Social Gradients in the Health of Indigenous Australians

Abstract

The pattern of association between socioeconomic factors and health outcomes has primarily depicted better health for those who are higher in the social hierarchy. Although this is a ubiquitous finding in the health literature, little is known about the interplay between these factors among indigenous populations. We begin to bridge this knowledge gap by assessing evidence on social gradients in indigenous health in Australia. We reveal a less universal and less consistent socioeconomic status patterning in health among Indigenous Australians, and discuss the plausibility of unique historical circumstances and social and cultural characteristics in explaining these patterns. A more robust evidence base in this field is fundamental to processes that aim to reduce the pervasive disparities between indigenous and nonindigenous population health.

It is an almost universal truth that indigenous peoples of the world have poorer health than their nonindigenous counterparts.^1,2 Although a lack of high-quality data limits an accurate assessment of the health disparities between indigenous and nonindigenous populations in many countries,³ the disparities in Australia, for example, are well documented and striking.^4,5 Life expectancy for Australian Aboriginal peoples is between 11 and 14 years lower than that for non-Aboriginal people,⁶ a signal that indigenous health problems in Australia are pervasive and potentially worse than those of indigenous populations in other developed countries.^7–10

A recent study highlights that socioeconomic variables (such as weekly cash income, source of cash income, and completed years of schooling) explain between one third and one half of the gap in self-assessed health status between Australian Aboriginal and non-Aboriginal people.¹¹ Although socioeconomic factors assume some significance in explaining these health disparities, they do not necessarily account for health differences within indigenous population groups.

The relationship between social factors and health has been discussed and acknowledged for centuries.¹² There is now a robust international literature that supports the notion that health inequities are the result of factors and processes that fall outside of the conventional domains of health. They are heavily influenced by the structures of society and the social conditions in which people grow, live, work, and age—or what are now popularly known as the social determinants of health.¹³

The pattern of association between social class (or status) and health is typically characterized by poorer health for those at lower levels of the social hierarchy^14,15—that is, health outcomes follow a social gradient. Importantly, social gradients reflect more than differences between the high and low ends of the distribution—at any point along this continuum, people will tend to have poorer health than those above them. This observation is not limited to a subset of measures, but extends to most measurable socioeconomic constructs (such as poverty, employment, occupational status, education, housing, and income)¹³ and across a range of health outcomes (including most aspects of physical and mental health).^14,16

Despite the ubiquity of these observations, providing an explanation for the social gradient has proven to be a challenge.¹⁷ Researchers continue to shed light on the pathways to disease and poor health and how these can differ between population groups. In particular, there is growing understanding of how psychosocial factors and the social environment (in addition to poor material conditions and health-related behaviors) can affect physical and mental health and resultant longevity.^13,18

Krieger outlines 3 causal frameworks that underpin the relationship between social inequalities and health outcomes each with a different emphasis on social and biological factors.¹⁹ Psychosocial theories focus primarily on factors in the social environment that influence susceptibility to disease and illness; they point to stress as the link between lower perceived social standing and behaviors and choices that pose risks to health.²⁰ Theories of the social production of disease place greater emphasis on economic and political determinants in which the most important influences on health tend to be more distal factors that shape material well-being and principally have an indirect effect on health outcomes. Ecosocial theories and frameworks attempt to integrate theories of the social production of disease with biological explanations of disease by considering the dynamic interrelationship among social, biological, and ecological attributes and their joint and cumulative impact on health.¹⁹ Although social gradients are clearly implicated in these theories and frameworks, no single theory accounts for the graded relationship between socioeconomic status (SES) and health.²¹

SOCIAL GRADIENT IN INDIGENOUS HEALTH AND WHY IT IS IMPORTANT

Amid the theoretical frameworks and emerging evidence, there is uncertainty whether the social gradients observed in the general population hold true for indigenous populations.^22–25 Indigenous status is typically used as a covariate to explain differences in population health by SES, and scant attention has been paid to the potential moderating effect of indigenous status on the SES–health relationship. Moreover, there are inherent difficulties in comparing indigenous outcomes across SES levels. Key among these is the overrepresentation of indigenous peoples in the lower levels of all constructs of SES, which reduces statistical power for comparing outcomes across SES levels and potentially obscures the nature of the SES–health relationship.

