infant deaths: demographic characteristics, ethnic...

17
Infant Deaths: Demographic Characteristics, Ethnic Reporting

Upload: others

Post on 10-Aug-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Infant Deaths: Demographic Characteristics, Ethnic

Reporting

Page 2: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint of 2005 paper Catherine Taylor, Anne Howard, Nicholas Thomson and Robert Didham1 Statistics New Zealand This paper reports on one of the very few empirical studies of ethnic mobility currently available in New Zealand. It is based on a research associated with the matching of death registration records of children aged under five years with their birth registrations. To provide a broader demographic context, a number of other characteristics of infant deaths are also discussed. The primary objective of the research was to identify differences between ethnicities recorded at birth and ethnicities recorded at death, to analyse the effect these differences may have on demographic indices, life tables and population estimates and projections and to determine what, if any, adjustments should be made to ethnic mortality rates to make the rates comparable with other data. A second objective was to analyse the relationship between birth characteristics (eg gestation) and deaths of children aged under five years. An earlier test carried out in 1999 formed the basis of this research because it provided an opportunity to observe change over time and undertake some analysis of the effect of changing questions, differing collection environments and a range of other aspects of social dynamics. Infant death matching test in 1999 Between 1984 and 1997, Statistics New Zealand undertook routine manual matching of birth certificates and death certificates for infants at the request of the Ministry of Health. The objective of this infant death matching was to improve the quality of the ethnic data collected for infant deaths. This was based on the assumption that the information on the ethnicity of the child recorded on the birth certificate was more accurate than that recorded on the death certificate. In 1995, a number of changes to the Birth and Death Registration Act were introduced. These included new ethnic questions, which were consistent with the ethnic question included in the 1996 Census of Population and Dwellings. Also, the period in which a birth could be registered was extended, allowing parents two years

1 The authors acknowledge the assistance, support and advice in the development of this paper of Bill Boddington, Michael Ryan, and other members of Demography Division of Statistics New Zealand.

1

Page 3: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

to register their child. These changes considerably complicated the matching exercise. Statistics New Zealand subsequently reviewed its collection of vital statistics and related research, and the infant death matching exercise was dropped, partly because of lack of demand for these results from external clients and partly because it was assumed that the new ethnic question (introduced in 1995) was performing much better than the old question. Although the new question was performing better, research was nevertheless needed to verify this assumption. Electronic capture and transfer of birth and death registration records from 1 July 1998 meant it was economically viable to reconsider infant death matching. The Demography Division of Statistics New Zealand undertook a trial electronic matching exercise involving infant deaths (under one year of age) registered between 1 July 1998 and 30 September 1999. The purpose of this exercise was to establish that matches could be done electronically, and that matching would reveal measurable inconsistencies between the reporting of ethnicities on the birth and death registration forms. Electronic matching involved attempting to match an infant's death record against its birth record. The variables considered include name, date of birth, sex and meshblock of residence. Matches were achieved by searching for possible matching birth records. Subsequently, visually checking the printed files of all the birth and death records was carried out to ensure the match was genuine. The number of inconsistencies in variables and minor variations in the spelling of names meant that this visual check was very important. To ensure the highest possible match rate, searches were carried out independently using date of birth, meshblock (derived from home address) and name (character strings down to three characters were searched for). Other variables such as the child's sex were also used to speed up the matching exercise. Table 1 Matched Infant Deaths, Mäori Ethnicity, July 1998–September 1999

Number of registrations Ethnicity Death registrations Birth registrations Sole Mäori 90 67 Mäori ethnic group (inc sole) 125 129 Not of Mäori ethnicity 177 173 Total matched infant deaths 302 302 Source: Statistics New Zealand The trial established that matching was possible and achieved an 81 percent match rate from a total of 374 infant deaths. The trial also highlighted differences between the ethnic data recorded on the birth registration form and that on the death registration form for these matched records. It is assumed that the 19 percent of infant deaths not matched would have very little effect on the pattern shown in the matched cases. The ethnicity of the 72 unmatched infant deaths was as follows: 20 sole Mäori, 28 Mäori ethnic group (includes sole), and 44 did not record Mäori ethnicity. This is virtually the same ethnic distribution as that recorded on the death certificates for the 302 matched infant deaths. Thus, it would be reasonable to conclude that, if the birth records had been found for the remaining 72 infant deaths, the pattern of response shift would be very similar to the pattern shown by those records already identified.

