Data source – NACS linked dataset
On this page:
The data set used for these analyses was created by the AIHW Data Linkage Unit by linking health and welfare data sets held by the AIHW. Approval for this project was provided by the AIHW ethics committee under project number EO2023/2/1413.
The final linked data asset is referred to as the NACS dataset and contains the following source datasets:
- National Death Index (NDI) January 2011 to May 2024
- Alcohol and Other Drug Treatment Services (AODTS) July 2012 to June 2023
- Commonwealth primary health datasets
- Pharmaceutical Benefits Scheme (PBS) January 2007 to February 2023
- Medicare Benefits Schedule (MBS) January 2007 to February 2023
- Specialist Homelessness Services Collection (SHSC) July 2011 to June 2023.
Data linkage
To create the study cohort, identifiers from Specialist Homelessness Service Collection (SHSC), Alcohol and Other Drug Treatment Services (AODTS) and National Death Index (NDI) were linked to AIHW Enhanced Medicare Spine (EMS). The EMS includes all individuals who have registered with Medicare since 1984, based on Medicare enrolment data.
The cohort was formed as a combination of all individuals (linked and unlinked) from SHSC, AODTS and NDI datasets for the period from 01/01/2012 to 31/12/2022 (or latest available data) and consisted of SLK-581’s (statistical linkage key 581). The SLK-581 is a concatenation of s235 (2nd, 3rd and 5th letters of surname), f23 (2nd and 3rd letters of first name), date of birth and sex. SLKs are used to preserve privacy in record keeping and during linkage. In general, the uniqueness of SLK-581 keys for a given Australian population is high.
For the SHSC and the AODTS, an initial stage of simple merging by common fields was undertaken. First, all unique matches by SLK and postcode were obtained, followed by all unique matches by SLK and jurisdiction. This stage was followed by a ‘dropout-1 linkage’ to complete the results. The ‘dropout-1 linkage’ has been found to be an effective approach to linkages using the statistical linkage key SLK-581 together with postcode. This approach allows for disagreement of a single SLK component or postcode in potential record pairs found multiple deterministic passes.
Linkage results – SHSC
A total of 1,262,977 records were linked from the SHSC cohort, which accounts for 79.25% of the 1,593,584 cohort SLKs. Previous experience with linking SHSC data to spine versions containing Medicare and DOMINO identifiers suggests that a significant portion of the remainder is likely composed of alternative SLKs for cohort members who have otherwise already found EMS matches but were not able to be linked without the benefit of DOMINO data. Despite this, since 1 in 5 records could not be matched, the number of people who died and had received support from SHS is likely to be an underestimate of the true number of deaths.
There were no notable differences in linkage rates between males and females. Unlinked records had a larger number of missing identifier fields and were more likely to have dummy day, month and year of birth, which are permitted within the SHSC.
There is some variation in the linkage rates across time (table 1), as expected. For years prior to 2017–18, SHSC data were weighted to account for data coverage issues such as agency non-response and invalid SLKs. Weighting was not required from 2017–18 onwards due to the high rates of agency response and SLK validity. Weighting was not applied to the SHSC data for the NACS project, and therefore, the total number of SHS clients for the periods prior to 2017–18 is an underestimate, which means the number of deaths among SHS clients is also an underestimate.
| Financial year of service | Percentage linked |
|---|---|
| 2011–12 | 85.7 |
| 2012–13 | 86.4 |
| 2013–14 | 87.3 |
| 2014–15 | 87.2 |
| 2015–16 | 87.6 |
| 2016–17 | 88.2 |
| 2017–18 | 88.4 |
| 2018–19 | 88.8 |
| 2019–20 | 88.2 |
| 2020–21 | 88.8 |
| 2021–22 | 88.8 |
| 2022–23 | 88.9 |
Note: Records in this table refer to a unique combination of SLK/year/state/postcode/suburb using client geography the week before presentation and the unique combination of SLK/year/state/postcode/suburb using the last permanent address.
National Death Index (NDI)
The NDI data provided for this project included all records, including those that were not linked to the EMS. Data cleaning was undertaken to remove potential duplicate records and any records that did not have a cause of death.
In the version of the NDI database used in the NACS, data were considered final for those deaths registered prior to 2021, revised for deaths registered in 2021, and preliminary for deaths registered in 2022 and 2023. Preliminary data are subject to further revision. Data for 2018–2022 reflect the updated ABS enhancements applied from 2024, as outlined in the Technical Note.
Age
Exact date of birth information is not available in the NACS dataset, with only month and year of birth included from base data sources. Date of birth is therefore approximated as the 15th of the birth month for all clients. Where available, the month and year of birth information is sourced from the Medicare Consumer Directory (MCD), and from either the SHSC or AODTS for clients who have not died. Date of birth information is not available for some clients due to data quality concerns.
Age at death is only able to be calculated for people who either link to the MCD or to either the SHSC or AODTS datasets. This excludes groups such as infants (who do not have an MCD listing as they are not yet enrolled in Medicare), overseas visitors and temporary residents, asylum seekers and refugees. More information about who is eligible for Medicare is provided at Enrolling in Medicare.
For the report about children who have died, the analysis was limited to children aged from 1 to 17. This is because the age at death calculation primarily relies on people being enrolled in Medicare, and infants who die before the age of 1 may not yet have an MCD record.
For the report about people who received specialist homelessness services and alcohol and other drug treatment services who have died, the analysis was limited to people who received SHS support aged 10 and older. This is because of the same age restriction among people accessing specialist alcohol and other drug treatment services.
Comparisons with other AIHW publications about deaths in Australia should be approached with caution due to potential differences in methodological approaches.
Medicare Benefits Schedule (MBS)
MBS data for the five years preceding the date of death was included for all people who died. MBS data was not included for people who were not linked to the NDI. Data from January 2017 to February 2023 were included in the final dataset.
Pharmaceuticals Benefit Scheme (PBS)
PBS data for the five years preceding the date of death was included for all people who died. PBS data was not included for people who were not linked to the NDI. Data from January 2017 to February 2023 were included in the final dataset.