Health service use in the year before death by suicide among ex-serving ADF members

The previous sections described overall health service use, including the proportion of ex-serving members accessing each health service, the types of health services used, and average frequency of health services used. However, these analyses considered each health service type in isolation and did not show how individuals used several services in combination or whether distinct patterns of health service engagement existed within the cohort.

This section reports the results from a modelling technique called latent class analysis (LCA). This analysis was conducted to identify subgroups (or groups) of health service users among ex-serving members who died by suicide. This data-driven analytical approach groups individuals based on their patterns of engagement in the year before death across seven indicators of health service use, derived from hospital (ED and admitted care), PBS/RPBS and MBS data. These indicators captured mental health, non-mental health and investigation-related services (such as diagnostic procedures, diagnostic imaging, and pathology from MBS), providing a more holistic view of health service use. 

Understanding these patterns provides insights into how ex-serving members interacted with the health system before death by suicide and where gaps in engagement may have occurred, despite the appearance of overall contact. It may also help identify opportunities for earlier intervention across a range of health service settings.

The analysis was limited to ex-serving members who died by suicide from 1 July 2014 to 30 June 2020 due to limitations in identifying mental health services from earlier ED data. However, the ex-serving population who died between 1 July 2011 and 30 June 2020 and the LCA cohort had similar demographic and service-related profiles, suggesting that the limitation to the cohort did not affect the representativeness of the patterns identified.

The LCA analysis identified five groups of health service users among ex-serving members who died by suicide. These groups reflect differences in the frequency, type and combination of health services used in the year before death by suicide. Labels were assigned descriptively, based on the relative frequency and type of health service use across the cohort rather than fixed thresholds, and are intended to summarise the dominant health service use characteristics within each group. Table 2 shows the predicted average health service use across the seven indicators for each group.

Each of the groups are described as follows:

• Minimal health service user: ex-serving members in this group represent the lowest users of health services, with lowest average use across all health services. This group accounted for 28.2% of the cohort.
• Low health service user: ex-serving members in this group showed slightly higher use than the minimal user group, but still relatively low across all health services. They represented 36.6% of the cohort.
• Moderate health service user (higher mental health): this group showed moderate health service use overall, with notably higher use of mental health services. On average, this group used 25 PBS mental health prescriptions and 12 MBS mental health services per person, substantially higher than even the high and very high health service user groups (described next). They represented 13.2% of the cohort.
• High health service user: this group demonstrated high overall health service use, driven primarily by non-mental health services (35 MBS non-mental health services and 25 PBS non-mental health prescriptions), while maintaining relatively lower mental health service use. This group comprised 16.2% of the cohort.
• Very high health service user: this group represented the highest health service users, showing elevated use across all health service types. They accounted for 5.8% of the cohort.

Source: AIHW Veterans Health Dataset (VHD), July 2014–June 2020

Which ex-serving members are using more health services?

The distribution of characteristics (demographics, service-related and health-related) across the five groups was examined to understand which characteristics were more commonly observed in each group, based on their relative proportions. 

AIHW performed multinomial logistic regression to further assess the relationship between each characteristic and allocation to a group. Figure 7 outlines the characteristics associated with being in the moderate user (higher mental health), high user and very high user groups compared to the minimal user (reference) group. These three groups were reported here due to their higher service use profiles. Full results for all five groups, are provided in the Supplementary table S7.1 and S7.3.