Potential Data Quality Issues Affecting External Causes of Death Data
Data quality is an important issue for every type of data collection, whether survey, census or administrative by-product, and the ABS makes every effort to achieve a high level of quality within the constraints of the resources available. Data quality depends on the procedures being followed at every stage of the collection and processing of the data. There are a number of factors which have the potential to affect data quality and, as far as possible, these can be minimised by the application of quality assurance measures.
The discussion in ABS Quality Framework and Causes of Death Collection provided an overview of potential quality issues in relation to ABS causes of death statistics. This section provides detail about particular issues currently relevant to external causes of death data.
CORONIAL PROCESSES
Consistency of practices across jurisdictions
There is currently a lack of standardisation in the way that coronial deaths are reported across Australia because different reporting formats, structures and forms are used in different coronial offices, states and territories. For example, coronial statements about the intent of a death are worded in different ways, and can be located anywhere in the coronial finding (meaning coders have to read the entire document). Some of the other areas where practices vary across jurisdictions are discussed further below.
The NCIS aims to support and encourage the national standardisation of certain coronial practices, in order to enhance the accuracy and efficiency of interpretation of coronial data. Similarly, the NCIS also aims to encourage the improvement of consistency and detail within various coronial documents which are made available via the NCIS. These include coronial findings and narratives of police reports.
Changes to certification practices
In general, coroners are responsible for certifying the causes of death for reportable deaths (i.e. those deaths which are reported to the coroner). However, procedures around reportable deaths are subject to change over time and across the states and territories. For example, recent efforts to minimise the risk of falls and fall-related injuries changes in Victoria have resulted in coronial involvement if a death results from a fall in a Victorian hospital. The effect of this in statistical terms is to ensure that any falls associated with a death are more likely to be mentioned on the original Medical Certificate of Cause of Death resulting in the fall being able to be coded as a cause of death (previously, only injury details such as fractured neck of the femur may have been mentioned).
In addition, certification of particular causes of death may be influenced by changes in general awareness or sensitivities, and/or by changes in medical understanding of particular causes of death. Over time, there have been changes in the pattern of certification related to different causes. Examples include 'old age (or senile debility)' which was once a leading cause of death but is now rarely used, or neoplasms (cancer) or Sudden Infant Death Syndrome (SIDS) which are used in more recent times.
Determination of intent
Coronial processes to determine the intent of a death (whether intentional self harm, accidental, homicide, undetermined intent) are especially important for statistics on suicide deaths because information on intent is necessary to complete the coding under ICD-10 coding rules.
However, coroners' practices to determine the intent of a death may vary across the states and territories. In general, coroners may be reluctant to determine suicidal intent (particularly in children and young people). In some cases, no statement of intent will be made by a coroner. The reasons may include legislative or regulatory barriers, sympathy with the feelings of the family, or sensitivity to the cultural practices and religious beliefs of the family.
For some mechanisms of death where it may be very difficult to determine suicidal intent (e.g. single vehicle accidents, drownings), the burden of proof required for the coroner to establish that the death was suicide may make a finding of suicide less likely.
CODING PROCESSES
Timing of data compilation
Causes of death statistics are released by the ABS on a calendar year basis and the current aim is to release the data within 12 months of the end of the reference period.
The timing of the compilation (including coding) and release of data is necessarily a compromise between two important aspects of quality - timeliness and accuracy. Recent issues with accuracy for external causes of death data have required a delay in the release of causes of death data to allow for further query action on reportable deaths to enable more specific causes of death coding in some cases. This has increased the time between the reference period and the release of the statistics to about fifteen months after the end of the reference period (for the 2004 and 2005 reference years).
Time required before coronial cases can be closed
Causes of death coding does not rely on all deaths registered in the reference year which were referred to the coroner, being investigated and the cases closed. The coronial process can take several years if an inquest is being held or complex investigations are being undertaken. However, any general increase in the length of coronial investigations (or in the workload of coroners) has the potential to affect data quality in terms of accuracy given that the need for timely information limits the amount of time available to wait for the findings of the longer cases. The fact that a case is still open limits the amount of information available to code causes of death, and may result in a less specific code being allocated consistent with ICD-10 coding rules.
