a School of Public Health, Queensland University of Technology, VictoriaPark Road, Kelvin Grove, Brisbane, QLD 4059 Australia. E-Mail: g.turrell{at}qut.edu.au
b Australian Institute of Health and Welfare, GPO Box 570, Canberra ACT 2601, Australia. E-mail: colin.mathers{at}aihw.gov.au
Abstract
Background Socioeconomic inequalities in mortality have been repeatedly observed in Britain, the US, and Europe, and in some countries there is evidence that the differentials are widening. This study describes trends in socioeconomic mortality inequality in Australia for males and females aged 014, 1524 and 2564 years over the period 19851987 to 19951997.
Methods Socioeconomic status (SES) was operationalized using the Index of Relative Socioeconomic Disadvantage, an area-based measure developed by the Australian Bureau of Statistics. Mortality differentials were examined using age-standardized rates, and mortality inequality was assessed using rate ratios, gini coefficients, and a measure of excess mortality.
Results For both periods, and for each sex/age subgroup, death rates were highest in the most disadvantaged areas. The extent and nature of socioeconomic mortality inequality differed for males and females and for each age group: both increases and decreases in mortality inequality were observed, and for some causes, the degree of inequality remained unchanged. If it were possible to reduce death rates among the SES areas to a level equivalent to that of the least disadvantaged area, premature all-cause mortality for males in each age group would be lower by 22%, 28% and 26% respectively, and for females, 35%, 70% and 56%.
Conclusions The mortality burden in the Australian population attributable to socioeconomic inequality is large, and has profound and far-reaching implications in terms of the unnecessary loss of life, the loss of potentially economically productive members of society, and increased costs for the health care system.
Keywords Socioeconomic status, mortality inequality, Australia, area-based measures
Accepted 21 December 1999
Despite marked improvements in the health of their populations, Great Britain, the US and many European countries continue to observe mortality inequalities between socioeconomic groups.13 There is now a large literature which shows that the socioeconomically disadvantaged experience higher mortality rates for most major causes of death, and this inequality exists for both males and females at every stage of the life-course. There is also growing evidence that mortality inequalities have widened over time in these countries.411 These increasing disparities appear to be due to faster declines in mortality among those of higher socioeconomic status (SES),4,12 although in some countries there is evidence of an actual increase in mortality rates for some conditions among the most disadvantaged.
Socioeconomic inequalities in mortality have also been repeatedly observed within the Australian population,1316 however, changes in the extent and nature of mortality inequality have received only limited coverage. This present study extends a mortality analysis conducted by Mathers who used an area-based measure of SES to examine mortality inequalities among children (014 years), young adults (1524 years) and working-aged adults (2564 years) for the period 19851987 and found that people living in disadvantaged areas experienced the worst health.1719 Using these data and results as a baseline, we employ an identical methodology and use the same life-course groupings to examine trends and inequalities in mortality for males and females over the period 19851987 to 19951997. Four issues are addressed. First, we determine whether the mortality inequalities in 19851987 were still evident in 19951997. Second, we investigate whether the differentials widened, consistent with the trend observed in many other countries. Third, we attempt to locate the basis of any widening inequality: were the increasing disparities due to faster reductions in the mortality rates of higher SES groups, or a worsening of the health status of the most disadvantaged, as reflected in an increase in mortality rates. Fourth, we estimate the extent to which the proportion of total mortality in the population attributable to socioeconomic disadvantage has changed over the 10-year period. Specifically, have there been significant changes in terms of how much of Australia's overall mortality can be attributed to variability between socioeconomic areas.
Methods
Measurement of socioeconomic status
This study uses a geographical measure known as the Index of Relative Socioeconomic Disadvantage (IRSD) developed by the Australian Bureau of Statistics (ABS) using data collected in the 1986 and 1996 Census to categorize areas on the basis of their social and economic characteristics.20,21 The IRSD is constructed using principal components analysis and is derived from attributes such as low income, low educational attainment, high levels of public sector housing, high unemployment, and jobs in relatively unskilled occupations. The IRSD is compiled initially at the Collector's District (CD) level, a census collection unit broadly equivalent in urban areas to a small group of suburban blocks, comprising approximately 250 dwellings (CD in rural regions usually contain fewer dwellings). This study uses IRSD scores for Statistical Local Areas (SLA), which in most cases correspond to council boundaries defined by Local Government Areas. In Australia, there are a total of 1315 SLA, with a mean population in each SLA of approximately 13 900 people (median 5400). The IRSD scores for each SLA are constructed by taking the weighted average, using population counts from the 1986 and 1996 Census, across all CD comprising the SLA. In aggregate, SLA cover the whole of Australia without gaps or overlaps. For the years 19851987 and 19951997, those deceased were classified into quintiles of socioeconomic disadvantage according to the value of the IRSD for their SLA of usual residence, with Q1 corresponding to the highest socioeconomic area and Q5 the lowest. The SLA were grouped into quintiles so that each contained approximately 20% of the total Australian population.
