Risk and Protective Factors for Suicide and Suicidal Behaviour: A Literature Review

Listen

CHAPTER THREE: RESULTS

Introduction

Chapter structure and notes on content

3.1 This chapter is divided into two main sections: the first section contains the evidence from systematic reviews of risk factors and the second, contains both review level and primary study evidence related to protective factors against suicidal behaviour. In total, there were 23 systematic reviews of risk factors, one systematic review of protective factors, and 44 primary studies relating to protective factors.

3.2 Papers appear more than once within the results section where they include data that are relevant to different sub-sections. Evidence tables encompassing all of these studies and reviews are listed in Annex 11. A full list of the included risk and protective factor references is provided in Annex 8. In order to avoid repetition in our summaries of the results, where reviews include data relevant to a number of sections, this is reported briefly under the appropriate headings rather than repeating the full results more than once. In the event that the reader requires more detail, they are directed to the evidence tables in the appendices, which contain results from each paper rather than results divided by risk or protective factor categories.

3.3 Attempts have been made, as far as is possible with studies that are often concerned with multiple protective factors and complex suicide risk groups, to set out the evidence firstly by individual, then by psychosocial and societal-level protective factors, followed by population sub-groups. Sub-groups may represent those known to be at elevated risk of suicide and suicidal behaviour, such as persons with schizophrenia who are at risk of non-fatal self-harm. Where the evidence allows, the sub-groups are further sub-divided according to life course stages.

3.4 Where the term deliberate self-harm ( DSH) appears, this refers to the author's own use of the term. We only report results of self-harm where authors have indicated that this is linked to suicidal intent, in accordance with our definition of suicidal behaviour (see chapter one).

3.5 To make the main text of this report more accessible to the lay reader, the full quantitative data relating to findings of included studies have been reported only in the evidence tables ( Annex 11). Not all of the included reviews and primary studies reported odds ratios (or relative risks) and confidence intervals in their results, we note where this is the case. Relative risk ( RR) is the ratio of the risk of an event among an exposed population to the risk among the unexposed. An Odds Ratio ( OR) is the ratio of the odds of an event in an exposed group to the odds of the same event in a non-exposed group. Therefore RRs can be interpreted as increased or decreased likelihood of an event between exposed and unexposed populations. In other words, when RR of suicide among the employed compared to the unemployed is 3.0, this can be interpreted as "suicide is 3 times more likely to occur in the unemployed population than in the employed population." However, ORs are difficult to comprehend directly and should not be interpreted (as is commonly done) as being equivalent to the RR. Thus, an OR of 2.5 of smoking among self-harmers compared to non-self-harmers cannot be interpreted as "self-harmers are 2.5 times more likely to be smokers than non-self-harmers." It is more appropriate to state that "self-harmers are more likely to smoke than non-self-harmers ( OR: 2.5; 95% confidence interval: X.X - Y.Y).

3.6 Rather than report the statistical significance values for each finding reported within the main text of the review, when a result is given as significant the p-value will be p=0.05 or less and confidence intervals will lie within 95% certainty.

3.7 Where possible, an indication of the strength of the evidence for a particular risk or protective factor is given when included in the reported reviews or studies. However, it was not possible to use set criteria (e.g. criteria such as number of studies) to provide a universal quantitative indicator of the strength of evidence across all sub-sections of this review. This was mainly due to the lack of consistency in type of studies included and the authors' acknowledgement of the imbalance between the amount of available evidence on risk factors and protective factors.

3.8 If papers referenced in the results (chapter 3) are not included in the systematic review proper (e.g. where it is an earlier paper updated and superseded by a paper included in the review), then the reference is given as a footnote.

Risk factors: the evidence

Individual-level factors: mental ill-health

3.9 Under the category of 'mental health' (with mental illness as a risk factor) the reviews below are further divided into three papers addressing mental ill-health in general, two papers on affective disorders, one review exploring borderline personality disorder and one review of schizophrenia as a risk factor for suicidal behaviour. There are a further two reviews that explore multiple risk factors within two high risk groups: schizophrenia and bipolar disorder.

General mental ill-health

3.10 Both of the reviews in this section explored completed suicides across a range of psychiatric diagnoses. Arsenault-Lapierre et al (2004) conducted a meta-analysis of overall and specific psychiatric diagnoses found in studies of completed suicides in order to explore potential gender and geographical differences in the distribution of psychiatric disorders among suicide completers. This population based review included 27 studies (14 from within Europe), with a total of 3275 completed suicides. Out of the total number of suicides, 87.3% had been diagnosed with a mental disorder prior to their death. Major gender differences were found. Diagnoses of substance-related problems, personality disorders and childhood disorders were more common among male suicides, whereas affective disorders, including depressive disorders, were less common among males. However, the gender differences were not completely clear-cut, as age was a mediating factor. Where there were significant differences, the female sample was older than the male one. Geographical differences were also likely to be present in the relative proportion of psychiatric diagnoses among suicides, although again this included a range of age groups. Psychiatric diagnoses were present in the majority of cases in all regions. This ranged from 89.7 % of the American suicides with at least one diagnosis, compared with 88.8 % of the European suicides, 83.0 % of the Asian suicides and 78.9 % of the Australian suicides. The authors concluded that, although a diagnosis of mental illness is clearly linked to suicide risk, gender and geographical differences are also apparent. However, these conclusions should be interpreted with caution, as the authors acknowledged that the results may have been confounded by the age variations across studies.

3.11 Fleischmann et al (2005) explored the role of specific mental disorders and their comparative importance for understanding suicide and its prevention in young people. Using a narrative approach the review synthesised evidence from 13 studies from seven countries (including one UK-based study). They included 894 suicide completers with an age range of 10-30 years. The majority of cases (88.6%) had a diagnosis of at least one mental disorder. Mood disorders were most frequent (42.1%), followed by substance-related disorders (40.8%) and disruptive behaviour disorders (20.8%). Of the 236 diagnoses that included information regarding the subcategories, 56.4% were major depressive disorder, 22.0% were mood disorders not otherwise specified, and 16.5% were dysthymia. Substance-related disorders were divided between drug use disorders/drug abuse and alcohol dependence/alcohol abuse. Of the 339 diagnoses, 53.7% were alcohol-related, and 46.3% were related to drug use disorders/drug abuse. Disruptive behaviour disorders' included conduct disorder, attention-deficit disorder, oppositional disorder and identity disorder. Information on the subcategories was available for 156 diagnoses, of which 66.0% were attributed to conduct disorder, 16.0% fell under attention deficit disorder, and 13.5% were disruptive behaviour disorders (without further specification). Only four of the studies used a case control design and provided odds ratios. The authors conclude that, in developing strategies for the prevention of suicide in young people, there is a need to broaden the notion of risk to include a wide range of psychiatric diagnoses that extends beyond the focus on depression.

3.12 Neeleman (2001) provides further evidence for mental ill-health as a risk factor for suicidal behaviour from a review of multiple risk factors (reported further below). Those with adult personality disorder, a psychiatric history, schizophrenia, bipolar disorder, depression or neurosis were 6.1-19.7 times more likely to die by suicide than those who were not mentally ill, with depression and bipolar disorder located at the higher level of risk.

Affective disorders

3.13 Bostwick and Pankratz (2000) conducted a review with the aim of both generating an alternative estimate of suicide risk than that reported by Guze and Robins (1970) 1, and to question the generalisability of their earlier estimate. This review included mainly observational studies and results were combined narratively, also drawing on a random effects model. Forty-one papers were included in the final meta-analysis with a total of 31,159 participants. They found that there was a hierarchy in suicide risk among patients with affective disorders. The estimate of the lifetime prevalence of suicide among affective-disordered patients who had a history of being hospitalised for suicidal behaviour was 8.6%, compared to 4.0% among hospitalised affective disorder patients who had no history of suicidality. The lifetime suicide prevalence for mixed inpatient/outpatient populations was 2.2%, and for those who did not have an affective illness, it was less than 0.5%. The case fatality prevalence of affective disorder inpatients significantly differed from that of both suicidal inpatients and affective disorder outpatients. The case fatality prevalence of the affective disorder outpatients and the suicidal inpatients also displayed a significant difference. Although patients with affective disorders had an elevated risk of suicide compared to the general population, no risk factor, including classification of diagnostic subtype, reliably predicted suicide. They claim that their findings demonstrated that there is a 'hierarchy of risk' closely related to the contexts and intensity of treatment and that the clinical decision to hospitalise offers a useful indicator of increased suicide risk.

3.14 One review explored the prevalence of completed suicide in depressed patients (Wulsin, Vaillant and Wells, 1999). This included 57 studies (132,128 participants) from a wide range of countries including Scotland and other parts of the UK. Twenty-nine (51%) of the studies showed a positive relationship between depression and increased mortality, 13 (23%) showed a negative relationship, and 15 (26%) had mixed results. The authors found that there were too few well-controlled, comparable studies to develop a reliable estimate of the mortality risk associated with depression. Only six studies controlled for more than one of the four major mediating factors: severity of physical illness, smoking, alcohol or suicide. Results were grouped as follows: group one included psychiatric patients identified via a psychiatric assessment or diagnosis; group two included a community dwelling sample identified by self report measures; and group three consisted of those medical or community samples assessed by structured interview, comparing depressed to non-depressed and controlling for physical illness. Suicide accounted for less than 20% of the deaths in the samples of patients identified by psychiatric diagnosis or assessment, and less than 1% in the medical and community samples. The authors stated that the lack of homogeneity and the variable methods used between studies made it impossible to take their meta-analysis further. Although the authors acknowledged that the studies included in this review were poorly controlled, they conclude that depression substantially increases the risk of death, especially death by unnatural causes and cardiovascular disease. However, clearly this paper raises the question as to how the risk of suicidal behaviour compares with the risk of death from other causes in persons suffering from depression.