Nevertheless, there is a theoretical basis for expecting that the association of traditional SES indicators with health will be different in an indigenous context. First, exclusion and discrimination, which are implicated in the production of relatively flat gradients among African American populations in the United States,²⁶ are often entrenched in the lives of indigenous peoples³ and may limit the health benefits that normally accrue from improved SES. Second, profound marginalization, which many indigenous cultures have faced over generations, can constrain human development, placing children at a disadvantage from the earliest stages of life and limiting the acquisition of skills that can be drawn upon for the benefits of health at every level of SES. Third, there may be social factors other than SES that exert a greater influence on indigenous health, including the well-being of the community and kinship network, cultural continuity, and connection to traditional lands that enables indigenous people to maintain spirituality central to the indigenous notion of health.^2,27,28

Knowledge of possible differences in the relationship between SES and health in indigenous populations has clear ramifications for both research and policy. For research, this knowledge will help broaden the scope of the field of social gradients in health with the recognition of social factors that may play a critical role in indigenous health but fall outside the traditional domains of social determinants of health. For policy, this knowledge can lead to more effective government decisionmaking. In Australia, for example, both federal and state governments have committed to closing the gap in key health and social indicators between mainstream and Aboriginal populations within a generation.²⁹ Although a worthy aspiration, this commitment is in part predicated on the assumption that the relationship between policies governing education, employment, and income transfers on the one hand, and health outcomes on the other, operate similarly in the Aboriginal and non-Aboriginal populations. If, in reality, there is a weak association between education and health among Aboriginal populations, then government investment in education, although generally beneficial, is unlikely to result in a significant improvement in Aboriginal population health or a substantial reduction in health disparities between Aboriginal and non-Aboriginal peoples. The implications in this scenario are that marginalized Aboriginal populations are likely to get trapped in poor health and that the policy expectation is unachievable unless efforts are devoted to addressing other, more salient, drivers of ill health.³⁰

DEFINITION OF INDIGENOUS STATUS

In the context of this review, it is important to recognize that there is no globally accepted definition of what constitutes an “indigenous” population. Nor is there agreement on whether a definition is even needed. The current view of the United Nations is that “a single definition will inevitably be either over- or under-inclusive, making sense in some societies but not in others.”^31(p6–7) Self-identification is therefore seen as a more relevant means of determining the indigenous status of an individual.^32,33 However, despite the ongoing debate, there is general agreement on the core aspects of the concept of “indigenous.” Most agree that indigenous communities and peoples are those that

1. demonstrate historical continuity (and have occupied land) before colonization or invasion,

2. consider themselves distinct from the societies that now prevail on ancestral land,

3. have a distinct culture and language,

4. tend to form nondominant parts of society and have a unique geographic dispersion, and

5. preserve and maintain their ancestral land and culture.^3,33,34

This description applies to hundreds of separate cultures, incorporating approximately 370 million people across 90 countries.³¹

We sought to assess the evidence for the direction and strength of social gradients in indigenous health and to comment on their potential implications for onward research and policy. We focused on Australian Aboriginal and Torres Strait Islander populations, which provide the most robust evidence base for the examination of this topic.

SEARCH STRATEGY AND SELECTION CRITERIA

We used a variety of information sources, including major citation databases and relevant Web sites (data providers, academic institutions, and reference sources). Searches were limited to articles published no later than April 2010 and were conducted with no language restrictions.

