This appendix provides information on the data used in the report and on issues relevant to interpreting the data.
Data on fatal injuries in this report are from the AIHW National Mortality Database (NMD). The NMD comprises cause of death unit record file (CODURF) data, which are provided to the AIHW by the Registries of Births, Deaths and Marriages and the NCIS and include cause of death coded by the ABS.
Data are presented according to the financial year in which each death occurred, rather than the calendar year in which the death was registered. There are 2 main reasons why the data are presented in this manner. Firstly, presenting data by year of occurrence provides a more meaningful interpretation of data, compared with presenting data by year of registration, because cases can be registered at a time significantly later (in some cases years later) than when death occurred. Secondly, reporting by financial year is in line with the practice in AIHW reports on injury morbidity, enabling deaths and hospitalisations to be compared for the same period.
Records that met the following criteria were included in this report:
- deaths that occurred between 1 July 2007 and 30 June 2016 and had been registered by 31 December 2016; and
- the UCoD was an external cause code in the range V01–Y36; or
- at least 1 MCoD was an external cause code in the range V01–Y36 and at least 1 other MCoD was a code for injury (S00–T75 or T79).
Deaths were excluded where date of death was unknown. The codes are from the WHO International Statistical Classification of Diseases and Related Health Problems, 10th revision (WHO 2016). The external cause codes are from Chapter XX External causes of morbidity and mortality and the injury codes are from Chapter XIX Injury, poisoning and certain other consequences of external causes.
Multiple causes of death (MCoD)
Box 1.1 provided standard definitions of the terms underlying cause of death (UCoD) and multiple causes of death (MCoD) codes. MCoD codes in this report relate to the causes of death that contributed to death and may or may not have been related to the underlying cause. For example, an elderly person falls and fractures their hip. This person’s advanced age, frailty and perhaps other comorbid conditions limits their capacity to tolerate injury, leading to their death. In this instance, this record may be assigned an UCoD of an external cause code for fall (W00–W19) and a MCoD code for hip injury (S72).
In another example, an elderly person might suffer a heart attack that results in a fall and a hip fracture. As with the first example, a combination of factors leads to death. In this instance, this record might be assigned an UCoD code for acute myocardial infarction (I21) and a MCoD of an external cause code for fall (W00–W19) and a MCoD code for hip injury (S72).
Both of these cases would be included in this report because the first example meets the second of the criteria listed above, while the second example meets the third criterion.
Coding of deaths data
The ABS obtains deaths data from the registries of births, deaths and marriages in each state and territory, which, in turn, obtain information from the doctor or coroner who certifies each death.
The ABS codes causes of death according to the 10th revision of the International Classification of Diseases (ICD-10) and, after de-identification, creates the Cause of Death Unit Record File (CODURF). Most of the coding is done using an automated coding system.
If a death was due to an injury, the ICD-10 requires coding of the ‘external cause’ of the injury (such as a car crash of a particular type) as the underlying cause of death (UCoD). Most injury deaths are certified by a coroner; for these deaths, the ABS seeks the additional information required to code external causes from the NCIS.
Some injury deaths, and most deaths from other causes, are certified by a medical practitioner. In these instances, ABS coders rely on information about causes of death that was entered onto the death certificate. Of the deaths included in this report, the most common type of injury in doctor-certified deaths is ‘Fall’.
The result of this process is a record in an annual ABS mortality data file that summarises characteristics of the person who died (for example, age, sex and Indigenous status) and characteristics of his or her death (for example, causes, date and place at which the person usually lived).
Certain aspects of the method used by the ABS have differed according to the registration year of deaths during the period covered by this report. The reasons for making the changes and their nature have been reported by the ABS (ABS 2009). The changes are described in more detail in a previous report dealing with trends in injury deaths (AIHW: Henley & Harrison 2018).
Data on SES quintiles for deaths registered between 2009 and 2015 are defined using the ABS’s Socio-Economic Indexes for Areas (SEIFA), Australia, 2011 (ABS 2013), while data on SES quintiles for deaths registered in 2016 are defined using the ABS’s Socio-Economic Indexes for Areas (SEIFA), Australia, 2016 (ABS 2018b).
The SEIFA data are generated by the ABS using a combination of Census data such as income; education; health problems/disability; access to internet; occupation/unemployment; wealth and living conditions; dwellings without motor vehicles; rent paid; mortgage repayments; and dwelling size. Composite scores are averaged across all people living in areas and defined for areas based on the Census collection districts. However, they are also compiled for higher levels of aggregation. The SEIFAs are described in detail on the ABS website.
