3303.0.55.002 - Microdata: Mortality, Enhanced Characteristics, Australia , 2011-12 Quality Declaration 
Latest ISSUE Released at 11:30 AM (CANBERRA TIME) 14/12/2016  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All RSS Feed RSS Bookmark and Share Search this Product

EXPLANATORY NOTES


INTRODUCTION

1 The Mortality, Enhanced Characteristics microdata product contains death registration records with added variables from the 2011 Census of Population and Housing. This has been achieved through a process of probabilistic linkage, whereby death registrations were linked to the 2011 Census of Population and Housing. The key feature of the linkage methodology used is the ability to locate records on two datasets that may refer to the same person, by evaluating the degree of commonality exhibited by a range of personal variables common to the two datasets. In total, 153,455 deaths were registered in the period from 10 August 2011 and 27 September 2012. The probabilistic linkage strategy linked 123,910 of these death registrations with records on the 2011 Census. For further information on the linkage strategy, see Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058).

2 The microdata product was produced as part of the Death Registrations to Census Linkage Project: A Linked Dataset for Analysis. The aim of the project was to re-create a linked death registrations to census dataset which had been previously conducted under the Census Data Enhancement (CDE) program. The methods employed for the microdata product involved using demographic variables common to both datasets, but which excluded name and address, to produce a single numerical measure of how well two records match. The original Death Registrations to Census Linkage Project was conducted to evaluate the consistency of Indigenous identification reported in the Death Registrations data and the Census data, and subsequently provide input into life tables and life expectancy estimates for Aboriginal and Torres Strait Islanders. The original project fell under the terms of the CDE program, which in conjunction with other personal characteristics, allowed for name and address information to be used in the data integration process to create a "gold standard" of data linkage. This file was not available for external use, but it did provide a benchmark for future linkage projects to be measured against, as is the case with the current microdata product.


DATA SOURCES

3 Mortality-related data items in the linked dataset come from the death registration process. Death registrations are provided to the Australian Bureau of Statistics by State and Territory Registrars of Births, Deaths and Marriages. As part of the registration process, information about the cause of death is supplied by a medical practitioner certifying the death, or by a coroner. Other information about the deceased is supplied by a relative or other person acquainted with the deceased, or by an official of an institution where the death occurred. For deaths which are certified by a coroner, the ABS also receives information pertaining to cause of death from the National Coronial Information System (NCIS). For further information about the death registration process, information sources and scope and coverage of death registrations in Australia, see the Explanatory Notes in Causes of Death, Australia (cat. no. 3303.0).

4 The 2011 Census of Population and Housing was held on Tuesday, 9 August, 2011. The Census provides a wealth of information about the Australian community. Further information about the 2011 Census and links to reference information, including the 2011 Census Dictionary, can be found on the 2011 reference and information page of the ABS website. The Census variables included in the linked dataset can be found in the Data Item List, in the downloads tab.


SCOPE AND COVERAGE

Scope of causes of death statistics in the microdata product

5 This data integration project includes 153,455 Death Registrations where the death occurred between August 10, 2011 and September 27, 2012. This reference period for the project was selected in order to capture as many deaths as possible of people who were counted in the 2011 Census of Population and Housing which was conducted on August 9, 2011.

6 Each year there are lags between the occurrence of a death and the registration of a death. This means that during a specified period, the number of deaths registered would be lower than the number of deaths that occurred, particularly deaths which occurred towards the end of a reference period.


Coverage of causes of death statistics

7 The ABS Causes of Death collection includes all deaths that occurred and were registered in Australia, including deaths of persons whose usual residence is overseas. Deaths of Australian residents that occurred outside Australia may be registered by individual Registrars, but are not included in ABS deaths or causes of death statistics.

8 The scope of causes of death statistics includes:

  • all deaths being registered for the first time;
  • deaths in Australia of temporary visitors to Australia;
  • deaths occurring within Australian Territorial waters;
  • deaths occurring in Australian Antarctic Territories or other external territories (excluding Norfolk Island);
  • deaths occurring in transit (i.e. on ships or planes) if registered in the State of 'next port of call';
  • deaths of Australian Nationals overseas who were employed at Australian legations and consular offices (i.e. deaths of Australian diplomats while overseas) where able to be identified; and
  • deaths that occurred in earlier reference periods that have not been previously registered (late registrations).

9 The scope of the statistics excludes:
  • repatriation of human remains where the death occurred overseas;
  • deaths overseas of foreign diplomatic staff (where these are able to be identified);
  • deaths occurring on Norfolk Island; and
  • stillbirths/fetal deaths (these are included in perinatal counts


Scope and coverage of 2011 Census of Population and Housing

10 The 2011 Census of Population and Housing was held on August 9, 2011. All people in Australia on Census Night are in scope, except for foreign diplomats and their families. Visitors to Australia are counted regardless of how long they have been in the country or how long they plan to stay. Australian residents not in the country on Census Night are out of scope of the Census.


