Mortality rate data for military personnel in the New Zealand Expeditionary Force (NZEF) were obtained from an electronic dataset (Roll-of-Honor) covering all deaths in these personnel during World War I (WWI) (Figure 1). This dataset was obtained courtesy of the compiler, Professor Peter Dennis (University of New South Wales at the Australian Defence Force Academy). Those reported as dying of disease or from no specified cause were studied further, and if dying during known pandemic periods, they were considered to have died from pandemic influenza (i.e., after excluding deaths that had a specific alternative diagnosis in the archival records). These pandemic periods were 1) August 27–December 31, 1918, for the second wave in the Northern Hemisphere); 2) January 1–March 31, 1919, for the third wave in the Northern Hemisphere; and 3) November 1–December 31, 1918, for the second wave in the Southern Hemisphere. There was no evidence for excess mortality rates in military personnel from any first or third pandemic wave in the Southern Hemisphere.

Military personnel were classified as having Māori ethnicity if they: had a first, second, or surname in the Māori language; had a parent with a Māori language name; were buried in a Māori cemetery or had a memorial in such a cemetery; or had a iwi (tribal) affiliation listed in the Cenotaph database covering NZEF personnel (purchased from the Auckland War Memorial Museum [7]). Personnel were classified as having Pacific person ethnicity if they came from a South Pacific island providing military personnel for the NZEF (i.e., Fiji, Gilbert and Ellis Islands, Niue, Samoa, Tonga, and the Cook Islands) and had a Pacific name or a parent with a Pacific name or came from a named village. The approach of using the language of a name for considering ethnicity has been used elsewhere in historical work in New Zealand (8). It is also used for identifying from the electoral roll potential Pacific persons as respondents to public health surveys in New Zealand (the Lexicon method used by Massey University) (9).

For denominator data, we extracted a random sample of 1,000 persons (≈1%) who served in the NZEF as detailed in the Cenotaph database. This denominator sample was then adjusted further to replace (with additional random selection) those who died in the prepandemic period. Ethnicity coding was then performed as for the numerator data.

A validation study was performed for the method of ethnicity coding for Māori; this study involved a University of Otago colleague with local history expertise (Dr George Thomson) who independently classified the ethnicity of WWI participants in a rural area in which he had performed historical research. The results indicated that the coding system we have used was underascertaining Māori ethnicity (sensitivity 73%, i.e., n = 11/15). Of note, however, is that the rural locality used in this validation study had a relatively high Māori population in the pre-WWI era, and intermarriage between Māori and New Zealand European persons was relatively common. As such, the underascertainment found would be a worst-case assessment if applied to New Zealand in general. In contrast, all of those classified as Māori by our coding system were also classified as Māori by Dr Thomson (specificity 100%, n = 17/17).

In addition to considering Māori mortality rates in official data identified in our literature search, we undertook an additional analysis of an online national database for deaths in New Zealand (10). Because ethnicity coding at this time in New Zealand history was poor, we again adopted the approach of using the language of the surname. We examined surnames in the Māori language by using the most common 50 surnames in the 2006 Māori electoral roll. Similarly, this examination was performed for the most common 10 surnames from the general electoral roll of 2006 (all of these were surnames of European origin). The periods considered were for August–September in 1956, 1957, and 1958; each of these months had maximum pandemic impact in 1957 (11). The 1956 and 1958 periods were used for comparison with the known pandemic period in 1957. This method was also used to identify the increased mortality rate for the 1890s pandemic for the European population because registration of Māori deaths was not routine in the 19th century. We did not study the Hong Kong influenza pandemic of 1968–1969 because we could find no published studies in New Zealand and no evidence for increased mortality rates in the annual mortality data (12).

A limitation with this surname method was that the number of Māori deaths involved was small, e.g., 23 and 28 deaths in August–September in 1956 and 1958 respectively, compared with 38 deaths in August–September in 1957 (using the top 50 surnames in Māori language). Furthermore, some persons with Māori surnames may not have been Māori (e.g., non-Māori women who married Māori men with Māori surnames). Also, a proportion of the total New Zealand mortality rate would have included Māori deaths because the most common 7 surnames on the Māori electoral roll (2006) are actually names of European origin that Māori have commonly adopted over the past 150 years (e.g., Smith, Williams, Brown, Wilson; Kingi, the first name in te reo Māori, is listed eighth). The latter factor is likely to dominate. Thus, our analysis is likely to have underestimated the true pandemic-related Māori mortality rate.