Mortality analysis
Mortality for the two periods is expressed using rates per 100 000 population directly age standardized (using 5-year age groups) to the total mid-year Australian population in 1988. Unit record mortality registration data were obtained from the ABS, where deaths were coded according to the Ninth Revision of the International Classification of Diseases. For males and females aged 014, 1524 and 2564 years, age-standardized rates were calculated for all causes, for specific causes that were major contributors to all-cause mortality (e.g. cancer) and for selected causes that contributed most to the specific-cause (e.g. lung cancer). Population data by age, sex and SLA were also supplied by the ABS and consisted of estimates of the population in 19851987 and 19951997 for each of the aforementioned sex/age subgroups and SLA grouped into quintiles of disadvantage. These population estimates were derived using data from the 1986 and 1996 Census, adjusted for under-enumeration.
Measures of mortality inequality
Rate ratios
The age-standardized mortality rate for the most disadvantaged quintile (Q5) is expressed as a multiple of the standardized rate for the least disadvantaged (Q1). Thus, for example, the rate ratio for all-cause mortality for males aged 014 years in 19851987 is 1.50 (RateQ1/RateQ5 = 125.6/83.8). Other researchers have also used the rate ratio to quantify the magnitude of socioeconomic mortality inequalities.11,12
Gini coefficient
In recent years, studies examining socioeconomic health inequalities have made increasing use of the Gini coefficient2224 particularly as an indicator of income inequality.25,26 The Gini coefficient summarizes inequality in the distribution across all subgroups of the population, unlike the rate ratio described above, which summarizes only the differences between the top and bottom groups. The Gini is based directly on the Lorenz curve, a graphical device for displaying the cumulative share of total income accruing to successive income intervals.25 In Figure 1, for example, the X and Y ordinates represent the proportion of people and income respectively, the 45° diagonal is the line of equality and the dashed line is the Lorenz curve. If no inequality exists, the Lorenz curve corresponds to the line of equality. As the extent of inequality increases, so does the area between the line of equality and the Lorenz curve. The Gini is defined as the area enclosed by the line of equality and the Lorenz curve expressed as a proportion of the area below the diagonal and is bounded to range from zero (complete equality) to one (complete inequality). In this present study we use a form of the Lorenz curve in which cumulative deaths are plotted against cumulative population across the five quintiles Q5 to Q1 (ranked in terms of decreasing disadvantage). Even if age-specific death rates were equal across all quintiles, there would still be inequality if population age structures differ across the quintiles (since there will be more deaths in older populations). To remove the effects of population age structure on the Lorenz curve we have plotted cumulative numbers of age-standardized deaths across quintiles. The corresponding Gini index measures the degree of mortality inequality across the quintiles of disadvantage, excluding inequality due purely to population age structure differences. The term Gini coefficient is used here to refer to a measure of mortality inequality based on population groups ranked by SES rather than health status. Wagstaff et al. have referred to these as health or ill-health concentration indices.27
|
Statistical testing
Although analytical solutions for 95% CI for the mortality inequality measures can be constructed, we used a simulation approach. Deaths were assumed to follow Poisson distributions, and Latin hypercube sampling was performed using the @RISK software program.29 This program allows input data (such as observed cancer deaths among males aged 2564 in Q1) to be specified as a distribution (in this case a Poisson distribution with a mean equal to the observed deaths) rather than as a single number. The program then calculates the output measures (i.e. rate ratios, Gini coefficients, excess mortality), many times taking random samples from each of the input distributions. The resulting distribution of the output measures can then be used to perform statistical tests and to estimate 95% CI. The use of such simulation measures avoids the difficult analytical task of deriving distributional parameters such as standard errors for complex indicators such as the Gini coefficient. All measures of mortality inequality shown in Tables 2 and 3 differ significantly from no inequality (1 for the rate ratio, and 0 for the Gini coefficient and excess mortality) at P < 0.001. Asterisks indicate the level of significance of the difference between the 19951997 and 19851987 values against the null hypothesis of no difference.
