* # 01 "2""%3

!"#$$%#&' ('% !)* + + ,-"*-.'%/*# 01*"2*""%3""*-%"*""$"45

1 23

Your article is protected by copyright and all rights are held exclusively by SpringerVerlag. This e-offprint is for personal use only and shall not be self-archived in electronic repositories. If you wish to self-archive your work, please use the accepted author’s version for posting to your own website or your institution’s repository. You may further deposit the accepted author’s version on a funder’s repository at a funder’s request, provided it is not made publicly available until 12 months after publication.

1 23

Author's personal copy Soc Psychiatry Psychiatr Epidemiol (2012) 47:1–9 DOI 10.1007/s00127-010-0306-z

ORIGINAL PAPER

The profile of suicide: changing or changeable? Fiona Judd • Henry Jackson • Angela Komiti Richard Bell • Caitlin Fraser

•

Received: 24 February 2010 / Accepted: 18 October 2010 / Published online: 30 October 2010 ! Springer-Verlag 2010

Abstract Purpose The aims of this study were to: (1) examine the role of psychosocial factors, physical and mental health in suicide; (2) to examine gender differences on those variables; and (3) determine whether there was a group who died by suicide who did not have a history of mental illness. Method Data were obtained from The Australian National Coroners Information System (NCIS) for all deaths classified as suicides from 2000 to 2004 in all Australian states. The NCIS is an internet-based system for storing and retrieving data on coronial cases. Results The overall results from the total sample reinforces many previous findings but also found some differences; importantly, psychiatric morbidity was less than generally reported, and comparable proportions of males and females used violent means to suicide. Using latent class analysis the study identified four clusters of people who had suicided. In two of those clusters mental illness appeared to be a significant factor; in one of those two clusters the mental illness was compounded by additional

drug and alcohol and relationship problems whilst the other was without such levels of comorbidity. The third group was predominantly male, older and physical illness seemed to be a significant factor. The final group was characterised by low rates of mental illness and treatment for the same, but marked by relationship and financial difficulties. Conclusions These data may suggest that the profile of suicide is changing or changeable. Certainly there has been a shift in the gender profile with comparable proportions of women and men. Whilst mental illness remains a major risk factor, perhaps greater emphasis needs to be placed on the broader psychosocial issues which may initiate or hasten the pathway to suicide. In addition, it may be that the relative contribution of mental illness and other factors is fluid in relation to both life stage and life circumstances. Suicide prevention programmes might usefully define a range of discrete areas of work. Keywords Suicide ! Mental illness ! Gender ! Psychosocial

Introduction F. Judd Centre for Women’s Mental Health, Royal Women’s Hospital, Parkville, VIC, Australia F. Judd ! A. Komiti (&) Department of Psychiatry, University of Melbourne, Parkville, VIC 3052, Australia e-mail: [email protected] H. Jackson ! R. Bell Department of Psychological Sciences, University of Melbourne, Parkville, VIC 3010, Australia C. Fraser Bendigo Health Care Group, Bendigo, VIC, Australia

Suicide occurs in approximately 16.7 per 100,000 persons per year, and is the 14th leading cause of death worldwide [1]. Whilst suicide rates vary between countries, two important consistencies have been noted; there is a higher rate of suicide among men than women, and the global rate of suicide increased between 1950 and 2004, especially for men [2, 3]. Studies of suicide vary in their objectives, scope and methodology. Many studies have attempted to identify risk factors for suicide. Broadly, areas of interest have included social and interpersonal factors, physical and mental health

