Main findings
Scottish male cohort trajectories demonstrated a pattern consistent with a statistically significant increase in suicide between 1960 and 1965 for those aged 20-39. The 1965 and 1970 cohorts peaked in suicide rate at age 35-39 and 25-29, respectively. It is possible that the 1965 and 1970 cohort are at greater mass vulnerability to suicide than earlier or subsequent cohorts, and that as they age there may yet be greater suicide rates in these cohorts as a consequence. Alternatively the period 1995-2004 may have posed increased risk to these cohorts; further research is needed to rule in or out explanatory discrete period effects of e.g. access to means of death which are recorded on death certificates and not available in this study for analysis. It is not possible to categorically establish if this pattern is one attributable to age, period or cohort, or a combination of all three factors.
One recent study has also now reported on this apparent cohort effect, and concluded that these cohort patterns were more pronounced for those living in the most deprived areas, and that the suicide rates in Scotland can be explained by a cohort effect consistent with exposure to neoliberal politics during the 1980s, an exposure experienced more in deprived communities [18].
Temporal patterns of suicide in Scotland
In Scotland post WW2 years, the burden of suicide was highest on older men (>44 years) with rates decreasing to the 1980s, before stabilising thereafter. The inverse was seen for younger men (<45 years) who had the lowest suicide rates which steadily increased and peaked by the late 1990s, with this group continuing to bear the burden of suicide until 2000. From 2000 to 2014 decreasing rates are seen for those aged 15-34 alongside increasing rates for 45-54 year olds, leaving the greatest impact of suicide on those aged 35-54.
During 1998-2004, 17 of the top 20 local areas with the highest male suicide rate in the UK were located in Scotland, with the Shetland Islands, Eilean Siar, Highland and Glasgow City having the highest rates at 47.5, 44.1, 43.3 and 41.6 per 100,000 population, respectively [5]. Rurality may explain some patterns of excess suicide - for Highland, an excess of male deaths appeared to be associated with access to more lethal means and rural occupation [24, 25]; a relative dearth of contact with mental health services in the month prior to suicide was found for those that lived in ‘remote rural’ or ‘remote small towns’ [26]; enforced social isolation has also been proposed in a conceptual model of rural suicide [27]. However excess suicide has also been observed in some urban areas. An analysis of suicides in Greater Glasgow to 2001 concluded that the East end of Glasgow formed a large geographical cluster of young adult suicide which persisted for two decades and which was likely not explained by ‘contagion effects’, but rather more likely by a concentration of deprivation [28].
Although there are many age and period risk factor interactions at play, it is not possible to explain in simple terms what drove the patterns observed, but it is likely that regional differences fuelled by deprivation are partly to explain.
In contrast, female suicide rates for all age groups in Scotland have converged and stabilised in recent years with evidence that women have not been affected by the huge impact of suicide as seen for men. Although there was a pattern for women with a limited number of successive cohorts showing peaks in younger age groups with each generation, this was unclear and no firm conclusions could be made due to sparser data.
Comparison with England & Wales
Temporal patterns of suicide for Scotland were markedly different to those in England & Wales, which had a much higher burden of suicide than Scotland in the 1950s. These patterns reversed over time so that Scotland became the country with disproportionate impact of suicide, compared to England & Wales. Men were affected more markedly than women over all years in both countries. In spite of the between country differences, the gender patterns within each country were similar, meaning gender is the bigger determinant with more overall predictive power.
The pattern reported for England & Wales by Gunnell et al. [12] of successive male birth cohorts from 1940 experiencing higher suicide rates peaking in younger age groups with each decade was also seen in Scotland. The differences in pattern with increased suicide rates for the 1965 and 1970 cohorts exist in both countries, but are more marked in Scotland.
Gunnell et al. explored the impact of discrete period events, and found that restricting access to lethal means (predominantly 1993 legislation on car exhaust emissions and the advent of catalytic converters) was effective in reducing suicide rates, i.e. when these period effects were controlled for, the pattern consistent with a cohort effect disappeared - there was no evidence that rates peaked earlier in later born cohorts (Additional file 3). We did not have access to the Scottish data by method of suicide and could not compare this discrete period effect. It has been reported elsewhere that deaths from motor vehicle exhaust fumes decreased in England alongside a corresponding increase in hanging deaths, whilst in Scotland hanging deaths were already increasing in men before deaths from motor vehicle exhaust fumes began to decline, with this increase being greater than the corresponding decrease [24]. Restricting access to means of suicide may have lasting impact in some countries, but for Scotland the discrete period effect of catalytic converter legislation did not achieve the same impact seen in England & Wales. Thus there may still be a greater risk in Scotland attributable to belonging to the 1965 and 1970 birth cohorts, for reasons which remain unclear.