2

Page 4: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Infant death matching, for the period 1 July 1998 to 30 September 1999, suggests that the number of Mäori ethnic group infant deaths is understated in death registrations by 3 percent. However, infant deaths for the sole Mäori population appear to be overstated in death registrations by 26 percent indicating that multiple ethnicity is less effectively reflected on death registrations and further reinforcing the caution against using ‘sole’ Mäori data without considerable care. These conclusions are based on the assumption that ethnic data is recorded more correctly on the birth registration form (completed by the parents) than on the death registration form (completed by the funeral director). Infant death matching in 2004 One of the reasons why it was considered that a new matching study was required related to the relationship between vitals data and other data with which it is frequently associated. In 2001, the Census of Population and Dwellings used a slightly different question from that used in 1996. However, birth and death registration forms retained the 1996 question. The perception was that the data from the 1996 and 2001 Censuses were not completely comparable and this implied that the birth and death registration forms may not be reflecting ethnicity in a manner consistent with the census. This is an important issue because many mortality and fertility measures depend on population estimates (derived in part from census data) for the subject population. Death registration records of New Zealand-born children aged under five years were matched with their birth registrations to identify differences between ethnicities recorded at birth and ethnicities recorded at death, to analyse the effect these differences may have on demographic indices, life tables, and population estimates and projections and to determine what, if any, adjustments should be made to ethnic mortality rates. The restriction to deaths to New Zealand-born children was to ensure that both birth and death records were potentially available to the electronic matching process. The electronic matching of records involved a number of steps. The initial series of steps associated death records with possible matching birth records. Variables such as first name, last name, date of birth, sex, and meshblock of residence were used to establish possible matches and subsequent visual checking of the printed files of these matches was done to ensure all matches were genuine. As in the 1999 test, inconsistencies and minor variations in some of the variables, for example the spelling of names and use of punctuation, meant that this visual check was very important. The variables were used independently of each other as well as in various combinations in order to ensure the highest possible match rate. The second part of the matching process involved the extraction of the data fields recorded on the birth registrations of the matched registrations that are relevant to the particular characteristic. For example for nuptiality, the ‘marriage date’ field is extracted from the relevant birth records and for multiple births the ‘multi birth code’ field is used.

3

Page 5: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Table 2 Dates of Birth Used in Matching Birth and Death Registrations Age at Death Birth Registration Dates Used Death Registration Dates Used Number of Years of Data Available

0 1 July 1998 to 30 June 2004 1 July 1998 to 30 June 2004 6 1 1 July 1998 to 30 June 2004 1 July 1999 to 30 June 2004 5 2 1 July 1998 to 30 June 2004 1 July 2000 to 30 June 2004 4 3 1 July 1998 to 30 June 2004 1 July 2001 to 30 June 2004 3 4 1 July 1998 to 30 June 2004 1 July 2002 to 30 June 2004 2

Source: Statistics New Zealand For infants, all death and birth registrations from 1 July 1998 were included. However, for older children not all deaths were included. For example, for four-year-olds, death records used were restricted to those registered after 1 July 2002 because a four-year-old who died before 1 July 2002 would have been born before the introduction of electronic records on 1 July 1998. The dates used for each age of death are illustrated in Table 1. In the case of named records (Table 3), the results achieved from the matching process were excellent. Table 3 Results of Matching Process from Named Records Age of Death (Years) Infants 1 2 3 4 Total number of death records 1784 147 55 23 15 Number matched to birth records 1572 142 49 22 15 Number remaining unmatched 212 5 6 1 0 Percentage matched 88% 97% 89% 96% 100% Percentage unmatched 12% 3% 11% 4% 0%

