Linking Death Registrations to the 2016 Census

This information paper describes the background and rationale for the 2016 Death Registrations to Census linkage project (previously referred to as the Indigenous Mortality Project). This project involved linking the Census with death registrations to examine differences in the reporting of Indigenous status across the two datasets in order to apply adjustment factors to mortality and life-expectancy estimates.

The 2016 Death Registrations to Census project involved linking twelve months of post-Census deaths data to the 2016 Census. Specifically, deaths date from 9 August 2016 until 28 September 2017, with a slightly longer range than 12 months to allow time for all relevant deaths to be registered and processed. The project attempted to link 177,380 death registrations records to 22,485,854 Census records, which led to 159,657 links, or a linkage rate of 90%. 

The aims of this project were to:

• assist in understanding the differences in recording of Indigenous status between death registrations and Census data; and
• assess the under-identification of Aboriginal and Torres Strait Islander deaths in death registrations records.

The 2016 Death Registrations to Census linkage project expanded on the methods used in the 2011 iteration of the same project. The main enhancements implemented for the 2016 project included:

• use of non-sequential probabilistic linking of 2016 Census data to death records (as opposed to sequential probabilistic linking used in 2011);
• use of alternative address information from Census and Death registrations to improve linkage of records between datasets;
• improved name repair processes, where the rarity of a name was used in evaluating the quality of links established; and
• an enhanced clerical review strategy resulting in higher quality links.

The Australian Bureau of Statistics has been funded as part of the Council of Australian Governments (COAG) Closing the Gap initiative and given a mandate to deliver information to improve the measurement of Aboriginal and Torres Strait Islander life expectancy. The intent of the funding was to ensure the ABS has the capability to deliver high quality linked data for statistical use. Through this investment in data linking capability, the project enables reporting against the COAG target to close the life expectancy gap within a generation.

The 2016 Death Registrations to Census linkage project enables an estimate of the under-identification of Indigenous status in death registrations to be produced. This allows for adjustments to the registered data when compiling the Life Tables for Aboriginal and Torres Strait Islander Australians, 2015-17 (cat. no. 3302.0.55.003) released on 29 November 2018.

Death registrations data from the State and Territory Registries of Births, Deaths and Marriages are used by the ABS to produce estimates of Aboriginal and Torres Strait Islander deaths. The information relates to all registered deaths including those referred to a Coroner. While there is some variation in practice among the jurisdictions, information supplied on both the Death Registration form and the Medical Certificate of Cause of Death (completed by medical practitioners) has been used where available to derive Indigenous status. Estimates of Aboriginal and Torres Strait Islander deaths are used as an input for calculating Aboriginal and Torres Strait Islander population and life expectancy estimates.

The Census is usually completed by a responsible adult answering for themselves or on behalf of another person present in the dwelling on Census night. In the standard Census form, Indigenous status is reported by the person completing the form and in some instances may not be answered. By contrast, Interviewer Household Forms are used in remote Aboriginal and Torres Strait Islander communities and in some urban areas. These forms are completed by a trained interviewer, who is recruited from the local community wherever possible. For further information on how the 2016 Census was undertaken please refer to Census of Population and Housing: Understanding the Census and Census Data, Australia, 2016 (cat. no. 2900.0).

In June 2017, the Report on the Quality of 2016 Census Data was released by the Census Independent Assurance Panel. The Panel determined that the 2016 Census data is of a comparable quality to previous Censuses, is useful and useable, and will support the same variety of uses of Census data as was the case for previous Censuses.

The report included a broad assessment of the key linking variables used in the Death Registrations to Census project; including name and date of birth. Although the quality of these variables was high for the 2016 Census, there was a decrease in the quality of this information relative to the 2011 Census. The Report noted the following:

• a substantial increase in the non-response rate for date of birth, increasing from 10% in 2011 to 19% in 2016;
• an increase in non-response for first name, from 49,000 persons in 2011 to 209,000 persons in 2016; and
• an increase in non-response for surname, from 127,000 persons in 2011 to 274,000 persons in 2016.

