Serving and ex-serving Australian Defence Force members who have served since 1985: suicide monitoring 1997 to 2022

In this section multiple service-related characteristics are assessed concurrently to more accurately identify risk or protective factors for death by suicide, by adjusting for the confounding effects among characteristics. This builds on the suicide monitoring analysis of previous chapters which reported suicide rates by single service-related characteristics in isolation. 

Summary of key findings

Service-related characteristics associated with death by suicide among the ex-serving population over time were assessed by modelling suicide rates. 

The following characteristics among the ex-serving population were significantly associated with a higher or lower rate of suicide on average at any point in time, holding all other available characteristics constant:

• Ex-serving females have a lower rate of suicide (0.6 times) compared with ex-serving males.
• Those who were officers at the time of separation have a lower rate of suicide (0.7 times lower) compared with those who are of ranks other than officer.
• Those who separated for involuntary medical reasons have a higher rate of suicide (2.3 times) compared with those who separated for voluntary reasons.
• Those who separated involuntarily for the reason of Retention not in service interest have a higher rate of suicide (2.0 times) compared with those who separated for voluntary reasons.
• Those who separated due to contractual or administrative changes have a lower suicide rate (0.5 times) compared with those who separated for voluntary reasons.
• Those with a length of service of less than 1 year or 1 to less than 10 years have a higher suicide rate (1.9 and 1.6 times respectively) compared with those with a length of service of 10 or more years.
• Those who served at least one day in the permanent forces have:
  • a higher suicide rate (1.6 times) compared with those who had served solely in the reserve forces, within the first 10 years after separation
  • similar suicide rates to those who served solely in the reserve forces, by 10 or more years after separation.
• Basic training completion status, year of hire, and year of separation were not found to have a statistically significant association with death by suicide once all other available factors were controlled for.

Overall, these results indicate that the rate of suicide is lower for: 

• ex-serving members who are female
• those who were officers at the time of separation
• those who separated for voluntary or contractual or administrative reasons
• those with a length of service of 10 or more years
• those who served solely in the reserve forces within the first 10 years of separation.

While these results can identify groups of ex-serving members who have a higher rate of suicide based on service-related characteristics, they cannot provide a causal link between specific characteristic and death by suicide.

What method was used to model suicide rates?

A form of multiple regression analysis known as Cox proportional hazards modelling was used to examine the association between suicide death and service-related characteristics. The Cox proportional hazards regression model (Cox 1972) is the most common method for studying the time it takes for an event to occur after an intervention, and how this is influenced by one or more predictor variables. 

In this report:

• the event measured is death by suicide,
• the intervention is separation from the ADF
• the predictor variables are the various available service-related characteristics.

For more information about Cox proportional hazard modelling and hazard ratios, see the Technical notes section.

This work is an update and extension to the previous modelling analysis presented in the 2021 suicide monitoring report Serving and ex-serving Australian Defence Force members who have served since 1985: suicide monitoring 2001 to 2019. Note that results may vary between reports due to the expanded population, monitoring period, and service-related characteristics assessed.

About the models

Two models are presented that differ in monitoring period and cohort for inclusion based on their year of separation from the ADF. 

As separation is when an individual becomes included in the model, the following models do not account for serving permanent or reserve members, only ex-serving members. Each individual was censored from further contribution at the point of their death (not by suicide) or at the analysis end date of 31 December 2022 for those alive.

The first model, the post-1997 model, includes ex-serving members who separated on or after 1 January 1997. This makes up 60% of all ex-serving members who had at least one day of service since 1 January 1985.

The second model, the post-2003 model, includes ex-serving members who separated on or after 1 January 2003. This makes up 44% of all ex-serving members who had at least one day of service since 1 January 1985. 

As the reason for separation from the ADF was only accurately recorded from 1 January 2003 onwards, only the second model contains analysis by the reason for separation. While the post-1997 model includes a larger cohort and therefore has a greater level of statistical robustness, the post-2003 model can consider the reason for separation as another interacting service-related characteristic

See Table 16 for a summary of the differences between the two models.

See demographics of the cohorts included in modelling analysis in Supplementary table S9.1.

Box 3 describes the interpretation of the hazard ratio measure from modelling analysis.

Modelling suicide rates 1997–2022

This analysis contains all ex-serving ADF members who separated on or after 1 January 1997 and models the rate of suicide over time by: 

• age group
• sex 
• rank
• service
• length of service
• year of hire
• year of separation
• prior service status
• basic training completion status.