Source: Statistics New Zealand Note: The number of years covered by data varies (see Table 2). Table 4 Results of Matching Process for Six Months Prior to Named Records (1 January 1998 to 30 June 1998) Age at Death (Years) Infants 1 2 3 4 Total number of death records 52 10 6 7 2 Number matched to birth records 27 3 1 4 0 Number remaining unmatched 25 7 5 3 2 Percentage matched 52% 30% 17% 57% 0% Percentage unmatched 48% 70% 83% 43% 100%

Source: Statistics New Zealand The same matching process was undertaken for the period 1 January 1998 to 30 June 1998 when the registrations, although recorded electronically, did not contain all details. The only details that could be used to match the two sets of records during this period were: sex, date of birth, and meshblock of residence. As expected, the success rate for matching this much smaller group of records was significantly less successful.

4

Page 6: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Internal migration As a consequence of the high level population mobility in New Zealand, many families change address frequently. Some aspects of internal migration are reflected in this data, but care is needed in interpretation of the patterns. The majority of cases considered here involved infant deaths occurring very shortly after birth. Even so, at least 13 percent of the children changed address at least once between birth and death. Of the 1572 infant records that were matched, 1224 (78 percent) recorded the same address on both the birth and the death registration forms; 1114 of these were matched using meshblock codes and 110 were matched manually. The manually matched records often contained rural addresses where area codes had been used, or the address given on the death registration matched that of the father on the birth record. Of the remaining 348 records, nearly a third (122 or 8 percent of the total) did not have a specified address or meshblock code on either the birth or the death record and therefore it was not possible to tell if a move had occurred between birth and death. A further 25 (2 percent) death records listed a hospital as the place of residence, meaning that the infant would probably not have left the hospital between birth and death. This does not enable us to ascertain, however, whether or not a parent, or parents, had moved during this time, though it would be extremely rare for this to happen in cases of this type given the very short period of time involved. The final 201 (13 percent) records showed a change of address between birth and death. Table 5 Children Who Changed Address Between Birth and Death Registration Age at Death (Years) 0 1-4 Number of deaths 1572 228 Percentage changing address 87% 51% Percentage not changing address 13% 49%

Source: Statistics New Zealand For those aged one to four years, there were 228 deaths recorded for the period 1 July 1998 to 30 June 2004. Of these, 100 (44 percent) had the same meshblock code detail on the matched birth records. A further 11 (4.8 percent) were able to be matched manually, three (1 percent) had unspecified death or birth addresses, and two (1 percent) had a hospital address as the home address on the death record. There were 112 (49 percent) records where the address at birth record was different from the address on the death record. One implication of this pattern is that children who die under five years of age have a very similar pattern of internal migration as the population as a whole. The families of children under one year of age are less likely to have moved than older children, but, as Table 5 shows, those aged one to four years at death have almost a 50-50 chance of having changed address in their lifetime. This is almost the same as for all children of that age. The records showing a change in the child's/infant's address between birth and death were analysed by urban area. If the urban area was the same from birth to death then it was a movement between similarly populated areas. If the urban area type had changed, it was determined whether the new residence was in an area of greater or lesser population density than that of the birth residence.

5

Page 7: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

The majority of children changing address between birth and death moved between areas of similar population density, accounting for 80 percent of moves among infants and 69 percent of those aged one to four years (Table 6). Of the remainder, a larger number moved to an area of smaller population rather than to one of greater population. Table 6

Internal Migration Between Birth and Death 0–4 years of age

By relative size of urban area

Age at Death (Years) Infant 1 2 3 4 1-4

Moved to similar urban area 160 (80%) 48 (71%) 17 (68%) 5 (45%) 7 (88%) 77 (69%) Moved to a more densely populated UA 19 (9%) 7 (10%) 3 (12%) 1 (9%) 0 (0%) 11 (10%) Moved to a less densely populated UA 22 (11%) 13 (19%) 5 (20%) 5 (45%) 1 (12%) 24 (21%) Total 201 68 25 11 8 112