As at 14 December 2006, 77% of cases for reportable deaths registered in 2005 were closed, according to the NCIS database. The lowest proportions of cases which were closed were recorded for New South Wales (58% of cases closed), Australian Capital Territory (69%), Western Australia (72%) and Queensland (73%). In the previous year, 73% of cases were closed at a similar stage (for deaths registered in 2004, as at 19 December 2005).
When 2005 causes of deaths statistics were finalised at the beginning of 2007, there were 337 coronial cases relating to deaths registered in 2005 still open or requiring more specific information. Of these, New South Wales had 109, Victoria had 29, Queensland had 96, South Australia had nil, Western Australia had 93, Tasmania and the Northern Territory had nil, and the Australian Capital Territory had ten.
Completeness of information available on the NCIS
As discussed above, not all coronial cases are closed when causes of death statistics relating to deaths registered in a particular calendar year are compiled at the end of the following calendar year. While the fact that a case is still open on the NCIS limits the amount of information available to code causes of death, there may be sufficient information available from documents recorded on the NCIS (such as autopsy, toxicology or police reports), or from the fields within the NCIS database, to code causes of death (except where information on intent is required).
In response to an increase in the number of coronial cases with incomplete information on the NCIS, the following actions were taken to reduce the number of such records used as the basis for the causes of death statistics for 2004 and/or 2005:
- additional time allowed for coronial cases to be finalised and documented before the finalisation of ABS causes of death statistics
- accessing of paper coronial files at selected coroners' offices by ABS staff
- additional scanning and attachment of hard copy police reports by NCIS
- employment of additional (casual) NCIS coders (coronial clerks) in selected coroners' offices
- employment of casual administrative support to enable trained NCIS coders to focus on coding current cases
- special training for coronial clerks in selected country areas to focus effort on current cases in terms of NCIS coding and case closure.
These actions involved jurisdictions/coronial offices where the additional effort was expected to be most effective, and were made possible by the full cooperation of the NCIS and coronial staff. While only the first two types of actions were possible for 2004 data, all of these measures were employed for 2005 data. They resulted in an increase in the amount of information available for coding causes of death than would otherwise have been possible for the 2004 and 2005 reference years
(footnote 6).
ICD-10 coding rules in cases of incomplete information
The ABS undertakes a large amount of query action to identify and minimise the number of cases with incomplete information for coding purposes. However, given that some coronial investigations will be lengthy, there will always be some cases which are still open at the point when statistics need to be finalised ready for timely release. In cases where, despite all efforts, the information sources are still incomplete at the processing cut-off date, coding is undertaken according to ICD-10 rules using the available information.
Even if a case is not yet closed (or the details are not yet all available on the NCIS), there may still be sufficient information available from the attachments on the NCIS (such as autopsy, toxicology or police reports) or from the fields within the NCIS database. For traffic accidents the details of the accident can usually be found in the police report or the autopsy report or within the database itself.
The causes of death code resulting from the application of the ICD-10 coding rules depends on the wide range of information which these rules require as input to the decision, and which will vary from case to case. A lack of critical information at the point of coding adds another level of complexity. However, it is possible to indicate in broad terms, the way in which the coding rules will direct the likely outcome depending on the nature of the information that is missing (noting that individual coding outcomes may differ). The following table indicates the level of coding which can be achieved according to ICD-10 rules when information for coding is incomplete.
INCOMPLETE INFORMATION FOR EXTERNAL CAUSES OF DEATH CODING, Coding outcomes
Required data elements |  |
| |
Whether external to
the body | Mechanism
of death | Intent
of death* | Coding outcome (following ICD-10 rules) |
|
Y | Y | N |
- Classified as an external cause.
- May be accidental, suicide or assault.
- Classified as Event of undetermined intent (Y10-Y34) where the coroner has clearly determined that the intent of the death cannot be established from the available evidence.
- Classified as an assault if the mechanism was stabbing.
- Otherwise classified as an accident
- nature of the accident is classified using the information about the mechanism
- in cases where information on the precise mechanism is not available, the case can be allocated to the correct group of codes for a particular mechanism but within this will be coded to an 'unspecified means' category. For example, if the case involved accidental poisoning, but the type of chemical or noxious substance was unclear.
|
|
Y | N | N |
- Classified as an external cause.
