Data sources

PMKeyS data

The Personnel Management Key Solution (PMKeyS) is a Department of Defence administrative database that contains demographic and service information on all people with Australian Defence Force (ADF) service on or after 1 January 2001 (when the system was introduced). The Department of Defence provided the AIHW with an extract (as at 11 August 2018) from the PMKeyS containing 115,551 records of all ADF members who have had at least 1 day of service on or after 1 January 2001 and who discharged after 1 January 2001.

Specialist Homelessness Services Collection data

The Specialist Homelessness Services Collection (SHSC) collects information about people who are referred to, or seek assistance from, specialist homelessness services (SHS) agencies. A specialist homelessness agency is an organisation which receives government funding to deliver specialist homelessness services to a client. These agencies collect data on an ongoing basis and are responsible for submitting these data to the AIHW on a monthly basis.

The SHSC has 2 parts—a Client collection and an Unassisted Persons collection. All data provided for this linkage project were from the Client collection. For this collection, the base unit is a person who presents to a SHS agency requesting services. A person becomes a ‘client’ once they receive a service(s). A ‘support period’ is the period of time a client receives assistance from a SHS agency. It relates to the provision of a service and/or supported accommodation. During a support period, a range of services additional to supported accommodation can be provided. Further detail on the types of services is available on AIHW’s website page about the SHSC.

On 1 July 2011, the SHSC replaced the Supported Accommodation Assistance Program National Data Collection (SAAP NDC), which began in 1996. As a result of this change, the scope of the data provided from the SHSC was homelessness service use from 1 July 2011 to 30 June 2017, which was the latest data available for inclusion in this study.

Homelessness services not provided by SHS agencies are outside the scope of the SHSC and therefore this service use is not captured in this report.

It is possible for SHS agencies to retrospectively add support periods in a particular financial year (for example 2013–14) which started in a previous financial year (such as 2011–12). If analysis of the national SHSC data in the future is conducted for the entire reporting period (2011–12 to 2016–17), it will include these updated support periods. However, in this report, data for individual financial years have not been updated in this way, and are current as at the end of each financial year. This difference only occurs for a very small number of records.

SHSC data used to calculate level of use

The data provided to calculate the level of use among Australian SHS clients was provided as part of an internal data request within the AIHW. Not all in–scope agencies submit data each year, and not all information sought from SHS clients is answered. This means data may not be completely representative of people receiving specialist homelessness services. How much this affects the representativeness of the data depends on how much information is missing, and how those people whose information was not collected are distributed among the study group. An imputation strategy has not been applied to correct for missing or incomplete data, therefore all SHSC data used for this analysis are unweighted.

Data quality issue with 2011–12 data

The clients in this study first appeared in the SHSC between 1 July 2011 and 30 June 2012. Previous analysis of the 2011–12 SHSC data identified some data quality issues. In particular, 90% of SHS agencies returned support period data in 2011–12, although many did not return data for all 12 months. This response rate has increased over time in 2014–15, 96% of agencies returned data for each month where they were expected to participate.

In 2011–12, the rate of invalid/‘don’t know’/missing responses was high for a number of data items. Data completeness has improved each year and this is important to consider when making comparisons between the cohorts within the study group as they differ based on the number of years for which specialist homelessness services were accessed.

The data quality statement for the 2016–17 SHSC is available on the AIHW website here. Data quality statements for previous years are also available at the AIHW Metadata Online Registry (METeOR).

Further information about the SHSC is in:

Specialist homelessness services annual report 2016–17

Specialist homelessness services Collection Manual

Department of Veterans' Affairs client data

The Department of Veterans’ Affairs (DVA) client database is an administrative database that contains demographic and service information on DVA clients. The extract received by the AIHW Data Linkage Unit contained 226,684 records of DVA clients who were alive as at 1 January 2001. Only those who also had a record in the PMKeyS were included in the contemporary ADF population (those who had at least 1 day of service on or after January 2001 who discharged after that date).