Source: Statistics New Zealand The majority of changes of location involved moves from one urban area to another rather than between urban and rural areas (Table 7). Interestingly, the distribution of moves varies from the distribution of all people aged under five years enumerated by the 2001 Census. In 2001, around 86 percent lived in an urban area. Table 7

Internal Migration Between Birth and Death 0–4 years of age

By urban/rural area type

Age at Death (Years)

Infant 1 2 3 4 1-4 Urban to rural 15 (7%) 8 (12%) 2 (8%) 3 (27%) 1 (12%) 14 (12%) Rural to urban 10 (5%) 4 (6%) 1 (4%) 0 (0%) 0 (0%) 5 (4%) Urban to urban 171 (85%) 53 (78%) 20 (80%) 8 (73%) 6 (75%) 87 (78%) Rural to rural 5 (2%) 3 (4%) 2 (8%) 0 (0%) 1 (12%) 6 (5%) Total 201 68 25 11 8 112

Source: Statistics New Zealand The level of movement and choice of destination would suggest that very few families have moved to be closer to health care providers. However, in the case of infant deaths, the time between birth and death is very short, so change of address in the intervening time would not be common. Gestation Gestation refers to the number of completed weeks between a woman's last menstrual period and birth. The average normal gestation period for human births is generally taken to be 40 weeks. Births with gestation of less than 37 weeks are considered premature, while those born after 42 weeks are considered post-term births2.

2 For further information see: http://www.nzhis.govt.nz/publications/coders33.html

6

Page 8: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Graph 1

Graph 2

Survival of premature babies varies sharply with gestation period. For births with a gestation period of less than 24 weeks, half resulted in an infant death. At 21 weeks, 74 percent of births resulted in an infant death. Although the chances of death

Percentage of Total Live Births that Die as Infants By Gestation

0

10

20

30

40

50

60

70

80

24 Weeks 28 Weeks 32 Weeks 36 Weeks More than 40 weeks

Gestation Period

Percent

Distribution of Gestation Period for Total Live Births and Deaths by Age

0

10

20

30

40

50

60

Under 25 Weeks 25-29 Weeks 30-34 Weeks 35-39 Weeks 40 Weeks and Over

Gestation Period

Percent

Age at Death = 0 Age at Death = 1-4 Total Births

7

Page 9: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

declined slightly between 21 and 23 weeks, for gestations of 24 weeks and longer, the chance of survival increased dramatically. At 24 weeks, only 31 percent of live births died as infants and by 29 weeks the proportion dying had dropped to 4 percent and remained under 4 percent from then on. Just over half (54 percent) of the infants whose death was registered between 1 July 1998 and 30 June 2004 had gestation periods of 35 weeks or more. Among those who survived beyond their first birthday, 85 percent of the one-year-olds, 94 percent of two-year-olds, 87 percent of three-year-olds and 86 percent of four-year-olds who died during this same period had been born after gestation periods of more than 35 weeks. For babies with very short gestation periods, the risk of dying is much greater. As Graph 2 shows, although only 0.2 percent of all births occur before gestations of 25 weeks, 24 percent of those who die as infants were born before 25 weeks. Although less than half (46 percent) of all births with gestation periods of less than 25 weeks survive beyond one year of age, almost all who do survive infancy survive to reach five years of age. A short gestation period is a significant factor in infant mortality. Birth weight Gestation period and birth weight are closely linked. In general the shorter the gestation period the lower the birth weight. Almost one-third of all infant deaths were accounted for by babies weighing less than one kilogram. Only 5 percent of the infants who died weighed more than four kilograms at birth. Seventy-eight percent of babies born with a birth weight of less than one kilogram died within their first year. However, more than 73 percent of these low-weight infant deaths were babies with a gestation period of less than 25 weeks. More than 70 percent of children who survived infancy but died under the age of five years weighed more than three kilograms at birth. The babies born with birth weights of less than one kilogram who did survive their first year tended to survive beyond five years of age, with less than 5 percent of of children who died aged one to four years weighing less than one kilogram at birth. The percentage of births resulting in infant deaths decreased sharply with increased birth weight. Less than 1 percent of all babies born weighing more than 1.5 kilograms at birth died as infants during the six years studied. Of the deaths investigated, 32 percent were infants whose birth weight was less than one kilogram, and 44 percent were infants whose birth weight was less than two kilograms. However, babies with such low birth weights are uncommon. During the six-year period, only 2 percent of live births resulted in a baby weighing less than two kilograms.