Anonymized mortality data were obtained from an official mortality review group that was charged by the New Zealand government with identifying deaths caused by pandemic (H1N1) 2009 (13). All but 1 of the 49 deaths included in our analysis occurred during June 1–September 30, 2009. The review group used prioritized ethnicity (i.e., where there are multiple ethnicities assigned, Māori ethnicity takes precedence, followed by Pacific persons, with the remaining persons included as European and others). Rates and RRs were calculated and age-standardized to the 2006 New Zealand Census with Māori as the standard population (the Māori population has a younger age structure than the European population and is more similar to Segi’s world standard population).

During the 1918 pandemic, the Māori mortality rate (4,230/100,000 population) was 7.3× the rate for the rest of the population (6), which was nearly all composed of settlers of European origin (Table; Figure 2). This work by Rice (14) was far more comprehensive than previous work that compared rates by ethnicity (15). Nevertheless, Rice noted limitations with data quality and considered that the Māori rate was still an underestimate of the true rate because some Māori deaths were not registered (14). This underestimation of Māori pandemic-related deaths was also a finding in a study of the pandemic in New Zealand’s largest city, Auckland (16).

The analysis of the effect of 2 pandemic waves among New Zealand military personnel found a substantially higher mortality rate among Māori personnel (2,501/100,000) than for European military personnel (Table). The rate for Māori soldiers was 2.3× higher than that for European soldiers.

In this pandemic, the Māori mortality rate (39.6/100,000) was 6.2× the European rate (Table; Figure 2). In our new analysis using selected surnames, an equivalent ratio 2.4× higher for Māori was found for 2 months when the pandemic was most active in New Zealand in 1957 (Table).

A national serosurvey after wave 1 of the pandemic found that Māori had evidence of higher seroprevalence of antibodies to the pandemic (H1N1) 2009 virus but not at statistically significant levels (Māori 36.3%, 95% CI 28.0–44.6, vs. other [mainly European] 25.9%, 95% CI 22.4–29.4) (18). However, multivariate analysis identified significantly higher seroprevalence rates for Pacific populations (49.5%; adjusted odds ratio 2.2, 95% CI 1.5–3.4) compared with the other ethnic group population (i.e., mainly European New Zealanders).

During the 2009 pandemic, Māori had relatively higher notification rates for pandemic influenza (age-standardized relative risk [aRR] 2.0, 95% CI 1.9–2.1) compared with Europeans and others (and similarly for Pacific persons, aRR 4.0, 95% CI 3.8–4.3) (19). The cumulative age-standardized hospitalization rate for Māori was 43.0/100,000 compared with Europeans and others at 14.1/100,000 (Māori aRR 3.0, 95% CI 2.9–3.2; Pacific persons aRR 6.7, 95% CI 6.2–7.1) (19). A local study also found higher hospitalization rates for Māori in the Wellington region, at 5× the rate for other ethnic groups (excluding Pacific persons) (20).

Intensive care unit (ICU) admissions were also significantly higher for Māori and Pacific persons compared with Europeans (13). More specifically, Māori and Pacific women who were pregnant were more likely to have an ICU admission compared with other pregnant women (RR 2.3, 95% CI 1.4–3.7) (21).

Our analysis of national mortality data found that the Māori rate was significantly higher (2.6×) than the rate for other New Zealanders (non-Māori and non-Pacific New Zealanders, largely European) (Table). Similarly, for Pacific persons, the rate was 5.8× higher than that for other New Zealanders (Table). An official post-pandemic review also reported elevated rates for both ethnic groups, but it did not perform age standardization of rates (13).

None of the work on these 3 pandemics systematically analyzed risk factors for adverse outcomes of pandemic influenza on Māori. Nevertheless, in 1918 the high rate of illness and death among adult caregivers is thought to have limited capacity to provide basic care for others, contributing to relatively high total Māori mortality overall (6,15). Furthermore, various supportive care services in towns and cities may have favored better health outcomes for the affected European population relative to Māori (6). Māori were largely rurally based in 1918 (22) and rural health services were limited at that time.

For the 2009 pandemic, it was reported that 86% of those who died had >1 concurrent or associated condition, particularly obesity (74%), morbid obesity (56%), and respiratory disease (49%) (13). Furthermore, 39% of those who died had a deprivation score of either 9 or 10 (most deprived) compared with the expected 20% of the population (using a New Zealand–specific small area deprivation index). Similarly, the ICU study identified risk factors for pandemic influenza–related admission, including pregnancy, body mass index >35, and preexisting asthma or other chronic pulmonary disease (23).