Are young female suicides increasing? A comparison of sex-specific rates and characteristics of youth suicides in Australia over 2004–2014

Consistent with data reporting conventions, we define young people as those aged between 10 and 24 years, and further categorize this group to include early adolescents (10–14 years), late adolescents (15–19 years) and young adults (20–24 years) [20]. The transition from adolescence to young adulthood marks a period of rapid change [21‐23]. In particular, high prevalence mental disorders, which are significantly associated with heightened risk of suicide, typically emerge during this period, and continue into young adulthood [24]. Indeed, depression and anxiety, which are common, show greater continuity into early adulthood among females compared with males [25]. As such, existing research supports the transition to adolescence as marking the beginning of a period of heightened suicide risk in young females.

Several OECD studies have documented recent growth in suicide rates among females as young as 10-years-old, and show rates increase across the adolescent years. Recent population studies in Finland, the US and Canada found rates have increased significantly in female adolescents and young adults over the past two decades. Canadian and US data further indicate that the greatest increase occurred among 10–14-year-old females, while overall rates remained significantly higher among 15–24-year-old females [6, 14].

Indigenous young people are consistently over-represented in Australian suicide statistics, raising the possibility that a rise in young female suicide rates may be driven in part by increases among young Aboriginal and Torres Strait Islander females in particular. National statistics indicate rates are four times higher among 15–24-year-old Indigenous females and three times higher among Indigenous males compared with their non-Indigenous peers [15]. Furthermore, between 2007-11 and 2012–16, Australian statistics showed crude mortality rates increased among young Indigenous females alongside declining rates among same-aged males [3, 18]. Increases among Indigenous females have also been documented for specific regions; over 1994–2007 in Queensland [16], and over 2005–14 in the Kimberley region of Western Australia [17]. Risk of suicide within Indigenous communities is underpinned by a myriad of systemic factors contributing to significant disadvantage in health and socio-economic outcomes (for a review see: Zubrick and colleagues [26]). Population studies have also shown that rates of depression and anxiety among Indigenous youth have increased [27]. Hence, a growing proportion of young Indigenous females may be experiencing poor social and emotional wellbeing which, in turn, may be contributing to increased suicide rates.

A further factor that may be contributing to possible increases in young female suicides relates to changes in suicide methods used by young females. Method lethality refers to methods which increase the probability of a fatal outcome [28], and is a well-documented risk factor for youth suicide. Typically, males have used more lethal methods of suicide (e.g. hanging and firearms [29]) as compared to females. Accordingly, males account for more completed suicides [15] while females account for more non-fatal attempts [1, 30]. The relationship between specific methods and the magnitude of suicide rates was illustrated by a New Zealand study which showed that sex differences in completed and attempted youth suicides were completely explained by males’ use of more lethal methods [31]. Thus, if young females have increased their use of more lethal methods this may be contributing to a rise in suicide rates. Indeed, research has shown a rise in the use of lethal methods of self-harm and suicide among young females [6, 14, 19, 32, 33], and hanging is now the leading method among female youth in both Australia and across the OECD [6, 7, 15, 19, 34, 35].

Whilst these factors have been investigated for their impact on youth suicide individually, the contribution of these factors in concert has never been examined in a nationwide cohort of young people to date. The present study therefore aimed to examine sex-specific suicide rates among 10–24-year-olds in Australia over the period 2004 to 2014, with a key focus on investigating whether or not key epidemiological risk factors (i.e., age, and Indigenous background) may explain changes in female suicide rates over time. Our secondary aim was therefore to examine if one explanation for the change in rates was increasing use of hanging (a highly lethal method).

This descriptive study employed a retrospective case series design, including all suicides and probable suicides in individuals aged 10–24-years in Australia, between 2004 and 2014.

Statistical analyses, including multinomial logistic regressions and negative binomial regressions, were conducted in Stata, version 15.1.

Negative binomial regression revealed rates for young males were significantly higher compared to rates among young females (Incident Rate Ratio [IRR] = 2.56, 95% CI: 2.30–2.85). There was no statistically significant change over time in the overall suicide rate (IRR = 1.01, 95% CI: 0.98–1.04), but when stratified by sex, the rate for females rose by 3% per year over the period (IRR = 1.03, 95% CI: 1.01–1.06). Using this model parameter, and assuming a linear relationship between suicide prevalence and time, this equates to a total increase of 38% (1.03^11) (see Table 2). Table 2 also shows the rate of suicide among Indigenous females was four times higher than for other females (IRR = 4.12, 95% CI: 3.41–4.98). Additionally, female rates differed significantly by age; compared with the 10–14-year-old age group, suicide rates were 6.80 (95% CI: 5.07–9.10) times higher among the 15–19-year-old age group, and 8.33 (95% CI: 6.24–11.11) times higher among the 20–24 age group (Table 2).

