In Ten to Men, we recruited a cohort of 15,988 males aged 10–55 in 2013/14 using a stratified, multi-stage, cluster random sampling strategy. We approached the full complement of households (n = 104,884) in 622 randomly-selected statistical areas (SA1s) from across Australia. SA1s are the smallest unit for which the Australian Bureau of Statistics will release census data, and have populations ranging from about 200 to 800. In total, we made contact with 81,400 (78%) of these households and identified 33,724 (32%) as having at least one male resident of the appropriate age. Ultimately, we identified 45,510 ‘in scope’ males in these households, so the 15,988 who agreed to be part of the cohort and returned usable data represented a response fraction of 35%. For the purposes of the current study, we focused on the 13,884 18–55 year-old men in the cohort.

At recruitment, we collected baseline (Wave 1) data from the 18–55 year-old via self-complete questionnaires. These men were asked questions that covered five broad domains (physical health, mental health and wellbeing, health behaviours, social determinants of health, and health service utilisation and health knowledge).

Our primary outcome was a single item taken from the Patient Health Questionnaire (PHQ-9) [23]. This question asks “Over the past two weeks, how often have you been bothered by thoughts that you would be better off dead, or of hurting yourself in some way?” Responses to this question are: (a) not at all; (b) several days; (c) more than half the days; and (d) nearly every day. Scores are assigned from 0 to 3, respectively. For the purposes of our analysis, we considered those who scored 1 (i.e., were bothered by these thoughts for several days or more) or more to be experiencing suicidal ideation. It is worth noting here that this single PHQ-9 item has been shown to have good specificity and reasonable sensitivity when compared to a structured clinical interview [24], and was recently shown to predict suicide risk in a large-scale register-based study of (predominantly male) Veterans Administration (VA) patients in the United States [25].

We measured individual-level socio-economic position by considering responses to questions on employment status and occupation type. The employment status question asked: “Are you currently: (a) employed/working for profit or pay; (b) unemployed and looking for work; or (c) neither working nor looking for work?” Those who indicated they were employed were asked: “What is your current occupation (in your main job)?” and given the opportunity to provide a free text response. These responses were then coded according to the Australian and New Zealand Standard Classification of Occupations (ANZSCO) by occupational skill level: (a) low (sales, machinery workers, and labourers); (b) medium (technical and trade workers, community and personal service workers, and clerical and admin workers); and (c) high (managers and professionals) [26]. We created a single variable that was designed to represent a gradient of individual-level socio-economic status by combining these responses into the following categories: (a) unemployed; (b) working in low skilled occupation; (c) working in medium skilled occupation; and (d) working in high skilled occupation.

We measured area-level socio-economic position using the Index of Relative Socio-economic Disadvantage (IRSD) [27]. The IRSD is one of four composite indexes in the Socio-Economic Indexes for Areas (SEIFA), created by the Australian Bureau of Statistics to summarise the social and economic conditions of areas (in our case, SA1s) using census information. The IRSD focuses on disadvantage. It is a multidimensional construct that reflects the disadvantage of individuals, families and dwellings within areas. It is constructed using data reduction techniques (principal component analysis) to combine information from a range of variables assessing the prevalence within areas of unemployment, employment in low skilled occupations, low income, low levels of educational attainment, overcrowding, and other social correlates of disadvantage. A low IRSD score is indicative of relatively greater disadvantage, and a high IRSD score denotes relatively lesser disadvantage. The Australian Bureau of Statistics distributes SA1s in Australia into quintiles based on the IRSD; SA1s in Quintile 1 are the most disadvantaged and SA1s in Quintile 5 are the least disadvantaged. We used these values for our study.

It is worth noting here that we deliberately chose our individual-level indicator of socio-economic position on the basis of it having a degree of consistency with our area-level indicator. As noted, unemployment and employment in low skilled occupations feature as measures that are used to build up the composite IRSD index.

We also controlled for the influence of a number of socio-demographic and clinical covariates, largely selecting these on the basis that they have been shown in previous studies to be associated with suicidal ideation in males. These were age, marital status, region of residence, level of social support, alcohol use and depression. The way in which each of these was operationalised is described below:

Age was classified into four broad groupings: (a) 18–29; (b) 30–39; (c) 40–49; and (d) 50–55.

We conducted univariate and multivariate logistic regression analyses to evaluate the strength of association between individual-level and area-level socio-economic position and suicidal ideation. In each multivariate analysis, we controlled for the alternative measure of socio-economic position and for each of the other covariates. Based on the regression model coefficients, we also calculated the probability of suicidal ideation for each grade of individual-level socio-economic position within each quintile of area-level socio-economic position.

All data were analysed using Stata (Version 13.1) [32].

As with other areas of health, it is easy to see how low individual-level socio-economic status might contribute to suicidal thinking. We know from elsewhere in the suicide prevention literature that life stressors can sometimes act as a trigger, particularly if underlying or longer-standing risk factors are already present [33]. Some of the ways in which individual-level socio-economic disadvantage manifests itself in relation to employment and occupational skill level—the indicators we used here—certainly qualify as life stressors. For example, those who are unemployed or working in low skilled occupations may experience financial strain, powerlessness, frustration and disrespect [34,35], all of which may act as potential tipping points.

