Interpretation of External Causes of Death Data
Given the data quality issues discussed in this publication, care should be taken in interpreting external causes of death data. This part summarises some issues to be aware of when interpreting the data.
CHANGES IN DATA OVER TIME
Year to year changes
It is more useful to look at long term trends than to focus on year to year changes. The data may fluctuate from year to year for a range of reasons including the extent to which these deaths are registered in the same year as they occurred, delays in finalising coronial processes, and actual fluctuations in the number of deaths from external causes. Data for the most recent year of registration should be interpreted in the context of the last five years or ten years of data. Another way of looking at the data is to combine data for several years (e.g. three or five years).
Longer term changes
The data may also change in the long term in response to systematic activities. These may be directly aimed at influencing the data (e.g. efforts to improve data quality) or aimed at addressing community issues (such as policy or legislative change).
DIFFERENCES BETWEEN JURISDICTIONS
In addition to the differences in practices across jurisdictions discussed earlier, the percentage of cases that are closed/completed on the NCIS varies widely by jurisdiction over time.
Changes between categories
Between years, the proportions of deaths coded to different categories may change in relation to others as a result of the level of incomplete source information. This may be represented by increases or decreases in one category of intent (such as suicide, or assault) compared to another (such as accidents) if only the information about intent is missing, or from any category to 'exposure to unspecified category' (X59) if information on both intent and mechanism are missing. In the former situation, investigating a particular mechanism of death under different categories of intent may be useful. These coding rules are summarised in Potential Data Quality Issues Affecting External Causes of Death Data and are discussed in relation to the data in External Causes of Death Data Analysis , earlier in this paper.
COMPARISON WITH OTHER SOURCES OF DATA
Comparison with other data sources can be valuable in order to validate data and assess the quality of different data source. However it is important that users of ABS external causes of death data are aware of the different bases upon which such estimates are arrived at by different organisations. As a result differences may result between ABS estimates of deaths from specific external causes and those estimates produced by other organisations.
Scope and coverage
Other organisations may apply different scope and coverage rules compared to the ABS.
The ABS process of producing estimates of numbers of deaths from external causes has been described earlier in this publication. The ABS uses a strict interpretation of the ICD-10 coding rules, relying on legal or medical documentation regarding the intent of a death. Other organisations, however, may produce valid alternate estimates of numbers of deaths from specific external causes (for example deaths from suicide) using different methods. These organisations may be using different classification rules from those used by the ABS. They may make different assumptions regarding intent - for example they may not rely solely on medical or legal documentation and may also be using different levels of 'burden of proof'.
A difference in the timing at which coding is undertaken also contributes to differences between datasets sourced from coronial information. The longer the period since the end of the reference period, the more cases which are closed by the coroner. Therefore, coding undertaken at a later point in time is likely to be more accurate (if less timely).
Multiple causes of death codes
Using multiple causes of death data may help to explain differences between the ABS and other sources of external causes of death data, particularly in relation to falls or the nature of injuries. When only a single underlying cause is selected for tabulating cause-specific statistics, other information provided on the death certificate is lost. For instance, causes and conditions that create linkage between the underlying and immediate causes of death and many other contributory causes that were involved, but did not directly influence the death, are unavailable on an underlying cause dataset.