123

Author's personal copy 2

and availability of means. Early studies emphasised sociological explanations of suicide. Durkheim’s three main aetiological types of suicide-altruistic, egoistic and anomic, representing different measures of the degree of integration in society, are a well-known example of this approach [4]. More contemporary research has focused on the stressdiathesis approach; identifying both the diathesis of longitudinal factors which either raise or lower the threshold to suicidal behaviour as well as stressors or triggers which precipitate that behaviour. Consistently identified risk factors include male sex, being an adolescent or older adult, past suicide attempt, stressful life events, presence of a psychiatric disorder, drug and alcohol abuse and access to lethal means [5–7]. Although the link between these factors and their relative importance remains poorly understood, most studies have focused on delineating the role of clinical aspects of suicide. Psychological autopsy studies have been widely used to identify the cause of suicide. Overall, they indicate that 80–90% of people who die by suicide had a diagnosable psychiatric disorder at the time of suicide [8]. Whilst these studies have been informative, they are beset by significant methodological problems [9]. First, case definition of suicide varies between countries. Second, whilst limited information may be available for all cases, more detailed information (from relatives, general practitioner and other health services) may only be available for 50–60% of cases [9]. Assessment of the presence of a psychiatric disorder or personality disorder by interview of relatives and friends may be unreliable for a variety of reasons, most notably through recall bias and/or the informant being unaware of certain factors or deliberately withholding information which may cast the deceased in a poor light [9]. In addition, these studies often involve small numbers of cases [8], or focus on particular groups, e.g., adolescents [10] or young adults [11]. An alternative approach is to examine a larger data set in a less detailed manner, such as using national databases. For example, Mortensen et al. [12] identified cases of suicide from the Danish medical registers on vital statistics which contains dates and causes of all deaths in Denmark and then linked person-identifiable data from two additional population-based registers to examine demographic, socioeconomic and psychiatric (admission status and

Soc Psychiatry Psychiatr Epidemiol (2012) 47:1–9

psychiatric diagnoses derived from a psychiatric case register) variables to explore risk factors for suicide. Using this approach, these researchers found the strongest risk factor was mental illness necessitating hospital admission. The aims of this study were to use a national database: (1) to examine the role of psychosocial factors, physical and mental health in suicide; and (2) to examine gender differences on those variables. A third aim was to identify clusters of people within the group to examine particular profiles in terms of demographics, psychosocial factors and mental and physical health factors. In particular, we sought to determine whether there was a group who died by suicide who did not have a history of mental illness. We thought this important as some have argued (e.g., [13]) that the role of mental illness in suicide is overstated.

Method An application was made to access The Australian National Coroners Information System (NCIS) for all deaths classified as suicides from 2000 to 2004 in all Australian states. The NCIS is an internet-based system for storing and retrieving data on coronial cases and is managed by the Victorian Institute of Forensic Medicine (VIFM). Since July 2000, information on all deaths reported to an Australian coroner has been recorded in the database for all states, with the exception of Queensland which started in January 2001. A license was granted to the authors to access all suicide cases for the required period for all states except Western Australia. Only the cases that were ‘‘closed’’, i.e., investigations had been completed, were considered for the purposes of this paper. Sociodemographic information and information regarding mechanism of death (Table 1) was available for each case. Some cases also had police and/or coroner’s reports available. These reports contained information on forensic (e.g., whether the deceased had a police record) and psychiatric histories, e.g., whether the deceased had past contact with psychiatric services, had been formally diagnosed with a mental illness, was currently receiving psychiatric treatment, or had a history of substance abuse. The Victorian data set was more complete than that from other states in that 99% of the cases listed had a

Table 1 Mechanism of death Blunt force, e.g., jumping from a height, person moving in front of a moving object such as train or motor vehicle, motor vehicle crash Piercing, penetrating force, e.g., shot with firearm, stabbed with knife or other sharp instrument Threat to breathing, e.g., hanging, suffocation by putting a plastic bag or pillow over one’s head, intentional drowning Exposure to chemical or other substances, e.g., intentional poisoning with tablets or by gaseous substances such as inhaled motor vehicle exhaust gasses Other mechanism of injury, e.g., electrocution

123

Author's personal copy Soc Psychiatry Psychiatr Epidemiol (2012) 47:1–9

coroner’s finding report attached and 90% also had a police report available. Importantly, these cases had thus all been subjected to the Victorian coronial process. This involves three stages in which qualitatively different types of information concerning the nature of the death are collected. First, basic demographics are recorded and information is collected about the circumstances of the death. Second, information such as medical history, autopsy report, and toxicology report is collected to determine the medical cause of death (e.g., lack of oxygen). Third, information is collected to determine the legal cause of death (e.g., lack of oxygen resulted from the deceased hanging themselves). It is during this stage that issues surrounding the intent of the deceased are examined. The Victorian coroner uses the words ‘intentionally and knowingly’ when they are convinced (beyond reasonable doubt) that the deceased did intend to take their own life (Victorian State Coroner’s Office). Whilst this minimises the chance of finding a case to be a suicide when it is not, it increases the chance of failing to recognise some cases of suicide. Table 2 Variables of interest studied in Victorian sample only