|
|
Mortality rates
Table 1 presents age-standardized mortality rates for males and females in the first, third and fifth quintiles of the IRSD (hereafter referred to as high, middle and low SES). To save space, data for the second and fourth quintiles are not presented: almost without exception, death rates for these quintiles were intermediate between high, middle and low SES. For the period 19851987 death rates were consistently highest for those living in the most disadvantaged areas. This pattern was evident for both males and females in each of the three age groups and was observed for all causes, for broad cause groups (e.g. circulatory system disease) and for individual causes (e.g. stroke). Despite marked overall declines in mortality rates between 19851987 and 19951997 for the majority of conditions, the socioeconomic differentials observed in the earlier period were still evident a decade later.
|
Mortality inequalities
Table 2 presents age-standardized rate ratios and Gini coefficients by SES area for males and females for the periods 19851987 and 19951997. When assessing changes in mortality inequality, we gave greatest weight to the Gini coefficient, as it reflects the degree of inequality across all socioeconomic quintiles. Less weight was given to the rate ratio, as it simply reflects the magnitude of the differential between the fifth and the first quintile. Among males, there was evidence of increased mortality inequality for all causes (014 and 1524 years), sudden infant death syndrome (SIDS), injury and poisoning (014 and 1524 years), motor vehicle traffic accidents (all age groups) and suicide (1524 years). For males aged 2564 increases were also evident for circulatory system disease (including coronary heart disease), cancer (including lung cancer), and asthma/emphysema. Mortality inequality remained relatively unchanged between 1985 and 1997 for perinatal conditions, stroke, diabetes mellitus, and for respiratory system disease classified at the broad cause level. Decreases in mortality inequality were evident for drug dependence, and among males aged 2564 years for all-cause mortality, injury and poisoning (including suicide), pneumonia/bronchitis and digestive system diseases.
Among females, increased mortality inequality was observed for SIDS, motor vehicle traffic accidents (1524 and 2564 years), and for coronary heart disease, diabetes mellitus, cancer (including lung cancer) and respiratory system diseases (including asthma/emphysema). There was little change in mortality inequality between 1985 and 1997 for injury and poisoning (014 years), and among those aged 2564 years for all-cause mortality, circulatory system disease classified at the broad group level, and digestive system diseases. Decreases in mortality inequality were evident for all causes (014 and 1514 years), perinatal conditions, injury and poisoning (1524 and 2564 years), drug dependence, and motor vehicle traffic accidents (014 years). For females aged 2564 decreases were also evident for stroke, suicide, and pneumonia/bronchitis.
Using data in Table 1 we also examined the basis of change in mortality over the 10-year period between the areas: in other words, were the declines in mortality similar or different across the SES quintiles (results not shown). For males aged 014 and 1524 years, declines in mortality rates tended to be greatest in the top quintile and smallest in the bottom, although there were a number of exceptions to this pattern (e.g. perinatal conditions, injury and poisoning). For females, declines in mortality for these two age groups were somewhat the reverse that of males: for the majority of conditions, the greatest declines occurred in the most disadvantaged quintiles, although again there were exceptions (e.g. SIDS, motor vehicle traffic accidents). Among males and females aged 2564 years, no discernible pattern was evident with respect to mortality declines. There were roughly equal instances where the declines were greater in the top and bottom quintiles, and for a number of conditions the rate of decline was similar for each quintile (e.g. stroke for each sex, circulatory system and coronary heart disease among women).