Borderline personality disorder

3.15 One review explored the suicide rate related to borderline personality disorder (Pompili, Girardi and Ruberto, 2005), drawing on the evidence from eight studies with a total of 1179 participants who had a diagnosis of borderline personality disorder. These studies were based in USA, Canada, Norway and Switzerland. The results across these studies were aggregated to calculate the mean N of completed suicides per year for 100,000 persons with borderline personality disorder. A meta-analysis of the included studies showed variable results between different cohorts, demonstrating nevertheless a higher rate of death by suicide in this patient group in comparison to the general population. The authors also found that higher completed suicide rates were associated with a shorter follow-up time, which they suggested could indicate that suicide risk was greater around the time of first diagnosis. They also state that suicide rates are highly heterogeneous both across and within cohorts of different studies and that there were several limitations to the study: they used a sample of data restricted to studies published in medical journals, the studies varied in terms of diagnostic criteria, and data on co-morbidities were unavailable.

Schizophrenia

3.16 One paper exploring schizophrenia as a risk factor for suicidal behaviour (Palmer, Pankratz and Bostwick, 2005) focused on developing a methodology for estimating lifetime suicide prevalence from published cohorts. The 61 included studies, with a total of 48176 participants, drawn from Europe (including UK based studies), North America and Asia. The studies were separated into those with participants recruited at a range of points in their illness (32 studies) and studies of those recruited at either initial diagnosis of schizophrenia or their first admission to hospital (29 studies). Regression models explored the intersection of proportionate mortality, which calculated suicide completion as a percentage of all deaths in this cohort, and the percentage of the total number of those who died by suicide. The authors estimated the lifetime suicide prevalence for the group who had been followed up from first diagnosis or first hospital admission was 5.6%. The group who had been recruited at different points in their illness (which included participants with a wide range of time since first diagnosis of schizophrenia) had an estimated lifetime suicide prevalence rate of 1.8%. The authors derived an overall estimate of the lifetime risk of dying by suicide in those diagnosed with schizophrenia as 4.9%. They emphasised that the risk was highest around the time of initial diagnosis or the first onset of symptoms. However, as Hawton et al. (2005a) argued (reported more fully below), studies should take account of the multi-faceted nature of risk, since other factors in combination with a diagnosis of schizophrenia, rather than simply schizophrenia alone, increased a person's risk of suicide.

Multi-faceted risk factors within mental ill-health risk groups

3.17 There were two papers that explored risk factors more fully within particular patient groups known to be at higher risk of suicide. The first paper explored risk factors related to schizophrenia (Hawton et al, 2005a) and the other focused on persons with bipolar disorder (Hawton et al, 2005b). These papers demonstrated the multi-faceted nature of risk. While mental illnesses such as schizophrenia and bipolar disorder are known risk factors for suicidal behaviour, persons with these diagnoses were also subject to further risk factors that in some cases were similar to those of the general population.

3.18 The first of these reviews (Hawton et al, 2005a) explored risk factors for suicidal behaviour in the schizophrenia patient group who were over 16 years of age. Twenty-nine cohort and case-control studies (37 papers) from a wide range of countries (including the UK) were synthesised in a meta-analysis. The core results focus only on those areas with stronger evidence as not all of the same outcomes were measured across all papers, making it difficult to conduct the synthesis. The authors identified a history of suicide attempts as a factor that contributed to elevated risk of suicide. A history of previous depressive disorders was also implicated in increased risk, as was drug misuse and agitation or motor restlessness. Anxiety or fear around deteriorating mental health was another risk factor, as was a lack of compliance in following treatment plans. Finally, a recent bereavement was also associated with elevated risk of suicidal behaviour. Conversely, a lesser or reduced risk was linked to people experiencing hallucinations. The authors acknowledge the limitations of the review, since odds ratios are sometimes drawn from only two studies, and some of the included studies were very small (e.g. only 11 suicides and 11 controls).

3.19 The second review in this section explored the main risk factors for suicide and nonfatal suicidal behaviour in patients with bipolar disorder (Hawton et al, 2005b). Outcome measures included completed and attempted suicide as well as non-fatal self-harm. The 37 studies (55 papers) varied in design (cohort, case-control and cross-sectional studies), with the majority from European countries (including UK) and North America. Reported study participants ranged from only three completed suicides matched with 136 controls to a very large study of 672 suicides and 14714 controls. Where more than one study reported the same variables, they were pooled and reported as odds ratios in accompanying tables. The main risk factors for suicide were reported to be a previous suicide attempt and hopelessness. The main risk factors for nonfatal suicidal behaviour included a family history of suicide, early onset of bipolar disorder, the extent of depressive symptoms (in one large study 137/219 cases compared with 230/429 controls), increasing severity of affective episodes, the presence of mixed affective states, rapid cycling, co-morbid Axis I disorders, and abuse of alcohol (2 studies) or drugs (2 studies). Suicide risk was higher in men than in women but there was no association with ethnicity, marital status or employment status. There was no gender difference in attempted suicide. The authors concluded that the prevention of suicidal behaviour in patients with bipolar disorder should pay attention to these risk factors in assessment and treatment aimed at reducing suicide risk. These results and conclusions should be accepted with caution, as the authors acknowledge the limitations of the evidence. Although one of the main risk factors was identified as hopelessness, this finding was drawn from only two studies, both with small numbers.

3.20 Across all age groups, genders and in a wide range of geographical locations, several diagnoses of mental illness have been established as risk factors for completed suicide, including: affective disorders (including depression, bipolar disorder etc), schizophrenia, personality disorders and childhood disorders. A history of psychiatric treatment in general is also a risk factor. In schizophrenia and borderline personality disorder suicide risk appears to be elevated around the time of first diagnosis. However, there is also evidence (for bipolar disorder and schizophrenia) that, while these diagnoses carry elevated risk, this is further exacerbated by other risk factors, such as a history of suicide attempts, other psychiatric diagnoses, drug or alcohol misuse, anxiety, recent bereavement, severity of symptoms and hopelessness.

Individual-level factors: self-harm

3.21 Neeleman (2001) conducted a review exploring the standardised mortality ratios for death by suicide, death by natural causes and accidental death in a total of 146 studies including a total of 1,179,126 participants. Studies were drawn from a wide range of countries including Western and Southern Europe. Results for deliberate self-harm (suicidal intent is not measured by the authors) from 14 cohorts and 21,385 subjects show that persons who self-harm are 24.7 times more likely to die by suicide compared with those who do not self-harm.

3.22 Those who deliberately self-harm have a much greater risk of dying by suicide compared with those who do not engage in this behaviour.

Individual-level factors: substance misuse

Alcohol misuse

3.23 Cherpitel, Borges and Wilcox (2004) investigated the link between suicidal behaviour (suicide and attempted suicide) and acute alcohol use in adults over 19 years of age. The 53 studies with over 10,000 participants were drawn from a wide range of countries, including Scotland and elsewhere in the UK. The majority of studies were purely descriptive, presenting the prevalence of suicide completers or attempters who tested positive for alcohol use. The results differed widely owing to the variation in approach to definitions, study designs, method of measuring blood alcohol levels, heterogeneity in terms of the focus of the studies (not necessarily just suicide and alcohol use) and so on. The percentage of completed suicides who were tested positive for blood alcohol ranged from 10-69%. For suicide attempters the figures for positive alcohol tests were similar (10-73%). The authors noted that the limitations of many of the studies included the lack of control groups, bias in selection and ascertainment, and small sample sizes. While there were over 10,000 participants across all of the studies combined, individual studies ranged in numbers from only 16 to over 6,000. There was also the problem that in many studies as many as 60% of deaths by suicide were not tested for the presence of alcohol. (In some cases this was not possible owing to the condition of the body or the length of time in hospital prior to death.) For a wide range of reasons, not all of the suicide attempters were screened for alcohol use. The results of a case-crossover pilot study indicated substantially higher risk of suicide during or shortly after use of alcohol compared with alcohol-free periods.

3.24 These findings are further supported by a review of multiple risk factors (Neeleman, 2001). Standardised mortality ratios for death by suicide in those misusing alcohol was 8.5 times higher than the comparison group.

General substance misuse

3.25 Wilcox, Conner and Caine (2004) explored substance misuse more generally (alcohol and illicit drug use disorders) and its impact on risk of completed suicide. This review aimed to update and expand on Harris and Barraclough's (1997) earlier review. 2 In total, there were 42 studies from a range of countries including Scotland and elsewhere in the UK. The authors estimated that heavy drinkers were more than three times more likely to die by suicide than the general population (standardised mortality ratio [ SMR] 3 = 351]) and those diagnosed with alcohol misuse disorder were nearly ten times more likely to die by suicide than the general population (standardised mortality ratio = 979). Standardised mortality ratios were also extremely high for those diagnosed with opioid use disorder ( SMR=1351), intravenous drug use ( SMR=1373) and mixed drug use ( SMR=1685). As found in the Harris and Barraclough review, there was a high degree of heterogeneity, with similar limitations, although they concluded that persons who have opioid use disorders and mixed intravenous drug use have a higher risk for suicide, and the risk is greater than that for alcohol misuse. They also confirmed that the association between alcohol use disorders and suicide is stronger among women than among men, as concluded in the review by Harris and Barraclough.

3.26 These findings are further supported by Neeleman's (2001) review of multiple risk factors in which standardised mortality ratios for death by suicide in illicit drug users was 10.1 times higher than that among non-drug users.

3.27 Substance misuse increases the risk of suicide attempt and death by suicide. The risk associated with opioid use disorders and mixed intravenous drug use is greater than that for alcohol misuse. The risk of suicide from alcohol misuse is greater among women than among men.