Time since separation is not a variable in these models as time while ex-serving is the factor that underpins this model.

Key requirements for this model

• The relative difference in suicide rates between groups of a characteristic remained approximately constant over time. See Technical notes for more information.
• Analysis by the service-related characteristic, prior service status, was further split into two time periods: less than 10 years since separation, and 10 or more years since separation. Distinct trends in the suicide rate were observed within each of these time periods between permanent ex-serving and reserve ex-serving members (see Time since separation, Figure 19).
• This time-dependent effect was similarly observed for the basic training completion status characteristic and adjusted accordingly.

Key findings from the post-1997 model

While holding all other analysed service-related characteristics constant, the rate of suicide in the ex-serving population at any point between 1997–2022 was:

• 0.6 times lower for females compared to males
• 0.7 times lower for those who separated as officer ranks compared to those of ranks other than officer
• 0.6 times lower for those who separated in the Air Force compared to those in the Army
• 1.9 and 1.6 times higher for those with a length of service of less than 1 year, and 1 to less than 10 years, respectively, compared to those with a length of service of 10 or more years
• 2.0 times higher for permanent ex-serving members compared to reserve ex-serving members within the first 10 years after separation
• 1.3 times higher for permanent ex-serving members compared to reserve ex-serving members at 10 or more years after separation.

The service-related characteristics age group, year of hire, year of separation, and basic training completion status were not found to have a statistically significant association with suicide once all other available factors were controlled for.

Modelling results are presented in Table 17 and Figure 25.

Note: TSS = Time since separation; CI = confidence interval.

Source: AIHW analysis of linked Defence historical personnel data–PMKeyS–NDI data 1985–2022.

Figure 25: Hazard ratios for suicides over time of the ex-serving population: 1997–2022, by survival modelling

Note: TSS = Time since separation; CI = confidence interval.

Source: AIHW analysis of linked Defence historical personnel data–PMKeyS–NDI data 1985–2022.

Modelling suicide rates 2003–2022

This analysis contains all ex-serving ADF members who separated on or after 1 January 2003 and models the rate of suicide over time by:

• age group 
• sex
• service
• length of service
• prior service status
• reason for separation.

Year of hire, year of separation, and basic training completion status were not included in the post-2003 model as these characteristics were not found to have a statistically significant association with suicide based on the previous post-1997 model.

Key requirements for this model

Like with the post-1997 model, prior service status was assessed within two time periods, less than 10 years after separation, and 10 or more years after separation, to meet the requirements for modelling. 

Key findings from the post-2003 model

While holding all other analysed service-related characteristics constant, the rate of suicide in the ex-serving population at any point between 2003–2022 was:

• 0.6 times lower for females compared to males.
• 2.3 and 2.0 times higher for those who separated for involuntary medical and for involuntary Retention not in service interest, respectively, compared to those who separated for voluntary reasons.
• 0.5 times lower for those who separated for contractual or administrative reasons compared to those who separated for voluntary reasons.
• 1.6 times higher for permanent ex-serving members compared to reserve ex-serving members within the first 10 years after separation. There was no statistically significant difference in the suicide rate between permanent and reserve ex-serving members at 10 or more years after separation.

The service-related characteristics age group, rank, service, and length of service were not found to have a statistically significant association with suicide once all other analysed factors were controlled for in this model.

Modelling results are presented in Table 18 and Figure 26.

Note: TSS = Time since separation; CI = confidence interval.

Source: AIHW analysis of linked Defence historical personnel data–PMKeyS–NDI data 1985–2022.

Figure 26: Hazard ratios for suicides over time of the ex-serving population: 2003–2022, by survival modelling

Note: TSS = Time since separation.

Source: AIHW analysis of linked Defence historical personnel data–PMKeyS–NDI data 1985–2022.

Differences between the post-1997 and post-2003 models

Note that differences in the results between the post-2003 and post-1997 models may be due to: 

• the addition of the reason for separation characteristic in the post-2003 model, or
• the smaller population size in the post-2003 model that increases the variance, or
• both.

Further investigation showed that the loss of statistical significance by rank was due to the smaller population size of the post-2003 model, rather than the addition of the reason for separation characteristic. This was demonstrated by performing the post-2003 modelling analysis without the reason for separation characteristic.

Data underlying all figures are available in Supplementary tables S9.2 and S9.3. See Data for a link to the tables.

Please note, data for more recent years are subject to change; see Technical notes for further detail.