Monday 25 March 2013

Schizophrenia and Violence: Delusional numbers?

Empiricists study the world as it is. Reformers try to create the world as it should be. These Accountants and Missionaries do not always see eye to eye.

The history of the treatment of the mentally ill makes painful reading, at least to modern sensibilities. The mentally ill have been abused, ridiculed, stripped of dignity and legal rights, and shunned by the public, who feared their madness. The mad were seen as dangerous and a nuisance. In their madness they might strike out at you, subject you to abuse, to unpleasant sights and embarrassing outbursts, and at the very least would need help and compassion while not being able to give much in return. For example, when we are depressed we have less to offer our friends, and more need of their support, encouragement and tolerance for our lack of contribution.

In the face of these historical abuses, in 1796 the Quakers’ York Retreat offered humane and moral treatment, on the Christian assumption that, whatever the sufferer’s behavioural degradation; the inner light of humanity could never be extinguished. Since that time we have become kinder, and more liberal and generous about disordered comportment. We know that not all mental illness is terminal, irreversible and dangerous. Nowadays we want to reduce stigma, that badge of shame and disgrace which some in society attach to those with characteristics, behaviours or beliefs they find disturbing. Hence a missionary movement among mental healthcare professionals to normalise mental distress, to understand the perspective of even the oddest point of view, and to break down the shame of admitting a mental affliction.

I was always somewhat surprised when, even in the privacy of my office, beset with problems and worries, my patient’s first words after bursting into tears were always “I’m sorry”. I would mumble something about not needing to apologise for having emotions, but in the purely social sense, my tearful patients were right. They had sent out a very powerful distress signal, generally without meaning to, had broken the taboo of emotional restraint and now had to justify this dramatic call for assistance. They felt they were being a drag, and didn’t want to be.

So, if every right thinking person in the psycho-business should be de-stigmatising mental illness, how do we deal with the fact that some of the mentally ill are dangerous? Ideally, we would follow the empirical tradition, and give an accurate and balanced message: “Most people with mental illness are no more dangerous than anyone else, but a minority are more prone to violence. It is difficult to predict dangerousness. Here are the facts and you can judge for yourself”.

What are the facts? In the behavioural sciences, facts are always a work in progress, but the overall picture is now reasonably clear: schizophrenics are more violent as a category, something of the order of 4 times more violent, perhaps even 5 times higher. Given that most people are not very violent at all, being four times more violent is still not very dangerous, but it is an appreciable increase. Drug takers are 10 times more violent. Perhaps shunning some people is a prudent policy. It is certainly within the discretion of any citizen to judge that a fourfold increase in even a very small risk is something worth avoiding, particularly if it can be done without too much bother.

At this point we need to cover just a few technical issues. The best technique to determine violence rates is a prospective study in which you define a population sample and study them long term. Birth cohorts are the gold standard. Follow this large sample, find out who gets schizophrenia, and then take objective measures of violence (arrests, cautions, convictions, jail sentences for example) and then you have your rates of violence, both absolute and relative to those who do not have schizophrenia. Here are some findings, which I have drawn from a paper I will discuss later on:

Hodgins (1992), in a 30-year follow-up of an unselected Swedish birth cohort, found that compared with those with no mental disorder, males with major mental disorder had a 4-fold and women a 27.5-fold increased risk of violent offences. No separate data were provided for schizophrenia. A later study using the same methodology revealed similar findings (Hodgins et al, 1996).

The first cohort study to demonstrate the quantitative risk of violent behaviour for specific psychotic categories followed an unselected birth cohort of 12 058 individuals prospectively for 26 years (Tiihonen et al, 1997). The risk of violent offences among males with schizophrenia was 7-fold higher than controls without mental disorder.

Arseneault et al (2000) studied the past-year prevalence of violence in 961 young adults who constituted 94% of a total city birth cohort. Three Axis I disorders were uniquely associated with violence after controlling for demographic risk factors and all other comorbid disorders: alcohol dependence, marijuana dependence and schizophrenic spectrum disorder.

Probably the most important study in the violence literature to date is that of Swanson et al (1990). Using a sample of 10 059 adult residents from Epidemiologic Catchment Area (ECA) study sites (Eaton & Kessler, 1985), the authors examined the relationship between violence and psychiatric disorder. Eight per cent of those with schizophrenia alone were violent, compared with 2% of those without mental illness. Comorbidity with substance abuse increased this percentage to 30%.

Two factors appear to discriminate those with schizophrenia at increased risk of committing violent acts: comorbid substance abuse and acute psychotic symptoms.

It is important to note that because there is an increase in violence risk in those without comorbidity, substance abuse merely increases the level of risk rather than causing it (Arsenault et al, 2000; Brennan et al, 2000). Hence, the risk from substance abuse appears to be additive.

