Technical notes
Data linkage, also known as data integration, is a process that brings together information relating to an individual from more than one data source.
Some sections in this report used deterministic and probabilistic linkage between 2 data sets: Defence Personnel data and the ABS 2021 Census of Population and Housing. These linked datasets have been integrated into the Person Level Integrated Data Asset (PLIDA) managed by the ABS.
The Department of Defence personnel data consists of an extract from current and historical personnel systems for ADF members who have served since 1 January 1985. This combines Personnel Management Key Solution (PMKeyS), Core HR system, D1, CENRESPAY (for reservists), ADFPAY (for permanent members) and other historical payment systems. Validation of the data extract was performed with other data sources. The Defence personnel data includes information on service characteristics. For ex-serving ADF members, service characteristics are reported as at date of separation from the ADF.
To allow linkage between the 2021 Census and the Defence personnel datasets to produce the final analysis presented in this report, processing was first conducted on Defence Personnel data:
- Initial data checking and cleaning by AIHW
- Linking the Defence Personnel data set to an AIHW-ABS interoperable spine by AIHW using probabilistic linkage techniques. This AIHW-ABS interoperable spine allows data held by both organisations (Defence Personnel by the AIHW, and the 2021 Census by the ABS) to be linked without the need to share any identifiable information.
- Secure transfer of the Defence Personnel data set to the ABS.
- Linkage of the Defence Personnel data set to the PLIDA Person Linkage Spine by the ABS.
Strict separation of identifiable information and analytical content data was maintained within the Data Linkage Units at both AIHW and the ABS, so that no one person or organisation will ever have access to both. This ensures that no individual can be identified by any persons involved in the data linkage process, nor in any analysis or reporting conducted as part of the content in this report.
Figure 10 summarises the linkage process undertaken for this report.
Figure 10: Linkage process
Where results in this report are created from the linked 2021 Census and Defence Personnel datasets, this is noted. Results are otherwise created from the interoperable 2021 Census of Population and Housing data set.
The veteran child linkage spine created as part of this report is the first of its kind and is a valuable foundation for future research. Notably, it could be used to explore other PLIDA data sources such as Medicare Benefits Schedule (MBS), Pharmaceutical Benefits Schedule (PBS), the National Health Survey (NHS), Deaths Registrations data and the Australian Early Development Census (AEDC). Analyses of these sources could provide insights into veterans’ children’s mental health service use, prescriptions of mental-health related medications, suicide, and early education outcomes.
Base 2021 Census analysis
To increase the accuracy of identification and analysis of veterans’ children in the 2021 Census of Population and Housing, some children were excluded based on their family characteristics. To be included in the base 2021 Census analysis, a child must have been living in a family where:
- There were no parents temporarily absent on Census night
- All parents had a recorded status for ADF service
- There were no parents under the age of 17
- The family resided in a private dwelling.
Counts of children are based on place of enumeration.
Children of single parents, and children who are adopted, step or foster children to the veteran parent, are included throughout all analyses in this report unless otherwise stated.
Linked 2021 Census to Defence Personnel analysis
For analysis involving linked data between the 2021 Census and Defence Personnel, further exclusionary criteria were applied. To be included in analyses using this data set, a child must have also been living in a family where:
- Only one of their parents was a veteran
- The veteran parent’s sex matched between the 2021 Census and Defence Personnel
- The veteran parent’s age matched between the 2021 Census and Defence Personnel (with a +/- 2-year leeway given to account for differing methods of reporting age in the 2 datasets).
The following limitations should be considered when interpreting the findings in this report.
1. Limitations of identifiable children
Children who were not living with their veteran parent(s) on Census night cannot be identified using 2021 Census of Population and Housing data, because parent-child relationships are not captured between households. This means that certain children are not captured, or are significantly underrepresented, in the analysis for this report:
Veterans’ older, non-dependent children who have moved out of home
This cohort will be underrepresented in the data, because they are less likely to live with their parent(s).
This may also mean that the cohort of older, non-dependent children who were identified in this report are a biased subpopulation and may not be representative of the health and welfare outcomes of this group at a population-wide level.
For example, the older, non-dependent children captured in this report may be biased towards those who have returned home in adulthood to provide care to their aging parent(s), or those who themselves require care from their family for disability and/or chronic health conditions.
Nonetheless older, non-dependent children were included for completeness and because there was no prior information to suggest a systematic difference between veteran and non-veteran families. This means that results about non-dependent children in this report – particularly those who are older – should be interpreted with caution.
Dependent children who were not living with their veteran parent on Census night
For example, children whose parents were separated, and the child was not in the same household as their veteran parent on Census night, or those whose veteran parent was on deployment on Census night, may inadvertently be captured within the comparator cohort of children with no veteran parents in this report.
This means that results about dependent children with a current serving veteran parent in this report should be interpreted with caution.
2. Small cell counts limited appropriate age-specific analysis for some sub-populations of children discussed in this report
When exploring the health of a population, corrections for age are often needed because older people are more likely to experience some health conditions than younger people. This is done either by producing age-specific rates, or by conducting age standardisation.
In this report, the underlying data did not meet the minimum criteria for age standardisation. Therefore, age-specific results have been produced where possible.
For some health-related results discussed in this report however, neither age standardisation nor age-specific results could be produced due to small counts. Instead, results could only be published by the type of child. The following data are impacted by this limitation:
- Number of long-term health conditions
- Types of long-term health conditions
- Rates of long-term mental health conditions by the veteran parents’ demographic and service characteristics.
While the risk of age-related bias is deemed lower for dependent children due to the inherent limitations of their age in their aggregation (that is, children under 15 and dependent students aged 15 to 24), significant age-related biases may exist for non-dependent children in these results, as this category includes people as young as 15, to older adults with no upper age cutoff. This may be the case particularly for the children of ex-serving veterans, as their parents are an older cohort compared with non-veterans, thus making it more likely that their children are also older.
Because of this, the above results - particularly those relating to non-dependent children - should be interpreted with caution.
3. Long-term health conditions are based on self-report data from the Census
The analysis of long-term physical and mental health conditions in this report is based on self-reported data from the 2021 Census. There are several limitations to using the Census as a data source on long-term health conditions:
- the data was collected for the first time in the 2021 Census so there is currently no time series data available
- the long-term health conditions data collected are based on a single question. ABS health surveys such as the NHS and NATSIHS have a detailed set of questions to capture prevalence of these conditions more accurately
- data relies on the respondent self-reporting health conditions, or in some cases reporting health conditions of other people in the household, in most cases without an interviewer present to help
- data is only available for the listed conditions above
- under-reporting may occur for some conditions compared with other ABS surveys. This is due to self-reporting and reporting on behalf of other members of the household, as well as potential sensitivities about individual health conditions.
Defence and DVA initiatives seek to make health care more affordable for the veteran populations examined. As such, differences in the proportion of people responding to a question about whether a doctor or nurse have provided advice about a long–term health condition may be influenced by potential differences in the affordability of health care in the populations examined.
The ABS recommends that the National Study of Mental Health and Wellbeing (NSMHW) be used as the main source of prevalence data as it uses diagnostic criteria rather than self-reporting. However, a limitation of the NSMHW is that it does not capture child-parent status.
For more information, see Comparing ABS long-term health conditions data sources.