- May be accidental, suicide or assault
- Classified as Event of undetermined intent (Y10-Y34) where the coroner has clearly determined that the intent of the death cannot be established from the available evidence.
- Otherwise classified as an accident
- nature of the accident cannot be classified
- classified as exposure to unspecified factor (X59) because the mechanism cannot be established.
|
|
N | N/A | N/A |
- Classified as a natural death.
- Classified as Ill-defined and unknown causes of mortality (R95-R99) as deaths that had no classifiable cause recorded and/or where no more specific cause could be attributed even after all the facts bearing on the case had been investigated by the coroner. This group includes unattended death (R98) and other ill-defined and unspecified causes of mortality (R99)
- in a small number of cases, the other ill-defined and unspecified causes of mortality (R99) category may be used if the coroner has not yet closed the case.
|
|
| * Usually only missing in cases which are still open and which may be suicide or assault, for which a coronial decision on intent is needed (however in some cases the coronial finding may not include any information on intent). |
In order to classify a death as suicide (intentional self-harm) the ICD-10 requires that specific documentation from a medical or legal authority be available regarding both the self-inflicted nature and suicidal intent of the incident. If this information is not available then the death must be classified as accidental. The case generally needs to be closed by the coroner to code a suicide as such, unless there is conclusive information in the police report or the autopsy report to show that the death was intentional. For example, if the NCIS contains evidence of a suicide note, the death would be coded as a suicide.
In general, the ICD-10 coding index defaults external causes to 'accidental' unless there is information to the contrary. An exception is for stabbing deaths for which the default is homicide.
CAUSES OF DEATH DATA FOR INDIGENOUS AUSTRALIANS
Accurate and timely information on the mortality of Aboriginal and Torres Strait Islander Australians is important for policy and program development. Causes of death statistics are particularly vital given the lower life expectancy of Indigenous Australians (ABS 2005b). However, there are known quality issues associated with the causes of death data for Indigenous Australians. Further information on issues relating to data quality of Indigenous mortality statistics can be found in
Recent Developments in the Collection of Aboriginal and Torres Strait Islander Health and Welfare Statistics, 2005 (cat. no. 4704.0.55.001) (ABS 2006b).
Identification (coverage)
While it is considered likely that most deaths of Indigenous Australians are registered, a proportion of these deaths are not identified as Indigenous by the family, health worker or funeral director. That is, whilst data is provided to the ABS for the Indigenous status question for 99% of all deaths, there are concerns regarding the accuracy of the data. The Indigenous status question is not always directly asked by the funeral director of relatives and friends of the deceased. The ratio of the number of Indigenous deaths registered to the number of expected deaths compiled from population projections is referred to as the 'implied coverage rate' and is used to assess the extent to which identification of Indigenous people occurs in the deaths collection. The table below provides current estimates of implied coverage rates for each of the States and Territories.
ESTIMATED COVERAGE(a) OF INDIGENOUS DEATHS, 2001-2005
| State | Implied coverage of Indigenous deaths(b)
% |
|
| New South Wales | 45 |
| Victoria | 31 |
| Queensland | 52 |
| South Australia | 64 |
| Western Australia | 70 |
| Tasmania | (c) |
| Northern Territory | 92 |
| Australian Capital Territory | (c) |
| Australia | 56 |
(a) Estimates of coverage are only indicative. (b) Calculated as the ratio of deaths registered to projected Indigenous deaths. (c) Not calculated due to small numbers. |
| Source: Deaths, Australia, 2005 (ABS cat. no. 3302.0) |
Coding
Another issue is the quality of the causes of death coding for those Indigenous deaths which are identified. The ABS has undertaken analysis regarding the quality of Indigenous causes of death data by identifying any differences in the level of detail of coding for the underlying causes of death for Aboriginal and Torres Strait Islander and non-Indigenous populations. Specifically, this analysis examined whether there were a greater proportion of deaths of Aboriginal and Torres Strait Islander people coded to less detailed causes of death including 'other ', 'not elsewhere classified' or 'unspecified'. Analysis of Indigenous status by external causes of death demonstrated that the level of poor specificity in coding is consistent across the Indigenous and non-Indigenous populations.
Print Page
Print All