ISI Web of Science and OVID platform databases (MEDLINE, EMBASE, Global Health, and PsycINFO) were the prime sources of academic literature. Generic keywords for the target population group included “Indigenous,” “Aboriginal,” “Aborigines,” and “Torres Strait Islander.” Population keywords were linked with a combination of subject matter terms, such as “gradient,” “social gradient,” “health inequality,” “socioeconomic,” “socio-economic,” “determinant,” “social status,” “social class,” and “health.”

We accessed the substantial body of gray literature on indigenous health issues via the Indigenous Australian HealthInfoNet, a range of index databases on Informit and other relevant Web sites. Consultation with experienced indigenous health researchers netted a number of other relevant published reports and unpublished work.

We included studies in the review if they (1) featured an examination of the relationship between at least 1 socioeconomic factor (preferably with at least 3 categories) and a health outcome, health risk factor, or health care action (i.e., seeking or accessing health care); (2) included some quantitative assessment of this relationship; and (3) described this relationship within an Indigenous Australian population group.

We focused on the nature of the association between health and SES and considered the direction of the association, the statistical significance of the original study findings, and, to a lesser degree, effect size. We present results as reported in the original study. The wide range of health and socioeconomic variables used in eligible studies precluded use of formal meta-analytic techniques. Instead, we provide a narrative synthesis of review findings, supplemented with an aggregate overview of effect estimates.

REVIEW FINDINGS

The electronic search of ISI Web of Science and OVID platform databases identified 774 articles. After screening titles and abstracts, we identified 61 articles as potentially relevant; 9 satisfied the criteria for inclusion in this review, 3 of which duplicated the findings of another study. We included another 9 articles after an electronic search of sources of indigenous research and gray literature. One article was sourced from the library of the author (S. R. Z.).³⁵ After we removed duplicates,^36–38 a total of 16 studies, reports, and books satisfied the criteria for inclusion in this review. Data on each study's design, sample, measurement of SES and health, and results are summarized in Table 1.

TABLE 1—

Summary of Results From Studies Examining the Relationship Between Health and Socioeconomic Status in Australian Aboriginal Populations

References	Study Type	Year(s) of Study	Indigenous Sample Sizea	Scope (Age, Years)	Outcome Variables	SES Variables	Pattern of Association Between SES and Healthb
Cass et al.39	Ecological	1993–1998	36 areas	All ages	End-stage renal disease (registry notifications)	Area-based measures (unemployment; household income; overcrowding)	Positive gradientc
Cunningham et al.40	Cross-sectional	2003–2005	777	15–64	Diabetes (oral glucose tolerance test)	Housing tenure; household income; employment; education	Positive gradient; inconclusive with 1 SES measured
Cunningham et al.41	Cross-sectional (secondary analysis)	1994	8782 (nationally representative)	≥ 15	General health (self-rated)	Labor force status; home ownership; education; household income	Positive gradient; inconclusive with 1 SES measured
Glover et al.42	Retrospective cohort	1997–1999	4378 (total population data in 4 states/territories)	All deaths	Mortality (registrations)	Area-based measure (relative disadvantage)	Positive gradientc
Gray and Boughton43	Cross-sectional (secondary analysis)	1994	15 700 (nationally representative)	0–14	Health care actions (self-report)	Parental education	U-shaped
Gray et al.44	Cross-sectional (secondary analysis)	1995	1536 (nationally representative—nonurban areas)	All ages	General health (self-rated)	Household income	Inconclusive
Oddy et al.45	Cross-sectional (secondary analysis)	2000–2002	5289 (representative of WA)	0–17	Chest, ear, and gastrointestinal infections and hospitalizations (self-report and administrative data)	Area-based measure (relative disadvantage)	Mixed findings (chest and gastrointestinal infections); inconclusive (ear infections)d
Australian Bureau of Statistics46	Cross-sectional	2002	9400 (nationally representative)	≥ 15	General health; disability; smoking; alcohol consumption (self-report)	Education; labor force status	Positive gradient (general health, disability, smoking); reverse gradient (alcohol consumption)c; some U-shaped associations with education
Australian Bureau of Statistics47	Cross-sectional	2004–2005	10 439 (nationally representative)	≥ 15	General health; long-term conditions; health care actions; obesity; smoking, alcohol consumption (self-report)	Labor force status	Positive gradient (general health, some health care actions, smoking, and a range of long-term conditions); reverse gradient (alcohol consumption); no association (asthma, cancer, obesity, some health care actions)c; inconclusive (back problems, ear or hearing problems)
Thomas et al.48	Cross-sectional (secondary analysis)	2002	9400 (nationally representative)	≥ 15	Smoking (self-report)	Household income, education, labor force status, financial stress, housing tenure	Positive gradientd
Hetzel et al.49	Various sources (secondary analysis)	1997–2001	Variouse	Variousf	Life expectancy (derived measure using death registrations); smoking in pregnancy (self-report)	Area-based measure (relative disadvantage)	Positive gradient (life expectancy); inconclusive (smoking in pregnancy)c
Jamieson et al.50	Cross-sectional (secondary analysis)	2002–2003	4414 (representative of NT)	4–13	Oral health (examination)	Area-based measure (relative disadvantage)	Positive gradientc
Hunter51	Cross-sectional (secondary analysis)	1994	3433 households (nationally representative)	≥ 15	Long-term health problems (self-report)	Household income	No association
Hunter52	Cross-sectional (secondary analysis)	1994	3433 households (nationally representative)	≥ 15	Long-term health problems (self-report)	Labor force status	Inconclusive
Titmuss et al.53	Retrospective cohort	2002	1706	All births	Birth weight (measured)	Area-based measure (relative disadvantage)	Inconclusive
Zubrick et al.35	Cross-sectional	2000–2002	5289 (representative of WA)	0–17	Mental health (validated measure)	Parental education; financial strain	Inconclusive