Parents’ age The distribution of ages for mothers and fathers of children who die under five years of age follow similar patterns to the ages of mothers and fathers of births in general. This is illustrated in Graphs 3 and 4. However, there is evidence that the parents of children who die under five years of age are younger than parents of live births in general. Births to mothers aged

8

Page 10: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

between 15 and 25 years account for 24 percent of total live births, but they account for the 35 percent of both infant deaths and deaths among children aged one to four.

Graph 3

Graph 4

Distribution of Live Births by Age of Mother

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

14 and Under 15-19 20-24 25-29 30-34 35-39 40-44 45 and Over Not Stated Age of Mother

Percent

Age 0 at Death Age 1-4 at Death Total Births

Distribution of Live Births by Age of Father

0.00

5.00

10.00

15.00

20.00

25.00

30.00

14 and Under 15-19 20-24 25-29 30-34 35-39 40-44 45 and Over Not Stated Age of Father

Percent

Age 0 at Death Age 1-4 at Death Total Births

9

Page 11: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Fathers are in general older than mothers because women tend to form partnerships with men who are older3. As with mothers, the fathers of children who die before the age of five are younger than fathers in general. Just over one-third (35 percent) of all babies were to fathers aged 15 to 30 years, but 42 percent of infant deaths and 50 percent of deaths among children aged one to four years had been babies with fathers of this age group. Nuptiality A nuptial birth is defined as 'a child born to parents who are legally married at the time of birth'. For this study a birth was considered to be a nuptial birth where a marriage date had been completed on the birth registration form. A birth was considered to be an ex-nuptial birth where this date was missing. Live births resulting in a death prior to five years of age had a higher rate of ex-nuptial births than live births overall. From 1 July 1998 to 30 June 2004, 59 percent of infant deaths were from an ex-nuptial birth, and 56 percent of child deaths aged one to four years were from ex-nuptial births. This is significantly higher than the proportion of ex-nuptial births (43 percent) among all live births. Table 8

Percentage of Nuptial and Ex-Nuptial Births 1 July 1998 to 30 June 2004

Nuptial Ex-nuptial Live births 57% 43% Age at death = 0 41% 59% Age at death = 1-4 44% 56%

A number of factors impinge upon data relating to nuptial births. Parents of the child may have married subsequently or may have been married at the time of birth but the marriage may have been not recorded on the birth registration form. Incomplete details may occur more frequently when a situation is particularly stressful, such as when an infant dies. There is also a link with parents’ age, in particular mother's age. Marriage tends to occur at older ages on average, while mothers of babies who die as infants tend to be younger. These two trends combine to produce higher infant mortality rates among women who are not married. Multiple births One of the factors considered as a contributor to infant and early childhood deaths was whether the child was a member of a multiple birth. One of the limitations in drawing conclusions from this, however, is that very small numbers of events are

3 For more information on age differences between partners, see Ryan and Boddington, “Age Differences Between Partners at the Time of Marriage: 1963, 1983 and 2003” in Demographic Trends 2004, Statistics New Zealand, Wellington (2005), http://www2.stats.govt.nz/domino/external/pasfull/pasfull.nsf/7cf46ae26dcb6800cc256a62000a2248/4c2567ef00247c6acc256fb10077cf90?OpenDocument