Contribution of socioeconomic status inequalities to population mortality
Table 3 presents estimates of the mortality burden that is attributable to variability in death rates across the SES quintiles. Interpretation of the estimates is straightforward. Take for example, lung cancer rates for males for the period 19951997. If quintiles Q2Q5 had the same rate as the highest SES quintile (Q1), mortality from lung cancer among males aged 2564 would be lower overall by approximately one-third (35%). This estimate has increased significantly since 19851987, where approximately one-fifth (23%) of overall lung cancer deaths among males of this age group were due to variability between the SES quintiles. For males, increases in mortality burden over the decade were evident for many other conditions, most notably, SIDS, motor vehicle traffic accidents (014 and 1524 years), suicide (1524 years), coronary heart disease, diabetes mellitus, cancer and asthma/ emphysema. The only apparent anomaly in an otherwise fairly consistent pattern were male deaths due to drug dependence: here the estimate for 19951997 (0%) reflects the fact that since 19851987 death rates due to drug use have tended to equalize across the quintiles. For males aged 014 years in 19951997, potential reductions in overall mortality ranged from 10% for perinatal conditions to 51% for SIDS. Among males aged 1524, potential reductions for the same period ranged from 25% for suicide to 44% for motor vehicle traffic accidents, and for those aged 2564, potential reductions ranged from 19% for cancer to 53% for asthma/emphysema. Among females, mortality burden due to variability across the quintiles was considerably higher than for males: this pattern was evident for both periods and for all conditions. The only notable changes in mortality burden between 1987 and 1997 for females was in terms of perinatal conditions (a reduction from 45% to 24%) and asthma/emphysema (an increase from 38% to 60%). The estimates of mortality burden for females suggested that substantial reductions in overall mortality would occur if all quintiles had a death rate equivalent to that of the highest SES. For females aged 014 in 19951997, this ranged from 24% for perinatal conditions to 66% for SIDS. For females aged 1524 the potential reductions were in the range 70% for all causes to 79% for injury and poisoning and motor vehicle traffic accidents, and for those aged 2564, potential reductions ranged from 23% for cancer to 84% for coronary heart disease. Finally, the estimates of mortality burden for drug dependence among females (41) reflects the fact that since 19851987 the relationship between SES and this cause of death has reversed direction. During the earlier period, the highest death rates were observed in the most disadvantaged areas, whereas for the period 19951997 rates were highest in the least disadvantaged.
Discussion
Methodological issues
Before discussing the study's findings, we need to consider a number of potential sources of bias in the mortality analyses, and in the use of the IRSD. First, death rates are calculated using numerator data that are collected as part of the mortality registration process, whereas the denominator data are derived from the population census. Mortality rates will be in error to the extent that deaths for a particular sex/age subgroup attributed to an SLA are not in fact drawn from that SLA.17 Quantifying the magnitude of bias resulting from these types of errors is difficult, however, our best estimates indicate that misclassification of deaths based on sex and place of residence is small, thus the overall impact on the mortality rate is likely to be minimal.30
Second, prior to calculating the mortality rate it was necessary to exclude death records where the identifier for the SLA of usual residence was missing, or where it was not possible to assign the SLA an IRSD score (because of small population numbers, IRSD was not calculated for a few SLA). These problems arose for approximately 3% and 0.4% of deaths in 19851987 and 19951997, respectively. The exclusion of these cases will have had little effect on the estimates of mortality inequality for the two periods.
Third, in assessing the mortality inequalities reported here, it should be remembered that the Australian population has been classified into quintiles using a small area index of socioeconomic disadvantage. Different estimates of mortality inequality would have been obtained if we had used either of the following: a different reference group (e.g. top decile rather than top quintile); a smaller more socioeconomically homogeneous unit of analysis (e.g. Collectors District instead of SLA); or an individual-level rather than an area-based indicator of SES (because the IRSD relates to the average disadvantage of all people living in an area). We used quintiles of socioeconomic disadvantage to ensure direct comparability with the methodology used at baseline (19851987), thus allowing us to assess whether inequalities had widened or narrowed during the ensuing decade. Further, Australian death data do not permit area-based analyses of mortality inequality to be conducted using smaller units such as collectors districts or characteristics of individuals. Importantly, it should be stressed that the data and measurement limitations that constrained our approach have produced estimates of mortality inequality that almost certainly understate the true extent of the mortality burden by level of socioeconomic disadvantage in Australia.
Finally, it should be noted that the IRSD for each reference period were calculated using data from the relevant population censuses and hence some SLA may have changed quintile between 1986 and 1996. Additionally, there are differences between some SLA boundaries for the two periods. Thus the corresponding quintiles for 19851987 and 19951997 do not consist of exactly the same areas, although for both periods, the top and bottom quintiles contain the 20% most advantaged and disadvantaged IRSD respectively.