Note. NT = Northern Territory of Australia; SES = socioeconomic status; WA = state of Western Australia.

^aNumbers refer to persons unless otherwise stated.

^bA positive gradient is defined as better health for those with higher SES (i.e., a positive association); a reverse gradient is defined as better health for those with lower SES (negative association); inconclusive is defined as a trend or effect that was not statistically significant or a study with too many methodological limitations to support definitive conclusions.

^cDefined as studies that established a positive association between SES and health in simple bivariate or cross-tabulation analyses, without controlling for other factors or confounders.

^dDefined as studies that accounted for at least some other confounding (usually demographic) variables in establishing a positive association between SES and health.

^eStudy estimates are generally based on data from government administrative sources and are representative of relevant populations in the state of South Australia.

^fLife expectancy estimates are based on mortality records for all ages; data for smoking during pregnancy is generally limited to women aged 15 years and older.

Population Groups and Study Designs

Most studies had at least 1 methodological limitation relating either to study design, scope, sample size, or analytic techniques. The majority of the study samples consisted of more than 1000 indigenous people (14 studies, or 88%) and, typically, were representative of populations of Aboriginal and Torres Strait Islander peoples at a national (10 studies) or state or territory (3 studies) level.

Seven studies examined outcomes for all adult age groups and 5 focused on children. Four studies examined outcomes for all age groups. Differences in target populations may be a source of heterogeneous results, as the literature suggests that the strength and shape of social gradients differ by age group.¹⁵

All studies used a data source with an observational design and most (13 studies) were cross-sectional, with 2 cohort studies (both retrospective) and 1 ecological study. The ecological study in this review was limited to assessing the associations between SES and health at an aggregate geographic level. Many of the cross-sectional data sources used in review studies have considerable breadth, and although their designs are unable to discount reverse causation, they enable adjustment for covariates in the analysis of SES–health relationships. Five cross-sectional studies adjusted for the effects of demographic (e.g., age and gender) and other known covariates (e.g., health service access and health history),^{35,40,41,43,45} whereas 2 adjusted for demographic variables only.^44,48 Only 1 of the 2 cohort studies accounted for covariates.⁵³ In the context of this review, it is important to note that overall only 2 studies assessed the impact of cultural factors in mediating the relationship between SES and health.^35,41 Hypothetically, multivariate analyses of variables that measure intrinsic characteristics of an indigenous culture can help to determine whether an observed social gradient is attributable to that culture or explained by more generic forces. By contrast, 9 studies (56%) exclusively examined bivariate relationships between SES and health variables using simple cross-tabulation or correlation techniques; studies from the gray literature were more likely to solely use these techniques (70%) than those sourced from the academic literature (33%).