Similarly, it is possible to see how the role of area-level socio-economic status may play out, with, for example, poorer areas having less in the way of infrastructure and services that may help to prevent individuals getting to the point where they view suicide as an option. For instance, areas of relatively greater socio-economic disadvantage may be less well served by mental health services. Perhaps even more importantly, other social, economic, cultural and physical features of these areas may also have a negative impact on the sense of wellbeing of their residents [36]. These features might include, for example, high levels of inter-generational poverty, community norms that foster stigma and discrimination, and built environments with poor amenity that promote social isolation.

Around the world, national and local suicide prevention strategies have placed insufficient emphasis on socio-economic position as a risk factor for suicide and non-fatal suicidal behaviours and thoughts. Platt argues for a redirection of policy focus in this regard, noting that care should be taken about the approach. In particular, he notes that governments should weigh up the relative merits of targeting individual-level or area-level disadvantage. His view is that targeting individuals within areas will yield the greatest benefits [3]. Our findings lend support to this assertion, which is consistent with the thrust of the recent World Health Organization report Preventing Suicide: A Global Imperative [37].

Area-level suicide prevention efforts are gaining traction because of the popularity of systems-based approaches delivered through initiatives like the European Alliance against Depression and the OSPI-Europe project [38]. In Australia, we have one such initiative, known as Lifespan (http://www.lifespan.org.au/). These initiatives take suicide prevention strategies for which there is a reasonable evidence base and deliver them to whole communities in a co-ordinated, integrated manner. The idea is that the whole is greater than the sum of the parts, and that by using a systematic approach—as opposed to a more disjointed, piecemeal one—greater benefits will ensue. Typically, areas that are selected for these systems-based initiatives are those which have high rates of suicide and/or suicide attempts; often, but not always, these will also be socially disadvantaged areas.

The suicide prevention initiatives in these area-level initiatives usually cover the full gamut of universal interventions (target whole populations, with the aim of favourably shifting risk factors across the entire population), selective interventions (interventions target those who are not yet manifesting suicidal thoughts or behaviours, but exhibit risk factors that predispose them to do so in the future) and indicated interventions (which are designed for those who are already beginning to display suicidal thoughts or behaviours, usually identified through screening programs or by clinical presentation). In the current context, there would seem to be opportunities for selective interventions that target those whose individual-level socio-economic position puts them at risk, doing so within area-level initiatives. This would mean that individuals with low socio-economic status would be targeted in socio-economically disadvantaged areas, but also, potentially, in better-off areas.

The way in which these policy and practice implications might be operationalised for men requires some thought. In areas of low socio-economic status, consideration might be given to how to increase the availability of mental health and suicide prevention services in general, and how to ensure that they are attractive and accessible to men in particular. Of course, efforts beyond the health and mental health systems will also be required in these areas. These might include urban regeneration or economic renewal programs (e.g., in areas where there have been widespread downturns in male-dominated industries like construction, car manufacturing or mining). They might also include efforts to improve the conditions in which men in these industries are working [39].

Unemployed men and men in low status occupations in these or other areas might be targeted through tailored programs. Again, some of these programs might be delivered through the mental health system; we know that men respond differently from women to some forms of psychotherapy, so gender-specific approaches may be required. Other programs may be broader in focus and aimed at addressing disadvantage itself. These might include training and education programs designed to upskill unemployed men or men working in relatively low skilled occupations.

Our study had several limitations which must be acknowledged here. Firstly, we used a measure of individual-level socio-economic position that comprised a composite of unemployment/employment and occupational skill level. As noted above, there are different ways of conceptualising socio-economic position, and our indicator is just one of these. Some might argue that our individual-level indicator and our area-level indicator were not sufficiently well aligned, and that they capture different components of the construct of socio-economic position. It is worth noting here, however, that unemployment/employment and occupation skill level are both used by the Australian Bureau of Statistics to build the aggregate area-level IRSD, creating some level of comparability.

Secondly, and on a related point, we relied on a single measure of individual-level socio-economic position, rather than exploring a fuller set of individual-level indicators. This means that we cannot comment on the extent to which unemployment/employment and occupational skill level act as a proxy for or behave differently from other individual-level indicators of socio-economic position. Having said that, it is worth noting that we conducted a sensitivity analysis (not reported here) which used a measure of financial stress as the individual-level indicator of socio-economic position, and that this analysis yielded similar results to our primary analysis.

Thirdly, we used data from a single wave of Ten to Men and therefore were only able to conduct cross-sectional analyses, which limits the extent to which we can ascribe causality to the observed associations between individual-level and area-level disadvantage and suicidal ideation. We have now completed our second wave of data collection and are poised to conduct our first set of longitudinal analyses.

Fourthly, there may have been issues in terms of our covariates. As noted, we chose these because they have consistently been shown to be predictors of suicidal ideation. Some may also have had a relationship with the exposure (i.e., our unemployment/employment and occupational skill level variable), which raises questions about whether our selected covariates acted as confounders or mediators. There is an argument that we may have over-controlled for some of those which acted as mediators, thereby underestimating the effect of unemployment and relatively lower skilled occupations on suicidal ideation.

Finally, there are a range of other limitations that relate to Ten to Men itself. These include a reliance on self-report, and the fact that the sampling strategy meant that males in remote areas and/or who were not proficient in English were under-represented.