Individual-level factors: chronic illness

Epilepsy

3.28 One review reported on suicide risk related to epilepsy (Pompili et al, 2005), comparing data related to completed suicides who had epilepsy compared with completed suicide in the general population. This review included 29 studies from a range of countries (including the UK), with a total of 50814 participants, 187 of whom died by suicide. Results obtained for each study were synthesised to calculate the mean number of suicides per 100,000 per year for individuals suffering from epilepsy and comparisons made with suicide rates for the general population within the relevant countries. There were inconsistent results between studies. Although the authors state that a meta-analysis was performed and that results demonstrated that suicide in patients with epilepsy was more frequent than in the general population (with a table illustrating results from individual studies), they do not provide the relevant statistical results showing the results of the pooled data.

3.29 The meta-analysis demonstrated that suicide in patients with epilepsy is more frequent than in the general population. However, a number of cohorts of epileptic patients had a suicide rate lower than that of the general population. There were also large discrepancies between studies included in the review. For instance, suicides in surgically treated patients had widely disparate figures. In based on a UK study from 1968, the reviewers calculated the mean number of suicides per 100,000 for individuals suffering from epilepsy at 833 (and at 7.9 for the general population) whereas calculations based on more recent UK studies were much more modest (in 1973 epilepsy calculated at 136 per 100,000 and general population 4.4 per 100,000, in 1994 epilepsy calculated at 16 per 100,000 and general population 14.1 per 100,000, in 2001 epilepsy calculated at 11.8 per 100,000 and general population 11 per 100,000. Calculations were made based on cohort size, number of suicides and length of follow-up, all of which varied across studies. Risk factors identified within the included studies showed that temporal lobe epilepsy and those with temporal lobectomies and surgical resections had an increased risk of suicide. One study found increases in depression after surgery. Another study found a greater risk of suicide after the suppression of seizures or full control of seizures. The authors conclude that, although there is a greater risk of suicide among epileptic patients compared with the general population, within epileptic conditions there are wide variations that may be related to the type and severity of epilepsy or co-morbidities. These results should be interpreted with caution as there were large variations in the results between studies that cannot be explained simply by reference to the possibility of cultural differences between study countries.

3.30 Neeleman (2001) adds to the above from a review of a wide range of suicide risk factors comparing standardised mortality ratios for suicide with accidental death and death from natural causes. Results from four cohorts totalling 4116 participants suggest that epilepsy patients are not more likely to die by suicide than accidental death or death from natural causes.

3.31 There is increased suicide risk associated with epilepsy. However, there is also evidence that this risk varies across different types of epilepsy and in relation to the degree of severity of the effects of the illness. Persons who have temporal lobe epilepsy or who have had temporal lobectomies or surgical resections have an even greater risk of suicide.

Individual-level factors: personality

3.32 Brezo, Paris and Turecki (2006) conducted a very large review including 90 studies with over 20,000 participants (countries of origin not stated), to explore the significance of personality traits in suicidal behaviours. Outcome measures included suicidal ideation, attempts and completed suicide. They explored hopelessness, neuroticism, impulsivity, anger, irritability, hostility, and anxiety. Although results were reported in tabular form for individual included studies, there were no reported overall odds ratios for this review. The authors identified hopelessness, neuroticism and extroversion as those personality traits with the strongest evidence that supports the presence of these traits as risk factors for attempted suicide. The authors indicated a need for further research to determine whether aggression, impulsivity, anger, irritability, hostility, and anxiety were also useful markers of risk for the prediction of suicidal behaviour. The authors concluded that these selected personality traits may provide useful indicators to inform the prediction of suicide risk, but also suggest that future research should explore the contribution of personality traits in relation to the environmental and genetic variations in different gender, age, and socio-cultural groups.

3.33 Neeleman (2001) adds to the above from a review of a wide range of risk factors, with findings that suggest that adolescent neuroticism is a risk factor that makes young people almost 2.3 times more likely to die by suicide than the general population.

3.34 One review explored the possible association between attention deficit hyperactivity disorder ( ADHD) and suicide in boys (James, Lai and Dahl, 2004). The review used completed suicides as an outcome measure (no indication of number of included studies or total numbers of participants) and narratively combined studies from Europe (including the UK) and the USA. The authors found that there was an association between ADHD and completed suicide, particularly for younger males. The suicide rate from studies of ADHD with long term follow-up was found to be 0.63-0.78%. In comparison to national suicide rates in the US, males aged 5-24 years with ADHD were nearly three times more likely to die by suicide. The authors concluded that males with ADHD have an elevated risk of suicide, since ADHD appears to increase the severity of co-morbidities such as conduct disorder and depression.

3.35 Another review explored suicidality in patients with eating disorders, obesity and weight concern (Pompili et al, 2006). Narrative synthesis was used to summarise the results of included studies (number of studies and participants not stated), although results were so heterogeneous that it was difficult to summarise the findings in a useful and succinct manner. Differences between studies reveal that there are ranges of severity and types of eating disorders that impact on suicide risk. Three of the included studies reported higher numbers of suicide attempts and DSH among bulimic patients where they exhibited purging behaviour. Co-morbidities and other factors were also implicated in suicide risk. For instance, one study of suicide in bulimia found that 20% of those who had attempted suicide had a major depressive disorder, and 11% were drug and alcohol misusers (further details can be found in Annex 11). There were similar results for anorexia nervosa, with higher suicide attempts among the binge/purge sub-type; and also co-morbid mental illnesses and/or drug or alcohol misuse. No suicidal behaviour data were provided for obesity and weight concern. The authors concluded that, although there is an assumption that death from exhaustion and lack of food is the major cause of death in eating disorders, their results illustrate that compared to the general population, people with eating disorders have a higher risk of suicide.

3.36 One review explored the potential relationship between deficiencies in social problem-solving skills and suicidal behaviour in young people (that is, children, young people and younger adults) in high risk groups such as young offenders and psychiatric patients compared with the general population (Speckens and Hawton, 2005). Outcome measures were suicidal behaviour, defined by the authors as suicide attempts, deliberate self-harm and parasuicide. Some papers included ideation as well as suicide attempts in their analysis; only results relevant to our review are reported below. The review included 22 studies from the USA, Canada, Norway and the UK, using cross-sectional, case control and longitudinal study designs. Most of these studies, which compared adolescent patients with suicide attempts to either non-suicidal psychiatric or normal controls, found evidence for problem-solving deficits in the attempters. A further three cross-sectional studies of young offenders found lower problem-solving skills in suicide attempters. The case control studies verified results for cross-sectional studies, although the comparative risk of suicidal behaviour was much higher when patients with a history of suicide attempt were compared with the general population rather than other patients with no history of suicide attempt. Longitudinal studies showed no significant difference in risk after controlling for hopelessness and depression. Because most of the studies were cross-sectional, it was not possible to establish whether deficiencies in problem-solving skills lead to depression and (therefore) the associated elevated suicide risk, or whether depression is the main factor which undermines problem-solving skills. The authors concluded that studies should control for other variables such as hopelessness and depression, in order to establish risk, although these results do suggest that low problem-solving skills may be linked to an increase in risk of suicidal behaviour.

3.37 There may be increased suicide risk associated with particular individual/ personality factors. The evidence is particularly heterogeneous in this section both within and between reviews, although it appears that the following statements can be made with reasonable confidence:

  • A wide range of personality traits is implicated in higher risk of suicide, including hopelessness, neuroticism, extroversion, impulsivity, aggression, anger, irritability, hostility, and anxiety
  • Persons with ADHD have an elevated risk of suicide
  • A higher risk of suicide is associated with eating disorders such as anorexia nervosa and bulimia
  • There is an association between low problem-solving skills and elevated risk of suicide attempt

These studies demonstrate the multi-faceted nature of risk.

Individual-level factors: genetic predisposition

3.38 Two systematic reviews investigated genetic determinants of suicide risk. The first review (Lalovic and Turecki, 2002) explored the association between suicidal behaviour and a biallelic intron 7 polymorphism in the tryptophan hydroxylase ( TPH) gene. The 17 controlled studies included numbers of participants ranging from 27 suicide attempters and 190 controls to 231 suicide attempters and 282 controls and the studies. Participants represented a variety of population groups: violent offenders and arsonists, and hospital inpatients such as schizophrenics, depressed (unipolar) or bipolar patients. Studies were based in Europe, North America and Asia with suicidal behaviour (either completed suicide or attempted suicide) as outcome measures. The authors attempted to do an analysis based on violent versus non-violent suicide attempters but very few studies classified suicide attempts in a way that allowed this. The results of two meta-analyses compared suicide attempters or completers with healthy controls; and suicide attempters with non-attempters. From the combined results of comparisons within both groups, the authors concluded that there was no overall association between suicidal behaviour and an intron 7 polymorphism of the TPH gene. They suggested a need for standard criteria for classifying suicide attempts that would include degree of violence, lethality and intent. This would enhance the pooling of data for meta-analysis.

3.39 The second review re-examined the data on the genetic link between 5- HT2A (a serotonin receptor) with schizophrenia and suicidal behaviour (Li, Duan and He, 2006). This review included 61 papers reporting on 73 controlled studies with almost 20,000 participants, based in Europe and Asia. There was little heterogeneity found between all studies and despite the large sample sizes, they did not find an overall significant association of the T102C polymorphism with either schizophrenia or suicidal behaviour. Only one significant association was found, between A-1438G (a polymorphism in 5- HT2A) and suicidal behaviour. The authors concluded that their results did not support evidence for the association of the 5- HT2A gene with either schizophrenia or suicidal behaviour.

3.40 Two reviews explored the evidence for genetic links to suicidal behaviour. There was no association between an intron 7 polymorphism of the TPH gene or for the 5-HT2A gene and suicidal behaviour.