10

Page 12: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

involved. Multiple births are relatively rare, accounting for around 3 percent of all confinements, and deaths at these young ages are also relatively rare events. A heightened risk of death for babies born in multiple births was observed, with 3.2 percent of all births being multiple births, but 14.3 infant deaths occurring among multiple births. For children of multiple births who survived beyond their first birthday, there is no clear indication that their death was directly connected with this aspect of their birth rather than an outcome of other risk unrelated factors. This implies that the shorter gestations and lower birth weights commonly associated with multiple births may be the prime factors for the slightly higher mortality levels. Although low birth weight or short gestation may also be an indicator of health issues that extend beyond infancy, the health disadvantage of multiple births is largely overcome in the first year of life. Parity Parity refers to the relationship between the current birth and previous births to the same mother. However, birth registrations collect information on the number of previous children born to the current partnership rather than the number of children ever born to the woman. A child may be recorded as the only child of the parents, but both or either parent may have previously had other children with different partners. It is also unclear whether birth registrations may not in fact frequently record all babies ever born to the mother rather than only those of the present relationship. The parity of children who died under five years of age was compared with the parity of all births. It had been surmised that first births might be more likely to result in an infant death than subsequent births to a mother. It was found that no significant differences in infant and early childhood deaths could be attributed to the parity of the birth. Ethnicity Ethnicity is not fixed. People in New Zealand, as in other countries, may change the way in which they identify themselves over time or identify themselves differently in different environments. Ethnic mobility occurs when people change their ethnic identification over time. For example, people's social environment may change leading to them identifying with additional or different ethnicities. People may also provide different responses depending upon the context or circumstances in which they are asked about their ethnicities. Ethnicity should ideally be collected by self-identification, but this is not always possible. People may identify their ethnicities differently from the way that a third party may identify them. Ethnic category jumping may occur because different people provide proxy responses to the ethnicity question. For example the ethnicity of babies and young children is generally identified by their parents. However, should a baby or young child die, a death registration may be completed by another family member or a funeral director. Ethnicities recorded in these circumstances may differ from the ethnicities identified by their parents at the birth of the child. Differences in ethnicity in the two different collections are of importance when measures of infant mortality by ethnicity are required. A key component of the matching of birth and death registrations was to understand the magnitude and effect

11

Page 13: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

of these differences in order to identify whether or not ethnicity was a factor in infant and early childhood deaths. Comparison of ethnicity at birth and at death The comparison of ethnicity between birth and death records looked at ethnicity at the 'major ethnic group' level (ie level one). There are currently five groups used at this level: European, Mäori, Pacific Peoples, Asian and Other. An individual can belong to one or more groups and they are counted in every ethnic group that they record as identifying with. Single response Mäori refers to records where Mäori was the only ethnic group stated. Table 9 Comparison of Ethnicities Recorded on Birth and Death Registrations for Deaths Under 5 Years of Age Age at Death (Years) Infants 1 2 3 4 Total under 5 Total number of matched records 1572 142 49 22 15 1800 Number with same ethnicities 1264 110 47 19 11 1451 Number with differing ethnicities 308 32 2 3 4 349

Of the 1800 death registrations matched to the relevant birth registrations, 81 percent recorded the same ethnicities in both places. However, 19 percent (349) differed (Table 9). Among infant deaths, the ethnicities recorded on the death certificates relative to the birth certificates showed that, in each year between 1999 and 2004, deaths among babies of:

• European ethnicities were understated (on average 13 percent) • Mäori ethnicity were understated (on average 8 percent) • Pacific ethnicities were understated (on average 21 percent) • Asian ethnicities were understated (on average 22 percent) • the single response Mäori group was overstated (on average 31 percent).

Note: ‘Understated’ indicates the proportion of individuals who stated the ethnic group on the birth registration but not on the death registration (‘overstated’ indicates the converse). The findings for early childhood deaths closely mirrored the patterns for infant deaths. For those aged one to four years at the time of death (registrations from July 1998 to June 2004) it was found that deaths among children of:

• European ethnicities were understated (on average 15 percent) • Mäori ethnicity were understated (on average 16 percent) • Pacific ethnicities were understated (on average 18 percent) • Asian ethnicities were understated (on average 13 percent) • the single response Mäori group was overstated (on average 37 percent).