This widening ‘suicide gap’ has been reported before, with the crossing over of increasing suicide rates in Scotland, and decreasing rates in England & Wales occurring in the 1960s, and differences since the 1990s being explained by a preference in the methods of hanging, suffocation or strangulation by young adult males [8].
It is plausible that regional differences in England & Wales data are obscured by using pooled national data, and that disaggregating these data would produce markedly different patterns. Analysis reported elsewhere concluded that between 1998 and 2004 large regional disparities existed between suicide rates in the countries of the UK and between local regions, and that deprivation as a risk factor fuelled these inequalities in suicide rates [5]. In 2011 the Office for National Statistics (ONS) published UK suicide data and estimated standardised male rates for 2011 ranging in England from 13.2 in London to 21.5 in North East England, with Wales being even higher at 22.5 [29]. Such regional variations have been noted within countries before and are typically associated with other risk factors [30]. Regional inequalities in suicide rates have also been characterised by markers of unemployment [31] and more recently, unemployment associated with the discrete period effect of the last UK recession [32, 33]; and low social integration, indicated by features of the proportion of single-person households, divorced people and population mobility [34].
A specific multilevel analysis exploring a range of factors between Scotland and England during 2001-2006, found that 57% of the excess suicide risk in Scotland was explained more by the area level measures of psychotropic drug prescriptions (proxy for poorer mental health), alcohol and drug misuse, with a relatively small contribution of deprivation and social fragmentation [35]. Therefore caution is required in reviewing national data patterns for men, and it is possible that Scottish data patterns observed may well find more concordance with specific regional English or Welsh patterns.
Public health and future research implications
Although Scotland has made substantive progress in reducing suicide in recent years alongside national suicide prevention initiatives, it is not possible to know whether there is a direct causal link between such suicide prevention strategies and relative period decreases achieved from the early 2000s [36]. The first ‘Choose Life’ initiative was a ten year plan introduced in December 2002 aimed at reducing suicides by 20% by 2013 – however it appears the decline in suicides may have started before ‘Choose Life’ had a chance to have an impact and it is not possible to categorically know which particular age groups would have been impacted on most. These national initiatives coincided with a period of perceived economic expansion and lower unemployment until the 2008 UK recession.
Other comprehensive social changes were going on aside from economic fluctuations that may also have compelling plausible explanations for the suicide impact of the 1965 and 1970 male cohorts, such as the continuing impact of de-industrialisation, related unemployment and persistent effects of deprivation as previously mentioned. Major reforms in education with increasing attention to well-being and increased average time in education may also possibly have had a long term impact. The increased use of mobile technology and its impact on social connectedness may also need to be considered in explaining recent or future trends. Important legislative and political changes that have impacts for certain minority groups e.g. immigration, equality legislation will also have had their effects. Therefore, further research at the individual level (e.g. means of suicide especially drug overdose, occupation, educational attainment) on the 1965 and 1970 cohorts of males who were aged 25-39 at death may shed some light on suggested potentiality for risk in later life in the same birth cohorts. Such research should seek to identify risk factors or ‘exposures’ that lead to mass vulnerability of cohorts so as to minimise the longer term impact of such exposures, and in the planning of building resilience in future generations in the very early years when such ‘exposure’ risks re-appear.
Strengths and limitations
One limitation of this comparison is that the data quality between countries is likely to differ, with potential variation in both coding and coding consistency across the years. We can also plausibly assume that data quality may be heterogeneous between age groups, with those younger age groups dying later having better quality data compared with their older counterparts dying a longer time ago. Graphical interpretations are reliant on the assumptions being made by each graph, and different assumptions would result in different interpretations. There was no matching data on specific suicide methods used, therefore it was not possible to compare the cohort trajectories using these different methods and our interpretation was guided by analyses reported elsewhere. In emulating the previous analysis for England & Wales by Gunnell et al. [12], the techniques used in this study were limited and the description of the methods expressed in terminology which takes account of known limitations. Nevertheless the utility of considering the factors of age, period and cohort in graphical terms has been demonstrated.
A strength of this study is the manual recovery of data prior to 1974 which permitted a long period of follow-up for the older cohorts. We were unable to do a comparison between the countries using a reference population given the historical data and the comparison with an earlier study; future research could standardise each country data to a standardised European population dataset to enable more robust comparisons to be made.