3

These Victorian data were analysed in more detail. A list of variables of interest, reflecting identified risk factors for suicide, was generated from conducting a review of the literature; these are listed and defined in Table 2. Three research assistants read the police and/or coroner’s reports for each case (where available) and extracted and recorded the presence or absence of each variable. The reports and the extracted variables were then reviewed by two of the authors (AK, FJ) to check the accuracy of the data.

Results Total sample The total sample comprised 8,244 individuals; 6,449 (78.2%) males and 1,795 (21.8%) females. One-third (33.3%) of the sample were married or in a de facto relationship, 28.3% had never been married and a smaller per cent were divorced (6.3%), separated (8.9%) or widowed (3.6%). More than one-third were employed (39.2%),

Variable

Definition

Past suicide attempt

Documented one or more previous suicide attempts (lifetime)

Diagnosis of mental illness

Formal diagnosis of mental illness, axis 1 or axis 11 made (lifetime)

Currently receiving treatment for mental illness

Receiving treatment from a primary care and/or specialist mental health practitioner/service for a mental illness prior to death Received treatment from a primary care and/or specialist mental health practitioner/service for a mental illness at some time prior to current treatment provision or current problems

Received treatment for mental illness in the past Living alone

Living alone prior to death

Relationship breakup/difficulties

Recent relationship breakup or difficulties prior to death (includes death of partner) Observed by family or friends to appear sad or low prior to death

Observed by others to be depressed History alcohol abuse

History of using alcohol to excess

History drug abuse

History of using illicit drugs

Forensic history

Past forensic history including past incarceration

Forensic pending

Police charges and/or waiting trial at time of death

Current physical illness

Any current physical illness at time of death

Receiving any medical treatment Financial problems

Receiving treatment from a primary care and/or specialist health care provider for a medical condition prior to death Financial problems prior to death

Gambling problems

Gambling problems prior to death

Previously threatened suicide

Had threatened to suicide prior to death (includes both immediately prior to death and at a more distal time-point)

Informed others of suicidal thoughts

Had informed others of suicidal thoughts prior to death

Informed others of clear suicidal plan

Had communicated a clear suicidal plan to others prior to death

Left a suicide note

Written suicide note, email, or SMS message

Alcohol use at the time of suicide

Toxicology report showed alcohol use close to time of death

Marijuana use at the time of suicide

Toxicology showed marijuana use close to time of death

123



one-fifth (20.8%) were unemployed and 23.3% were retired or a pensioner. The most common cause of death (47.1%) was threat to breathing, followed by poisoning (31.3%), and death by blunt force (17.3%). Death by penetrating force occurred in 2.2% of the sample. Chi-square analyses were performed to examine differences in categorical data. Differences were considered as reaching significance level (i.e., P \ 0.05) if the adjusted residuals C1.9. There were gender differences with regards to sociodemographic variables. Males (M = 42.7 years, SD 17.2), were younger than females (M = 43.8 years, SD 17.4), t(8,230) = 2.21, P = 0.03. Although the difference was significant at the 0.05 level, the effect size was negligible (Cohen’s d = 0.06). There were significant differences between males and females with regards to marital status (v2(4) = 1.16, P \ 0.0001). In this study sample, females were more likely than males to have been: married (44.2 vs. 40.6%), divorced (10.3 vs. 7.2%), and widowed (8.5 vs. 3.4%), and less likely than males to have been single (28.0 vs. 37.2%) or separated (9.0 vs. 11.6%). There were also significant gender differences (v2(5) = 8.91, P \ 0.0001) with regards to employment status. A greater proportion of males (46.7%) than females (30.2%) were employed at the time of death. More females were involved in home duties (13.5 vs. 0.1%), more women were retired/pensioners than men (28.9 vs. 24.8%), and more women than men were students (7.0 vs. 3.7%). There were also clear differences in the cause of death between males and females (v2 = (8) = 1.27, P \ 0.0001). Threats to breathing (M:F, 48.6 vs. 42.0%) and blunt force (M:F, 18.5 vs. 13.1%) were more common in males; poisoning was more common in females (M:F, 28.5 vs. 41.3%).