Health Outcomes

Because a number of studies reported multiple outcomes, the 16 in-scope studies provided findings on 60 separate associations between SES and health. Most of these associations (42, or 70%) examined a health outcome, with 13 (22%) focused on a health risk factor and 5 (8%) on a health care action measure. Health outcomes were predominantly an aspect of physical health (40 associations) as opposed to mental health (2 associations).

Many (62%) of the health outcome measures were derived from self-reports that included measures of general health, disability and long-term illness, respiratory problems, gastrointestinal infections, arthritis, diabetes, kidney disease, cancer, back pain, hearing and sight limitations, mental health, and problems with heart and circulatory system. Many of the health measures were simple indicators of the presence or absence of a disease or an event and did not include any information on severity, duration, or age of onset or occurrence.

Socioeconomic Status Measures

The studies identified by this review used a wide range of SES indicators, including those that measure the SES characteristics of individuals (7 studies), families and households (7 studies), and neighborhoods and communities (6 studies). Few studies examined multiple SES indicators simultaneously, and only 1 used a multilevel framework to adequately measure the effects of SES at various levels.³⁵ We noted that most (n = 10) in-scope studies used only indirect markers of material well-being (e.g., education and labor force status) and 2 focused solely on absolute measures (e.g., income and home ownership); 4 studies made use of both types of measures.

Evidence of Social Gradients in Indigenous Health

The majority of studies (10 of 16) reported a positive gradient in some aspect of health—that is, better health was associated with higher SES. Two studies highlighted a U-shaped relationship between education and health. Three studies found no relationship between health and SES, and 2 reported inconclusive evidence (e.g., a trend that was not statistically significant).

There were 33 separate associations that exhibited a statistically significant positive gradient. Most of the associations with general health (62%), health risk factors (62%), and indicators of mortality and morbidity (53%) displayed a positive gradient. Only 2 of the 5 effect estimates for health care actions (40%) were in a positive direction (Figure 1).

FIGURE 1—

Nature of the associations between SES and health in Australian Aboriginal populations, by domain of health indicator.

Note. SES = socioeconomic status. No association (a) indicates no relationship was found between SES and health, the trend was not statistically significant, or there were too many methodological limitations to support definitive conclusions; reverse gradient (b) indicates better health was associated with lower SES; and positive gradient (c) indicates better health was associated with higher SES.

The majority of estimates based on multivariate regression models exhibited a positive gradient (61%)—that is, a positive effect remained after control for at least 1 additional variable; adjusting for additional variables generally diluted the strength of the association between SES and health—or “flattened” the social gradient. By comparison, 51% of bivariate associations displayed a positive gradient. Objectively measured health variables more commonly revealed a positive relationship with SES (69%) than self-reported measures (52%).

There is evidence to suggest that social gradients in indigenous health exist at all 3 levels of SES: individuals, families or households, and neighborhoods or communities (Figure 2). When no association was found between SES and health, the SES indicator was more often an indirect marker of material well-being than an absolute measure.

FIGURE 2—

Nature of the associations between SES and health in Australian Aboriginal populations, by type of SES measure.

Note. SES = socioeconomic status. No association (a) indicates no relationship was found between SES and health, the trend was not statistically significant, or there were too many methodological limitations to support definitive conclusions; reverse gradient (b) indicates better health was associated with lower SES; and positive gradient (c) indicates better health was associated with higher SES.

SUMMARY OF ORIGINAL STUDY FINDINGS

In this section, we provide a narrative synthesis of review findings, using original study results.

