Sample

Ten to Men recruited 15,988 males aged 10–55 years from across Australia in 2013/14 using a stratified clustered random sampling strategy. Males were included if they were resident in a private dwelling, were an Australian citizen or permanent resident and had sufficient proficiency in English to complete the questionnaire/interview. Sampling, recruitment and data collection methods are described elsewhere [6]. For this analysis males 18 years and older were included (n = 13,884) as they are exposed to a different and wider set of life stressors than younger males.

Measures

Lifetime suicidal thoughts and behaviours were ascertained by three standard questions: Have you ever seriously thought about killing yourself?; Have you ever made a plan about how you would kill yourself?; Have you ever tried to kill yourself?. The item ‘Thoughts that you would be better off dead or of hurting yourself in some way’ from the Patient Health Questionnaire (PHQ9) was used to capture suicidal ideation in the past 2 weeks [7], which we deemed to be current suicidal ideation. Respondents indicated the frequency of such thoughts in the past 2 weeks as ‘not at all’, ‘several days’, ‘more than half the days’, and ‘nearly every day’. Self-report psychiatric illness was ascertained by two questions: ‘Has a doctor or other health professional ever told you that you had this condition?’ for lifetime prevalence, and ‘Have you been treated for or had any symptoms of this condition in the past 12 months?’ for 12 month prevalence. Mental health conditions available for endorsement were depression, post-traumatic stress disorder (PTSD), other anxiety disorders, and schizophrenia. This format for capturing self-reported long-term conditions was based on the Australian Health Survey [8]. Alcohol use was assessed using the Alcohol Use Disorders Identification Test (AUDIT) in which harmful or hazardous alcohol use is defined as an alcohol dependence score of 8 or more [9]. Life events in the prior 12 months were ascertained using a life events checklist comprising 24 items. The majority of the items (19) had been used in the Australian Longitudinal Study of Women’s Health surveys, based on their adaption of the Social Readjustment Scale [10]. Modifications included two instances of combining multiple items into one (serious personal injury, illness or surgery; break-up of a serious relationship/divorce/separation). Two items were modified to be more general (being pushed, grabbed, shoved, kicked or hit modified to victim of physical violence, being forced to take part in unwanted sexual activity modified to being a victim of sexual assault) one item broadened in scope (loss of a job to loss of a job or looking for work unsuccessfully for a long time) and one revised to be male specific (birth of your first child to becoming a father for the first time). Five additional items were included, four to capture important transition points: starting your first job, retirement, bankruptcy (personal or business), buying a house, becoming a carer for someone, and one from the Social Readjustment Scale (serious illness, injury or assault to a close relative/friend) [11]. There was no rating of the severity of stress associated with the event.

Participant socio-economic status was based on the Australian Bureau of Statistics, Socio-Economic Indexes for Areas - Index of Relative Socio-Economic Disadvantage (SEIFA-IRSD). This is general socio-economic index produced from census data summarising a range of indicators of economic and social disadvantage [12]. Lower IRSD scores indicate greater disadvantage.

Statistical analyses

Frequencies for clinical factors and suicidal behaviour were produced using data weighted to the general population age and regional distribution (for details on design and use of sample weights see Spittal et al. [13]). All 24 life events variables were entered into a logistic regression model with current suicidal ideation as the outcome variable, and socio-demographics (10 year age group, marital status, socio-economic status), and clinical factors (12 month depression, 12 month anxiety and 12 month harmful/hazardous alcohol use) as covariates. This regression did not adjust for the sample design or for sample weighting. Life events that were not significant at the p < 0.05 level were removed and the model was re-estimated.

To examine stress/diathesis effects we constructed variables representing combinations of stressors and putative elements of the diathesis (12 month depression, 12 month anxiety, 12 month harmful/hazardous alcohol use). We constructed three variables representing each diathesis element. Each variable could take one of four values and had the form: (1) no stressors present and no diathesis present; (2) stressors present but no diathesis present; (3) no stressors present but diathesis present; and (4) both stressors and diathesis present. We defined the stressor variable as the participant reporting “yes” on any of the life events that remained significant in the multivariable model described above. We then conducted separate logistic regressions for stressor/depression, stressor/anxiety and stressor/alcohol use with current suicidal ideation as the outcome variable (also without adjustment for sample design or weighting). We also included the socio-demographic variables described above and the two other clinical variables not included in that model’s ‘diathesis’ as covariates.

Prevalence of suicidal behaviour, depression, anxiety and alcohol use

Suicidal ideation and life events

Stress-diathesis

Suicidal behaviour

The rates of suicidal thoughts and behaviour in Ten to Men adult respondents are higher than reported in the 2007 National Survey of Mental Health and Wellbeing (NSMHWB), the benchmark population data collection on non-fatal suicidal thoughts and behaviours in adults in Australia. Lifetime suicidal ideation was reported by 18.8 % of Ten to Men adult respondents compared to 11.5 % of males in the NSMHWB, and lifetime suicide attempts were reported at more than twice the rate in Ten to Men adult respondents (5.4 % compared to 2.2 %) [14]. Likewise lifetime and 12 month depression was reported more frequently in the Ten to Men adult cohort (depression: 20 % lifetime and 12.8 % 12 month) than in the NSMHWB (depressive episode: 8.8 % lifetime and 3.1 % 12 month) [14], although anxiety was less frequently reported (Ten to Men: 13.4 % lifetime and 8.8 % 12 month; NSMHWB: 20.4 % lifetime and 10.8 % 12 month) [14]. These differences in rates likely reflect methodological differences including the use of self-report questionnaires in Ten to Men versus face-to-face interview in the NSMHWB and the use of clinical interview to assess depression and anxiety in the NSHWBS compared to a general question about lifetime diagnosis in Ten to Men. While the study design aimed to obtain a representative sample and sample weights adjusting for age and region have been applied to the analysis [13], Ten to Men is not designed as a cross-sectional prevalence study but rather a longitudinal study aimed at investigating the causes of and pathways to health outcomes within individuals. As such, a cohort enriched for suicidal behaviour and psychiatric disorder offers greater scope to investigate the complex causality of suicidal behaviour.