CLASSIFICATIONS

11 Classifications used in the mortality and Census datasets include:


International classification of diseases

12 Causes of death statistics are coded to the International Classification of Diseases 10th Revision (ICD-10). The ICD is the international standard classification for epidemiological purposes and is designed to promote international comparability in the collection, processing, classification, and presentation of causes of death statistics. The classification is used to classify diseases and causes of disease or injury as recorded on many types of medical records as well as death records. The ICD has been revised periodically to incorporate changes in the medical field.

13 The ICD-10 is a variable-axis classification meaning that the classification does not group diseases only based on anatomical sites, but also on the type of disease. Epidemiological data and statistical data is grouped according to:
  • epidemic diseases;
  • constitutional or general diseases;
  • local diseases arranged by site;
  • developmental diseases; and
  • injuries.

14 For example, a systemic disease such as sepsis is grouped with infectious diseases; a disease primarily affecting one body system, such as a myocardial infarction, is grouped with circulatory diseases; and a congenital condition, such as spina bifida, is grouped with congenital conditions.

15 For further information about the ICD refer to WHO International Classification of Diseases (ICD).

16 The records in the linked dataset were coded according to the 2006 version of the ICD-10, with the exception of two more recent updates applied by the ABS. The first update was applied in 2007 and relates to the use of mental and behavioural disorders due to psychoactive substance use, acute intoxication (F10.0, F11.0...F19.0) as an underlying cause of death. If the acute intoxication initiated the train of morbid events it is now assigned an external accidental poisoning code (X40-X49) corresponding to the type of drug used. For example, if the death had been due to alcohol intoxication, the underlying cause before the update was F10.0, and after the update the underlying cause is X45, with poisoning code T51.9. The second update implemented from the 2009 reference year was the addition of Influenza due to certain identified virus (J09) to the Influenza and Pneumonia block. This addition was implemented to capture deaths due to Swine flu and Avian flu, which were reaching health epidemic status worldwide.


Types of death

17 All causes of death can be grouped to describe the type of death, whether it be from a disease or condition, from an injury, or whether the cause is unknown. These are generally described as:
  • Natural Causes - deaths due to diseases (for example diabetes, cancer, heart disease etc.) (A00-Q99, R00-R98)
  • External Causes - deaths due to causes external to the body (for example intentional self-harm, transport accidents, falls, poisoning etc.) (V01-Y98)
  • Unknown Causes - deaths where it is unable to be determined whether the cause was natural or external (R99).


CAUSES OF DEATH IN THE LINKED DATASET

Underlying cause of death and multiple causes of death

18 Causes of death are most commonly presented based on the Underlying Cause of Death (UCOD), which is the disease or injury that initiated the morbid events leading directly to death. Accidental and violent deaths are classified according to the external cause, that is, to the circumstances of the accident or violence which produced the fatal injury rather than to the nature of the injury. In the linked dataset, the underlying cause of death is represented by the 'UCOD' variable.

19 Statistical analyses can also be conducted using multiple causes of death. Multiple causes of deaths refer to all morbid conditions, diseases and injuries entered on the death certificate. These include those involved in the morbid train of events leading to death which were classified as either the underlying cause, the immediate cause, or any intervening causes, as well as those conditions that contributed to death but were not related to the disease or condition causing death. For deaths where the underlying cause was identified as an external cause (for example suicide, a fall, homicide etc.) multiple causes include the circumstances of injury/poisoning and the nature of injury/poisoning as well as any other conditions reported on the death certificate.

20 Multiple causes of death can be selected in the linked dataset using the variables MCOD_1 to MCOD_14. No death registrations in this dataset recorded more than 14 conditions on the death certificate. In principle, MCOD_1 represents the underlying cause of death. This means, the ICD-10 code in the field for UCOD and MCOD_1 in the detailed microdata product for a record should be the same. The MCODs following MCOD_1 (MCOD_2, MCOD_3 etc.) are listed alphabetically, according to their ICD-10 code and have no relationship to ordering of conditions on the death certificate.

21 There are a small number of records in the linked dataset for which the UCOD and MCOD_1 do not match. The mismatch in these variables can be divided into two types. Firstly, the UCOD does not equal MCOD_1, but appears in the MCOD_2 to MCOD_14 fields. For such cases, the UCOD should be considered as the official underlying cause, and all other codes should be considered as associated causes. The second case is where the UCOD does not appear in any of the MCOD1-14 filelds. In these cases it is recommended that multiple cause analysis not be performed.


