By continuing to browse this site you agree to us using cookies as described in About Cookies
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of February 24 & 25 and will be live on February 26, 2018. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/
Correspondence to: Pauline Jane Gulliver, Injury Prevention Research Unit, Dunedin School of Medicine, PO Box 913, Dunedin, New Zealand; e-mail: firstname.lastname@example.org
Objectives: Counts of mortality and morbidity based on routinely collected national datasets have undercounted Māori, the indigenous people of New Zealand. To correct for the undercount, when estimating fatal and serious non-fatal injury incidence, the ‘ever-Māori’ method has been used. This study sought to determine how well the ever-Māori method corrects for the undercount.
Methods: Trends in frequencies and age-standardised rates for fatal injury indicators were compared using: (a) ever-Māori classification; (b) New Zealand Census Mortality Study adjustment ratios applied to Total Māori counts from the Mortality Collection; and (c) Total Māori counts from the Mortality Collection. For serious non-fatal injury, trends using ever-Māori were compared with Total Māori from hospital discharge data.
Results: The absolute number of injuries attributable to Māori varied depending on the method used to adjust for ethnicity status, but trends over time were comparable.
Conclusions and implications: At present, there is no optimal method for adjusting for the undercount of Māori in routinely collected health databases. Reassuringly, trends in fatal and serious non-fatal injury are similar across the methods of adjusting for the undercount.
Internationally, there remains a disparity between the health of indigenous and non-indigenous peoples.1 Accurate measures of ethnicity are required to understand the true magnitude of the disparity, and to assess progress towards reducing the impact of health conditions on indigenous peoples.2 In New Zealand, ethnicity is defined as “the ethnic group or groups that people identify with or feel they belong to … self perceived and people can belong to more than one ethnic group”.3
Historically, when deriving counts of death, using the New Zealand Ministry of Health's (MoH) Mortality Collection, and hospitalisation, using the MoH National Minimum Data Set of Hospital Discharges (NMDS), there has been an undercount for Māori, the indigenous people of New Zealand. The undercount resulted from a lack of self-reported ethnicity data in both the Mortality Collection and the NMDS.4 In contrast, since 1996, ethnicity for the population as a whole has been self-reported and is derived from five yearly Census questions. These differences result in a numerator-denominator bias when rates of death or hospitalisation are calculated for the New Zealand population,4 and an underestimate of the magnitude of the impact of health issues for Māori.
The New Zealand Census Mortality Study (NZCMS) has provided an estimate of the undercount for Māori deaths for the periods 1996–19995 and 2001–2004.6 The estimates presented for the period 1996–1999 showed differences in the magnitude of the undercount depending on age and cause of death.5 More recent estimates of the undercount have revealed improvements in the way ethnicity is recorded on death certificates, with “negligible differences between mortality and census data” for the period 2001–04.6 Such improvements have not, however, been reflected in ethnicity recorded in the NMDS.7
In 2003, the New Zealand government introduced the New Zealand Injury Prevention Strategy (NZIPS), an inter-agency strategy aimed at reducing the burden of injuries. To measure the impact of the strategy, Cryer and colleagues developed serious injury outcome indicators to measure trends in injury for the New Zealand population.8 Injuries were regarded as serious if they resulted in either: (a) death or (b) admission to hospital with an injury associated with at least an estimated 6% threat to life. A Chartbook of serious injury outcome indicators for Māori was produced in 2007.9 The primary data sources for the indicators were the Mortality Collection and the NMDS.10,11
Using adjustment ratios for the Serious Injury Outcome Indicators
For Māori, the Chartbook presented frequencies and age standardised rates for the period 1996–2004. However, as described above, if ethnicity was taken directly from variables in either the Mortality Collection or the NMDS, this would have resulted in an undercount of injuries and an underestimate of the age-standardised rates. As the NZCMS adjustment ratios have been calculated only for deaths, they could not be used to adjust for an undercount in the NMDS. For the injury indicators, an alternative method was required. The ever-Māori method12 was recommended and selected.9 The ever-Māori method relies on National Health Index (NHI) numbers, unique identifiers assigned to individuals by the Ministry of Health and used in national administrative health data collections. The ever-Māori methodology assigns Māori ethnicity to an individual if their NHI had been assigned Māori ethnicity in any of the following databases: NMDS (1982–2007); Cancer Registry (1948–2007); Primary Health Organisation collection (2006–2007); and the Mortality Collection (1988–2005).
Other methods are also available for adjusting for the undercount, such as that used in Hauora IV. Hauora IV is an examination of the disparities between Māori and non-Māori health status (fourth in a series established in the 1980s and funded by the Health Research Council of New Zealand and the Ministry of Health). To adjust for the undercount of Māori in the NMDS, the authors of Hauora IV linked the NMDS, the Mortality Collection and Housing New Zealand data (New Zealand government state housing agency) to derive adjustment factors for the NMDS for the period 2003 to 2005.7
There is a question of how well the ever-Māori method corrects for the undercount and the numerator-denominator bias between numerators produced from the Mortality Collection and the NMDS, and denominators derived from the New Zealand Census data.7 Concern exists that this method may overcompensate for the undercount in Māori as false negatives as well as false positives are assigned Māori ethnicity using this method. This paper seeks to investigate this by comparing the adjustment ratios and fatal injury trends derived for injury deaths obtained from NZCMS and ever-Māori methods. Similarities in injury trends derived using the ever-Māori method and ethnicity derived directly from the NMDS will also be examined, as will similarities between adjustment ratios for the ever-Māori method and those derived for Hauora IV.