The picture was more complex for males, as shown in Table 2. Among the 10–14-year-old age group (IRR = 0.98, 95% CI: 0.91–1.05) and the 15–19-year-old age group (IRR = 0.98*1.03 = 1.01, 95% CI: 0.99 to 1.03) there was no evidence of change over time. For the 20–24-year-old age group, the rate decreased by 3% per year (IRR = 0.98*0.99 = 0.97, 95% CI: 0.96 to 0.99). Once again, assuming a linear relationship between suicide prevalence and time, this model parameter equates to a 28% decline overall for this group. The suicide rate was more than three times higher among Aboriginal/Torres Strait Islander males than for other males (IRR = 3.42, 95% CI: 3.05–3.83).

Most young people used hanging (67.1% females, 66.5% males); whereas a smaller proportion used drug-poisoning (9.4% females, 3.2% males), or other methods (23.5% females, 30.3% males). Table 3 presents the results of separate multinomial logistic regression models for youth and each sex by suicide method.

This study identified that whilst overall youth suicide rates did not increase significantly in Australia between 2004 and 2014, rates among females did, and this was consistently the case across all age groups and regardless of Aboriginal/Torres Strait Islander background. Rates were consistently higher among Aboriginal/Torres Strait Islander young people, among males, and older age groups (20–24-year-olds).

Throughout the study period, hanging was the most commonly used suicide method, and females and males were equally likely to have used hanging. Overall the odds of using hanging increased significantly over the period, although this was not the case among females.

In contrast to earlier reports [11], this study found that overall rates of youth suicide did not increase significantly in Australia between 2004 and 2014. However, broadly stable rates among youth masked significant increases in young females. This finding adds to a growing body of research concerning the epidemiology of suicide among young females, which demonstrates increasing rates across OECD countries [4‐8]. Importantly, the current study extended these findings to Australian females as young as 10 years old. Significant rate increases among younger females (10–14-year-olds) in this study are consistent with findings from the US and Canada, suggesting that an increasing number of early adolescent females are dying by suicide [6, 14]. This trend is concerning, given our findings that 10–14-year-old females showed greater odds of dying by hanging; a highly lethal method [40] that is difficult to restrict in community settings [41]. As such, these findings hold implications for prevention efforts.

Given crude rates in 10–14-year-olds were higher among females than males over several years, the gap in suicide rates typically observed between the sexes was not as evident in this younger age group. Rather, the rate ratio between the sexes appeared to widen with increasing age. This may be understood in light of earlier findings regarding sex and age differences in lethality of suicide attempts, which has shown that lethality tends to be lower among females than males for all methods [42], and elevated among younger age groups overall [43]. Therefore, sex differences in attempt lethality might depend on age. Future research linking sex-based trends in suicides with attempt data might thus elucidate whether the transition to adolescence marks the onset and peak lethality of suicide-related behavior in young females.

Contrary to expectations, no variable studied accounted for increased rates among females. Research highlights recent growth in the prevalence of modifiable risk factors that may explain an overall rise in young females’ vulnerability to suicide across groups, particularly when co-occurring within individuals [44]. These include concurrent increases in rates of depression and self-harm [30], alcohol misuse and related harms [45], and declines in vocational participation [46]. As this study focused on epidemiological risk factors coded in the NCIS, examining clinical risks and their cumulative effects was beyond our scope, but ought to be prioritized in future research to inform targeted interventions.

An unexpected finding was that females’ use of hanging showed no change over time, despite increasing among males and youth overall. This contrasts with prior studies that have consistently reported increases in the use of hanging among young females [19, 32, 33]. However, such studies did not report on younger-age females separately to older groups [32], and examined longer time-periods resulting in a larger number of cases [32, 33]. Our finding of no change in females’ use of hanging may be attributable to small annual counts over a constrained study period.

A novel finding was the emergence of distinct profiles of females who died by specific methods. Use of highly lethal methods (hanging) was more likely among 10–14-year-old and Aboriginal/Torres Strait Islander females, while methods less likely to be fatal (drug-poisoning) were more likely to have been used by 20–24-year-olds and non-Indigenous females by comparison. This is important as young people who self-poison are more likely to present to services and survive an index attempt [42, 47], and are more likely to later die by drug-poisoning than other methods [47]. Taken together, opportunities to intervene following an index attempt may be greater among young adult and non-Indigenous females presenting to services following drug-poisoning. In such cases, the risk of dying by suicide following health service contact for an index attempt ought to be carefully evaluated and followed up by health professionals.

Groups demonstrating relatively higher odds of using hanging indicate that suicide attempts are more likely to be fatal in pre- and early-adolescent compared with young adult females, late adolescent compared with young adult males, and Indigenous young people of both sexes compared with their non-Indigenous peers. This might be particularly salient for younger Indigenous females, in whom suicide rates are both high and rising. The higher risk of cluster suicides among Indigenous youth [11] make this broad sub-group of young females a clear public health priority. Supporting local Indigenous communities to improve the social and emotional wellbeing of young people is crucial [48].