Table 3 Sociodemographic characteristics for males compared to females in the Victorian sample

Victorian data As described above, more extensive information about individuals who had died was available for those who had been resident in Victoria (n = 2,137; 1,617 males and 520 females). There were significant differences between the Victorian group and the other states with regards to proportion of gender as well as marital and employment status. There were more males in the other states compared to the Victorian sample (79.1 vs. 75.6%) but more females were represented in the Victorian sample (24.4 vs. 20.9%). The Victorian group were also different from the other states with regards to marital status, with a higher proportion of the former being married (50.8 vs. 39.0%) and less likely to have been divorced (4.3 vs. 8.8%), or single (18.2 vs. 39.4%). With regards to employment status, significantly more Victorians in this study sample had been employed at the time of death than those from other states (48.5 vs. 41.2%) and were less likely to have been unemployed (17.8 vs. 24.8%). Sociodemographic differences between males and females were then explored within the Victorian sample (see Table 3 below). As shown in Table 3, there was no difference as regards the age of males and females and there were few differences between males and females with regards to marital status. Similar proportions of males and females were divorced, married, or single, although significantly more males were separated (22.3 vs. 17.4%) at the time of death compared to females. In addition, a significantly lower proportion of males at time of death were widowed, compared to females (4.7 vs. 8.3%). In terms of

Variable

Males (n = 1,617) %

Females (n = 520) %

v2 or t test

Mean age in years (SD)

43.4 (17.5)

43.7 (16.6)

0.27

0.79

9.48

0.05

1.96

\0.001

P

Marital statusa Married/de facto

50.2

52.6

Divorced

4.1

4.9

Separated

22.3

17.4

Widowed

4.7

8.3

18.7

16.8

Employed

52.2

37.0

Unemployed

18.3

16.3

Retired/pensioner Home duties

24.8 0.2

28.5 12.2

4.4

6.1

Never married Employmenta

a

The numbers do not always add up to 100% because of missing data

123

Student


5

employment status at the time of death, significantly more males were employed, compared to females (52.2 vs. 37.0%). Cause of death for males and females in the Victorian study group (see Table 4 below) reflected cause of death in other states. Cause of death for males was more likely to be by penetrating force (2.5 vs. 0.8%) whereas females were more likely to die by poisoning (41.9 vs. 33.9%). These differences were statistically significant (v2 = (4) = 15.57, P = 0.004). Clear gender differences in identified risk factors for suicide were evident in this study group (Table 5). More women than men had a diagnosis of mental illness, were currently receiving treatment and/or had received treatment in the past for mental illness, or had previously attempted

Table 4 Cause of death for males and females who died by suicide in the Victorian sample Cause of death

Males Females Total (n = 1,617) % (n = 520) % (n = 2,137) %

Threats to breathing 44.0

40.4

43.1

Poisoning

33.9

41.9

35.8

Blunt force Penetrating force

17.6 2.5

15.2 0.8

17.0 2.1

1.9

1.7

1.9

Other

suicide. Women were less likely than men to have a history of alcohol abuse, alcohol use at the time of death, forensic issues or financial problems. Establishing groups from the Victorian data set There was an attempt to identify groupings from the Victorian data set and the application of latent class analysis derived four groups. Using Mplus version 5 [14] we fitted two through five latent class models to the data. Table 6 shows the fit indices. We used several statistical fit indices: likelihood ratio chi-squared (LRv2), Akaike information criterion [15], Bayesian information criterion (BIC [16]), and sample-size adjusted BIC (SSABIC [17]). We also used the Lo–Mendell–Rubin’s adjusted LR test (LRT [18] to compare competing models, a non-significant value (where P = 0.05) would suggest that the model with one fewer classes should be accepted. Thus, although all models fitted the data, the three-class model fitted better than the two-class and the four-class model fitted better than the three, but the five-class model did not fit better than the four-class model. The conditional probabilities are shown for each variable for each of the four groups in Table 7. It can be seen from inspection of Table 7 that the conditional probabilities are low across all four groups for a number of