Note: ‘Understated’ indicates the proportion of individuals who stated the ethnic group on the birth registration but not on the death registration (‘overstated’ indicates the converse). As Table 10 and Table 11 indicate, the most common difference occurred where a child was registered at birth as being of both European and Mäori ethnicities, but at death was registered with either European or Mäori ethnicity. This occurred in 126

12

Page 14: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

cases for those aged under one at death (410 percent) and 12 cases for those aged one at death (38 percent). Table 10 Differences in Ethnicities Recorded on Birth and Death Registrations Among Infant Deaths

Ethnicity on Birth Record Ethnicity on Death Record Percentage (N=316) Mäori 2.5 European/Mäori 1.3

European

Not Specified 4.4 European 2.5 Mäori European/Mäori 3.8 European 1.3 Mäori 0.9 Mäori/Pacific 2.5

Pacific

Not Specified 2.5 European 14.2 Mäori 25.6

European/Mäori

Not Specified 0.9 European 2.8 Mäori 2.8 Pacific 1.6

European/Pacific

Not Specified 1.3 European 1.3 European/Asian Asian 0.9

Mäori/Pacific Mäori 4.1 Pacific/Asian Pacific 1.9

European 1.3 Mäori 2.8 Pacific 1.6 European/Mäori 1.3

European/Mäori/Pacific

Mäori/Pacific 1.3 European/Mäori/Asian Mäori 0.9 Other combinations not elsewhere mentioned 11.7 Table 11 Differences in Ethnicities Recorded on Birth and Death Registrations Among Deaths Aged 1 to 4 Years

Ethnicity on Birth Record Ethnicity on Death Record Percentage (N=43)

European Not Specified 7.0 European 11.6 Mäori 25.6

European/Mäori

Not Specified 7.0 Mäori/Pacific Mäori 7.0 European/Mäori/Pacific Mäori 4.7 Other combinations not elsewhere mentioned 37.1 The above results cover the matched records. However, unmatched records may not follow the same pattern and may have been sufficiently different to affect conclusions on ethnic reporting. The 212 unmatched records for infant deaths were studied. The differences between ethnicity recorded on the birth and death registrations for the matched records followed a similar pattern for each year studied. Unmatched death

13

Page 15: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

registrations also followed similar patterns each year but the patterns observed among the unmatched records differed from those found in the matched records (Table 12). Table 12 Percentage of Infant Deaths by Ethnicity for Matched and Unmatched Death Records by Year of Registration Matched Death Records (N=1572) Unmatched Death Records (N=212)

Year Ended June European Mäori SR Mäori European Mäori SR Mäori 1999 58 39 28 31 43 31 2000 56 36 24 29 42 39 2001 52 37 25 21 67 58 2002 52 38 27 30 58 45 2003 56 33 23 22 57 39 2004 48 39 30 19 49 43

Failure to match a death registration to a corresponding birth registration may reflect incomplete or inconsistent information being recorded in the key fields used to match records, though lifestyle, living arrangements, housing and mobility and cultural factors may also contribute to the problem. For example it is possible that a birth registration was not completed in the study period. Late registrations do occur, especially when other events, such as the death of the child, intervene. Cultural factors may influence the birth registration of a congenitally ill child or a child who has since died. Disproportionately, many Mäori deaths were not able to be matched, and this pattern was more marked for Mäori who gave a single response. Since approximately 78 percent of mixed ethnic births belong to the Mäori ethnic group and 60 percent of Mäori children have multiple ethnicity this suggests that unmatched records may further understate mortality rates among babies of European, Pacific and Asian ethnicities. However, the small number of records (212) involved suggests that this would not either influence greatly or invalidate the results from the matched records. Ethnic adjustment factor In order to derive sound values for important measures such as infant mortality, data used from different sources to derive the rates should be as consistent as possible. It is rarely possible to have perfect consistency, but this study provides a means of directly measuring the magnitude of the effect of the inconsistencies. Using the differences between the ethnicities registered on the birth and death records over a six-year period (1998 to 2004), an adjustment factor was calculated. This adjustment factor was then used to adjust the infant mortality rates (IMR) for each ethnicity. The underlying assumption was that the ethnicities recorded at birth more accurately reflected the ethnic identity intended by the parents of the baby. The percentage change between the IMR based on death records and the IMR based on birth records was calculated for each of the six years. The first adjustment factor used was an average of these six years. The second adjustment factor used is an average of only the four mid-values (the highest and lowest values being excluded from this adjustment) to eliminate any outliers that may have skewed the data. The final adjustment factor for each ethnicity was derived from these first two measures (Table 13).