For further information on the quality of particular Census variable, please refer to Census of Population and Housing: Understanding the Census and Census Data, Australia, 2016 (cat. no. 2900.0).

The data linking methodology used in the 2016 Death Registrations to Census linkage project can be generalised into the following steps:

• data standardisation
• data preparation
• record pair comparison
• decision model.
   

This project benefited from advances in methodology and technological resources to deliver improved integrated statistical outputs including:

• Use of a combination of deterministic and probabilistic linkage techniques;
• Adoption of a non-sequential approach to probabilistic linking designed to link a high quality dataset that is representative of the Australian population. The sequential approach used for the 2011 linkage removed accepted links after each probabilistic pass; in comparison, a non-sequential approach allowed for:
  • All records to be given an opportunity to link in every probabilistic pass;
  • All possible links from all passes assessed together to identify the best quality links overall for the dataset;
  • Prevention of poorer quality (and potentially inaccurate) links from earlier passes being accepted, where a higher quality link could be found in a later pass;
  • Quality of links could be assessed consistently across probabilistic passes;
• Use of alternative address information from Census and Death registrations to improve linkage of records between datasets;
• Improved name repair processes, where the rarity of a name was used in evaluating the quality of links established;
• An enhanced clerical review strategy resulting in higher quality; and
• Potential links that were rejected after review were assessed for alternative links on each record in the rejected linked pair.
   

Before records on two datasets are compared, the contents of each need to be as consistent as possible to facilitate comparison. This process is known as 'standardisation' and includes a number of steps such as verification, recording and re-formatting variables, and parsing text variables (i.e. separating text variables into their components). Additionally, some variables such as name may require substantial repair prior to standardisation. 

Some variables differ between the two datasets in a predictable way, and an adjustment is required to account for this variance. Variables may also be recoded or aggregated in order to obtain a more robust form of the variable. Standardisation takes place in conjunction with a broader evaluation of the dataset, in which potential linking variables are identified.

The standardisation procedure for the Death Registrations to Census linkage project involved coding imputed and invalid values for selected variables to a common missing value. These variables included name, address, day of birth, month of birth, year of birth, age, sex, year of arrival and marital status. Standardisation for hierarchical variables involved collapsing at higher levels of aggregation to allow for potential differences in the recording and coding of the variable. This was done to improve the quality of the linkage data for the purpose of increasing the likelihood that a link would be made. An example of this is country of birth. On the Death registration record a person may have been coded to 'Northern Europe' (two digit level of country of birth), while on the 2016 Census they may have reported a specific country such as 'England' or 'Norway' (four digit level of country of birth). If left in its original state, a comparison between 'Northern Europe' and 'England' would not agree, even though one is a sub-category of the other. To account for this all 2016 Census country of birth responses were coded to the two digit level to allow for accurate comparison.

An additional data preparation technique was used in this linkage for Census records where multiple responses had been provided for key linking variables. A record may have had multiple responses for a single linking variable in the following situations:

• a name that required repair had more than one possible repaired name value; or
• the respondent reported different locations for address of usual residence and enumerated address (2016 Census records only).
   

The process for allowing the use of multiple responses for a linking variable involved restructuring the data for affected records; multiple rows were created for the affected record, with the number of rows generated equal to the number of different combinations that could be created from the linkage information. This is demonstrated in Tables 1a and 1b below. A respondent with two different anonymised first name values and two different mesh blocks would have four permuted rows generated. Meanwhile, the information that only had one stated value (in this example surname and date of birth) was duplicated across all of the generated rows. Structuring the data in this manner allowed for all combinations of a respondent's linkage information to be considered in a highly efficient manner while increasing the likelihood of finding the true link for the record.

Death registrations data and the 2016 Census were brought together using a combination of deterministic and probabilistic data linkage techniques. Deterministic linkage methods were initially used to identify matches that could be used as part of a training dataset for the creation of m and u probabilities for probabilistic linking (see Section 3.4.2 Probabilistic Linking for further information). Probabilistic linking was then used to link records that would be accepted for the final linked file.