Table 5 Victorian sample—social and clinical circumstances of suicide for males and females Males (n = 1,617) %

Females (n = 520) %

Total (n = 2,137) %

v2

P


20.2

39.4

24.9

77.59

\0.001


37.3

64.4

43.9

1.18

\0.001

Currently receiving treatment for mental illness Received treatment for mental illness in the past

36.6 24.7

60.6 45.0

42.4 29.7

92.51 77.42

\0.001 \0.001

Living alone

9.2

10.2

9.4

0.50

0.48

32.1

28.3

31.2

2.69

0.10

Observed by others to be depressed

35.1

31.9

34.3

1.79

0.18

History alcohol abuse

14.6

9.6

13.4

8.42

0.004

History drug abuse

14.5

12.7

14.1

1.10

0.29

Forensic history

2.4

0.4

1.9

8.59

0.003

Forensic pending

5.9

2.7

5.1

8.48

0.004

25.0

24.6

24.9

0.03

0.87

Receiving any medical treatment

4.4

3.3

3.6

1.51

0.22

Gambling problems

2.8

2.1

2.6

0.69

0.41

Relationship difficulties


Financial problems

10.1

4.8

8.8

13.89

\0.001


13.9

14.6

14.0

0.19

0.66


19.0

18.5

18.9

0.09

0.77

4.9

4.8

4.9

0.02

0.90

40.9 25.1

46.5 18.3

42.3 23.4

5.17 10.25

0.02 0.001

4.1

3.7

4.0

0.24

0.62

Informed others of a clear suicidal plan Left a suicide note Alcohol use at the time of suicide Marijuana use at the time of suicide

123



Table 6 Fit indices for two through five latent class models for the Victoria suicide data Model

LRv2

P

AIC

BIC

SSABIC

LMR LRT

2

4,353

1.00

35,305

35,548

35,412

3

4,632

1.00

35,053

35,421

35,215

4

4,087

1.00

34,869

35,362

5

4,048

1.00

34,774

35,392

Table 7 Latent class-derived groups in Victorian sample

P

Entropy

2,487

0.000

0.929

294

0.000

0.826

35,085

227

0.003

0.798

35,045

172

0.462

0.769

Variable

1

2

3

4

N in each cluster

921

232

282

702

% of people in each cluster

43.1

10.9

13.2

32.8


0.132

0.094

0.472

0.348


0.000

0.024

0.874

0.952

Observed by others to be depressed

0.375

0.496

0.310

0.267

Relationship difficulties

0.408

0.120

0.445

0.181

Alcohol use at the time of suicide

0.290

0.095

0.482

0.086

Marijuana use at the time of suicide

0.051

0.019

0.096

0.006

Living alone

0.084

0.156

0.087

0.090


0.132

0.289

0.273

0.191

Informed others of a clear suicide plan

0.039

0.034

0.087

0.050


0.113

0.107

0.255

0.133


0.101

0.901

0.215

0.245

Gambling problems

0.035

0.015

0.053

0.005

Financial problems Currently receiving treatment for mental illness

0.130 0.051

0.040 0.114

0.103 0.805

0.042 0.841

Received treatment for mental

0.029

0.038

0.565

0.611

Illness in the past Receiving any medical treatment

0.004

0.178

0.034

0.031

History of drug abuse

0.113

0.000

0.427

0.085

History of alcohol abuse

0.119

0.022

0.501

0.012

Suicide note

0.464

0.485

0.326

0.393

Forensic history

0.023

0.000

0.053

0.004

0.087

0.009

0.068

0.010

Male (%)

Forensic pending

86

80

72

62

Median age in years

38

66

36

42

variables, e.g., the forensic variables and having a detailed suicide plan. From Table 7 it can be seen that Group 1 is the largest of the four groups (43.1% of the total Victorian sample), has a median age of 38 years and is 86% male. The conditional probabilities are low for most variables. The most striking findings are that 47% left a suicide note, for 41% a relationship breakup/difficulties was reported, 13% had financial problems, alcohol use at the time of suicide was reported for 29 and 38% were observed to be depressed. Even more striking is the total lack of a formal diagnosis of mental illness (0%) and the low rates of treatment for mental illness (3–5%).

