Applying Indigenous identity definitions in official health statistics: a case study using linked cancer registry data on stomach cancer

There are two conceptual approaches to defining Indigeneity in the New Zealand official statistics system—ethnic affiliation and Māori descent. Ethnicity and descent are two different concepts, with different theoretical foundations and legislative frameworks, as well as different applications by researchers, policymakers and Māori organisations. The concepts define closely related yet distinct populations in New Zealand, who may experience different health outcomes.

Ethnicity is a measure of self-reported cultural affiliation with an ethnic group and is widely used in research to define populations. In New Zealand, the official ethnicity classification standard defines ethnic groups as consisting of people with some or all of a) a proper name, b) one or more elements of common culture, c) unique community interests, feelings and actions, d) a shared sense of common origins or ancestry, and e) a common geographic origin.¹

Māori descent is based on biological ancestry and genealogy using definitions that are defined in legislation, including those that attribute specific constitutional or legal relationships. Numerous legislative acts, including the Electoral Act 1993 and the Māori Land Act 1993 (Te Ture Whenua Māori Act 1993), state “‘Māori’ means a person of the Māori race of New Zealand; and includes any descendant of such a person.”^2,3 Māori descent is the concept that the Māori population use to construct identity, with the cultural concept of whakapapa (genealogy or descent) being foundational to individual and collective identities.⁴ Providing information by descent in addition to ethnicity is also consistent with the United Nations’ Declaration on the Rights of Indigenous Peoples, which outlines Indigenous peoples’ right to determine their identity and that their cultures, traditions, histories and aspirations “shall be appropriately reflected in education and public information.”⁵ In New Zealand, the census is a key data source for Māori descent information, and census questions on iwi (tribal nation) affiliation follow for those who indicate they are of Māori descent. Methods enabling analysis of outcomes by Māori descent may therefore also provide a precursor to examining outcomes by iwi affiliation.

The Māori descent population (19.6% of the total 2023 census usually resident New Zealand population) is larger than the Māori ethnicity population (17.8% of the total 2023 census usually resident population) in New Zealand, as not everyone of Māori descent identifies as Māori ethnically and vice versa. In the 2018 census (used as a data source alongside the 2013 census in the present study), 96% of the Māori population by ethnicity (775,836) also identified as being of Māori descent. However, a smaller percentage (86%) of the Māori descent population (869,850) also identified as being of Māori ethnicity.⁶ In other words, about 14% of the total Māori descent population who do not identify as Māori ethnically is excluded when analysing outcomes by Māori ethnicity.

Although Māori ethnicity is very widely collected in New Zealand’s official statistics, Māori descent is only collected in a few official statistical sources, namely the census, birth and death records and the Electoral Commission.⁷ For this reason, social and health outcomes for Māori are routinely reported using the Māori ethnicity variable, and it is largely unknown how they might differ if reported on using Māori descent. However, there has been increasing use of Māori descent to define populations, especially in surveys focussed on or designed by Māori such as Te Kupenga and the Māori Disability Survey.^8,9 Defining the Māori population by descent, which is based on ancestry, is also relevant when studying health outcomes associated with genetic causes, as the concept of descent is more aligned (but still imperfectly) with genotype than the social affiliation-based concept of ethnicity.

This paper looks at how health outcomes can be reported by Māori descent using existing information within the official statistics system in New Zealand. We examine the feasibility of reporting on cancer by Māori descent, using gastric cancer as an example. In New Zealand approximately 400 people are diagnosed with gastric cancer per year.¹⁰ There are large and persistent ethnic disparities in gastric cancer statistics in New Zealand,¹¹ with gastric cancer incidence and mortality rates being more than three times higher in Māori compared to non-Māori.¹⁰ Helicobacter pylori (H. pylori) infection is the largest contributor of high gastric cancer prevalence among Māori, and is estimated to contribute to up to 61% of the excess cases among Māori compared to NZ European men.¹² However, CDH1 gene mutation is also a significant contributor to advanced disease in the Māori population, being associated with an estimated 6% of advanced gastric cancers, and 13% of advanced diffuse-type gastric cancers among Māori.¹³ There is also evidence of H. pylori infection being strongly associated with the expression of CDH1 mutations in diffuse gastric cancer.¹⁴

We construct a Māori descent variable for gastric cancer registrations using linked administrative records in the Stats NZ Integrated Data Infrastructure (IDI). While Māori descent information is not recorded in the New Zealand Cancer Registry (NZCR), it is recorded in several linked administrative datasets available in the IDI; namely, the 2013 and 2018 census and Department of Internal Affairs (DIA) birth and death records. We document the percentage of the 1995–2021 gastric cancer registrations that can be linked to Māori descent information in the IDI, and where that information is located. We then compare the number of Māori descent gastric cancer registrations and the degree of overlap with the Māori ethnicity cancer registrations and consider the quality of the Māori descent information for potential future reporting on gastric cancer outcomes.