Individual factors: biomedical/physical determinants

3.41 Two systematic reviews provided evidence related to biomedical or physical determinants of suicide risk in women. One of these reviews investigated the possible associations between phases of the menstrual cycle and suicidal behaviour (Saunders and Hawton, 2006). Outcome measures used in the 44 included studies (sample sizes ranging from 13-3110 participants (countries not stated)) included suicide attempt, ideation, completion and DSH. These studies employed a range of research designs, which may explain the variation in the conclusions reached across these studies. In suicide completers, the percentage of women in the menstruating phase ranged from 15-100%, therefore providing insufficient (or indeed contradictory) evidence for this association. The authors found that the higher quality, methodologically more rigorous studies provide evidence of a positive relationship between aspects of the menstrual cycle and attempted suicide, and that there is limited evidence that the first week of the menstrual cycle is more commonly associated with attempts. However, once again the results differ widely between studies. Suicide attempts appears to be associated with phases of the menstrual cycle that have lower oestrogen levels (such as the late luteal and follicular phases) and also in women who suffer from pre-menstrual syndrome. Despite the methodological limitations of the included studies, the authors conclude that the interplay between oestrogen and the serotonergic system may account for the association between the menstrual cycle and non-fatal suicidal behaviours.

3.42 Another review explored the causes of pregnancy-associated death in maternal homicide and suicides (Shadigian and Bauer, 2005). There were 28 included studies, from the USA, Mozambique, Finland, Zimbabwe, India, Bangladesh, Sweden, and Australia, with a total of 349 completed suicides. Although the leading cause of death in pregnant and recently pregnant women was found to be homicide, suicide is also a significant cause of pregnancy-associated death. Twenty-one of the included studies reported on pregnancy associated suicide. However, synthesis of the results across studies was difficult because of the variation in inclusion criteria between studies, with one study only exploring suicide during pregnancy and others defining the post-partum period as anywhere from 42 days to 8 years (with one year postpartum being the most common). The authors state that, although suicide is less prevalent than death by homicide and despite the variation in study types, all studies showed that suicide accounts for a significant proportion of pregnancy-associated mortality. The two included case control studies revealed a 3-6 times higher rate of suicide in women who had an abortion compared with pregnancies carried full term, however careful analysis and replication of these findings is required and any confounding factors such as abuse rates or mental illness (not made available in the review) should be examined.

3.43 Two reviews explored the evidence for biological/physical links to suicidal behaviour in women. The risk of suicide attempt may increase in phases of the menstrual cycle which have lower oestrogen levels and in women who suffer from pre-menstrual syndrome. Pregnancy was also identified as a period during which women may experience elevated risk of suicidal behaviour.
Furthermore, there is limited evidence (2 studies) that suicide rates are higher in women who have abortions compared to those who carry the baby to full term. Careful analysis and replication of these findings is required and any confounding factors such as abuse rates or mental illness (not made available in the review) should be examined.

Psychosocial-level factors: work and unemployment

3.44 Platt and Hawton (2000) conducted a systematic review that explored the relationship between conditions of the labour market and suicidal behaviour. 'Suicidal behaviour' was defined by the authors as completed and attempted suicide, parasuicide and deliberate self-harm. This review combined (by narrative synthesis) a large number (n=165) of studies from a wide range of countries including Scotland and other UK countries. The total number of participants was not stated. The authors acknowledge that variations in study design and lack of methodological rigour in some of the studies may have been responsible for the sometimes inconsistent results that they reported. The authors found an increased risk of suicide and DSH among the unemployed, although the magnitude of the risk varied by study design. Individual cross-sectional studies showed an increased rate for both suicide and DSH, while individual longitudinal studies showed a double or triple rate of suicide, but inconsistent evidence for DSH. Aggregate-level cross-sectional and longitudinal studies showed either no evidence or heterogeneous results for this association. They also state that evidence from UK studies shows an association between unemployment, suicide and DSH in the 1970s but either a negative or non-significant association in the 1980s. There was no strong evidence to suggest female labour force participation rates have led to increased suicide rates. Once again, there were inconsistent results both within and between groups of studies of different research designs. Social class and suicide (and DSH) were linked: the lower the social class, the higher the risk. The authors found that the highest proportional mortality rates for suicide were found amongst those working in medical and allied occupations, farmers (males only), nurses, health, education and welfare professionals and personal service workers. This list includes professions (i.e. doctors and other health professionals) which would be compatible with a higher social class which would run contrary to findings on social class. The authors suggested that this may be because there is a lower mortality rate for other causes in these groups, there may also be a link to access to means.

3.45 Two further reviews pick up the theme of elevated risk of suicidal behaviour in particular professions or occupational groups. Lindeman et al (1996) provides substantiating evidence for the above statement regarding health professionals. This review explored the variations in estimates of risk (comparing absolute and relative mortality rates) for the medical profession, with particular focus on gender difference. A narrative synthesis of 14 studies containing almost 1,000 participants in total (from the UK, other European countries, USA and South Africa) revealed that the estimated relative risk was almost double for women, varying from 2.5 to 5.7 times more likely than the general population, while the range for male doctors was 1.1 to 3.4 more likely in comparison to the general population. In another comparison between medical doctors and other professions, the estimated relative risk for male doctors ranged from 1.5 to 3.8 more likely and from 3.7 to 4.5 more likely in female doctors than the general population. The authors found that the crude suicide mortality rate was approximately similar in male and female doctors. They concluded overall that the suicide rates among doctors were both higher than those in the general population and also higher than other professional occupational groups.

3.46 Hem, Berg and Ekeberg (2001) also explored suicide mortality rates and occupational group, in this case focusing on the police force. The twenty studies included in their review were drawn from North America, Europe (including UK studies) and Australia. Narrative synthesis of the results showed that, across the recent national studies, police did not have an elevated risk of suicide compared with the general population. The largest (nationwide) study was conducted in France, with a total of 749 police suicides. Adjusting for age and gender, the suicide rate in police was 34.8 per 100,000 per year compared to 35.4 per 100,000 per year in the general population. Another nationwide study in Germany also demonstrated no higher risk of suicide among the police force. In a population based study in England and Wales (covering 1982-96), police had the lowest occupational suicidal mortality ratio. Although the proportional mortality ratio increased from 61 to 79 in the later period included in the study, the trend was not statistically significant. Other studies showed inconsistent results. One study found that many countries do not keep records of suicides in the police force and few countries have gathered statistics on suicides. However, from the available evidence, the authors concluded that employment in the police force is not a risk factor for suicide.

3.47 Neeleman (2001) adds context to the above from a review of a wide range of risk factors comparing standardised mortality ratios for suicide with accidental death and death from natural causes. Results from three cohorts containing 26330 subjects show that those with lower socio-economic status ( SES) and the unemployed are 2.2 times more likely to die by suicide than those from higher socio-economic groups or those who are employed.

3.48 Unemployment is linked to elevated risk of suicide. Occupational social class and suicide (and DSH) are inversely linked: the lower the social class, the higher the risk of suicidal behaviour. However, the highest proportional mortality rates for suicide are found in medical doctors and farmers, with female doctors having a higher risk of suicide than male doctors. Employment in the police force is not a risk factor for suicidal behaviour.

Psychosocial-level factors: poverty

3.49 One review (Rehkopf and Buka, 2006) explored the association between local area-level suicide rates and socioeconomic advantage/disadvantage ( SES). The authors conducted a meta-analysis on 86 studies from a range of countries in Europe and beyond. Total numbers of participants included within the review were not stated. They found that the level of aggregation had an important effect on results. Analyses conducted at the community-level (that is, in a smaller local area) were significantly more likely to demonstrate lower rates of suicide among higher socio-economic areas than studies using larger areas of aggregation. Seventy per cent of the significant results showed an inverse relationship between higher socio-economic status and suicide, i.e. higher SES was associated with lower suicide rates. Neighbourhood-level aggregates produced an inverse relationship in 95% of the studies. Study results also varied according to the measure of SES used. Measures of area poverty and deprivation (using indexes such as Townsend/Carstairs) were, in 95% of studies, inversely associated with suicide rates. Median income was least likely to be inversely associated with suicide rates. Analyses using measures of unemployment, education or occupation were equally likely to demonstrate inverse associations as 73% of the included studies achieved such results using these measures. There was a trend towards an increase in the inverse association among the more recent studies (0% inverse association in years 1941-1960, 57% inverse association in years 1961-1980, and 76% inverse association in years 1981-2004). However, study results did not vary significantly by gender or by study design. The authors suggest that the heterogeneity of associations is mostly accounted for by study design features that have largely been neglected in this literature. Enhanced attention to size of region and measurement strategies in this review provided a clearer picture of how suicide rates vary by region. The authors reveal the importance of taking account of relative poverty or deprivation since this enables the researcher to provide a context for SES and how this impacts on the individual at community or neighbourhood level.

3.50 Poverty and deprivation are linked to suicide risk at an ecological (area) level. Areas with greater levels of socio-economic disadvantage (lower SES) are more likely to have higher suicide rates.

Protective factors: the evidence

Individual-level factors: problem solving

3.51 Three studies relevant to the review identified problem-solving skills as protective against suicide and suicidal behaviour within different population groups.

3.52 Elliott and Frude (2001) used a cross-sectional interview study to explore the relationship between level of hopelessness and stressful life events (measured across the two years preceding suicide attempt) and coping strategies, among a sample (n=80) of people aged 18+ years who had self-poisoned in Wales. Their results showed that hopelessness was a strong predictor of suicide risk (r = 0.6) but that problem-focused coping strategies had a mediating effect on this. They found that problem-focused coping strategies had a negative correlation (r = -0.34) with suicide attempt and that the higher patients scored on the hopelessness scale, the less they tended to employ problem-focused coping (although this finding only approached statistical significance).

3.53 Chapman, Specht and Cellucci (2005) set out to explore the association between risk and protective factors and suicide attempts among a sample (n=105) of a female prison inmate population using a cross-sectional study with a key focus on hopelessness as a risk factor for suicide attempt. Survival and coping beliefs and problem-focused coping strategies were negatively correlated with suicide attempt. A secondary objective was to test whether these protective factors remained when controlling for hopelessness, that is, exploring the role of hopelessness in mediating protective factors. Conversely to Elliott and Frude, they found that, when controlling for hopelessness, the protective effect of survival and coping beliefs and problem-solving coping was not statistically significant.

