarily going to help you.
So when a clinician says to me, when they ask me that question, like, so what should I do to help protect me from being sued? What are your bullet points?
Oh, look, documentation, documentation and documentation. And I mean, when the thing doctors don't do, and presumably psychologists, is we collect a lot of information, but we often don't explain why we've done something. So I think it's quite OK to have yourself in the notes.
You know, I've considered this person could be classified under some schemes as having an elevated risk of suicide, but, you know, they're mentally competent. I'm aware that the, you know, positive predictive value of a high risk categorization over the next week is less than one in a thousand. Patient really doesn't want to be in hospital.
I can provide some further outpatient care. I've discussed this with their mum. You know, that's the standard of mental health practice is to be a reasonable person.
And so I think my one bullet point really would be to inject your decision making process into your documentation. And that's much more powerful than a form. And in fact, when you see these cases, sadly, often they've had hundreds of risk assessments and then the patients are equally deceased.
And no one's explained how they used it, how to inform their treatment. And often I don't think it did. It's a little it's a little ritual that we do that may reduce our anxiety a little bit, stop us from thinking about the uncertainties that the patient faces.
And I mean, life's pretty uncertain for all of us. And I think one of the problems with suicide risk assessment, it does give a sort of sense of otherness, whereas we all stand a definable risk of suicide. I was in a lecture the other day, and I had a period when I wasn't well many decades ago, and I did my own Columbia Suicide Severity Rating Scale.
And on the basis of suicidal thoughts, years and years ago, I came up as moderate. I was thinking, that's ridiculous. I've got absolutely zero suicidal ideas.
I've been a very happy person for 20 or 30 years. So I don't know if that's helpful to you to talk about that particularly, but I do think you'd need to be sceptical about a simple solution to a very complex and ultimately centrally human problem.
Yeah, and I've also read folks who make the argument that if you engage in doing a safety plan, doing means restriction counseling, those sorts of things, and note that that is also, you can note the things that you are doing.
Yes, of course. And I think the evidence for safety planning is not as good as some people make it out to be, but obviously it shows that you're interested in the patient and helping them think it through. And so I'm a firm advocate of safety planning.
I think safety planning has been in advance. Means restrictions are very, you know, it's a very important population measure. We're very lucky in Australia in that we've had, you know, very rigorous regulation of handguns and rules about storage and use of long guns.
And I mean, female suicide by firearm has disappeared as a category of Australian suicide. Suicide by firearm is a rare cause. It's no more common than jumping.
And it has been associated, I think, with a genuine decrease in the suicide rate. Most of our suicides are by hanging. It's very difficult to regulate hanging.
I don't think we've really thought about how that might happen outside a hospital situation. But, you know, absolutely.
Well, Matthew, I can't tell you how much I appreciate this conversation. I think that when you said that the journal said, oh, nobody really buys this risk assessment thing, I was sort of like, wow, that's, you know, perhaps the readers of that journal. I don't know if that's true.
It's not true across the board at all.
No, it isn't. It isn't. But I do, you know, if you want to read something, you can read some of the stuff that I've written, or stuff Nav Kapur or Greg Carter has written, or anything written by Franklin or Ribeiro is very well, you won't waste a minute reading them.
There's a great paper, and I think it's in German Psychiatry by Belscher from a few years ago. Don't take it from me. Just actually, you know, as a mental health responsible professional, you have a responsibility to know the literature, and the literature is pretty clear now.
Yeah, yeah, absolutely. It also struck me when you were talking about folks who have developed risk assessments. And I think a lot of very well-meaning and intelligent people have put a lot of time and energy into the development of these tools.
And I think that there's something to just, you know, I mean, there's some cost, right? And I think that it's difficult for folks to let go of that. And I understand that at the human level.
I really do. And at the same time, and there is an intuitive appeal, there's a logic to it. But then when you sort of, to me anyway, and part of it was reading your work, reading other people's work, listening to people talk.
When you peek under the hood of suicide risk assessment, and that in fact, you need barely any data at all, that this really is very much, there's a logical argument here, which is that this is such a low probability event that there's no way that I'm going to be able to collect data from a handful of questions to be able to predict this incredibly low probability event is going to happen in the next 24 hours. It's just, it's irrational.
It can't be done. It can't be done. Look, I still do my own suicide risk assessment studies.
I'm not opposed to studying risk assessment. I think that we have a lot to be thankful for, for the people who've spent a lot of time trying to work this out. We're only at the position of knowing about the general weakness of suicide prediction because of the efforts of a really large number of people in lots of different countries in the last 50 years.
And of course, I'm very grateful for those researchers.
Well, this has been great. I really appreciate it. Let's do a real quick fake sign off here.
And then I'll just chat with you for a minute on the back end if that's okay with you.
That's a wrap on today's conversation with Dr. Matthew Large, to whom I want to send my sincere appreciation. I thought that it was a fascinating conversation, and I hope that you learned as much as I did. Don't forget to subscribe to the podcast for more episodes like this and to share it with anyone who might enjoy it as well.
I really appreciate your support. Until next time.