Life stress and suicidal ideation

In the initial multivariable analysis, elements from both the stressor and diathesis domains increased risk for suicidal ideation. Depression, anxiety and alcohol use, all putative elements of the diathesis for suicidal behaviour, were associated with increased risk for suicidal ideation. Depression in the past 12 months was the most robust predictor with more than twice the effect size of any other risk factor in the model. In the stressor domain six of the twenty-four life events assessed made an independent contribution to risk for suicide ideation. While a smaller effect than depression, these life events contributed to risk in the same magnitude as anxiety and harmful/hazardous alcohol use, suggesting that the stressor domain has an important role and is thus a potential target for intervention in suicide prevention.

The six life events associated with increased odds of suicidal ideation were serious family conflict, break-up of a relationship, difficulty finding a job, legal troubles, major loss of property and serious personal injury. Of these, events in the employment and interpersonal domains had somewhat stronger effects. Liu and Miller in their recent systematic review found some evidence that relative to other types of stressors interpersonal difficulties were more often associated with suicidal thoughts although they did not present data specific to males [5]. In this study, the variable nature of the six significant life events and the modest differences in effect sizes between them makes it difficult to identify particular domains or types of life events that are higher risk than others. Moreover, while certain life events were found to be more strongly associated with increased risk for suicidal thoughts than others, the likely collinearity between life events and the low frequencies of certain events (e.g., loss of a child) may have meant that events which are highly stressful were not retained in the final models. Thus no general conclusions can be drawn about the salience of particular life events over others with respect to risk for suicidal thinking. The variability may also indicate that the type of life event is less important to risk for suicidal ideation than the broader context in which any particular life event takes place, including the presence of elements of the diathesis.

While we cannot elucidate causal pathways in a single wave of data, we can investigate if the co-occurrence of a stressor and elements of the diathesis increases risk for suicidal behaviour. In the case of depression, we found this to be the case. The co-occurrence of life events and depression in the past 12 months resulted in an almost four-fold increase in risk for suicidal ideation over life events with no co-occurring depression, as well as a 35 % increase in the odds of suicidal ideation compared to depression in the absence of life events. While this elevated risk for suicidal ideation associated with co-occurring stressors and depression is consistent with a stress-diathesis model, our analysis also indicates that life events plus depression does not fully encapsulate the stress-diathesis pathway. Independent effects for both life stress and depression persisted in the model and, moreover, the magnitude of increase in ORs associated with co-occurring life stressors and depression did not demonstrate the multiplicative relationship as might be expected if life events plus depression alone constituted the stress-diathesis pathway.

The independent risk associated with life events, depression and, to a lesser degree, anxiety and alcohol use, likely reflects the multi-dimensional nature of both the stressor and diathesis domains. Thus, in the absence of depression other diathesis elements such as deficits in cognitive function, maladaptive stress response, aggression, impulsivity etc. not included in this analysis may be pathways to suicidal ideation in response to life stressors. Likewise, in the stressor domain we assessed only a limited range of life events on the more severe end of the spectrum, and among those in the ‘depression alone’ group reporting suicidal thoughts other life stressors may be occurring. Given the multi-determined nature of suicidal ideation and behaviours it is also likely that there are multiple phenotypes and causal pathways. More complex modelling that includes a broader range of stress and predisposing factors is required to elucidate these. However, this limited analysis of more commonly occurring stressors and psychiatric disorders does indicate that the co-occurrence of life stress and depression substantially increases the risk of suicidal ideation. It also indicates a need for further research to identify predisposing factors in non-depressed males which may increase their vulnerability to suicidal ideation and behaviour when encountering fairly common life stressors in order to identify targets for prevention in this group.

Limitations

The outcome measure of suicidal ideation was drawn from a composite scale designed to assess depression rather than from a suicide assessment instrument and, as such, may lack specificity. Moreover, given that suicidal ideation is a symptom of depression, individuals who were positive for current depression were more likely to have endorsed current suicidal ideation and 12 month depression. While this may overestimate the magnitude of the association between depression and suicidal ideation, the finding that several life events remained significantly associated with suicidal ideation after controlling for depression indicates the robustness of that association. Other limitations are those found in life event studies using self-report checklists regarding reliability of recall of past life events [15]. While the relatively short recall period (past 12 months) and the serious nature of the life events assessed may reduce recall error in Ten to Men this remains a limitation of the checklist method for capturing life event information. Moreover, as with all cross-sectional studies a single wave of data can only discern correlations. Thus we cannot rule out reverse causality whereby individuals who are currently depressed and suicidal may differentially recall more adverse events compared to those who are not, as has been reported in some studies, but not all [16, 17]. Finally, this study does not include any assessment of early-life stressors. Early life trauma has been associated with suicidal behaviour [18] and a number of putative pathways implicated including neurodevelopmental deficits resulting in impaired stress response later in life and greater vulnerability to psychiatric disorder [19]. This study focussed on proximal life stressors (past 12 months) as putative ‘triggers’ to current suicidal ideation, however it is likely that for some individuals early-life stressors contribute to the association between suicidal thoughts and life events observed here, and that analyses examining both distal and proximal stressors are required to fully elucidate the relationship between life stress and suicidal ideation. Over its longitudinal course, Ten to Men will be better able to track the timing of life stressors and identify causal pathways between proximal and distal life events, psychiatric illness and suicidal behaviour.