Counts of fatal and serious non-fatal injury were drawn from the Ministry of Health's (MoH) Mortality Collection (fatalities) for the years 1996 to 2005, and National Minimum Data Set of hospital discharges (NMDS, serious non-fatal injuries) for the years 1996 to 2007. Population data was sourced from the Statistics New Zealand estimated resident population, as at June 30 of the relevant years.
Since 1996, MoH and Statistics New Zealand protocols require that the person concerned should identify their ethnic affiliation wherever feasible. A person can belong to more than one ethnic group and the ethnicities with which a person identifies can change over time (http://www.nzhis.govt.nz/moh.nsf/pagesns/399?Open, accessed November 2008).
For this investigation, Māori ethnicity was taken from the ‘Total response’ (‘Total Māori’) reporting option, where each person is counted in each ethnic group reported. Because individuals who indicate more than one ethnic group are counted more than once, the sum of the ethnic group populations will exceed the total population of New Zealand.13
The operational definition of injury used for this investigation included all those cases recorded with International Classification of Diseases, revision 10 (ICD10) injury diagnosis or external cause codes. Injury fatalities were identified as those cases recorded in the Mortality Collection with an underlying cause of death in the range V01-Y36. Serious non-fatal injuries were those recorded in the NMDS with a principal diagnosis in the range S00-T78 and a first-listed external cause code in the range V01-Y36. ‘Serious’ was defined in terms of the ICD based injury severity score (ICISS).14 Only those cases with an ICISS score <0.941 (indicating a threat to life of 5.9% or greater) were included.15
Adjustment ratio estimates
Mortality records for persons who died aged 1–74 years within three years of the 1996 or 2001 Census were anonymously and probabilistically linked with their corresponding censal records.5,6 Counts for Māori from the mortality collection were compared with those from the Census. Adjustment ratios for the period 1996–1999 and 2001–2004 were derived separately from these comparisons.
Māori ethnicity was assigned to an individual according to whether their NHI number had an associated record assigned as Māori in any of the following databases: NMDS (1982–2007), Cancer Registry (1948–2007), Primary Health Organisation collection (2006–2007), and the Mortality Collection (1988–2005). Adjustment ratios were calculated by comparing the counts for Total Māori from the Mortality Collection or NMDS with those derived using the ever-Māori method.
The NMDS was linked to the mortality collection for the period 2000–2004, and Housing New Zealand Corporation tenant data for the period 2003–2005, using encrypted NHIs.7 Adjustment ratios were estimated by calculating weighted averages of Housing NZ and Mortality Collection linkage ratios for five-year age bands.
Adjustment ratio comparison
The ever-Māori adjustment ratios for fatal injuries were compared with NZCMS adjustment ratios for ‘injury/suicide’ (1996–1999, ICD-9 external cause codes E800–E999)5 and ‘external causes’ (2001–2004, Unintentional injury V01–X59, suicide X60–X84).6 Age-specific adjustment ratios were compared for the age groups 0–14 years, 15–24 years, 25–44 years, 45–64 years, 65–74 years (1996 to 1999)5 and 0–14 years, 15–24 years, 25–44 years, 45–64 years, 65–74 years, 75–84 years and 85+ years (2001 to 2004), ‘all causes’– no age specific breakdowns were available for ‘external causes’ for 2001 to 2004).6 Ever-Māori adjustment values for serious non-fatal injuries were compared with those published in Hauora IV7 for the period 2003 to 2005, by five-year age band.
Trends in the frequencies and age-standardised rates for the fatal injury indicators were compared for the period 1996 to 2005 for counts derived using:
b. NZCMS adjustment ratios applied to MoH Total Māori counts from the Mortality Collection; and
c. Total Māori counts from the Mortality Collection.
Trends in serious non-fatal injury using ever-Māori and Total Māori from the NMDS were compared.
Fatal injury – adjustment ratio comparisons
For the period 1996 to 1999, when compared with the NZCMS adjustment ratios, the ever-Māori method over-compensated for the undercount for people aged 65 and over, and under-compensated for people aged 15–44. It is apparent, for the period 2001 to 2004, that the ever-Māori method over-compensated for all age groups when compared with the NZCMS adjustment ratios (Table 1).
Table 1. Adjustment ratio comparison: fatal injuries.
a. Adjustment ratio for all causes.
b. Adjustment ratio for “external causes”.
Fatal injury – trend comparison
Frequencies and age standardised rates for fatal injuries were higher post-2000 using the ever-Māori method than when applying the NZCMS adjustment ratios or using Total Māori from the Mortality Collection. The trends showed similar increases for the period 2001 to 2004 for each of the three methods (Figure 1).