Mortality rates
Between 19851987 and 19951997 mortality rates for the majority of causes declined markedly for all quintiles. Despite these overall improvements in health, however, both periods were characterized by large socioeconomic inequalities. Death rates were typically highest in the most disadvantaged areas for both males and females in each age group. Moreover, mortality rates very often fell in a continuous linear gradient from the most to the least disadvantaged quintile. Thus in Australia, as elsewhere,31 mortality inequalities are not confined to differences between the rich and poor, but rather, are observed across the entire socioeconomic spectrum.
Mortality rates for a number of conditions increased (or remained relatively stable) between 1985 and 1995. Among both sexes this was evident for suicide, drug dependence, diabetes mellitus and asthma/emphysema, and for lung cancer among women. Increases in rates of suicide are consistent with that reported in other Australian studies,32 and when viewed in conjunction with a rise in drug deaths, suggest a worsening in the mental and psychosocial health of some sections of the Australian population, most particularly, adolescents and young adults. Increases in mortality rates for asthma/emphysema and lung cancer among women were limited mainly to the middle and low SES quintiles and presumably reflect a number of interrelated factors directly linked to tobacco consumption. These may include higher smoking rates in disadvantaged areas, limited reach and uptake of anti-smoking campaigns and hence lower cessation rates, and increases in cigarette smoking among young women.33,34
Mortality inequalities
A number of studies from Britain,4,5,35,36 the US,7,8 and Europe1012,37,38 have examined socioeconomic trends in mortality inequality for all causes. The general finding is that during the last few decades, all-cause mortality inequalities have widened, although the extent of the increase appears to differ by gender and age group. The evidence, however, is not entirely consistent: studies have reported little change in socioeconomic mortality inequality for women,38 and a narrowing of differentials among younger men38 and infants.39 The results of the present study also show a somewhat mixed picture in terms of mortality inequality for all causes by gender and age. Among females, there was a decrease in all-cause mortality inequality between 19851987 and 19951997 for each age group. Among males, all-cause mortality inequality increased over the reference period for those aged 014 and 1524 years, and decreased slightly for those aged 2564 years.
In terms of mortality inequality for specific conditions, we are again presented with a picture that varies by gender and age. Among males aged 014 and 1524 years, mortality inequality increased for each specific cause except perinatal conditions and drug dependence. Increases in mortality inequality for females in these age groups were less common than those found for males: over the reference period, mortality differentials increased only for SIDS and motor vehicle traffic accidents. Given the recent interest in Australia in drug use among youth, the mortality trends due to drug dependence warrant some comment. The decrease in inequality for drug deaths was due to a faster rate of increase in illicit drug deaths in higher socioeconomic areas, and is likely to reflect the increasing use and availability of heroin in Australia across all socioeconomic groups. Among males and females aged 2564, mortality inequality increased for six and eight specific conditions, respectively, with widening differentials being evident for a number of the major contributors total mortality in Australia, such as circulatory system disease and cancer.
Overseas studies have reported that increases in mortality inequality among socioeconomic groups are due mainly to greater declines in death rates among those of high SES.4,12 The results of this present study generally concur with this evidence. Most of the significant increases in mortality inequality between 19851987 and 19951997 were associated with a greater decline in death rates in the high SES areas. Importantly, not all increases occurred for this reason: for diabetes mellitus and asthma/emphysema among both sexes, and lung cancer among women, increases in mortality inequality were due to an actual increase in death rates in the most disadvantaged quintile.
The simultaneous occurrence of widening, narrowing and unchanging mortality inequalities which characterizes our results is difficult to explain, especially if one attempts to do so solely on the basis of broad ranging societal-level explanations.40 It may also be necessary to seek explanations that are more narrowly conceived, that are specific to a particular cause or group of causes with a similar aetiological profile, and that occur over a similar timeframe. Injury and death due to accidents for example, are likely to be a consequence of events occurring within a narrow timeframe, and are likely to be due to the direct impact of material and physical conditions in the wider environment. Coronary heart disease, by contrast, is likely to be due mainly to the cumulative impact of behavioural and psychosocial factors occurring over many decades. It may also be necessary to focus our explanatory lens on a particular sex/age subgroup. The mechanisms and processes underpinning increases in mortality inequality for motor vehicle traffic accidents among males aged 014 years are likely to be qualitatively different from those that contributed to decreases in mortality inequality for suicide among females aged 2564 years.