123

The second group is made up of 10.9% of the total sample and is 80% male. It is much older than the other three groups with a median age of 66 years. The striking features for this group are the high percentage with physical illness (90%); 18% were receiving medical treatment—higher than in the other three groups. Exactly half (50%) were observed by others to be depressed, 29% had informed others of suicidal thoughts, and 49% left a suicide note—a rate similar to that seen in the first group. Other factors that are noteworthy are the very low rates of a formal diagnosis of mental illness (2%), and treatment for mental illness (4–11%).


The third group was made up of 282 people, i.e., 13.2% of the total Victorian group. The group had a mean age of 36 years and was 72% male. This group is best described as a ‘mentally ill’ group with a striking 87% formally diagnosed with mental illness, 81% currently receiving treatment for mental illness, and 57% having ever received treatment for mental illness. Nearly half (47%) had made a previous suicide attempt and 43–50% had a history of drug abuse or a history of alcohol abuse. Alcohol use prior to suicide was identified in 48% and a relationship breakup/ difficulties was reported in 45% of the group. The fourth group is made up of 702 individuals constituting 32.8% of the total sample. It has a median age of 42 years and is 62% male. It is also a ‘mentally ill’ group and in some ways looks very similar to the third group with a high percentage of members carrying a formal diagnosis of mental illness (95%), a high percentage being treated for mental illness (61–84%). But it differs dramatically from the third group with a low percentage reported for relationship break-up/difficulties (18%) and alcohol use prior to suicide (9%). Even more noticeable are the very low percentages reported for either a history of drug abuse (9%) or alcohol abuse (1%).

Discussion The total sample was predominantly male (78.2 vs. 21.8%) consistent with the male: female ratio generally reported for suicide [19]. The most common cause of death was threat to breathing, followed by poisoning. Death due to blunt force (such as jumping from a height or moving in front of a car or train) occurred in approximately 1 in six people with comparable frequency among men and women dying by suicide. Death due to penetrating force was less frequent, consistent with the decreased rate of death due to firearms [19]. Consistent with previous findings [8, 11, 12] being unemployed, not married or in a de facto relationship, previous suicide attempt, physical illness, and alcohol or drug abuse were common amongst those who died by suicide. Stressors including relationship difficulties, financial problems and pending forensic issues were also common. There were two striking findings. First, psychiatric morbidity was less than that generally reported [8, 20]. Less than half of those who died had a diagnosis of mental illness or were currently receiving treatment for mental health problems at the time of death. Less than one-third of the sample had received treatment for mental health problems in the past. Second, contrary to many reports, comparable proportions of women and men who died by suicide, used violent means (e.g., threats to breathing and blunt force). This is