The two datasets were linked in a way that was independent of reported Indigenous status so that any future analysis (including use in compiling the Life Tables for Aboriginal and Torres Strait Islander Australians - 2015-17 (cat. no. 3302.0.55.003)) would not be affected by bias introduced in the linking process. For this reason, Indigenous status was not used as a linking variable.

Deterministic data linkage, also known as rule-based linkage, involves assigning record pairs (i.e. potential links) across two datasets that match exactly or closely on common variables. This type of linkage is most applicable where the records from different sources consistently report sufficient information to efficiently identify links. It is less applicable in instances where there are problems with data quality or where there are limited characterisitics.

Initially, a deterministic linkage method was used to identify links to create a training dataset that could be used to inform the creation of m and u probabilities. This involved using selected personal and demographic characteristics (first name(anonymised), surname (anonymised), sex, date of birth/age, geography, year of arrival, marital status and country of birth), to identify record pairs.

Probabilistic linking allows links to be assigned in spite of missing or inconsistent information, providing there is enough agreement on other variables to offset any disagreement. In probabilistic data linkage, records from two datasets are compared and brought together using several variables common to each dataset (Fellegi & Sunter, 1969).

A key feature of the methodology is the ability to handle a variety of linking variables and record comparison methods to produce a single numerical measure of how well two particular records match, referred to as the 'linkage weight'. This allows ranking of all possible links and optimal assignment of the link or non-link status (Solon and Bishop, 2009).

The strategy employed for linking the 2016 Death Registrations to Census project builds on the 2011 linking strategy, using developments in linking methodology, software and available data to improve the approach. For further details on the 2011 linkage refer to Information Paper: Death registrations to Census linkage project - Methodology and Quality Assessment - 2011-12 (cat. no. 3302.0.55.004).

Table 2 displays the blocking and linking variables applied in this linking project for each pass.

In probabilistic linking, once record pairs are generated and weighted, a decision algorithm determines whether the record pair is linked, not linked, or requires further consideration as a possible link. The generation of record pairs from probabilistic linking can result in the records on one dataset linking to multiple records on the other, resulting in a file of ‘many-to-many’ links. The first phase of the decision process involves assigning a record to its best possible pairing. This process is known as one-to-one assignment. Ideally (and often true in practice) each record has a single, unique best pairing, which is its true match.

The 2011 Death Registrations to Census project used an auction algorithm to assign probabilistic links optimally from the pool of all possible links. The auction algorithm maximises the sum of all the record pair comparison weights through alternative assignment choices, such that if a record A1 on File A links well to records B1 and B2 on File B, but record A2 links well to B2 only, the auction algorithm will assign A1 to B1 and A2 to B2, to maximise the overall comparison weights for all record pairs.

For the 2016 project, a change was made to the assignment algorithm. Using the previous example, A1 may still link to B1, but A2 would only link to B2 if it was considered a better quality link than A1 to B2. This change ensured that links would only be assigned when they are the absolute best option for both records in the link, which subsequently improved the quality of the links output at this phase. The modified algorithm was also far more efficient than the auction method, with the assignment process completed in a matter of minutes compared to several hours or days when using the auction algorithm.

An additional change made for the linkage was that the one-to-one assignment was generated using the combined many-to-many results from all passes in the linkage (i.e. non-sequential approach), rather than running the assignment over the results from each pass individually and accepting links before moving to the next pass (sequential approach). This allowed the best links from all passes to be obtained from a single assignment procedure.