3.54 Using a case control study, Donald et al (2006) investigated risk and protective factors for medically serious suicide attempts among young Australian adults using a sample (n=475) of young adults (18-24 years) admitted to a hospital emergency department following a suicde attempt (n=95) and matched controls from a population survey (n=380). Using a social-ecological protective factor model that considers how individuals interact with their social and environmental contexts, they found that locus of control and problem-solving confidence protected against suicide attempts.

3.55 Problem-solving skills may be protective against suicidal behaviour among those who have attempted suicide. There is conflicting evidence on the interplay between the suicide risk factor of hopeless and problem-solving based coping skills. One study shows that problem-solving coping may mediate against hopelessness among adults who have attempted suicide while another demonstrates that hopelessness can mediate against the protective effect of problem-solving-based coping.

Individual-level factors: self control of behaviour, thoughts and emotion

3.56 Several studies provide evidence on protective factors that centre around elements of perceived self control of behaviour which contribute to resilience. The studies cover three population groups: young people/adolescents; those with depression or borderline personality disorder; and women who have been exposed to domestic violence.

Young people

3.57 In a qualitative interview based study of resilience in sample (n=13) of previously suicidal female adolescents (15-24 years), Everall, Altrows and Paulson (2006) set out to understand how these adolescents had overcome their suicidal behaviour. From their results, the authors identified four main domains of resilience (based on the dynamic and multi-dimensional process represented by resilience), three of which are relevant in this section of the review:

  • cognitive processes, that is, having the control to gain a better perspective on their lives through positive future thinking and focusing on the positive rather than negative aspects of self and the present rather than the past
  • purposeful and goal directed action, that is, a sense of control and self-efficacy enhanced by taking action to change their situations, in turn developing their confidence and self esteem
  • emotional processes, that is, the ability/willingness to face difficult emotions or anger and motivation towards recovery

3.58 Piquet and Wagner (2003) used a case control study to compare the coping responses of 23 hospitalised adolescent (13-18 years of age) suicide attempters in the USA with those of 19 hospitalised non-attempters matched on diagnosis and demographics. Although the sample size in this study was small (n=42), the results demonstrated that coping effectiveness was significantly higher among the control group than among the suicide attempters, with suicide attempters using more automatic coping (involuntary responses such as approach and avoidance) responses than effortful coping responses (processes in which individuals regulate their attentional and behavioural response tendencies). Greater coping effectiveness in suicide attempters was also linked to a decline in suicidal symptoms (although no statistical test of this association was reported). The authors concluded that suicide attempters may be more exposed to stressful situations that are not controllable which may explain their increased likelihood to respond using automatic e.g. avoidance coping strategies.

3.59 Apter et al (1997) carried out a case control study of suicide attempting inpatient and non-attempting inpatient and non-patient adolescents (12-19 year olds) in Israel (n=223) to examine defence mechanisms in suicidal behaviour. They found that sublimation (the internal process of transforming emotional dynamics that are (usually) considered to be unpleasant into socially acceptable attitudes and states of mind and good qualities of character) correlated negatively with both suicidal and violent behaviours.

3.60 Chandy, Blum and Resnick (1997) took a sub-sample of boys reporting to have been sexually abused and a control group from participants in the National Longitudinal Study of Adolescent Health ( US School Grades 7-12) to explore the protective factors that help male victims of sexual abuse to overcome vulnerability to a number of associated factors (including suicide). They found that sexual abuse, while associated with a higher risk of negative behaviours (including suicidal behaviour), is mitigated by protective factors, including being in control of behaviour, thoughts and emotion.

Adults

3.61 In a cross-sectional study, Kelly et al (2000) explored the relationship between recent life events, social adjustment and suicidal behaviour in a sample (n=80) of adult (over 18 years of age) patients with major depression or borderline personality disorder in the USA, 53 of whom had attempted suicide. In particular the authors aimed to explore the potential protection that high levels of social adjustment (defined as a broad measure of functioning in work, personal relationships, family life and available social support) might provide against suicidal behaviour. The authors found that high levels of social adjustment within immediate and extended families and overall may be protective against stress-related suicidal behaviour.

3.62 Meadows et al (2005) carried out a case control study to examine the role of protective factors (hope, spirituality, self-efficacy, coping, social support-family, social support-friends, and effectiveness of obtaining resources) against suicide attempts among economically, educationally, and socially disadvantaged African-American women (18-59 years) who had experienced recent intimate partner violence and who had attempted suicide. The sample (n=200), included 100 women with the above characteristics who had presented at a large urban trauma centre following a suicide attempt and 100 controls who had no history of suicide attempt and had presented at a walk-in clinic for non-emergency medical problems. Women with high levels of coping, high self-efficacy and high effectiveness in obtaining resources were more likely to have attempted suicide than were women with lower levels of these factors, respectively. A greater number of protective factors in combination was significantly associated with lower likelihood of having attempted suicide.

3.63 A number of coping skills requiring an element of self control including self-efficacy, instrumentality , social adjustment skills, positive future thinking and sublimation appear to be protective against suicidal behaviour particularly among adolescents and/or at times of stressful life events,. Being in control of emotions, thoughts and behaviour can mediate against suicide risk associated with sexual abuse among adolescents.

Individual-level factors: hopefulness, reasons for living and optimism

3.64 Several studies explored factors related to individual-level reasons for living, optimism and hopefulness as protective against suicide among a diverse range of population groups.

Women who have experienced domestic violence

3.65 Meadows et al (2005) (also reported above) found that one of the main protective factors against suicide attempts among economically, educationally, and socially disadvantaged African-American women (18-59 years) who had experienced recent intimate partner violence and who had attempted suicide was hopefulness. Those with high levels of hopefulness were less likely to have attempted suicide than those with low levels of hopefulness.

People with depression

3.66 Malone et al (2000) used a cross sectional study (n=84) to test the hypothesis that 'reasons for living' might protect or restrain US patients (18-80 years of age) with major depression from making a suicide attempt. Patients who had not attempted suicide scored higher on the following items in the reasons for living inventory: feelings of responsibility toward family, fear of social disapproval, moral objections to suicide, greater survival and coping skills, and a greater fear of suicide. Clinical suicidality was inversely correlated with reasons for living (canonical correlation = -0.64). Neither objective severity of depression nor quantity of recent life events differed between those who had and had not attempted suicide. This suggests that reasons for living can mediate against suicide attempt at times of risk such as severe depression or stress. The protective effect of reasons for living may be more relevant to how suicidal behaviour is expressed than how often stressful life events occur.

3.67 Hirsch et al (2006) conducted a cross-sectional study in the USA with a sample (n=202) of depressed adult in- and out-patients aged between 50 to 88 years. The objective of the study was to test the hypothesis that future orientation (defined as the ability to think about the future, the cultivation of a general positive outlook and mood about the future, the development of strategies to achieve goals and the presence of reasons for living) is associated with lower levels of suicide ideation and lower likelihood of suicide attempt in patients being treated for depression. They found that higher scores for future orientation were associated with a lower probability of a history of attempted suicide in the past than lower future orientation scores. However, future orientation was not associated with current suicide attempt status. Older participants were less likely to have attempted suicide, but there were no differences related to gender.

3.68 Oquendo et al (2005) undertook a cross-sectional study in the USA to determine whether the Reasons for Living Inventory ( RFLI) might capture protective factors against suicidal behaviour in a sample (n=460) of hospitalised Latinos and non-Latinos with diagnosis of major depression, bipolar disorder or schizophrenia aged 18 - 80 years. Protective factors against suicide attempts were found to include moral objections to suicide and survival and coping skills; these factors had a stronger relationship to suicide attempt than ethnicity. Latinos were found to be less likely than non-Latinos to have made lethal suicide attempts and scored much higher on the Reasons for Living Inventory than non-Latinos, suggesting that moral objections and survival beliefs are protective against suicide for this group.

Previous suicide attempters

3.69 Chesley and Loring-McNulty (2003) undertook a cross sectional study in the US with a sample (n=50) of community-based adults who had attempted suicide in the past to understand the experience of the suicidal individual and to identify factors that contributed to survival following a suicide attempt. The question around what was preventing future attempts gained the following responses (percentages are of the total sample): 14% said health professional intervention; 10% developing a sense of self-empowerment; and 10% achieved personal / professional success; 10% new outlook on life. In response to the question 'Has someone or something made a difference in keeping you alive?', 18% said children; 15% treatment by health professional; 15% self-empowerment; 14% spirituality; 22% personal relationships; 5% personal and professional success; 3% change in attitude; 2% medication; 2% support groups; 2% sobriety; 1% structure in daily life; 1% lifestyle change.

The authors interpreted this as meaning that most of their sample had identified positive reasons for living and had developed successful coping strategies which had contributed to their survival following a suicide attempt.

3.70 High levels of reasons for living, future orientation and optimism protect against suicide attempt among those with depression and those exposed to stress. Hopefulness is protective against suicide among African-American women exposed to poverty and domestic violence. There is some evidence that those who have previously attempted suicide can develop positive coping strategies to protect themselves against future suicidal behaviour. Resilience factors such as those above have been found to be better predictors of suicidal behaviour than the amount of exposure to stressful life events.