In Dunedin, New Zealand, 94% of a total city birth cohort were followed up at age 21 years. Without considering comorbidity, just over 10% of past-year violence committed by these young adults was attributable to schizophrenic spectrum disorders

So, how has psychiatry responded to this awkward junction of facts and missions? The Royal College of Psychiatrists currently gives the following statement as a key fact about schizophrenia: Many people think that schizophrenia makes people violent. This is the exception, not the rule. People with schizophrenia are more likely to be victims of violence by others.

Comment: Well, this presentation is not entirely balanced. Gang members are also likely to be victims of violence by others. They are also violent themselves. Schizophrenia makes people four times more violent. Their violence is an exception, but these exceptions happen four times more frequently.
So, what other exhibits should we look at?

One of the better known papers in the field is Walsh, Buchanan, and Fahy (2002) Violence and schizophrenia: examining the evidence, from which I had drawn the above papers on violence rates.

They say: It is now accepted that people with schizophrenia are significantly more likely to be violent than other members of the general population. A less acknowledged fact is that the proportion of societal violence attributable to schizophrenia is small.

Comment: So, they are more violent, but make a small contribution to “societal” violence? How can that be so? Now you see it, now you don’t? Or is there some real reason why their increased violence does not have an impact on society?

Here is their argument, in the expanded form of a subsequent book chapter: Criminal and violent behaviour in schizophrenia by Walsh and Buchanan in (Eds) Murray, Jones, Susser, van Os and Cannon (2003) The Epidemiology of Schizophrenia, Cambridge University Press.

It is now generally accepted that people with schizophrenia, albeit by virtue of the activity of a small subgroup, are significantly more likely to be violent than members of the general
population, but the proportion of societal violence attributable to this group is small.

They continue thus:

To prevent unnecessary stigmatization of the seriously mentally ill, with all the attendant difficulties, it is the duty of researchers to present a balanced picture. By neglecting to report measures of both relative and absolute risk a skewed picture may emerge. An example of a balanced report found that men with schizophrenia were up to five times more likely to be convicted of serious violence than the general population (Wallace et al., 1998). Results also presented indicated that 99.97% of those with schizophrenia would not be convicted of serious violence in a given year and that the probability that any given patient with schizophrenia will commit homicide is tiny (approximate annual risk 1:3000 for men and 1:33000 for women).

Comment: Although intending to be balanced, this is another misleading presentation. Fahy did better in earlier papers in which he said that schizophrenics were about 4 times more violent than ordinary members of the public, but that the absolute rates were very low. Let us study some of the assertions:

“99.97% of those with schizophrenia would not be convicted of serious violence in a given year”.

Comment: Presumably this is intended to be reassuring. What does 99.97% mean? It sounds like 100%.  In his marvellous “Reckoning with Risk” (2003) Gerd Gigerenzer showed that percentages with decimal points were almost impossible for most people to understand. (I had to count the zeros, and got it wrong the first time round). The best way to make the numbers transparent is to convert them to natural frequencies, ordinary numbers without fractions or decimal points. Using this approach we can say that 9997 schizophrenics in 10,000 won’t be convicted of violence, but 3 will. So 3 in10,000 schizophrenics (10,000 minus 9997 = 3) will get convicted of serious violence. Using the usual yardstick for rare events, that means that 30 per 100,000 schizophrenics are convicted of serious violence each year.

What is the rate of convictions per thousand in the general public? It has not been given for comparative purposes, so the percentage figure is difficult to assess. However, it has been admitted that schizophrenics are 5 times more violent. Thus we can calculate that the rate for mentally well citizens would be roughly 6 per 100,000. One popular proposal to make statistics easier to understand is to place the frequency statistics in the context of villages, towns and cities. Therefore, if you spent a year in a town of 100,000 schizophrenics you might not be at too much risk yourself, but there would be about 30 violent crimes in that town. You might find this somewhat alarming. A neighbouring town of ordinary citizens would have 6 violent crimes to contend with. People will be people, you may say, but from a civilized point of view, every violent crime is unnecessary. It is likely that the reputations of these towns would differ significantly. Which town would you wish to live in, if you were free to choose?

(Paradoxically, the Press in the first town would never bother to report that the accused was a schizophrenic. It would be redundant, so in that town the media could not be accused of stigmatising mental illness.)

The other issue, never covered in these discussions, is that the overall risk rate is not a good predictor of the perceived personal risk rate. Violence is most likely to come from someone you know, with whom you are in a dispute, for whatever reason, usually because of tangled love affairs, business deals and, most of all, criminal activities. Most people know this, and know the sort of people they ought to avoid. The fear caused by schizophrenia is that it is associated with totally motiveless crimes. There is no way you can protect yourself by regulating your behaviour or your friendship patterns. You might be sitting on a bus, minding your own business, when a passing stranger stabs you to death in a matter of fact way, as happened recently to a school girl. It may seem odd to a statistician, but that sort of thing bothers us. In Kahneman and Tversky’s phrase, it has “salience”. To die at the hand of a jilted lover is bad enough, but to die because of a delusion is senseless, and gives us no chance of security. We fear more than statistics.