Key strengths of this study included comprehensive sampling of suicides among youth in Australia, including probable suicides, over an 11-year period. Additionally, this study reported on suicides among individuals as young as 10-years-old, and used standardized data thereby minimizing bias in reported estimates. Use of stratified mid-year population estimates facilitated standardization of population-stratified rates, controlling for annual variations in population distributions. Finally, this study employed statistical methods appropriate for empirically testing time trends for count data (negative binomial regression [49]), while multinomial logistic regressions examined trends in the use of specific methods.

Several limitations need to be acknowledged. Our model parameters assume a linear relationship between suicide prevalence and time whereas, in fact, suicide rates in both males and females fluctuated over time (Fig. 1). Therefore, whilst we can be confident that suicide rates increased over this time period in females, but not males, the observed magnitude of this increase may be different to that predicted by our model, highlighting the challenges in using linear prediction models to estimate changes in suicide rates. Relatively small annual counts may also have contributed to under-powered analyses in some instances, particularly among females. In particular, analyses relating to temporal changes in suicide rates for females between the ages of 10-14 years may have been under-powered given the model parameter estimates observed [50]. Therefore, whilst consistent with findings from other jurisdictions [6, 14], results for this group should be considered illuminative. A longer study period may facilitate aggregating counts over several years and improve statistical power for testing interaction effects.

The variables under study comprised a small proportion of established risk factors for youth suicide. Therefore, the scope of this study did not address potential contributions of clinical risk factors, including mental disorders and self-harm history. This, in turn, may have contributed to the discrepancy between our model parameters and the magnitude of change in suicide rates observed for both males and females over this time period. However, we restricted data to that reliably recorded by the NCIS to clarify the magnitude of rates and the contribution of key epidemiological risk factors. Additionally, by limiting our models to those factors that have been implicated in previous work as underlying the increase in suicide rates in females, we ensured our models were protected from over-fitting.

A final limitation relates to data coding. We included cases coded with ‘undetermined intent’ to capture probable suicides, as is conventionally reported in national statistics [50]. These cases were included in order to minimize previously reported underestimations, as coronial determination of intent is influenced by multiple factors including legal definitions and jurisdictional processes [37], as well as social and cultural sensitivities [51]. Prior to Australian data coding reforms in 2007, deaths that did not clearly meet criteria for a ruling of suicide, coded ‘accidental’, were later found to result in underestimations [37]. Additionally, data for 2013 and 2014 were incomplete at the time of writing [15] as data for equivocal deaths remain open for several years throughout the revisions process. Therefore, suicide counts for study years prior to 2007, and for 2013–2014, may be conservative. Variations in data collection are also reflected in incomplete data for Indigenous origin, which may be unavailable or not reliably reported [51]. We combined cases with Indigenous origin coded ‘unlikely to be known’ and ‘non-Indigenous’ into a single category in order to guard against inflating Indigenous suicide estimates, while retaining the total sample for analysis. However, this likely resulted in an underestimation of the true number of Indigenous suicide deaths, which could not be verified using the available coded data. Relatedly, the absolute number of suicides among Indigenous young people was low (although the relative risk was high), meaning that our estimates of risk for this group may be measured imprecisely.

This study highlights the importance of broadening current conceptualizations of youth suicide within public policy from that of a male problem to one increasingly involving young females. National policy on suicide prevention benefits from a greater understanding of recent time trends in young female suicides stratified by age and Indigenous background, which informs the need to target prevention efforts from a younger age and across both Indigenous and non-Indigenous groups. Current national approaches to suicide prevention in Australia may have overlooked preventative opportunities among young females, given females are more likely to seek help than males [30, 52].

A substantial body of research has shown that young females more often seek help for suicide-related behavior compared to males, including professional and non-professional sources of help [53]. Preliminary research within Victorian emergency departments has also shown that, among 12–24-year-olds, females make up the greater proportion of presentations for suicide-related behavior, and over half are sent home without a mental health assessment or referral [Donaldson A, Hetrick S, Redlich N, Spittal MJ, Robinson J: Youth Emergency Department Presentations for Self-Harm: A Retrospective File Audit Study, in preparation. Unpublished], despite heightened risk of suicide within 30 days post-discharge [54]. As such, females’ greater propensity for help-seeking, which can facilitate access to effective treatments [53], presents an important target for service reform. Such findings underscore a clear imperative for policy-makers to advocate for a more coordinated response to suicide-related behavior in young females, and to resource services accordingly.

At a population-level, restricting access to lethal means has proven to be an important and largely effective universal prevention strategy [55]. Although research highlights the potential for means restriction to reduce the population-level burden of suicide, difficulties with restricting hanging in community settings [41] necessitates the use of multiple evidence-based prevention strategies that target young people. Access to lethal means is a well-documented environmental precipitant that increases the risk of a fatal suicide attempt among young people [24], and hanging is both highly accessible and lethal. Combining alternative, evidence-based prevention strategies might include a combination of school-based awareness programs, gatekeeper training and screening, and cognitive behavioral and dialectical behavior therapies [56‐59]. Such programs ought to be made available to younger age groups in recognition of the growing evidence that suicides are increasingly occurring among early adolescent females.