Individual protective factors: perceptions of positive health and participation in sporting activities

3.71 Two studies explored the relationship between positive perceptions of health and suicide attempt in adolescents. Chandy (1996) undertook a case control study with a sub-sample (n=2022) of girls reporting to have been sexually abused and a control group from participants in the National Longitudinal Study of Adolescent Health ( US School Grades 7-12) to explore the protective factors that help female victims of sexual abuse to overcome the vulnerability (to a number of factors including suicide) associated with this. They found that those with a history of sexual abuse had a significantly higher suicidal involvement (defined as past suicide attempts) (30.5%) than controls (16.6%). There was evidence that protective factors, including a perception of themselves as healthier than others, mitigate against negative outcomes. (Other protective factors from this study are reported below.) In another cross-sectional study, Tomori (2003) used school-based survey data (n=200) to explore the role of sport in relation to self-reported suicide attempts among adolescents, specifically examining sport and physical activity as possible protective factors in relation to Slovenian adolescent (14-19 years old) suicide attempts. Key findings included: among girls and boys, the attitude towards sport as a healthy activity was associated with lower likelihood of suicidal behaviour; among girls, non-attempters turned to sport as a coping behaviour in distress more frequently than attempters, and among boys non-attempters reported a significantly higher frequency of engagement in sport and physical activity than attempters. However the authors caution against these findings being interpreted in a way that allows engagement in sport to be used as a predictor for suicidal behaviour amongst adolescents and advise careful assessment of potential confounding variables.

3.72 There is some evidence that an attitude towards sport as a healthy activity and participation in sporting activity is protective against suicidal behaviour among adolescents. A perception of yourself as healthier than others may be protective against suicide among females who have experienced sexual abuse.

Psychosocial-level factors: family relationships

3.73 Eleven studies investigated the role of a range of aspects of family-based relationships as protective factors against suicide. Most of the studies focused on the experiences of adolescents.

Adolescents

3.74 Using a case-control study design (n=64), Israelashvili et al (2006) explored whether suicidal behaviour among female adolescents (12-18 years) in the USA, attending medical emergency rooms because of first-time suicidal behaviour, is an imitation of their mothers' tendency to escape active and problem-focused coping. The paper suffers from a lack of methodological detail (e.g. was group matching taken into account in the statistical tests?) and misreporting of results. A significant difference between suicidal and non-suicidal (control) subjects was found on only one of the 14 COPE (multidimensional coping inventory) scales. There were no significant differences between mothers of suicidal adolescents and mothers of controls on the COPE scales, nor between adolescent groups in the Active Coping Test ( ACT) mean score. There were three significant correlations between suicidal adolescents' and their mothers' scores on three COPE items (one of which was negative) and six significant correlations between non-suicidal adolescents' and their mothers' scores on six COPE scales (all positive). Significant between-group differences in the size of the correlation were found for eight items; in six of these the correlation was higher in the control group. The results do not support the study hypothesis.

3.75 In a UK based cross-sectional survey of a sample (n= 2560) of 14-18 year olds in schools and youth groups, Flouri and Buchanan (2002) tested the hypothesis that perceived parental involvement is negatively associated with self-reported suicide attempts in adolescence, after controlling for both risk and protective factors. Their findings suggested that adolescents who reported higher parental involvement, characterised by a number of factors including emotional support, engagement, responsibility and accessibility, were less likely to have made a suicide attempt. This effect was not weaker when family structure had been disrupted than when young people lived in intact two-parent families.

3.76 Chandy, Blum and Resnick (1996), also reported above, explored the protective factors that help adolescent female victims of sexual abuse to overcome the associated vulnerability to a number of factors including suicide. Living with biological parents who cared about them was identified as a protective factor for this group. In a later case-control study, Chandy, Blum and Resnick (1997) took a sub-sample (n=740) of boys reporting to have been sexually abused, and a control group from participants in the National Longitudinal Study of Adolescent Health ( US School Grades 7-12), to explore the factors which protected against vulnerability to a number of factors including suicide. They found that, while sexual abuse is associated with a higher risk of negative behaviours (including suicidal behaviour), this is mitigated by protective factors, the most powerful of which were maternal education beyond high school and the perception that their parents cared about them.

3.77 Husler, Blakeney and Werlen (2005) carried out cross-sectional research in Switzerland with a sample (n=1028) of 'at risk' adolescents (e.g. school drop-outs and substance misusers) to test a model of adolescent risk and protective factors including mental illness, suicidality, use of tobacco, alcohol and cannabis, and secure self and family relations as interacting outcome measures. For girls good relationships with their families and good parental relationships were found to be marginally protective against suicide.

3.78 Svetaz, Ireland and Blum (2000) carried out in-depth interviews with a sub-sample (n=1301) of those who participated in the National Longitudinal Study of Adolescent Health to identify differences in emotional well-being among adolescents ( US School Grades 7-12) with and without learning disabilities, and to identify risk and protective factors associated with emotional distress. Adolescents with learning difficulties who had experienced emotional distress reported eight times the number of suicide attempts than those without emotional distress. Family connectedness was associated with lower suicide risk.

3.79 O'Donnell et al (2004) set out to explore and understand the growing problem of suicidality (suicidal ideation and suicide attempts) in African-American and Latino teenagers (average age of 17) through a questionnaire to a sample (n=879) of school pupils from deprived backgrounds. They found that family closeness was a strong protective factor against suicide attempts.

War Veterans

3.80 Benda (2003) attempted to determine which factors discriminate between homeless, substance-misusing Vietnam veterans who were non-suicidal, those who had suicidal thoughts, and those who had attempted suicide. A sample of (n= 600) was recruited from those attending a substance misuse facility for veterans. Data were collected on several factors based on attachment theory, including caregiver attachment, sexual abuse, physical abuse, resilience, self-efficacy, and self-esteem. A range of protective factors were found among those who were non-suicidal homeless substance abusers, one of which was 'caregiver attachment'.

Adult women

3.81 Meadows et al (2005) (also reported in the individual protective factors section above) found that high levels of social support from family was a main protective factor against suicide attempts among economically, educationally and socially disadvantaged African-American women (18-59 years) who had experienced recent intimate partner violence and attempted suicide and a control group. Those with high levels of social support from their family were less likely to have attempted suicide than those with low social support.

3.82 Two studies examined the protective role of having children for women. Driver and Abed (2004) used records of completed suicides (n=60) in Rotherham ( UK) to assess the effect of having offspring, dependent offspring (<18 years), non-dependent offspring (>18 years) and no offspring, on suicide rates in women. The results provided evidence that having children, per se, does not protect against suicide but that having dependent children living at home (< 18 years) and children over 18 years of age living at home mitigated against the risk of suicide. This protective effect is lost when the offspring leave home. In another study by Cooperman and Simoni (2005) which explored the prevalence, timing, and predictors of suicidal ideation and attempted suicide in a sample of HIV-positive women (over 18 years of age) in New York City, there was no evidence of the protective effect of having children. Contrary to Cooperman and Simoni's original hypothesis, those with children were significantly more likely to attempt suicide than those who did not

3.83 Chesley and Loring-McNulty (2003) (also reported above) found that personal relationships (22% of sample) and children (18% of sample) were factors that protected against repeat suicide attempts.

3.84 Good relationships with parents mitigate against suicide risk, especially in adolescents and including those who have been sexually abused. Positive family relationships also provide a protective effect for adolescents including those with learning disabilities. Further evidence suggests that positive maternal coping strategies can have a protective effect on female adolescents. Having children living at home is protective against suicide for women; however, another study indicates that this protective effect may not exist among women who are HIV-positive.

Psychosocial-level factors: marriage and partnership

3.85 Five studies provide evidence on the protective effect of the commitment of marriage and same sex partnerships.

3.86 Using completed suicide data (n=25476) from regional, national and urban (Austria, Belguim, Denmark, Finland, Switzerland, Turin, Madrid and Norway) longitudinal mortality registers linked to census data, Lorant et al (2005) explored whether being married is protective against socio-economic inequalities in suicide, and whether any such buffering effect varies between countries. Being married had a buffering effect against inequalities in suicide risk arising from low educational qualifications (except for those 65 and over) and among those who do not own their houses. The buffering effect of being married was stronger among men than among women. The protective effect of marriage was not affected by the level of social capital at the country level.

3.87 Nisbet (1996) undertook a secondary analysis of epidemiological data (n=16477) to evaluate whether a model of social support could help explain the low suicide rate of Black females over 18 years of age in the USA. The research used data from a cross-sectional US study of the incidence and prevalence of major psychiatric disorders and the utilisation of health and mental health services undertaken between 1980 and 1985. Women had a higher suicide attempt rate than men, but the difference between black women and white women was not significant. The authors also modelled the data to examine the relationship of background characteristics including marital status and the number of children on attempted suicide as mediated by emotional state. Marriage was found to have a protective effect, but the effect was stronger for white females than for black females. Seeking support from friends and family was also found to be protective.

3.88 Kraut and Walld (2003) set out to compare the relationships of unemployment, part-time work, non-labour force participation, and full-time work with attempted suicide among residents of Manitoba, Canada aged 15-64 who made use of health services (n=43,188). The results related to employment are reported below in the employment section. Other results relevant to this section on marriage include the findings that residential stability and marriage were protective against suicidal behaviour.

3.89 Benda (2003) attempted to determine which factors discriminate between homeless, substance misusing veterans who were non-suicidal, those who had suicidal thoughts, and those who had attempted suicide (n=600). A number of protective factors were found (see employment, religion and parental relationships), including the commitment of marriage.

3.90 Mathy, Kerr and Lehmann (2003) used survey data (n=38,204) to explore the combined effects on mental health of marriage as a protective factor and homosexuality as a risk factor in USA and Canada. They found a significant association between suicidality and sexual orientation in both countries, with homosexual and bisexual people being at higher risk of suicide ideation and attempts than heterosexual people. Marriage appeared to mediate this risk in some, but not all, instances. Relationship status was only significant in protecting against suicide attempt among US and Canadian men, but not among Canadian or US women. The authors point out that married men were less likely than non-married men to have problems with drugs, sex and gambling and having used or currently using psychiatric medication.