7

consistent with findings reported in an earlier study by Kposowa and McElvain [21]. Previously, the male preponderance of use of violent means of suicide has been attributed to greater suicidal intent, aggression, knowledge regarding violent means, and less concern about body disfigurement [3]. These data suggest an important change in suicidal behaviour amongst women. Using latent class analysis we identified four groupings in which the role of physical and mental illness in suicide appear to differ. Two groups were ‘familiar’ with histories of mental illness (Groups 3 and 4). In Group 3 the male: female ratio approximated that of the total sample. This group also had high rates of alcohol and drug abuse. In contrast, Group 4 had a greater than expected proportion of females. The most likely reason is that this group had very low percentages of members with a history of alcohol and drug abuse. The third group (Group 2) was predominantly male and of older age (66 years) with high levels of physical illness and treatment for physical illness. Half were observed to be depressed and 29% informed others of suicidal thoughts. It is recognised that depression is common in the medically ill, and the problems of detection and under-diagnosis of depression in this population are well known [22, 23]. The final group (Group 1) is the most striking given they do not appear to have mental illness and have not come to the attention of mental health practitioners. They have high rates of stressors including relationship difficulties, financial problems and forensic issues. Almost one-third used alcohol at the time of suicide. This group challenges our emphasis on mental illness as a key determinant of suicidal behaviour and reminds us that social factors have long been regarded as important in the causation of suicide. For example, could this group reflect what Durkheim [4] termed anomic or egoistic suicide? Durkheim [4] proposed that suicide was associated with the degree of social integration in society. Anomic suicide was the result of a sudden and unexpected change in a person’s social position creating a new situation with which s/he is unable to cope. Contemporary examples might include financial ruin or some ‘shameful’ forensic charges such as viewing of child pornography. Egoistic suicide resulted when an individual is not properly integrated into society, with the suicidal act occurring when the degree of social integration is weakest. For example, an unemployed male, estranged from his extended family, who experiences a relationship breakup. These data may suggest that the profile of suicide is changing or changeable. Certainly there has been a shift in the gender profile. In this study, comparable proportions of men and women who died by suicide used violent means. It should, however, be noted that given the overall higher rate of suicide in males, women are far less likely to die in this

123


manner than men. Whilst mental illness remains a major risk factor, perhaps greater emphasis needs to be placed on the broader psychosocial issues which may initiate or hasten the pathway to suicide. In addition, it may be that the relative contribution of mental illness and other factors is fluid in relation to both life stage and life circumstances. So when we consider the early intervention and prevention paradigm [24], the focus should not just be on symptoms, or severity or duration of symptoms and prevention of psychiatric disorder but also on relationship problems and financial issues which may act as catalysts for symptom emergence or in fact bypass symptom emergence altogether and lead the person to taking a more impulsive and rapid course of action. Clearly ensuring aggressive follow-up of people with a mental illness, ensuring treatment is optimal, and that such people are compliant with attendance for appointments with medical and mental health staff, their medication and therapies remains a key intervention for suicide prevention. Alcohol and abuse histories as found in Group 3 require special attention. However, a range of additional interventions are required. The first is improving access to care for those who are experiencing mental health problems but who are not receiving treatment. Strategies include increasing mental health awareness through mental health literacy, i.e., universal public health messages (including pictorial messages using male actors), raising awareness of depression and suicidality particularly in older males and especially in those with physical illness, and talking about the appropriate avenues of care and ensuring people assist their ‘at risk’ friends and family members into appropriate care. This must be followed with action at the primary care level. Here we would ask GPs to pay special attention to screening for depression and suicidality and ensuring that those with depression are treated optimally with antidepressants or medication, or referred for psychotherapies of choice, e.g., CBT or IPT. Second, attention is needed to address what seems to be increasing acceptability of violent (and lethal) means of suicide. Availability of these means (e.g., jumping from a bridge, standing in front of a train, hanging) is difficult to control (compared for example, with firearms legislation), so the focus needs to be media and other influences to combat acceptability. Third and most challenging, is targeting individuals such as those in our Group 1. Their distress that results in suicide may be related to longstanding psychosocial issues (e.g., financial problems) which could be addressed with the appropriate resources. This assumes these individuals recognise the need for assistance, that this assistance is available and are prepared to seek this. More problematic are the ‘unexpected’ social problems (e.g., relationship

123


break-up, forensic problem). Here, a universal approach to suicide prevention might afford the best way forward. This would take the form of health promotion messages amongst the general public emphasising the need for care, the availability of, and appropriate pathways to care. Important limitations of this study were that the survey did not allow a fine-grained approach to the problem. The variables used were coarse, and were collected at a single time-point. As with psychological autopsy studies [9] the accuracy of reported information is, in some case, uncertain. For example, reporting that a person appeared depressed assumes that the informant knew the deceased individual well enough to recognise that they might be depressed, and that the informant is able to detect depression. The cluster analysis was applied to data from a single state of Australia and we cannot be sure that the same four clusters would be obtained in other states of Australia; it is important in this regard to note that there were differences between Victoria and other Australian states on the variables of gender, marital status and employment. It is also important to note that this study is not meant to be an epidemiological piece but rather attempts to describe the characteristics of a definitive data set of suicide cases within a circumscribed time period in the state of Victoria.