General Health

Four studies examined general health status, with all using a self-rated measure.^41,44,46,47 Three studies found a positive relationship with SES for adults, which included measures of education, labor force status, and home ownership.^41,46,47 The evidence for household income was weaker. Gray et al. showed that indigenous persons in high-income families generally had slightly better health after adjusting for age, although the differences were not statistically significant.⁴⁴ The results of Cunningham et al. also highlighted a positive gradient with household income, although this association was attenuated after adjustment for demographic, socioeconomic, and cultural factors.⁴¹ In summary, although self-rated general health is consistently associated with education, labor force status, and home ownership, the evidence for household income is inconclusive.

Mortality

Two studies examined indicators of mortality, with both finding a positive association with area-based measures of SES.^42,49 One study examined administrative mortality data across 4 states and territories of Australia; it found that Aboriginal people living in the most disadvantaged areas had higher death rates than Aboriginal people living in the least disadvantaged areas: 1.52 times higher (P < .001) for males and 1.61 times higher (P < .001) for females.⁴² The other study focused on life expectancy and reported a positive association (although not a continuous gradient) in bivariate analyses.⁴⁹

Physical Morbidities and Birth Weight

Three studies used nationally representative samples to examine disability and long-term health conditions in adults, with mixed results.^46,47,52 There was no association with self-reported household income in a study by Hunter.⁵² One study found a positive gradient with labor force status⁴⁶; another suggested that labor force status was largely unrelated to the presence of disability or a long-term condition, although these findings are suggestive of variation by gender and geographic location.⁵² One study reported a generally positive gradient by education, although those with a nonschool qualification had an elevated likelihood of this outcome relative to those who had only completed their secondary schooling.⁴⁶

Two studies examined kidney disease, with both finding a positive association.^39,47 Registry notifications of end-stage renal disease were strongly correlated with household income (r = 0.71, P < .001), overcrowding (r = 0.84, P < .001), and a composite index of relative disadvantage (r = 0.88, P < .001) in an ecological study.³⁹ A national cross-sectional survey highlighted that the self-reported prevalence of this relatively uncommon outcome was higher among those not engaged in the labor force (4.1%; 95% confidence interval [CI] = 3.0%, 5.2%) than among employed adults (1.7%; 95% CI = 1.1%, 2.3%).⁴⁷

The 2 studies on ear and hearing problems were inconclusive.^45,47 One study found the prevalence of self-reported ear and hearing problems to be similar across labor force status categories in a simple cross-tabulation.⁴⁷ The other study showed that both the occurrence of recurring ear infections (parent reported) and hospital admissions for ear infections (objectively reported) had a weak positive association with area-based SES in multivariate analyses, although the effects were statistically insignificant.⁴⁵

The 2 studies that examined diabetes showed consistent evidence of a positive gradient.^40,47 Cunningham et al. demonstrated strong associations between an objective test of diabetes and self-reported measures of housing tenure, household income, employment status, and an area-based index of disadvantage among urban indigenous people in the city of Darwin, and a weaker, statistically insignificant, positive association with education.⁴⁰ The broader, national study of the self-reported prevalence of diabetes highlighted a positive gradient with labor force status.⁴⁷

Two studies examined respiratory infections and conditions with mixed results.^45,47 A large study of indigenous children in the state of Western Australia showed a positive, but not continuous, gradient between area-based SES and the prevalence of both parent-reported recurring chest infections and objectively reported hospital admissions for nonwheezing lower respiratory infections; no association was found with hospital admissions for either upper respiratory infections or wheezing lower respiratory infections.⁴⁵ The other study reported no association between asthma and labor force status in a nationally representative sample of indigenous adults.⁴⁷

One study examined oral health and highlighted a positive gradient among indigenous children in the Northern Territory of Australia.⁵⁰ The study analyzed the number of decayed, missing, and filled teeth in the deciduous and permanent dentition of children aged 4 to 13 years on the basis of dental examinations, with consistent patterns in the bivariate relationship with area-based SES.⁵⁰