Serious non-fatal injuries – trend comparison
Frequencies and age-standardised rates for serious non-fatal injuries were higher between 1996 and 2007 using the ever-Māori method than when using Total Māori from the NMDS. The trends relating to each method, however, were similar (Figure 2).
Serious non-fatal injuries – ever Māori vs Hauora IV
Compared with Hauora IV, the ever-Māori method resulted in higher adjustment ratios for all age groups. The magnitude of the difference was more pronounced for the 50–54 and the 75+ age groups. In comparison with the Hauora IV peak 1.16 (95% CI=1.12–1.19) adjustment ratio in the 85+ age group, the ever-Māori method reached a peak of 1.46 (95% CI=1.21–1.75) in the 85+ age group. Because of the small numbers on which this estimate is based, the confidence intervals for the ever-Māori method were wide.
Although the absolute number of injuries (fatal or non-fatal) occurring to Māori varied depending on the method used to adjust for ethnicity, this investigation showed that the trends over time were comparable irrespective of the method chosen. Adjustment ratios calculated using the ever-Māori method were similar in 1996–1999 to those identified by NZCMS, and higher in 2001–2004. The adjustment ratios calculated using the ever-Māori method, in comparison with those produced for Hauora IV were higher.
Harris and colleagues have indicated that although the ever-Māori method produced ‘reasonable estimates’ of deaths in the period 1996–2001, application of the method to more recent Mortality Collection data (2001–2005) resulted in an over-count of Māori cases.7 Nevertheless, trends in fatal injury from 1996 to 2005 using NZCMS adjustment ratios and the ever-Māori method show similar patterns. As additional years of data are used to generate the ever-Māori indicator, the chances of false positives increase because, once assigned Māori ethnicity, either correctly or incorrectly, this ethnicity is retained for the NHI. Such an effect could lead to increasing numbers of cases identified as Māori in later years. Similar trend patterns presented for this investigation are reassuring as it suggests that the ever-Māori method is not over-estimating Māori to the extent that it causes substantial trend deviations in later years.
Serious non-fatal injuries
Figure 1 reveals lower age standardised rates when using the ethnic group classification from the NMDS record (‘Total Māori’) than when using the ever-Māori method. The ever-Māori method also produced higher adjustment ratios than those calculated for Hauora IV. Reassuringly, even with these differences, the trends presented in Figure 1 show similar patterns when injury rates are estimated using the ever-Māori method as opposed to using ‘Total Māori’ as recorded in the NMDS.
One of the main advantages of the ever-Māori method is that adjustment occurs at the individual level. Therefore, differences by age group, year and diagnostic category are implicitly taken into account. However, there is the potential to over-count. Once a National Health Index number is assigned Māori ethnicity (correctly or incorrectly), using the ever-Māori method, this ethnicity is retained. As ethnicity is self-determined in New Zealand, all people (not just Māori) have a right to identify with a particular group at any time. The factors that influence these choices are many and varied.16 Therefore, it is possible that any method of counting Māori, or any other ethnicity in New Zealand, will produce results different to what is the population snapshot at the time.
There was no bias in the trend when estimating the rates and frequencies of death and serious non-fatal injury when using the ever-Māori method. Bias in the slope of trends is the most important consideration in the context of the NZIPS indicators, as the indicators have been designed to be valid indicators for monitoring the change in injury incidence and rate over time. Bias in the slope may reduce our ability to detect change over time or may result in spurious estimates of change.
One of the weaknesses of the ever-Māori method is the apparent over-compensation for the undercount of Māori, which could lead to an over-estimation of the impact of injuries on Māori. Over-estimates could lead to inappropriate conclusions being drawn from comparisons between Māori and non-Māori in the same way that an undercount has led to inappropriate conclusions being drawn in the past. As such, it appears the ever-Māori method is not an acceptable method to use when the aim of an investigation is to provide a cross-sectional comparison between Māori and non-Māori.
The results presented suggest that the ever-Māori method does over-compensate for the undercount of Māori in the Mortality Collection and the NMDS in the period 2001 to 2005. The over-compensation does not result in substantial trend deviations in later years. Trends in fatal and serious non-fatal injury are similar irrespective of the method of adjusting for the undercount. As such, the ever-Māori method is an appropriate method to use for the investigation of trends in serious injury incidence for the New Zealand Injury Prevention Strategy.
The Accident Compensation Corporation (ACC) provided the funding for this research project. Views and/or conclusions in this paper are those of the project team and may not reflect the position of the ACC.
The authors would like to thank Craig Wright (Ministry of Health Public Health Intelligence), Tony Blakely and Bridget Robson (Wellington School of Medical and Health Sciences), and Joanne Baxter and John Broughton (Dunedin School of Medicine) for their advice during the initial development of the Serious Injury Outcome Indicators for Māori. We would also like to acknowledge the help of Craig Wright for providing the ever-Māori indicator. Finally, we would like to thank Emma Wyeth her helpful comments on a draft of this paper.