The second phase of the probabilistic decision rule stage takes the output of one-to-one assignment and decides which pairs should be retained as links, and which pairs should be rejected as non-links. The simplest decision rule uses a single ‘cut-off’ point, where all record pairs with a linkage weight at or above the cut-off are assigned as links, and all those pairs with a linkage weight below the cut-off are assigned as non-links. A more sophisticated decision rule was used in the 2016 Death Registrations to Census linkage project, employing lower and upper cut-offs. Record pairs with a weight at or above the upper cut-off were declared links while those with a weight below the lower cut-off were declared non-links. In order to establish the upper and lower cut-off values, a sample of the record pairs identified by the assignment algorithm was clerically reviewed. The upper cut-off was then set at a weight value such that no false links had been detected above the cut-off in the sample. The record pairs with weights between the upper and lower cut-offs were clerically reviewed to determine which links to retain for the final linked dataset.

Each record pair was manually inspected to resolve its match status (i.e. if the link was 'true' or 'false'). As part of this process, a clerical reviewer was often able to use information which could not be captured in the automated comparison process, but could be identified by the reviewer, such as common transcription errors (e.g. 1 mistaken as 7) or transposed information, such as the day of birth reported as the month or vice versa.

In addition to the linking variables, supplementary information was also used to confirm a link as true. This included:

• non-linking variables such as ancestry, occupation, schooling and qualification; and,
• reviewing the dates of birth and country of birth of parents (when available) for child records that had been linked.
   

These supplementary variables helped to inform difficult decisions, especially on record pairs belonging to children, allowing for greater insight into whether a record pair was an actual match or just contained similar demographic and personal characteristics for two different individuals. 

Clerical review was performed on 62,115 links, resulting in the confirmation of 35,820 matches. Initially, reviewers assessed the 'best' option for a link, that is, where Death registrations were matched to Census records based on the greatest level of agreement on linking variables. However, for Death registrations where the best option was rejected, subsequent clerical review also assessed the second-best and, if relevant, third-best option. This was further supplemented by a specific investigation into Aboriginal and Torres Strait Islander links. Following the inspection of first, second and third options for Death registrations, reviewers also assessed the remaining potential links identified as Aboriginal and Torres Strait Islander on either Census or Death registrations datasets. By this late stage of clerical review, fewer than 10% of those potential links had an agreement weight comparable to the other accepted links.

While the 2011 project applied high standards for precision, the 2016 linkage placed an even greater emphasis on ensuring as many links as possible in the final set of results were 'true' (i.e. the linked records do in fact belong to the same individual). This was achieved through the following processes:

• A more extensive sampling review process. About 17,000 links were sampled in 2016 to assess precision of the 2016 links, compared to approximately 3,000 links in 2011. The sample size was much larger to ensure adequate agreement patterns were found in the links sampled. This involved sampling at least 5% of links from each individual weight value generated in the linkage run. All of the sample 2016 links were reviewed twice by different clerical review staff to ensure reliability of the precision estimates deduced from sampling. For more information refer to 3.5.1 Quality Assurance of clerically reviewed record pairs; and
• A more conservative approach to confirming links in clerical review. Links were only confirmed when there was a high degree of confidence that the link was true. Staff were instructed to reject links where a '50/50' decision had to be made (i.e. there were valid reasons to both confirm and deny the link.). Emphasis was given to finding sufficient agreement in key linkage fields (i.e. name, address, date of birth) to confirm links. Staff were instructed to reject links where two or more key linking fields did not agree, and there was no available evidence that explained the fields' disagreement (e.g. examining the Census form to find typographical errors in the data).
   

Clerical review relies upon judgment by a well-trained individual, therefore, while efforts are taken to minimise the risk, it is possible for a link to be incorrectly assigned as a match or non-match.

Quality assurance (QA) techniques were applied to clerical review to assess the accuracy of the clerical review decisions. The QA process involved having a sample of the clerical record pairs reviewed a second time by a different reviewer. If the decision for a record pair made by the QA reviewer conflicted with the decision made in the original clerical review, this was identified as an 'adjudication' pair. Adjudication results were used to update the original decisions made on clerically reviewed links.