3.91 Marriage is a protective factor against suicide (although more so for white females than black females in the USA). There is also evidence that marriage has a buffering effect against socio-economic inequalities found in suicide, particularly for men. Homosexual and bisexual people are at higher risk of suicide ideation and attempts than heterosexual people. and marriage-like partnership was found to be protective of homosexual men and not women. It is important to consider other confounding variables including the finding that married men were less likey than non-married men to have problems with drugs, sex, gambling and having used or currently using psychiatric medicine.

Psychosocial-level factors: social relationships and social connectedness

Positive school experiences and school connectedness

3.92 A number of studies explored protective factors related to a sense of social belonging and connectedness often arising from a combination of social sources, emphasising the need to understand protective factors against suicidal behaviour as interactive social processes.

3.93 Kidd et al (2006) explored the impact of social relations on suicide attempts in a longitudinal study of adolescents (grades 7 -11, mean age of 16) in the USA (n=9142). Adolescents who felt more connected to their parents were less likely to commit suicide ( OR = .60 (no CIs stated)). Parent relations were the most consistent protective factor, and among boys with prior suicide attempts, school relations augmented the effects of parent relations when peer relations were poor. Similar findings were gained in a study by Pharris, Resnick and Blum (1997) which sought to identify factors which protect against the adverse health correlates of sexual abuse (including suicide) in reservation-based American Indian and Alaskan Native adolescents (12-21 years) using a cross-sectional school based survey. Factors associated with an absence of suicide attempts for females were family attention, parental, family, and adult caring, parental expectations that were not too high, and belief that school people care. For males, the only factor associated with absence of suicide attempts was family attention.

3.94 Svetaz, Ireland and Blum (2000), previously reported in the section on 'good family relationships', found school connectedness to be associated with lower suicide risk among adolescents with learning disabilities.

3.95 Chandy, Blum and Resnick (1996), also reported above, explored the protective factors that help female victims of sexual abuse to overcome the associated vulnerability to a number of factors including suicide. The presence of a school nurse or clinic was one of a number of protective factors identified. In a later study, Chandy, Blum and Resnick (1997) took a sub-sample (n=370) of boys reporting to have been sexually abused and a control group (n=370) from participants in the National Longitudinal Study of Adolescent Health ( US School Grades 7-12) to explore the protective factors that help male victims of sexual abuse to overcome the vulnerability to a number of factors including suicide. They found that a perceived supportive school was a protective factor against suicidal behaviour.

3.96 Two studies examined the experiences of adolescents identifying as lesbian, gay, bisexual or transgendered ( LGBT). The combined protective effect of family and school was also found by Eisenberg and Resnick (2006) who examined the association between four protective factors (family connectedness, teacher caring, other adult caring, and school safety) and suicidal ideation and attempts among gay, lesbian and bisexual ( LGBT) young people in the US using a cross-sectional survey design (n=2255). The study found that those identifying as LGBT reported significantly lower levels of each of the protective factors examined than their non- LGBT peers. LGBT as a suicide risk factor is significantly mediated by the protective factors of family connectedness, adult caring, and school safety.

3.97 Fenaughty and Harre (2003) explored resiliency to suicide among young (under 26 years of age) gay men in New Zealand recruited to a small scale (8 participants) qualitative study through 'youth networks'. Fenaughty and Harre describe a complex interplay between risk and resiliency for this risk group ('seesaw' model), with perceptions of gay sexual orientation as the pivot. Important factors that participants felt contributed to increased resiliency to suicidality included positive stereotypes or representations of gay men, positive family acceptance of homosexuality, GLB friendly schools and school peer support, gay support network participation, high self esteem gained through having a positive perception of gay sexuality and coping mechanisms such as problem-solving coping. According to Fenaughty and Harre, 'coming out' is one of the most stressful experiences for gay youth and this study demonstrates that differences in suicidality may be more related to the amount of resiliency that individuals have than the amount of stress they experience. This is similar to observation by Malone et al (2000) that resiliency to stressful life events is a better predictor of suicide than the amount of stressful life events experienced.

3.98 Supportive school environments, including access to health care professionals at school are important protective factors among adolescents including those who have experienced sexual abuse, those with learning disabilities and those who identify as lesbian, gay, bisexual or transgendered.

General social support

3.99 The protective effect of social support against suicide was explored in a number of studies which cover a diverse range of population groups.

3.100 In a small-scale (n=13), qualitative study of resilience among suicidal female adolescents (15-24 years) Everall, Altrows and Paulson (2006) set out to understand how these adolescents had overcome their suicidal feelings (n=7 had attempted suicide). The authors found four main domains of resilience. Three of these (purposeful and goal directed action, cognitive processes and emotional processes) are reported in individual-level factors section above. The fourth domain was 'social processes', that is, a significant relationship that provided social support, usually within a social setting with which respondents felt a sense of belonging.

3.101 Two US-based studies explored protective factors among black Americans. Kaslow et al's (2005) hospital clinic-based case-control study examined the effect of several potential risk factors (life hassles, partner abuse, partner dissatisfaction, and racist events) and potential protective factors (effectiveness of obtaining resources, social embeddedness, and social support) for suicide attempts among a sample (n=200) of adult (18-64) African-American suicide attempters and a control group of non-suicide attempters. They concluded from their results that suicide attempter status could be predicted by two independently significant social variables, including one risk factor, life hassles, and one protective factor, social support. Nisbet (1996) undertook a secondary analysis of epidemiological data to evaluate whether a model of social support could help explain the low suicide rate of Black females over 18 years of age in the USA (reported in the marriage section above). Seeking support from friends and family was found to be protective against suicide attempt.

3.102 Coker et al (2002) undertook a cross-sectional study of a sample of female victims of domestic abuse seeking medical help (n=1152). The study aimed to determine associations between intimate partner violence (defined as sexual, physical, or psychological abuse) and mental health outcomes, and to assess the protective roles of abuse disclosure and social support on mental health among abused women. Coker found that higher levels of social support reduced the risk of adverse mental health outcomes among the abused women by almost one half, and higher social support scores were associated with reduced risk of suicide attempts.

3.103 Donald et al (2006) investigated risk and protective factors for medically serious suicide attempts among young Australian adults (18-24 years) in a case control study (also reported above in the individual-level protective factors section). The findings revealed a trend towards social connectedness being more protective of those with high depressive symptomatology than those with low depressive symptomatology and among smokers rather than non-smokers. Immediate family support was not found to be protective against medically serious suicide attempts.

3.104 In a study of African-American women (reported above) Meadows (2005) found that those with high levels of social support from friends, were less likely to attempt suicide than those with low levels of social support.

3.105 Social support and connectedness in general is protective against suicide among a range of population groups, including black Americans and women who have experienced domestic abuse, young adults with severe depression and smokers.

Psychosocial-level factors: religious faith and spirituality

3.106 One systematic review and thirteen primary studies explored religious faith and spirituality as a protective factor against suicidal behaviour. This section is reported under the main themes of religious participation, moral objections to suicide and differentiating factors within religious groups.

Religious participation

3.107 Van Ness and Larson (2002) conducted a systematic review of the evidence concerning religiosity/spirituality and mental health in persons over 65 years of age. We report the section of these results that linked religious participation/observance to suicide rates. One of the included studies attempted to explore critically Durkheim's theory of religion as an aspect of 'organic solidarity'. The study found that the percentage of residents in a region participating in religious organisations was inversely proportional to the rates of suicide in that region. A further two studies contributed to this evidence, with one of these distinguishing between family support and that provided by religious organisations as they recognised that social integration could be a confounding factor. However, yet another study included in the review found that religious involvement and suicide had a non-significant association when measures of social integration were added to the multivariate regression models.

3.108 Tubergen, Grotenhuis and Ultee (2005) also set out to explore critically Durkheim's study of suicide, particularly to investigate the support provided by religious networks and/or religion-based moral sanctions on suicide. This cross-sectional study used data from the Netherlands (1936-73) of Catholic, Protestant and non-churchgoing suicide completers (n=14744). They found that suicide rates decreased among populations with rising proportions of church attendees in a community.

Moral objections to suicide

3.109 The study above provides evidence for the community norms theory: that high levels of church attendance is associated with strong prohibitions against suicide across the whole community (rather than among only those attending church). The community norms hypothesis was also tested by linking suicide rates to the overall decline in church attendance over time. Findings suggested that religious communities have a protective effect because of the prohibition on suicide, but with increasing secularisation and the waning of religious participation, community norms related to the religious community are losing impact.

3.110 Three further papers provided supporting evidence that religious participation/religious communities can generate protection against suicide because of the moral sanctions on this behaviour. In a cross-sectional study in the USA, conducted with Latinos and non-Latinos with a diagnosis of major depression, bipolar disorder or schizophrenia from 18-80 years of age, Oquendo et al (2005) (also reported above) found that Latinos were less likely than non-Latinos to have made lethal suicide attempts and scored much higher than non-Latinos on the Reasons for Living Inventory (which included a moral objections to suicide factor). In a psychiatric hospital-based case control study (n=357) Dervic et al. (2006) explored the potentially protective role of moral objections to suicide against suicidal behaviour in patients with cluster B personality disorders ( CBPD) or depression. They found that moral objections to suicide or religious beliefs may have a protective effect against suicidal behaviour in depressed patients with co-morbid cluster B personality disorder, as suicide attempters were less likely to have religious affiliation than non-attempters. One further study (Malone et al, 2000, also reported above) suggests that moral objections to suicide may lead to less lethal methods of suicide attempts. In exploring the lethality of attempts, moral objection to suicide was the only reason for living that was significantly stronger in the subjects with low-lethality suicide attempts than in those with high-lethality attempts.