Now we go on to the murder statistics: The probability that a schizophrenic will commit murder each year is 1 in 3000.

Comment: Why so high? Can that be right? (I have checked the publication to make sure).That translates to a murder rate of 33 per 100,000. In Britain the annual murder rate is 1.2 to 1.4 per 100,000. Out of charity I will take the higher estimate of 1.4. (The more peaceable a society, the more it makes sense to give the figures per million, and 14 per million is the usual British estimate). This suggests that schizophrenics are 23 times more murderous. Something wrong here, I fear. Are these authors really arguing that 1:3000 is a low rate? It would be like moving from the UK to South Africa or Columbia, currently among the most murderous nations on the planet. I hope there is something wrong with their figures, because if they are accurate the implications are alarming. There must have been a miscalculation, because if the figures above are correct in showing that there are 30 violent crimes per 100,000 schizophrenics, then fewer of them will go on the whole way to murder, and the homicide figures should be much lower.

A few years ago the President of the Royal College of Psychiatrists was reported as having said that schizophrenics caused only about 10% of the violence. Given that schizophrenics are at most 1% of the population (and probably only 0.7%) this was tantamount to saying they were at least 10 times more dangerous than normal. Four or five times are the more accurate figures.

Looking at the more recent Royal College of Psychiatry report: Rethinking risk to others in mental health services Final report of a scoping group (June 2008) provides us with fresh estimates of dangerousness.

It is estimated that 5% of homicides are committed by people with a diagnosis of schizophrenia. (page 18)

Comment: A truthful statement. That is exactly what one would expect, if schizophrenics were 1% of the public and committed 5 times more violent crimes. Their greater rate of violence translates directly into a proportionately greater share of “societal” violence.

The incidence of mental illness among those remanded for acts of violence is relatively high: Taylor & Gunn (1984) found psychosis in 11% of those remanded for homicide and 9% of those remanded for other acts of violence. Similarly, violence in mental health services is not infrequent. The UK700 study (Walsh et al, 2001) found physical assaults had been committed by 20% of patients over a 2-year period and 60% had behaved violently over the same period. Taking the figure of 1 homicide per 20 000 patients with schizophrenia per annum, over the 20 years of a typical patient ‘lifetime’ (assuming active disease from the age of 20 to 40 years) the risk per patient is 1 in 1000 (Maden, 2007). The occurrence of a homicide by a patient with a mental disorder also has potentially devastating implications for the professionals involved. (page 20).

Comment: The homicide would have devastating implications for the victim and the victims’ family and friends. It would have serious implications for the professionals, whose careers can be damaged, but they would not face capital punishment. Having had friends in this category, I can testify that it is an awful process for them, but most of their colleagues know it could happen to any of them doing front line psychiatric work:  “There but for the grace of God go I”.  

The next point that this report gives the homicide risk as 1 per 20,000 patients per annum, not the 1 in 3000 each year given by Walsh and Buchanan. Taking these new figures, they suggest that the lifetime risk of homicide over 20 years is: the addition of 1 in 20,000 risks per annum for 20 years of active exposure, namely 20 in 20,000, or 1 in 1000. In comparison, the risk of being murdered by an ordinary member of the public is 1.4 per 100,000. Over 20 years that comes to 28 per 100,000 or 1 per 3,571. Thus, these figures suggest schizophrenics are 3.5 times more violent, which is within the published range. These figures look reasonable.

So, can we detect and prevent violent events?

For example, it has been calculated – using the average of all the tests assessed by Buchanan & Leese (2001) – that if 5% of the patient population were within a high-risk category, use of the tests would correctly identify 8 people out of every 100 in the group who would go on to commit acts of violence but misidentify as violent the other 92. In fact, fewer than 1% of community patients will commit serious violence over a period of a year, which means that the tests would correctly identify only 3 patients out of 100.  Homicides occur at a rate of 1 in 10 000 patients suffering from a psychosis, per annum, which makes prediction impossible (Shergill & Szmukler, 1998; Dolan & Doyle, 2000). (page 23)

Comment: Two problems here. First, this is significantly different risk estimate: in that on page 20 it was 1 homicide per 20 000 patients with schizophrenia per annum and on page 23 it is 1 in 10 000 patients suffering from a psychosis, per annum. This is confusing, but presumably it refers to the symptom of psychosis, rather than the diagnosis of schizophrenia, and other conditions can induce psychotic behaviour, so we should probably put this estimate aside for the moment, though it is twice as high as the former figure.