Performing QA on clerically reviewed record pairs enabled a basic measure of quality, referred to as a 'clerical review consistency rate' (CR), to be obtained. This rate is calculated by dividing the number of adjudication pairs against the total number of record pairs that were quality assured. Note that the CR is not strictly an estimate of clerical review accuracy, rather it is a measure of the level of consistency with which different coders applied decisions to record pairs. The QA results were not used to supplement the final linked results. The quality assurance process produced a clerical review consistency rate of 95%, indicating the clerical review process was of high quality.

Of the 177,380 death records, 159,657 (90.0%) records were linked to one of 22,485,854 eligible Census records. Of the 3,246 Aboriginal and Torres Strait Islander death records, 2,315 (71.3%) were linked.

Examination of the characteristics of the links identified and the unlinked records can be found in 4.3 Characteristics of linked and unlinked Death registrations.

The following quality measures were calculated for the linkage and indicate a good level of overall quality:

• The linkage rate, 90%, being the proportion of Death registrations linked to a 2016 Census record; and
• The estimated proportion of correctly linked records, otherwise referred to as 'linkage precision'.
   

Not all record pairs assigned as links in a data linkage process are a true match, that is, a record pair belonging to the same individual. While the methodology is designed to ensure that the vast majority of links are true, some are actually false, i.e. the records in the link belong to different people rather than the same person. The linkage strategy used for the project was designed to ensure a high level of accuracy. Accordingly, the strategy was restrictive and conservative.

One of the key measures of linkage quality is the proportion of links in the dataset that are false. The number of false links is able to be estimated through the use of methods such as clerically reviewing a sample of links, or by using modelling techniques. Once an estimate of the number of false links is obtained, a 'precision' can be calculated. The precision is an estimate of the proportion of links that are matches (i.e. belonging to the same entity).

                                         Precision = (Total links - False link estimate)/Total links

Once the precision of the dataset is estimated, the false link rate is easily calculated.

                                          False link rate = 1 - Precision

The estimated link precision of the 2016 Death Registrations to Census linkage dataset is 100% as the decision model did not allow for any false links. As previously discussed, the upper cut-off was set such that it was estimated there were no false links above the cut-off while the clerical review process only accepted links for which there was sufficient evidence to support them being accurate matches. In reality, there will be a small number of false links due to a slight degree of inconsistent decisions between clerical reviewers. While the number of false links is not able to be quantified precisely, the proportion is expected to be very small.

The number and percentage of death records linked to Census records by selected characteristics of deceased persons are presented in Table 4. A slightly higher linkage was achieved for females (91.4%) compared with males (88.6%). The linkage rate varied considerably by age, being lowest for 0-14 year old deceased persons (63.4%). This may be due to the comparatively high Census undercount rate in this age group. The linkage rate was highest for 75 years and older deceased persons (92.9%).

The linkage success varied by state of usual residence as reported on the death registration form. Rates were highest for South Australia (91.5%) and lowest for the Northern Territory (72.9%). All other states and territories had linkage rates between 88.8% and 90.9%. The low linkage rate for the Northern Territory reflects comparatively low linkage rates for both the Aboriginal and Torres Strait Islander and non-Indigenous populations. The linkage rate was similar for married and widowed persons (92.2 and 92.6% respectively). The linkage rate was lower for deaths which occurred within six months of the Census (88.8%) than those which occurred beyond six months after the Census (91.0%).

The linkage success also varied by Indigenous status recorded on the death registration form. People of non-Indigenous origin on the death registration form had a considerably higher linkage success (90.4%) compared with people of Aboriginal and Torres Strait Islander origin (71.3%). A more strict approach to implementing the 2016 linkage clerical review resulted in a lower, but more accurate linkage rate than in 2010-2012.

There were two main reasons why death registrations were not linked to a Census record:

1. Records belonging to the same individual were present in the Death registration and Census datasets but these records failed to be linked because they contained missing or inconsistent information; or
2. A link was not possible because there was no Census record corresponding to the death registration as the person was missed from the Census. Proximity of death to Census night is a significant factor in the ability for a link to be achieved.
    

This release previously used catalogue number 3302.0.55.004.