Data linkage methodology and outcome

Data linkage is a process that identifies an individual from more than 1 source and merges that information for analysis. Identifying that an individual is the same person in different data sets can be done by matching identifying information about the individual (such as name and address) or a using a person identifier (such as a client or patient number). In some cases where directly identifying information is not available from data sets, AIHW uses a statistical linkage key (SLK). This is created at the point of data collection and is drawn from the person’s name, sex and date of birth (see the statistical linkage key 581 data element cluster for more information). The SHSC uses this approach.

For the PMKeyS, names and addresses were matched to data from the DVA client database. The PMKeyS and DVA client data sets were matched probabilistically, that is, where close matches are likely to be the same person (see Methods section below). Information from the DVA client database was used to update name and address information from the PMKeyS, where available.

Following this, an SLK was generated for each individual in the PMKeyS data, which was then directly matched to the SHSC SLK. This yielded 1,427 distinct matches with the SHSC.

All data linkage was carried out by the Data Linkage Unit at the AIHW, an accredited Commonwealth Integrating Authority. This accreditation requires the AIHW to adhere to stringent criteria and abide by the National Statistical Service High level principles for data integration involving Commonwealth data for statistical and research purposes and Best practice guidelines. As well as these guidelines, data linkage at the AIHW is carried out under the protections of the Privacy Act 1988 and the Australian Institute of Health and Welfare Act 1987 (which carries additional privacy protections for companies and deceased people).

Strict separation of identifiable information and content data is maintained within the Data Linkage Unit in accordance with the AIHW linkage protocols, so that no 1 person has access to both types of data. Individually identifiable information is not released, and no individual can be identified in any reporting. The AIHW also does not release SLKs. The linked data set created for this study will be stored securely on site at the AIHW for 10 years.

Scope of analysis data set

The scope of the analysis data set is homelessness support received between 1 July 2011 and 30 June 2017, which was accessed, after discharge, by ex-serving ADF members with at least 1 day of service on or after 1 January 2001. 

Scoping of contemporary ex-serving population in the PMKeyS 

The PMKeyS extract provided by the Department of Defence contained 115,551 records for ex-serving ADF members, which was thereafter scoped using the process outlined here.

To be included in the final PMKeyS data set, a record must have a termination date (date of discharge) after 1 January 2001 and before 1 July 2017, and at least 1 day of service between the hire date and the termination date. In the PMKeyS data set, 6,726 records were excluded from analysis because they had: 

  • a termination date after 30 June 2017


  • a termination date on or before 1 January 2001


  • a reason for discharge of ‘Military—Irregular enlistment’ or ‘Military—Failed to enlist’ and less than 1 year of service. Individuals discharged for these reasons were considered out of scope as they do not meet the criteria of having served at least 1 day in the ADF. Individuals within this group complete the enlistment process and do not attend the first day of service; therefore, they never complete 1 day in service


  • an exact duplicate record of another record


  • a missing age


  • less than 1 day of service.

Unique individuals with multiple records in the PMKeyS data were processed by combining key information from both records where appropriate to create a unique record for analysis.

From the original 115,551 records provided in the PMKeyS extract, the scoping processes resulted in 108,825 unique in-scope records for ex-serving ADF members in the PMKeyS data set. 

Scoping of analysis population in the SHSC

SHS use was in-scope if the services were accessed between 1 July 2011 and 30 June 2017 by an ex-serving ADF member. As such, the termination dates of clients were compared with the financial years of SHS use to determine if any records were out of scope. A record of SHS service use was considered in-scope if it occurred in the same financial year as the client’s termination date, or occurred in a later financial year. 

Before scoping, there were 1,425 distinct matches between the PMKeyS and the SHSC, which corresponded to 3,462 records (there are multiple records per client in the SHSC data set). After SHSC records were compared against a client’s termination date to determine whether they were in-scope for analysis, 3082 records remained, which represented 1,215 individual clients who linked between the PMKeyS and SHSC.