Contribution of socioeconomic status inequalities to population mortality
The final section of this study's analysis estimated the extent to which socioeconomic inequalities in death rates contributed to the total mortality burden in the general population. In other words, if it were possible to reduce death rates among the SES areas to a level equivalent to that of the least disadvantaged quintile, what would be the potential savings in premature mortality? The results of this analysis showed that the mortality burden attributable to socioeconomic inequality was large, and that for many conditions, the burden had increased significantly between 1985 and 1997. Among males in the three age groups for example, excessive all-cause mortality in the mid-to-late 1990s was 22%, 28% and 26%, respectively. Among females, the corresponding excesses were 35% (014 years), 70% (1524 years) and 56% (2564 years). When expressed in terms of the absolute numbers of premature deaths, the mortality burden attributable to socioeconomic inequality is particularly stark. For the period 19951997, a reduction of 26% in all-cause mortality among males aged 2564 would have resulted in a saving of approximately 12 418 premature deaths. The corresponding figures for circulatory system disease and cancer were 4190 and 2944 deaths respectively. For females in the same age group, the number of premature deaths attributable to socioeconomic inequality was approximately 14 532 for all causes, 3504 for circulatory system disease, and 3038 for cancer. The size of the mortality burden attributable to variability among the quintiles of area disadvantage in Australia clearly has far-reaching implications: not only in terms of the unnecessary loss of life, but also in terms of the loss of potentially economically productive members of society, and added costs for the health care system and other public sectors more generally.41 In so called civil societies such as Australia, therefore, there is ample justification on humanistic, social, and economic grounds to strive towards the reduction and eventual elimination of all socioeconomic health inequalities.
References
1 Acheson D. Independent Inquiry into Inequalities in Health. London: The Stationery Office, 1998.
2 Drever F, Whitehead M (eds). Health Inequalities: Decennial Supplement. London: The Stationery Office, 1997.
3 Pamuk E, Makuc D, Heck K, Reuben C, Lochner K. Socioeconomic Status and Health Chartbook. Health, United States, 1998. Hyatsville, MD: National Center for Health Statistics, 1998.
4 Marmot M, McDowall ME. Mortality decline and widening social inequalities. Lancet 1986;ii:27476.
5 Marang-van de Mheen PJ, Davey Smith G, Hart CL, Gunning-Schepers LJ. Socioeconomic differentials among men within Great Britain: time trends and contributory casues. J Epidemiol Community Health 1998;52:21418.[Abstract]
6
McCarron PG, Davey Smith G, Womersley JJ. Deprivation and mortality in Glasgow: changes from 1980 to 1992. Br Med J 1994; 309:148182.
7
Pappas G, Queen S, Hadden W, Fisher G. The increasing disparity in mortality between socioeconomic groups in the United States 1960 and 1986. N Engl J Med 1993;329:10309.
8 Feldman JJ, Makuc DM, Kleinman JC, Cornoni-Huntley J. National trends in educational differentials in mortality. Am J Epidemiol 1989;129:91933.[Abstract]
9 Dunleep HO. Measuring socioeconomic mortality differentials over time. Demography 1989;26:34551.[ISI][Medline]
10 Borrell C, Plasencia A, Pasarin I, Ortun V. Widening social inequalities in mortality: the case of Barcelona, a southern European city. J Epidemiol Community Health 1997;51:65967.[Abstract]
11 Regidor E, Gutierrez-Fisac JL, Rodriguez C. Increased socioeconomic differences in mortality in eight Spanish provinces. Soc Sci Med 1995;41:80107.[ISI][Medline]
12 Diderichsen F, Hallqvist J. Trends in occupational mortality among middle aged men in Sweden 19611990. Int J Epidemiol 1997;26: 78287.[Abstract]
13 McMichael AJ. Social class (as estimated by occupational prestige) and mortality in Australian males in the 1970s. Community Health Stud 1985;15:32127.
14 Siskind V, Najman JM, Copeman R. Socioeconomic status and mortality revisited: an extension of the Brisbane area analysis. Aust J Public Health 1992;16:31520.[ISI][Medline]
15 Bennett S. Socioeconomic inequalities in coronary heart disease and stroke mortality among Australian men, 19791993. Int J Epidemiol 1996;25:26675.[Abstract]
16 Burnley IH. Inequalities in the transition of ischaemic heart disease mortality in New South Wales, Australia, 19691994. Soc Sci Med 1998;47:120922.[ISI][Medline]
17 Mathers C. Health Differentials Among Adult Australians Aged 2564 Years. Australian Institute of Health and Welfare; Health Monitoring Series No. 1. Canberra: AGPS, 1994.