3.111 Four studies included religious participation or spirituality as a protective factor against suicidal behaviour among a range of other factors. Each dealt briefly with this topic and did not investigate what role religion might play in the lives of participants or which aspect of religious participation offered protection against completed or attempted suicide. Svetas, Ireland and Blum (2000) carried out interviews with adolescents ( USA school grades 7-12) with and without learning disabilities (also reported above). Religious identity was associated with lower risk of suicide. This was also the case in another USA study (Chandy, Blum and Resnick, 1996, also reported above) of girls who had been sexually abused, in which spirituality or religious participation was found to protect against that higher risk. A study of homeless, substance misusing veterans drew on a resilience approach to determine which factors distinguished those who had and had not attempted suicide (Benda, 2003). The author found multiple protective factors, including religiosity, were associated with the resilient group. Meadows et al (2005), reported above, found spirituality protective against suicide among African-American women who had experienced inter-partner violence. Those with high levels of spiritual well-being were less likely to attempt suicide compared with those with lower levels of spiritual well-being. Finally, Chesley and Loring-McNulty (2003) (reported above) found that 14% of their participants reported spirituality to be a factor that protected against repeat suicide attempts.

3.112 A further four studies which explored the role of religious faith and spirituality as a protective factor demonstrated no evidence of reduced risk of suicidal behaviour related to religious participation or spirituality. These studies covered a diverse range of population groups, including abused women (Coker et al, 2002), HIV-positive women (Cooperman and Simony, 2005), lesbian and bisexual women (Mathy and Schillace, 2003) and Inuit youth in Canada (Kirmayer, Boothroyd and Hodgins, 1998). In the latter study, church attendance ceased to be significant as a protective factor when psychiatric illness was included in the analysis.

Differentiating factors within religious groups

3.113 As Kirmayer et al's (1998) study suggests, the protective effect demonstrated by this field of study may include multiple underlying factors at play beyond simply adherence to religion or being an active member of a religious group. A study involving North American First Nations (Garroutte et al, 2003) showed that there may be interplay between different religious forms observed by members of the same community and that it is necessary to look more closely at these in order to attribute protective effects. Drawing on data (n=1465) from the American Indian Service Utilization, Psychiatric Epidemiology, Risk and Protective Factors Project, the cross sectional study distinguished between participation in Christian churches and participation in ritual/cultural practices associated with American Indian traditions. Those with a high level of American Indian ritual/cultural orientation had a reduced prevalence of suicide compared with those with low level of ritual/cultural orientation; there was no association between Christian practice/affiliation and suicide attempts. According to the authors, these results suggest that a positive cultural identity may be protective against suicidal behaviour in American Indian populations.

3.114 Nonnemaker, McNeely and Blum (2003) explored the association between public and/or private domains of religiosity and a range of health-related outcomes, including suicidal behaviour, among adolescents (school grades 7-12). They used existing data from the Longitudinal Study of Adolescent Health to identify a sample of adolescents who had expressed some religious affiliation (n=16306). Public religiosity was defined as church attendance or participation in church organised groups, while private religiosity was considered to be individual, private prayer. While public religiosity was associated with lower emotional distress, private religiosity was not. Private religiosity was significantly associated with a lower probability of having had suicidal thoughts or having attempted suicide, while public religiosity was not.

3.115 Distinctions between the protective effects of different forms of religious observance were also explored in a cross sectional study by Molock et al (2006). The study questioned whether involvement in public religious observance and using different religious coping strategies protected African-American teenagers (13-19 years) from suicidal behaviours using a school-based sample (n=212). Different styles of religious coping, based on the model of Pargament et (1988) 4 model, were identified: self-directed coping, where the person is active in problem solving and God/'Higher Power' is passive; collaborative coping style, where the individual involves God as a partner in problem solving; and deferred coping, where the person is passive and expects God to solve their problems. A person might draw on all three styles at different times depending on the context/stressor. There was a significant relationship between self-directed coping and suicide attempts: those using this style were more likely to report having attempted suicide than those using other types of coping. According to the authors the results may indicate that increased church attendance and church involvement have an influence on coping style and thereby represent a protective factor.

3.116 There is a wide range of evidence to suggest that religious participation may be a protective factor against suicidal behaviour. However, this may vary according to the level of secularisation within a country or community. Moral sanctions against suicide promoted by members of a religious community may have wider protective effect on the non-religious members of a community where the religious members are in the majority. Religious observance does not confer equal protection on individuals. Other factors, such as the observance of traditional cultural rituals, may have a stronger protective effect. The manner in which individuals relate to their God (in terms of religious coping style or private versus public expressions of religiosity) may further highlight different levels of protective factors within a single religious community.

Employment

Three studies provide research evidence on the potential of employment as a protective factor against suicide and suicidal behaviour.

3.117 Kraut and Walld (2003) carried out a cross-sectional study to compare the relationships of unemployment, part-time work, non-labour force participation, and full-time work with attempted suicide among residents of Manitoba, Canada aged 15-64 who made use of health services (n=43188). Unemployment was associated with a higher likelihood of attempted suicide and those who worked part-time (1-15 weeks, 26-51 weeks) and those not working at all had an elevated likelihood of attempted suicide when compared to those working 52 weeks (that is, full time). These findings suggest that full-time employment is protective against suicide attempts.

3.118 Benda (2003) attempted to determine which factors discriminate between homeless, substance misusing veterans who were non-suicidal, those who had suicidal thoughts, and those who had attempted suicide, this study is also reported above. A number of factors were found to be protective among those who were non-suicidal homeless substance abusers, including employment.

3.119 Cooperman and Simoni (2005) (reported above in 'having children' section) set out to explore the prevalence, timing, and predictors of suicidal ideation and attempted suicide in a sample (n=207) of HIV-positive women (over 18 years of age) in New York City. Contrary to the authors' original hypothesis, those who were employed were significantly more likely to attempt suicide.

3.120 There is some evidence that employment, especially full-time, has a protective effect against suicide. However, employment was not found to be protective among women who were HIV-positive.

Exposure to suicidal behaviour

3.121 Mercy et al (2001) carried out a case control study in the USA to determine the association between nearly lethal suicide attempts and exposure to the suicidal behaviour of parents, relatives, friends, or acquaintances and to accounts of suicide in the media in a sample of young people (13-34 years of age) who had attempted suicide and non-attempters (n=666). The authors found that, although exposure to the suicidal behaviour of a parent or a non-parent relative was not significantly associated with nearly lethal suicide attempts, both exposure to the suicidal behaviour of a friend or acquaintance and exposure to accounts of suicidal behaviour in the media (i.e., having seen any movies, watched any television shows or videos, read any news articles, or read any books or stories during the 30 days prior to the suicide attempt for case subjects or interview for control subjects) were associated with a lower risk of nearly lethal suicide attempts.

3.122 The authors note that these findings are contradictory to many previous studies which have found exposure to media accounts to be a risk factor and suggest a number of potential explanations for their finding. These include that the study was conducted at a time when the nature of media stories or popular perceptions about suicide had changed from that of earlier research, that at the time of their study media stories may have portrayed suicide in more realistic and less glamorous terms, that they examined the effects of media exposure over a 30-day interval, rather than immediately after exposure which may have more of a risk impact or that suicide attempters are more socially isolated than other people and may be likely to be exposed to suicide models in their social networks or in the media. It is important to recognise that there is a growing body of research based evidence of the risk factors associated with media reporting of suicidal behaviour such as the review by Stack (2003) 'Media Coverage as a Risk Factor in Suicide' and that by Hawton and Williams (2001) 'The connection between media and suicidal behaviour warrants serious attention.' Unfortunately these reviews were not reported as systematic reviews and could not be critiqued as part of the review of reviews for risk factors in this report. The Hawton review, although focussed on the risks of media reporting does indicate that responsible approaches to the portrayal of suicidal behaviour in the media, that is, voluntary restraints on reporting suicides by specific lethal methods have been shown to result in statistically significant reductions in deaths by those methods can save lives, a finding that is not at odds with that of Mercy et al.

3.123 One study found that exposure to accounts of suicidal behaviour in the media and, to a lesser extent, exposure to the suicidal behaviour of friends or acquaintances may be protective against nearly lethal suicide attempts. However, it is important to note that there is also a body of evidence of the suicide risks associated with media reporting.

Social values

3.124 Lam et al (2004) examined specific individualistic and traditional values in the context of suicidal ideation and behaviours in Hong Kong among adolescent youths (14-18 years) through a cross sectional school survey (n=2427). Among both boys and girls who attempted suicide but did not require medical attention, those who endorsed individualistic values were less likely to make a serious suicide attempt. However, when controlling for family relationships and symptoms of depression, the value of individualism had no influence on girls' suicidal behaviour, while the value was associated with lower rather than higher risk among boys. The reverse was true for the traditional values of obedience and respect for elders, which was protective among girls but not among boys.

3.125 Traditional social values may have a protective effect against suicidal behaviour among adolescent girls, while individualistic values may have a protective effect among adolescent boys.

Access to treatment by a health professional

3.126 Chesley and Loring-McNulty (2003) (also reported above) found that when their sample was asked 'Has someone or something made a difference in keeping you alive?' 15% of the sample stated that treatment by a health professional was one of a number of factors that protected them against repeat suicide attempts.

3.127 Access to treatment by a health professional may be protective against repeat suicide attempts.

Importance of recognising the presence of multiple protective factors

3.128 Fergusson (2003) examined factors that influenced both the vulnerability and resiliency to suicidal ideation and suicide attempt among depressed young people (15-21 year olds) with depressive disorders in New Zealand (n=1063). Participants were selected from those involved in Christchurch Health and Development Study, which was a longitudinal birth cohort followed over a 21 year period. The majority of their sample did not go on to attempt suicide, leading the authors to conclude that vulnerability and resiliency to suicidal behaviour among their participants was influenced by a complex interplay of factors. These included having a family history of suicide, a history of childhood sexual abuse, personality factors, peer affiliations and success at school. Positive aspects of these factors appeared to enhance resiliency, whereas negative configurations increased vulnerability.

3.129 The interplay between a number of risk and protective factors at individual and psychosocial levels must be taken into consideration when attempting to understand which factors promote resiliency and vulnerability to suicide and suicidal behaviour.