Conclusions The overall results from the total sample reinforces many previous findings but also found some differences; importantly, psychiatric morbidity was less than generally reported, and females in this study sample, were just as likely as men to use violent means to suicide. Using latent class analysis the study identified four clusters of people who had suicided. In two of those clusters mental illness appeared to be a significant factor; in one of those two clusters the mental illness was compounded by additional drug and alcohol and relationship problems whilst the other was without such levels of comorbidity. The third group was predominantly male, older and depression seemed to be a significant factor as did physical illness. The final group was characterised by low frequency of mental illness and treatment for the same, but marked by relationship and financial difficulties. We suggest that different groups may require different courses of intervention and outline some courses of action consistent with the early intervention and prevention paradigm. Acknowledgments We wish to acknowledge the following organisations that supported the study: Australian National Coroners Information System (NCIS) and Bendigo Health Care Group.


References 1. World Health Organisation (2007) Suicide prevention (SUPRE). http://www.who.int/mental_health/prevention/suicide/suicideprevent/ en/ 2. Bertolote JM, Fleischmann A (2002) A global perspective in the epidemiology of suicide. Suicidology 7:6–8 3. Hawton K (2000) Sex and suicide. Gender differences in suicidal behaviour. Br J Psychiatry 177:484–485 4. Durkheim E (1951) Suicide: a study in sociology. Free Press, New York 5. Harris EC, Barraclough B (1997) Suicide as an outcome for mental disorders: a meta-analysis. Br J Psychiatry 170:205–228 6. Moscicki EK (1997) Identification of suicide risk factors using epidemiological studies. Psychiatr Clin N Am 20:499–517 7. Nock MK et al (2008) Suicide and suicidal behaviour. Epidemiol Rev 30:133–154 8. Cavanagh JTO et al (2003) Psychological autopsy studies of suicide: a systematic review. Psychol Med 33:395–405 9. Hawton K et al (1998) The psychological autopsy approach to studying suicide: a review of methodological issues. J Affect Disord 50:269–276 10. Houston K, Hawton K, Shepperd R (2001) Suicide in young people ages 15–24: a psychological autopsy study. J Affect Disord 2001:159–170 11. Appleby L et al (1999) Psychological autopsy studies of suicide by people aged under 35. Br J Psychiatry 175:168–174

9 12. Mortensen PB et al (2000) Psychiatric illness and risk factors for suicide in Denmark. Lancet 355:9–12 13. House A (2008) The first cut is not the deepest: understanding the link between self-harm and suicide. Australasian Society for Psychiatric Research Conference, Newcastle 14. Muthe´n LK, Muthe´n BO (2007) Mplus user’s guide, 5th edn. Muthen & Muthen, Los Angeles 15. Akaike H (1987) Factor analysis and the AIC. Psychometrika 52:317–332 16. Schwartz G (1978) Estimating the dimension of a model. Ann Stat 6:461–464 17. Sclove SL (1987) Application of model-selection criteria to some problems in multivariate analysis. Psychometrika 52:333–343 18. Lo Y, Mendell N, Rubin DB (2001) Testing the number of components in a normal mixture. Biometrika 88:767–778 19. Australian Bureau of Statistics (2008) Causes of death, 2006. Australian Bureau of Statistics, Canberra 20. Mann J et al (2005) Suicide prevention strategies. A systematic review. JAMA 294:2064–2074 21. Kposowa AJ, McElvain JP (2006) Gender, place, and method of suicide. Soc Psychiatry Psychiatr Epidemiol 41:435–443 22. Kelly BJ, Turner J (2009) Depression in advanced physical illness: diagnostic and treatment issues. Med J Aust 190:S90–S93 23. Rodin G, Voshart K (1986) Depression in the medically ill: an overview. Am J Psychiatry 143:696–705 24. Gunnell D, Lewis G (2005) Studying suicide from the life course perspective: implications for prevention. Br J Psychiatry 187:206–208

123

* # 01 *"2*""%3

* # 01 *"2*""%3

* # 01 "2""%3

* # 01 "2""%3