RELIABILITY OF ESTIMATES

22 Error in estimates produced using the mortality enhanced characteristics data file may occur due to false links and missed links.

Link rate

23 A link rate of 81% was achieved between the census and mortality datasets, resulting in a dataset of 123,910 records. However this link rate was not achieved uniformly over all possible sub-populations of interest within the mortality dataset.

24 For example, link rates were lower than average for the overlapping subpopulations of:
  • Aboriginal and Torres Strait Islander People;
  • People living in remote and very remote regions;
  • Residents of the Northern Territory; and
  • Persons aged under 50 years.
Among causes of death, the link rates for intentional self harm were lower than those of the other selected causes. A weighting strategy has been used to partly mitigate for this: See below.

25 Subpopulations with low link rates should be analysed with caution. See the Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058) for considerations and recommendations for these subpopulations in the dataset.


WEIGHTS - STAGE 1 AND STAGE 2

26 In total, 153,455 deaths were registered in the period from 10 August 2011 and 27 September 2012. The probabilistic linkage strategy linked 123,910 of these death registrations with records on the 2011 Census. To increase the representativeness of the linked file, a two stage weighting strategy was applied. The Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058) alerts users of this data to pertinent issues that arise in the analysis of linked data. Familiarity with these issues will assist users to frame suitable research questions, select appropriate weighting options, and make sound inferences.

27 The linkage rate of 81% (123,910 linked deaths/153,455 death registrations) is not recognised uniformly over all subpopulations of interest. For example, people living outside of capital cities had a lower linkage rate than people living within capital cities. To adjust for relative under or over representation of selected subpopulations in the linked dataset, Stage 1 weights can be applied to the microdata product. The demographic categories used to define subpopulations for the Stage 1 weights are shown in the table below:


Sex
Age Cohort
Migrant City

M= MaleAO= Born after 1945MO= Born in AustraliaCO= Does not live in a major capital city
F= FemaleA1= Born after 1946M1= Born in Europe (incl. UK)C1= Lives in a major city
M2= Born elsewhere overseas



28 Stage 1 weights can also be modified to derive Stage 2 weights. Specifically, the Stage 2 weights enable analysts to restore the weighted totals of selected aggregates, which is useful for analysis, especially when measuring against external sources. This second weight can be applied to the microdata product in two scenarios. Firstly, a need may exist to match record counts for selected categories in the original Deaths Registrations dataset. For example, 15 selected causes of death have a Stage 2 weight available for application, and when these are summed with the weight applied, will amount to the total Death Registrations. Secondly, there are a small number of examples where the Stage 1 weight alone does not restore the representativeness of a variable in all dimensions. For detailed information regarding the cause of death weightings and other issues relating to the weighting strategy see: Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058).

29 The multiple cause of death data was an addition to the Microdata: Mortality, Enhanced Characteristics, 2011-12, product after the weighting strategy had been applied. Therefore, it is recommended that weights are not applied when analysing the multiple cause of death dataset.

30 As the weights are designed to adjust for under and over representative of subpopulations in the microdata product, the numerical value assigned to the weight do not equal a whole rounded number (e.g. the Stage 2 Weight for blood and lymph cancers is approximately 0.95). This means that when a weighting is applied to a variable in the dataset, the total number will not be a total (e.g. 110.5). It is recommended that weighted numbers be rounded in output.


DATA CONSIDERATIONS

31 Please note that this microdata file contains 123,907 records. This is three fewer than the 123,910 records in the original probabilistic-linked dataset and supplementary tables presented in the Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058). This is a result of three death registrations being recognised as out of scope for linkage purposes subsequent to the linkage process.

32 As noted in the Research Paper: Death Registrations to Census Linkage Project - A Linked Dataset for Analysis, Mar 2016 (cat. no. 1351.0.55.058), the weighting strategy applied to the dataset was experimental, with the aim being to provide weights which would be fit for purpose for analytical purposes. There may be potential for future products to consider other weighting strategies if deemed necessary for analytical purposes.

CONFIDENTIALITY

33 In accordance with the Census and Statistics Act 1905, all extracted data will be subjected to a confidentiality process before release. This process is undertaken to minimise the risk of identifying particular individuals, families, households or dwellings in aggregate statistics, through analysis of published data.

FURTHER INFORMATION

34 It is recommended that users of the microdata product consult the glossary, explanatory notes and technical materials included in Causes of Death, Australia (catalogue number 3303.0) and the reference and information relating to the 2011 Census of Population and Housing: 2011 reference and information.