The single study on gastrointestinal infections was inconclusive.⁴⁵ Although it showed a positive association with an area-based index of disadvantage, the pattern was neither continuous nor statistically significant.⁴⁵

The single study on birth weight was inconclusive.⁵³ This study used data collected at a large urban hospital and lacked sufficient power to compare values across the full spectrum of SES. Although it reported a lower mean birth weight for babies in families living in the most disadvantaged areas (for lowest quintile, mean = 3101 g; 95% CI = 2868 g, 3333 g) compared with all others (mean = 3413 g; 95% CI = 3254 g, 3572 g), the finding was not statistically significant.⁵³

The evidence for other physical morbidities was sourced from a single study.⁴⁷ This national study focused on the labor force status of indigenous adults and reported a positive gradient for self-reported arthritis, eye and sight problems, and heart and circulatory problems; inconclusive evidence for back pain and problems; and no association with the relatively rare outcome of cancer.⁴⁷

Mental Health

The single study on mental health was inconclusive.³⁵ There was no association between parental education and a child being at high risk of clinically significant emotional and behavioral difficulties. There was a positive gradient for family financial strain that was mostly explained by demographic factors, the physical and mental health status of the primary caregiver, and exposure to people experiencing problems with alcohol.³⁵

Health Risk Factors

Four studies assessed the relationship between SES and smoking,^46–49 3 of which showed a consistent positive gradient with self-reported smoking status.^46–48 Two of these studies applied different analytic techniques to the same nationally representative data source and revealed that unemployed persons and those with less education were the most likely be smokers.^46,48 One study showed a general trend for higher rates of smoking during pregnancy among those living in more disadvantaged areas, although the statistical significance of the effect was not reported.⁴⁹

Both of the studies that examined alcohol consumption found the lowest prevalence of risky alcohol consumption among adults who were not engaged in the labor force, suggesting a reverse association between this behavior and SES.^46,47 One study calculated body mass index based on self-reported height and weight and found that the prevalence of overweight or obesity was unrelated to labor force status in a simple cross-tabulation.⁴⁷

Health Care Actions

Two studies examined health care actions with mixed results.^43,47 One study found a U-shaped relationship between (1) any of 8 health care actions taken for children and (2) the education of the mother after controlling for demographic factors, health status indicators, and objective measures of health service access. In this study, health care actions were highest among indigenous children whose mothers had less than 14 years (odds ratio [OR] = 1.55; P < .05; reference category = 14 years) and 17 or more years (OR = 1.40, P < .05) of formal education.⁴³ The other study assessed 4 separate actions among indigenous adults in bivariate analyses and found that persons in the labor force were more likely than others to have been admitted to a hospital and to have visited a general practitioner or specialist; there was no association between visiting a casualty–outpatient service or a dentist and this measure of SES.⁴⁷

The Impact of Cultural Factors

Only 2 studies assessed the joint impact of cultural factors and SES on health,^35,41 and they reported contrasting results. Cunningham et al. broadly showed that cultural factors affected the general health of indigenous peoples in Australia above and beyond the effects of SES. Identifying with a clan, tribe, or language group appeared to be protective of health for men, whereas recognizing an area of land as traditional country or a homeland (among men only) and being taken away from the family as a child (among women only) were associated with worse health.⁴¹ Zubrick et al. highlighted that the mental health of indigenous children in the state of Western Australia was not independently associated with either SES, the language spoken by the primary caregiver, or children's participation in cultural activities.³⁵