The SEIFA Index of Relative Socio-Economic Disadvantage (IRSD) is one of the ABS’s SEIFA indexes. The relative disadvantage scores indicate the collective SES of the people living in an area, with reference to the situation and standards applying in the wider community at a given point in time. A relatively disadvantaged area is likely to have a high proportion of relatively disadvantaged people. However, such an area is also likely to contain people who are not disadvantaged, and people who are relatively advantaged.
Mortality rates by SES, were generated by the AIHW using the IRSD scores for the SA2 of usual residence of the patient, reported or derived for each separation. The ‘1—Lowest’ group represents the areas containing the 20% of the national population with the most disadvantage, and the ‘5—Highest’ group represents the areas containing the 20% of the national population with the least disadvantage. These socioeconomic groups do not necessarily represent 20% of the population in each state or territory. Disaggregation by socioeconomic groups is based on the area of usual residence of the deceased person.
The following labels for each socioeconomic group have been used throughout this report:
||Socioeconomic status group
|1 – Lowest
||Second most disadvantaged
||Second least disadvantaged
|5 – Highest
Deaths per 100,000 population are reported as directly age-standardised rates based on the Australian population as at 30 June of the year of interest. The Australian population as at 30 June 2001 was used as the reference population (ABS 2001). Age-standardisation of rates enables valid comparison across years and/or jurisdictions without being affected by the differences in age distributions.
Where possible, rates were calculated using the final estimated resident population (ERP) as at 31 December in the relevant year as the denominator (for example, 31 December 2015 for 2015–16 data). Where tables of 31 December ERPs were not available, but tables of 30 June ERPs were available, population denominators were calculated as the average of 30 June estimates for adjacent years.
Due to issues with the availability of data, SEIFA population data for the period from 2009–10 to 2015–16 was a combination of Australian Standard Geographical Classification (ASGC) based estimates for the period from 2009–10 to 2010–11 and Australian Statistical Geography Standard (ASGS)-based estimates for the period from 2011–12 to 2015–16.
Although numerator data for the period from 2009–10 to 2015–16 were ASGS based, investigations indicated that the use of ASGC-based population estimates for the first 2 years of the period would still provide a reasonable estimate of rates for each socioeconomic group over the period from 2009–10 to 2015–16.
The Australian Statistical Geography Standard (ASGS) provides a framework of statistical areas used by the Australian Bureau of Statistics (ABS) and other organisations to enable the publication of statistics that are comparable and spatially integrated. First introduced in 2011, the ASGS replaced the Australian Standard Geographical Classification (ASGC) that had been in use since 1984. The ASGS provides users with an integrated set of standard areas that can be used for analysing, visualising and integrating statistics produced by the ABS and other organisations.
Estimated change in rates over time
Estimated trends in rates of deaths were reported as annual percentage change, obtained using negative binomial regression modelling using Stata 14.2 (StataCorp 2018).
The use of the terms ‘significant’ or ‘significantly’ throughout this report indicates an outcome that was statistically significant (p < 0.05).
Population-based rates of injury tend to have a similar value in 1 year and the next. Exceptions to this can occur (for example, due to a mass-casualty disaster) but are unusual in Australian injury data. Some year-to-year variation and other short-run fluctuations are to be expected, due to unknown and essentially random factors, and so small changes in rates over a short period normally do not provide a firm basis for asserting that a trend is present.
However, the period covered by this report is long enough for noteworthy changes to occur. The fundamental questions concerning a series of annual estimates of population-based rates are whether they show a statistically significant rise or fall over the period and, if so, the average rate of change. Analysis in this report is limited to those characteristics of change.
For each type of injury for which estimates of change were made:
- age-adjusted annual case numbers were obtained by multiplying age-adjusted unscaled rates by the Australian population in the corresponding year
- negative binomial regression, a method suitable for count-based data, was run with the adjusted case numbers as the dependent variable; year (as an integer, from 0 to the number of years of data) as an independent variable; and annual population as the exposure. The relevant outputs are a modelled rate for each year and a model-based estimate of average annual change in rate and its 95% confidence interval (CI).
If the 95% CI around the point estimate for trend is entirely above zero then the rates have tended to rise; if the 95% CI is entirely below zero then the rates have tended to fall— otherwise it cannot be said with useful confidence that the age-standardised rates tended to rise or to fall in the period considered.