Second, low base rates do not necessarily make prediction impossible, but simply very difficult in the absence of valid indicators. Phenylketonuria occurs in 1 in 10,000 to 15,000 newborns (higher in the US) and most cases are detected by screening.  Tandem mass spectrometry is claimed to be 100% sensitive and 98% specific, which means that virtually all cases are picked up and the test rarely misclassifies other conditions as being phenylketonuria. Treatment is started promptly, and as a result, the severe signs and symptoms of the classic condition are rarely seen.

Have pity on Psychiatry. As the quotation above reveals, current tests do have the sensitivity and specificity requires to usefully predict violence, let alone murder.  The best test that forensic professionals have is a detailed interview leading to a risk assessment, though there is still debate about precisely what this risk assessment should contain. There must also be checking of records, and good links with social services and the Police. Often, this still defeats the current organisation of services. Resource constraints make the problem more difficulty. Poorly trained or poorly motivated ward staff sometimes do not follow risk protocols, even when they are written in to the patient notes.

Note that this unsatisfactory state of affairs does not have to continue for ever. One good indicator seems to be patients trying to re-admit themselves to hospital. Failure to take medication, and over-indulgence in alcohol and drugs is another indicator. It is very hard to detect the signal from among the noise, but we have to keep trying. Perhaps if we could monitor electronically, as if the person had diabetes, we could eventually predict with less error. What would we monitor? Thoughts? Arguments? Drugs in the bloodstream?  

So, how can one argue that schizophrenia makes people more violent, but that this does not cause much “societal violence”? This boils down to a particular statistical technique, the calculation of population-attributable risk per cent (PAR%): the percentage of violence in the population that can be ascribed to schizophrenia and thus could be eliminated if schizophrenia was eliminated from the population. Stand back from any calculations, and you already know the answer. If 1% of the population has schizophrenia, and they account for 5% of the violence, then curing them of the condition (or just the violent aspects) would reduce societal violence by 5%. It is a matter of perspective whether you think that this accounts for relatively little societal violence. To me, this line of argument seems like trying to wish away a finding by changing the currency of account. I think it is better for researchers to publish their results on the risks of violence in schizophrenia, which is higher in relative terms but low in absolute terms, without excursions into concepts of “societal violence” as a false comparator. It is a bit like saying that someone who drives a gas-guzzling Hummer has little societal impact because they contribute only a very tiny fraction of the world’s pollution. Don’t move from one metric to the other for rhetorical purposes.

So, to summarise what psychiatrists are telling us about schizophrenia and the risks of violence: there is general agreement that schizophrenic men are four to five times more violent than the rest of the population, but the absolute rates of all violence are mercifully low. The public should make up their own minds how they wish to respond.


  1. "To prevent unnecessary stigmatization of the seriously mentally ill, with all the attendant difficulties, it is the duty of researchers to present a balanced picture."

    That was either a silly truism or a call to abandon science and indulge in propaganda. I'm guessing it was the second.

  2. It has drifted towards the second, worse luck.

  3. I have twice had to deal with a young man becoming schizophrenic - that's "deal with" as a university tutor, not as a medical man. Although neither seemed (at least at the stage that involved me) to carry any threat of violence, it was nonetheless a most unsettling and painful business to witness. "There but for the grace of God ..." you think, and "Dear God, there's nothing useful I can do for the poor sod".

    The bewildered, horrified reaction of the parents to the sudden decline of their golden boys was awful to behold.

    A question in two parts, if I may: I'm struck by your five-fold conclusion for males. I take it that the ratio of the risk of violence from the two sexes is, for sane people, greater than five-fold? Is the ratio also greater than five-fold in any racial comparisons - I have in mind particularly blacks and whites in the USA? In other words, I'm wondering whether there are any natural comparators for the five-fold result.

  4. A second question: "violence" might be rather a widely-defined category. Is the ratio still five-fold if the violence is homicidal? If violence to self is excluded?

  5. Another "natural comparator" thought: men 15-30 versus men 50-65?

  6. Men more violent than women about 11 times (from memory).
    Black men more violent than white men about 7 times (US data).
    Younger than 30 year men 4 times more violent than the older than 45 year old men.
    Yes, all this depends on definition of violence, but these are rates where the definition is the same within the study. Self violence not included here.

  7. "all this depends on definition of violence, but these are rates where the definition is the same within the study": fair enough, but are the acts of violence that are recorded comparable? For example suppose someone said that he'd record an act of violence only if it drew blood. That means that a couple of drunken punches may join the same category as a decapitation.

  8. No, the usual metric is convictions for crimes of violence, compared against convictions for the same crimes on the part of the general public.

  9. Then that's my point made for me: a mere conviction for a crime of violence makes no distinction between a drunken brawl and a decapitation by a madman - it's too crude a tool for the use it's being put to. In other words performing an act of violence may be five-fold higher, but it tells us nothing about how violent the acts are.