14

Page 16: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Table 13 Published and Adjusted Infant Mortality Rates, 1998–2004(1)

Total Population

European Mäori Pacific Asian Single Response Mäori

Births registered 336,902 239,450 94,059 49,018 29,655 38,806 Deaths registered 1796 918 708 306 100 511 Infant mortality rate 5.33 3.83 7.53 6.24 3.37 13.17 Adjustment factor ... 1.13 1.06 1.13 1.19 0.82 Adjusted infant mortality rate 5.33 4.33 7.98 7.05 4.01 10.80

(1) Per 1,000 estimated mean population. Symbol: ... not applicable Conclusion The electronic availability of named birth and death records enabled a high level of matching between death and birth registrations to be achieved. This same high level of matching was not possible when the only details provided were meshblock, date of birth and sex. Full named records provide sufficient relevance or accuracy to enable adjustment to demographic indices, projections and estimates. The Demography Division of Statistics New Zealand currently only publishes mortality rates for the total population and for the Mäori population. This matching project has ascertained that infant mortality for the Mäori ethnicity group was consistently the most accurate over the six years investigated. The European and Pacific infant mortality rates were the next most accurate. However, large fluctuations in accuracy from year to year were seen in the Pacific and, to a greater extent, the Asian ethnic groups because of the very small number of deaths. The infant mortality rate for the single-response Mäori group was shown to be the most inaccurate. Given both the small Pacific and Asian populations and the instability of the single-response Mäori data, continuing to limit publication of mortality rates to total population and Mäori ethnic group is in the best interests of accuracy and reliability. The trial matching exercise in 1999 concluded that it was unnecessary to apply adjustment factors when deriving infant mortality rates for the Mäori ethnic group. Results from this study, which draws on a larger sample of infant mortality, are less conclusive. This outcome, together with subsequent changes, such as a change in questions used, implies that a return to routine matching may be necessary. Should calculated adjustment factors change markedly it will become necessary to apply adjustments in order to maintain a valid time series. There is an even stronger argument for using adjustment factors if, in future, infant mortality rates are produced for Pacific, Asian and European groups.

15

Page 17: Infant Deaths: Demographic Characteristics, Ethnic Reportingarchive.stats.govt.nz/~/media/Statistics/browse... · Infant Deaths: Demographic Characteristics, Ethnic Reporting Reprint

Glossary Live birth The birth of a child, who breathes or shows any other evidence of life, such as beating of the heart, pulsation of the umbilical cord or definite movement of voluntary muscles, whether or not the umbilical cord has been cut or the placenta is attached. All live-born infants should be registered and counted as such, irrespective of length of gestation or whether alive or dead at the time of registration. If they die at any time following birth, they should also be registered and counted as deaths. Still birth A dead foetus that either weighs 400g or more when issued from its mother; or is issued from its mother after the 20th week of gestation. The Births, Deaths and Marriages Registration Act 1995, which took effect from 1 September 1995, redefined what constituted a stillbirth. Before the new Act, a stillbirth was defined as a child born dead after 28 weeks of gestation. This change in definition means stillbirths from September 1995 onwards are not directly comparable with earlier years. Confinement A pregnancy resulting in either liveborn or stillborn children. Such an event is counted as one confinement irrespective of whether a single or multiple birth results.

16