TECHNICAL NOTE CONFIDENCE INTERVALS FOR INDIGENOUS CHILD MORTALITY RATES


INTRODUCTION

1 In 2008 there were 1,200 infant deaths (deaths of children less than one year of age) and 230 deaths of children aged 1-4 years, comprising 0.9% and 0.2% of all deaths in Australia respectively. Disaggregation of these data to more detailed levels - in particular, by Indigenous status and state/territory of usual residence - will necessarily lead to smaller numbers.

2 As a result of such numbers, derived mortality rates may have high levels of uncertainty associated with them. This Technical Note illustrates levels of uncertainty for infant and child (1-4 years) mortality rates for 2006-2008, by state/territory, sex and Indigenous status of the child, through the calculation of confidence intervals.


VARIABILITY IN DEATH RATES

3 Death rates in this publication are based on the mortality experience of an entire population, rather than from a sample. As there is no sampling error associated with this data, it might be considered that standard errors are not relevant to the question of variability in death rates.

4 However, observed death rates are commonly treated as estimates of an underlying death rate that gives the probability of death for individuals in a population. This presupposes a statistical model in which each person is subject to some risk of dying over a given period, and whether each person in fact dies is a random event. Considered in this sense, the underlying death rates are unknown, and the observed death rates measure them with some sampling error.

5 The standard errors used to derive confidence intervals presented in this Technical Note are calculated to measure the extent of such random variations in numbers of deaths, using the methods described in Chiang (1984).


INFANT MORTALITY RATES

6 Infant mortality rates (IMRs) by Indigenous status and state/territory of usual residence for 2006-2008 are presented in the following graph (for males and females combined).

7 Confidence intervals associated with non-Indigenous IMRs are relatively small, apart from in the Northern Territory.

8 For Indigenous IMRs the range is larger, especially for the Northern Territory and South Australia, for which confidence intervals for males and females combined range from 10.3-16.9 infant deaths per 1,000 live births and 3.3-9.5 infant deaths per 1,000 live births respectively. Rates for males and females separately show wider ranges (see table below).

Infant mortality rates, 2006-2008
				95% confidence intervals
		Infant deaths	Infant mortality rate(a)	Lower limit	Upper limit
INDIGENOUS
Males
	NSW	48	8.3	6.0	10.6
	Qld	53	8.4	6.2	10.7
	SA	9	6.8	2.4	11.2
	WA	37	11.5	7.8	15.2
	NT	37	15.1	10.3	19.9
Females
	NSW	39	7.1	4.9	9.4
	Qld	45	7.4	5.3	9.6
	SA	7	5.9	1.5	10.2
	WA	29	8.8	5.6	12.0
	NT	27	11.9	7.4	16.4
Persons
	NSW	87	7.7	6.1	9.4
	Qld	98	7.9	6.4	9.5
	SA	16	6.4	3.3	9.5
	WA	66	10.1	7.7	12.6
	NT	64	13.6	10.3	16.9
NON-INDIGENOUS
Males
	NSW	653	4.9	4.5	5.3
	Qld	448	5.3	4.8	5.8
	SA	97	3.5	2.8	4.2
	WA	116	2.9	2.3	3.4
	NT	15	4.4	2.2	6.6
Females
	NSW	465	3.7	3.3	4.0
	Qld	322	4.1	3.6	4.5
	SA	87	3.2	2.6	3.9
	WA	121	3.2	2.6	3.7
	NT	11	3.3	1.3	5.2
Persons
	NSW	1 118	4.3	4.0	4.6
	Qld	770	4.7	4.4	5.0
	SA	184	3.4	2.9	3.9
	WA	237	3.0	2.6	3.4
	NT	26	3.8	2.4	5.3
(a) Infant deaths per 1,000 live births.

AGE-SPECIFIC DEATH RATES, CHILDREN AGED 1-4 YEARS

9 Age-specific death rates for children aged 1-4 years by Indigenous status for NSW/Qld combined and SA/WA/NT combined for 2006-2008 are presented in the following graph (for males and females combined).

10 Similar to non-Indigenous IMRs, confidence intervals associated with death rates for non-Indigenous children aged 1-4 years are relatively small.

11 However, confidence intervals for Indigenous children aged 1-4 years (males and females combined) show very large ranges for both NSW/Qld combined and SA/WA/NT combined, such that the lower limits are lower than the age-specific death rates for non-Indigenous children. Such wide ranges severely question the usefulness of these rates. Confidence intervals for males and females separately are wider still (table below).

Age-specific death rates, 1-4 years, 2006-2008
				95% confidence intervals
		Deaths, children aged 1-4 years	Age-specific death rate(a)	Lower limit	Upper limit
Indigenous
NSW/Qld
	Males	24	51.3	15.8	86.8
	Females	15	33.6	4.2	62.9
	Persons	39	42.6	19.5	65.8
SA/WA/NT
	Males	13	55.2	3.3	106.9
	Females	13	56.9	3.4	110.4
	Persons	26	56.0	18.8	93.2
Non-Indigenous
NSW/Qld
	Males	199	23.9	18.2	29.7
	Females	145	18.4	13.2	23.6
	Persons	344	21.2	17.4	25.1
SA/WA/NT
	Males	67	24.4	14.3	34.5
	Females	36	13.9	6.0	21.7
	Persons	103	19.3	12.8	25.7
(a) Deaths per 100,000 population.

CONCLUSION

12 The death rates presented above for Indigenous Australians have large associated confidence intervals and should be interpreted with a high degree of caution. Making comparisons over time (for example, year to year comparisons), between jurisdictions or between Indigenous and non-Indigenous Australians is therefore problematic due to the high level of uncertainty. In particular, confidence intervals for death rates of Indigenous children aged 1-4 years are very large as a result of the very small numbers of deaths in these ages.

13 While methods such as aggregating data over jurisdictions and/or longer time periods may help to reduce the width of the confidence intervals, this may limit the usefulness of the data for comparisons over time. In the case of Indigenous children aged 1-4 years, aggregation has not improved the confidence intervals to a level at which the rates could be interpreted with confidence.

14 The confidence intervals presented in this Technical Note relate to random variations in numbers of deaths only. No other sources of errors or deficiencies in either deaths or population data have been considered in these calculations. For example, variations in Indigenous death rates may arise from uncertainty in the recording of Indigenous status on the death registration form (in particular, under-identification of Indigenous deaths) and in the Census, from which population estimates are derived (for more information see Chapter 3: Deaths of Aboriginal and Torres Strait Islander Australians). Confidence intervals would differ to those presented here if all these sources of variations were considered.