Method

Stats NZ’s IDI

Stats NZ’s IDI is a large database containing linked then de-identified individual-level microdata about people and households in New Zealand.¹⁴ The IDI consists of a range of government and non-government administrative and survey data sources connected to a central “spine”, including people who have ever been a resident in New Zealand.¹⁵ This enables analysis of populations of interest across government sectors (e.g., health, social services, housing, justice, education and training).¹⁴ Data in the IDI is periodically updated throughout the year in distinct “refreshes”. This study analysed data available in the June 2023 IDI refresh, using SAS Enterprise Guide version 7.1. Data was output from the IDI according to Stats NZ’s confidentiality rules, which means random rounding to base 3 has been applied to all count data in this study and counts of less than 6 have been suppressed.

Gastric cancer registrations and Māori ethnicity in the NZCR

The NZCR is a population-based register of all primary malignant diseases diagnosed. The register collates details about patients and their cancer diagnosis, including sex, ethnicity, address, date of diagnosis and type of cancer, as well as information about cancer staging and grading.¹⁵

For this study, records on all gastric cancer registrations (i.e., ICD-10 codes C160 to C169) from 1 January 1995 to 31 December 2021 were available.

Māori ethnicity was sourced from the NZCR, which records up to three ethnicities per person. An ethnic identification is allocated by an algorithm process if it is present on at least 20% of a person’s records across the National Health Index ([NHI] person-level health identifier) database, the Mortality Collection and the National Minimum Dataset of hospital discharges. The algorithm is periodically re-run to account for any new records across these sources.¹⁶

Māori descent in the IDI

Information on Māori descent in the IDI can be obtained from the 2013 and 2018 censuses and in New Zealand birth and death records (see Table 1).⁷ Māori descent data held by the Electoral Commission and iwi organisations are not available within the IDI.⁷ Māori descent information is taken from September 1995 onwards (with much improved coverage from 1998 when the collection of birth and death data was digitised) due to significant definitional issues of Māori descent and resulting missing records prior to 1995.¹⁷

View Table 1–5, Figure 1–3.

Information about Māori descent for the individual, as well as their parents, was present on birth records. Where the descent information was missing for the individual’s birth registration but was present for either one of the parents, the parental information was used. Descent information could also be sourced from the birth registrations of children of those with a cancer registration, based on an individual’s record as parent 1 or parent 2 on the child’s birth record.

Information on Māori descent from the different data sources was combined into one Māori descent variable. If a person had identified as being of Māori descent in at least one of the 2013 census, the 2018 census, New Zealand birth records or death records, they were classified as being of Māori descent.

Results

A total of 10,575 people were diagnosed with gastric cancer between 1995 and 2021 in New Zealand. Of these, 1,833 (17.3%) had Māori recorded as one of their ethnic group affiliations in the NZCR.

Figure 1 summarises the data obtained on Māori descent from IDI sources for the gastric cancer registrations. New Zealand death records contained information on Māori descent for 65.2% (6,900) of the gastric cancer registrations, representing the highest coverage among IDI sources. The 2013 census and 2018 census contained Māori descent information for 34.3% (3,630) and 26.0% (2,745) of the gastric cancer registrations respectively (and 4,179 or 39.5% of registrations combined). Only 0.3% (30) of the registrations had Māori descent information from their own birth record. However, 6.1% (648) had descent information from their child/children’s birth registration (i.e., they were listed as parent 1 or parent 2 to a child). In total, combining information across these sources provided Māori descent information for 8,655 (81.8%) of gastric cancer registrations, of which 1,614 (18.6%) were recorded as being of Māori descent in at least one of New Zealand death records, 2013 or 2018 census, or New Zealand birth records.

Although the extent to which information on Māori descent for gastric cancer registrations was recorded in multiple sources was low, when it was recorded in two sources the level of agreement of the information recorded was high. As shown in Table 2 and Table 3, the percentage of gastric cancer registrations identified as being of Māori descent in a given source (New Zealand death records, New Zealand birth records, or the 2013/2018 census) who were also recorded as being of Māori descent in a corresponding source was between 87–98%. Consistency was generally highest with and between the 2013 and 2018 censuses, and lowest for New Zealand death records. Specifically, a lower percentage of individuals with gastric cancer registrations recorded as being of Māori descent in the 2013/2018 census (87%; Table 2) or New Zealand birth records (89%; Table 3) were also recorded as being of Māori descent in New Zealand death records.

For the 2018 census, missing descent records were obtained where necessary and possible from the 2013 census, which may increase the degree of alignment between these sources. We therefore also looked at alignment between the 2013 and 2018 censuses when excluding administrative data–sourced enumerations from the 2013 census for Māori descent in the 2018 census (but retained enumerations from other administrative data sources). As shown in Table 4, this made little difference to the consistency in descent information between the 2013 and 2018 censuses.