18 Mathers C. Health Differentials Among Australian Children. Australian Institute of Health and Welfare; Health Monitoring Series No. 3. Canberra: AGPS, 1995.
19 Mathers C. Health Differentials Among Young Australian Adults. Australian Institute of Health and Welfare; Health Monitoring Series No. 4. Canberra: AGPS, 1996.
20 Australian Bureau of Statistics. Socio-economic Indexes for Areas. Cat. No. 1356.0. Canberra: AusInfo, 1990.
21 Australian Bureau of Statistics. 1996 Census of Population and Housing: Socioeconomic Indexes for Areas. Cat. No. 2039.0. Canberra: AusInfo, 1998.
22 Leclerc A, Lert F, Fabien C. Differential mortality: some comparisons between England and Wales, Finland and France, based on inequality measures. Int J Epidemiol 1990;19:100110.[Abstract]
23 Carr-Hill R. The measurement of inequities in health: lessons from the British experience. Soc Sci Med 1990;31:393404.[ISI][Medline]
24
Kennedy BP, Kawachi I, Prothrow-Stith D. Income distribution and mortality: cross sectional ecological study of the Robin Hood index in the United States. Br Med J 1996;312:100407.
25 Kawachi I, Kennedy BP. The relationship of income inequality to mortality: does the choice of indicator matter? Soc Sci Med 1997; 45:112127.[ISI][Medline]
26 Creedy J. Measuring income inequality. The Australian Economic Review 1996;2nd Quarter:23646.
27 Wagstaff A, Paci P, Doorslaer EV. On the measurement of inequalities in health. Soc Sci Med 1991;33:54557.[ISI][Medline]
28 Kunst A. Cross-national Comparisons of Socioeconomic Differences in Mortality. Thesis. Rotterdam: Erasmus University, 1997.
29 Palisade. @RISK: Advanced Risk Analysis for Spreadsheets. New York: Palisade, 1996.
30 Lee SH, Smith L, d'Espaignet E, Thomson N. Health Differentials Among Working Age Australians. Australian Institute of Health and Welfare. Canberra: AGPS, 1987.
31 Adler NE, Boyce T, Chesney MA et al. Socioeconomic status and health: the challenge of the gradient. Am Psychol 1994;49: 1524.[ISI][Medline]
32 Dudley MJ, Kelk NJ, Florio TM, Howard JP, Waters BG. Suicide among young Australians, 19641993: an interstate comparison of metropolitan and rural trends. Med J Aust 1998;169:7780.[ISI][Medline]
33 Patton GC, Carlin JB, Coffey C, Hibbert M, Bowes G. The course of early smoking: a population-based cohort study over three years. Addiction 1998;93:125160.[ISI][Medline]
34 Hill DJ, White VM, Scollo MM. Smoking behaviours of Australian adults in 1995: trends and concerns. Med J Aust 1998;168:20913.[ISI][Medline]
35
Phillimore P, Beattie A, Townsend P. Widening inequality of health in northern England, 198191. Br Med J 1994;308:112528.
36 Drever F, Bunting J. Patterns and trends in male mortality. In: Drever F, Whitehead M (eds). Health Inequalities: Decennial Supplement. London: The Stationery Office, 1997, pp.95107.
37
Jozan P, Forster DP. Social inequalities and health: ecological study of mortality in Budapest, 19803 and 19903. Br Med J 1999;318: 91415.
38 Dahl E, Kjaersgaard P. Trends in socioeconomic mortality differentials in post-war Norway: evidence and interpretations. Sociol Health Illness 1993;15:587611.[ISI]
39
Whitehead M, Drever F. Narrowing social inequalities in health? Analysis of trends in mortality among babies of lone mothers. Br Med J 1999;318:90812.
40 Wilkinson RG. Unhealthy Societies: The Afflictions of Inequality. London: Routledge, 1996.
41 Woodward A, Kawachi I. Why Should We Reduce Health Inequalities? Reasons for Acting on the Social, Cultural and Economic Factors that Cause Ill-health. National Health Committee: Health Determinants Programme: Background Paper 2. Wellington, New Zealand, 1998.