- Joined
- Mar 7, 2016
- Messages
- 4
- Reaction score
- 3
- Points
- 4,531
- Resident [Any Field]
Advertisement - Members don't see this ad
For context, I'm a second year psych resident/registrar outside of the US. And my apologies, this turned in to a bit of a rant. I'm reading around risk at the moment but have struggled to find much in the way of supporting evidence.
Risk assessment, particularly regarding suicide risk, takes up a large fraction of work time: mostly after hours, but also tacked on to a lot of routine visits. The problem I have with it is that:
a) There's very little/no predictive value in these assessments (as far as I know)
b) Acute treatment (i.e admission) isn't proven to do anything long term (as far as I know)
c) The focus seems to be on minimizing legal/professional risk by asking all these questions (and documenting them), rather than altering patient outcomes.
I think it has parallels to trying to predict road fatalities. They're both rare events, with quite complex aetiology and (arguably) a lot of chance involved. Certain populations have higher risk that others. But how relevant is that on an individual level? If a young male on a learners permit with a history of a few crashes is pulled over by police, they don't have their licence suspended for ticking a lot of the "high risk" boxes. That would be absurd. Their absolute risk is still very similar to anyone else on the road.
I know the comparison is flawed, particularly as one of these is a deliberate act and the other isn't. My point is that we seem to go along with these risk assessments, possibly because they *seem* intuitive, but actually have very little statistical or even logical basis. This also goes for less obviously poor predictors such as plan or intent. If we were able to just add up all these risk factors to predict suicide, then there should be some evidence that a structured interview/inventory/whatever covering them is able to predict suicide with a useful amount of sensitivity/specificity, but there doesn't appear to be any studies that support this. E.g the Beck's hopelessness scale is given as an example, but the very low sensitivity combined with the rarity of suicide makes it clinically useless as far as I can see?
Another issue is that of "inevitability" or true individual lifetime risk. If the only prevention for MI was admission to hospital, then admitting someone with atypical chest pain because they're a middle age diabetic overweight smoker might seem like a good idea. But if that prevention only continues as long as they are admitted, you're ultimately fairly unlikely to greatly extend their life by a large amount (unless they want to live in hospital). It may be that the actual lifetime risk of suicide for some individuals is just absurdly high, but you have no way of knowing.
My question then is: does anyone have any high quality research that argues against the above?
I.e that we can predict suicide in any meaningful way, or that inpatient admission does anything at all.
P.S Don't get me wrong, there are many situations where it's clear that the individual is at an acute high risk (perhaps equivalent to driving drunk in the above car metaphor). But I wonder if we should be more honest about our ability to predict and treat acute suicidal risk... does there have to be an investigation every single time? When someone dies from a medical illness it's usually some random intern that will do the death cert. But when someone dies from suicide there's a massive circus involved with internal reviews, coroner investigations, the staff feel awful etc.
Risk assessment, particularly regarding suicide risk, takes up a large fraction of work time: mostly after hours, but also tacked on to a lot of routine visits. The problem I have with it is that:
a) There's very little/no predictive value in these assessments (as far as I know)
b) Acute treatment (i.e admission) isn't proven to do anything long term (as far as I know)
c) The focus seems to be on minimizing legal/professional risk by asking all these questions (and documenting them), rather than altering patient outcomes.
I think it has parallels to trying to predict road fatalities. They're both rare events, with quite complex aetiology and (arguably) a lot of chance involved. Certain populations have higher risk that others. But how relevant is that on an individual level? If a young male on a learners permit with a history of a few crashes is pulled over by police, they don't have their licence suspended for ticking a lot of the "high risk" boxes. That would be absurd. Their absolute risk is still very similar to anyone else on the road.
I know the comparison is flawed, particularly as one of these is a deliberate act and the other isn't. My point is that we seem to go along with these risk assessments, possibly because they *seem* intuitive, but actually have very little statistical or even logical basis. This also goes for less obviously poor predictors such as plan or intent. If we were able to just add up all these risk factors to predict suicide, then there should be some evidence that a structured interview/inventory/whatever covering them is able to predict suicide with a useful amount of sensitivity/specificity, but there doesn't appear to be any studies that support this. E.g the Beck's hopelessness scale is given as an example, but the very low sensitivity combined with the rarity of suicide makes it clinically useless as far as I can see?
Another issue is that of "inevitability" or true individual lifetime risk. If the only prevention for MI was admission to hospital, then admitting someone with atypical chest pain because they're a middle age diabetic overweight smoker might seem like a good idea. But if that prevention only continues as long as they are admitted, you're ultimately fairly unlikely to greatly extend their life by a large amount (unless they want to live in hospital). It may be that the actual lifetime risk of suicide for some individuals is just absurdly high, but you have no way of knowing.
My question then is: does anyone have any high quality research that argues against the above?
I.e that we can predict suicide in any meaningful way, or that inpatient admission does anything at all.
P.S Don't get me wrong, there are many situations where it's clear that the individual is at an acute high risk (perhaps equivalent to driving drunk in the above car metaphor). But I wonder if we should be more honest about our ability to predict and treat acute suicidal risk... does there have to be an investigation every single time? When someone dies from a medical illness it's usually some random intern that will do the death cert. But when someone dies from suicide there's a massive circus involved with internal reviews, coroner investigations, the staff feel awful etc.