Comparing Māori descent and Māori ethnicity among those with complete data for both measures (see Table 5), a high proportion of cases (93.9%) of those with Māori ethnicity recorded in the cancer registry were also of Māori descent. A slightly lower percentage (91.4%) of the Māori descent gastric cancer cases were also of Māori ethnicity.

As shown in Figure 2, the percentage of missing Māori descent information among gastric cancer cases decreased in more recent diagnosis periods. Missing descent information was highest among those diagnosed from 1995–2001 (43.6%), but lower among those of Māori ethnicity, and noticeably lower among those diagnosed during 2012–2016 (5.4% of total cases) or 2017–2021 (4.6% of total cases). During these years, the number of cancer registrations of Māori descent was larger than the number of registrations by Māori ethnicity (see Figure 3).

Discussion

In total, information on Māori descent was found for 81.8% of the gastric cancer registrations (N=8,655). For registrations with available Māori descent information, 18.6% were recorded as being of Māori descent in one or more of New Zealand death records, 2013 or 2018 census, or New Zealand birth records. This is in line with what was found for the general population in the 2013 census (i.e., 17.5%) and the 2018 census (i.e., 18.5%).^6,19 However, because of missing descent data, only 15.3% of the total population gastric cancer registrations were identified as being of Māori descent, less than the percentage of Māori ethnicity (17.3%), and in contrast to the relatively smaller size of the Māori ethnic population compared with the Māori descent population.

Most of the information on Māori descent was retrieved from New Zealand death records, whereas comparatively less information about Māori descent was found in the 2013 or 2018 census and New Zealand birth records. There are several explanations for this finding. Gastric cancer is predominantly a disease affecting older ages,¹⁰ meaning the vast majority of individuals with cancer registrations were born well before 1995 (when information on Māori descent was not routinely collected in a standardised format in New Zealand birth records). A major factor is the poor survival rate of gastric cancer in New Zealand,²⁰ meaning many cancer cases died prior to the 2013 and/or 2018 census. Even for those for whom census records were found, Māori descent was “not stated” or answered with “don’t know” for 9.2% of the cases in the 2013 census and 22.9% of cases in the 2018 census. Furthermore, the 2013 and 2018 censuses both undercounted the Māori population (with a 4.2% net undercount of the descent population in 2018).^21,22 This means some individuals with cancer registrations may have simply been missed by the censuses.

The results indicate it would be problematic to explore gastric cancer rates and outcomes by Māori descent for all cancer registrations due to the amount of missing Māori descent data (18.2%). However, examining outcomes by descent is feasible among cases diagnosed from around 2012 onwards due to significant reductions in missing data from this point in time to around 5%. This is likely due to a combination of a) a higher proportion of the population being born after 1995 meaning greater coverage of descent information on birth records, and b) more cancer cases having census records (i.e., the 2013 and 2018 censuses). The proportion of the gastric cancer cases that could be identified as of Māori descent also exceeds the proportion identified as of Māori ethnicity from 2012, matching the relative sizes of these Indigenous groups in the total New Zealand population. However, a reliance on death records for Māori descent information poses issues. It may introduce bias in survival outcomes among cases of Māori descent, as surviving cases of Māori descent would be more likely to have missing Māori descent information, potentially resulting in inflated mortality rates. Death records could be removed as a source of Māori descent information for analyses where this would be an issue, although this will reduce the number of cases with Māori descent information.

There are other limitations with the methods used here to source Māori descent information. For example, Māori descent identification can vary between sources and over time based on factors such as knowledge of ancestry and willingness to report one’s descent in a particular context. However, information was combined here across sources regardless of the time the data was collected, relative to diagnosis date. Descent information can also be self-reported in some sources (e.g., censuses) but not others (e.g., birth and death records). Finally, it is also important to note that changes in methods for collecting descent information in some sources could lead to changes in incidence rates reported by Māori descent. For example, the 2023 census uses additional sources of data for administrative enumerations of Māori descent information (including electoral roll data),²³ which could lead to increases in the size of the Māori descent due to a combination of both improved population coverage and population growth.

This research demonstrates the utility of the IDI for applying different Indigenous identity definitions in official health statistics. As we made use of all available Māori descent data sources in the IDI at the June 2023 refresh, the methods could be applied to other administrative datasets to examine other outcomes by Māori descent as an additional method of reporting alongside ethnicity. Routinely collected Māori descent information in healthcare data itself, such as through NHI numbers, could reduce potential issues and sources of bias when joining descent information across different data collections (e.g., mortality records). The issues identified here may apply to other outcomes and should be considered when designing or assessing any research focussed on health or social outcomes for the Māori descent population.