Identifying risk factors in first episode psychosis: results from two new meta-analyses


A first episode of psychosis is a higher risk time for acts of self-harm and of violence, and a chance to engage people, modify risk factors and change outcomes.

Two related meta-analyses by Large, Neilssen and Challis are presented in a letter in the Australian and New Zealand Journal of Psychiatry.  They summarise two studies with overlapping author groups which both look at the risks facing patients suffering from their first episode of psychosis. Specifically they investigate the risk factors for self-harm and for violence to others, and then compare and contrast these to make suggestions on identification and management of those at risk.

Risk of self-harm

The first meta-analysis by Challis, Neillssen, Harris and Large reports on studies looking at factors influencing deliberate self-harm (DSH) before and after engagement in treatment for those with first episode psychosis (FEP). They highlight that previous research shows 5% of those with a diagnosis of schizophrenia die by suicide, and as many as 50% have at least one episode of self-harming. The aims of Challis’ paper are to:

Previous research shows 5% of people with a diagnosis of schizophrenia die by suicide

Previous research shows 5% of people with a diagnosis of schizophrenia die by suicide

  1. Estimate how many people with FEP presenting for treatment have a history of DSH
  2. Look at how many self-harmed in the period after starting treatment
  3. Identify demographic and clinical factors associated with DSH in FEP
  4. Determine any differences in factors before treatment, compared with after starting treatment.

To do this they used a standard method for meta-analysis of studies (Preferred reporting items for systematic reviews and meta-analyses – PRISMA); they searched 3 large databases (Medline, Embase and PsychInfo) for papers up to March 2012 that looked at FEP, schizophrenia or early schizophrenia, and suicide attempts or DSH. Studies looking at frequency of self harm at or before initial contact with services were deemed pre-treatment, while those describing groups of patients already in contact with services were post-treatment. Studies looking only at completed suicide were not included. This strategy yielded 276 studies, of which 23 were suitable for inclusion in the meta-analysis. 18 contained data on pre-treatment FEP and 13 had data on post-treatment cases. Hand searching of references, and a trawl through the top 1,000 hits on Google Scholar did not bring up any new articles.

They analysed the data using a statistical software package to convert the odds ratios into effect sizes of the different factors associated with self-harm. They also made calculations to ensure that the data had a high probability of being a true result (p<0.01), that heterogeneity of the sample was taken into account, and that publication bias was not skewing the results (using a failsafe N calculation).

The main conclusions were that:

18.4% of people with first episode psychosis had self-harmed at some point prior to treatment

18.4% of people with first episode psychosis had self-harmed at some point prior to treatment

  • 18.4% of FEP have self-harmed at any time prior to treatment (i.e. some may have self harmed before psychotic symptoms began) (95%CI = 14.4 to 23.3)
  • 9.8% self-harmed during a period of untreated psychosis (95%CI = 6.7 to 14.2)
  • 11.4% went on to self-harm during treatment (95%CI = 8.3 to 15.5)

There were 22 overlapping factors (found in >5 studies each) associated with DSH. Factors increasing the risk of DSH included clinical factors like depressed mood, suicidal ideation and increased insight into the psychosis. They also included demographic factors such as younger age, alcohol or drug use and longer duration of untreated psychosis. Previous DSH predicted higher likelihood of future DSH post-treatment. Other factors like gender, marital status employment, level of symptoms, compulsory treatment and level of functioning were not associated with DSH.

Risk of violence

Most people with a mental illness do not become violent, but that there is an association between psychosis and violent offending

Most people with a mental illness do not become violent, but that there is an association between psychosis and violent offending

In the second paper (by just Large and Neilssen this time) the same type of methodology had been used to look at the factors associated with violence in patients with FEP.

They are quick to point out that most people with a mental illness do not become violent, but that there is an association between psychosis and violent offending.

Again the authors searched for articles in the way outlined above, this time including studies if they were on groups of people with FEP and compared those who had been violent with control cases that had not been violent. Out of 900 initial references they found only 9 studies suitable for inclusion in this meta-analysis, but again applied the statistical measures to ensure their calculations were giving robust and accurate results.

The main conclusions made this time were:

    • 35.4% of people with FEP committed some act of violence, but only 16.6% caused harm to another
    • 0.6% caused serious harm leading to hospitalisation or permanent injury, and 1 in 630 cases committed a homicide during their first episode of psychosis
Only 0.6% of people with first episode pyschosis caused serious harm leading to hospitalisation or permanent injury

Only 0.6% of people with first episode psychosis caused serious harm leading to hospitalisation or permanent injury

  • Factors associated with violence overlapped with those known for the general population: young age, lack of education, prior offending and substance misuse
  • Increased mood, severity of psychosis and hostile affect were associated with an act of any level of violence
  • Greater severity of violence was associated with greater length of untreated psychosis
  • Compulsory treatment was strongly associated with violence during the treatment phase (but the authors note that most legislation requires the person to be a risk to themselves or others before compulsion can be used)


Limitations for both of these two meta-analyses are the small number of papers identified, and the inability to use co-variance to analyse whether different grouping of factors account for a greater proportion of the risk than others. The authors were also unable to examine some of the factors already known to be associated with violence or DSH (for example cognitive factors, a history of trauma or childhood abuse, presence of personality disorder) as this was not included in the studies available for meta-analysis.

Comparison and conclusions

In the letter bringing these two papers together, the authors comment that some of the factors in FEP were risks for one type of harm, but protective against other forms of harm. Some factors increased the risk of both, and some were not associated with either. This is summarised in the table below.

Factor Risk of self-harm Risk of violence
Depressed mood Increased Decreased
Insight Increased Decreased
Compulsory treatment Decreased Increased
Male gender Decreased Increased
Young age Increased Increased
Substance use Increased Increased
Failure to complete school Increased Increased
Duration of untreated illness Increased Increased
Alcohol use Increased Increased
Self-harm Increased Increased
Schizophrenia Decreased No change
Positive symptoms No change No change

The implications then are that the picture is complicated, as by reducing risk factors for DSH you may be increasing the risk of violence for that person.

The authors conclude that the most effective targets for intervention would be the shared factors that increase risk for both outcomes, in particular substance misuse and minimising the duration of untreated psychosis through early detection and engagement in treatments. By doing so one could genuinely hope to change the course of people’s lives by preventing them from committing an act of violence that would otherwise lead to incarceration or forensic unit admission.

One could also make an impact by preventing harm to self and thereby reducing the stigma of carrying scars, need for surgical interventions, and the elevated risk of suicide attached to these actions.

Using the data available to determine some of these target factors sets the scene for further work to improve awareness of psychosis in the general population, so that there is a greater chance that those in the very early stages can be identified and helped to recover.


Challis S, Neilssen O, Harris A and Large M, (2013) Systematic meta-analysis of the risk factors for deliberate self-harm before and after treatment for first episode psychosis. Acta Psychiatrica Scandinavica 127:442-454. [Abstract]

Large M, Dall B and Nielssen O, (2013) Risk assessment for violence and self-harm in first episode psychosis and the need for early psychosis intervention services. Australian and New Zealand Journal of Psychiatry 00(0) DOI: 10.1177/0004867413489176. [PubMed record – no abstract]

Large M and Neilssen O, (2011) Violence in first-episode psychosis: A systematic review and meta-analysis. Schizophrenia Research 125: 209-220. [PubMed abstract]

Share on Facebook Tweet this on Twitter Share on LinkedIn Share on Google+