EVALUATION

Our review of the empirical evidence on the relationship between SES and health in Australian Aboriginal populations leads to 3 primary conclusions. First, there is a dearth of research to date that has specifically focused on this topic and, on the basis of the limited research and varied findings across available studies, we are unable to make strong assertions about the nature and strength of the SES–health relationship. The mixed findings partly reflect the wide array of health and SES measures and a diversity of indigenous population groups and analytic techniques within a small number of eligible studies. Second, there is, however, consistent evidence supporting a positive social gradient in mortality, kidney disease, diabetes, and smoking status. This effect was also shown in single studies on arthritis, eye and sight problems, oral health, and heart and circulatory problems. Although general health status tended to exhibit a positive social gradient, the effects were not always statistically significant. Third, there are a number of methodological issues that make it difficult to interpret the study results and assess differences between them. There is also the potential that weak gradient effects merely reflect low variability in the distributions of SES and health measures in indigenous populations. Overall, the review findings call for continued efforts to improve the quantity and quality of research to provide more insights into the gradient effect (or absence of it) among indigenous population groups. The discussion that follows provides more detail on the limitations of review studies, the implications of the findings for policy, and directions for future research.

Our understanding of whether and to what extent the social gradient in health exists in Aboriginal Australia is primarily hampered by a scarcity of research. Although there is a need to improve the quantity of data that can be used by researchers to adequately examine this topic, existing data sets have been underused and should be investigated in more detail.

Study Limitations and Measurement Challenges

Low variability in the distribution of SES and health is a pertinent limitation. Indigenous peoples are vastly overrepresented in the lower levels of all constructs of SES used in review studies. For example, Oddy et al. reported that almost two thirds of indigenous children in Western Australia lived in the lowest quartile of disadvantaged areas.⁴⁵ Notwithstanding this, some measures of household income and poverty may have underestimated the extent of indigenous marginalization by not properly accounting for the size and structure of indigenous families and households or the nature of sharing of economic resources between extended family members.⁵⁴ The skew in the distribution of SES measures in indigenous populations can reduce statistical power for comparing outcomes across SES levels, particularly if conventional groupings are used (e.g., quintiles), and potentially obscure the nature of the SES–health relationship. Although this skewing is generally applicable to our review, a number of the studies we reviewed were based on population-level data with sufficient power to potentially detect an effect across the full spectrum of SES categories, despite an uneven distribution.

A deeper understanding of the SES–health relationship can be attained by the simultaneous use of SES variables at individual, household, family, and community levels.^16,55,56 The importance that indigenous peoples place on social connections with family and community, relative to the needs of individuals, suggests that a multilevel analysis that includes SES indicators at a contextual or community level may shed some light on the nature of social gradients in indigenous health.

Our findings highlighted that there was often no definitive evidence of an association with self-reported, or subjective, measures of health. Self-reported measures of morbidity have been criticized in the past as being misleading, particularly among socially disadvantaged people who may underreport or understate poor health outcomes.⁵⁷ This can give rise to a flatter health gradient compared with results that rely on objective assessments of health.⁵⁸ Future studies in this field will benefit from collecting information on both objective and self-reported health measures and comparing the patterns of their association with SES.

There are potentially many pathways through which SES influences health. Most of the studies examined in this review (and research more generally into indigenous–nonindigenous health inequalities) have not examined the range of psychosocial and environmental factors that define these pathways, or the factors that characterize indigenous cultures.³⁹ The results of the 2 review studies that incorporated cultural factors in their analytic framework suggest that they can influence, although not invariably, the relationship between SES and health. Strategies for testing social gradient effects need to consider the conceptual basis on which mediating variables are included in multivariate analyses. Researchers should report the effect of SES on health with and without mediating factors, so that the total, direct, and indirect effects of SES on a health outcome can be estimated.⁵⁸

Health and Social Determinants in Indigenous Contexts

In addition to the range of methodological limitations in review studies, there are substantive social, cultural, and historical factors that may contribute to the mixed findings. Aboriginal Australia is not a homogenous group; as Bell states, “Aboriginal Australia is a network of interconnected Aboriginal nations, with their own languages and ways of life.”^59(p4) Health determinants may therefore differ by region or along cultural lines,² and these differences may predict variation in health outcomes within each group. Health determinants also differ between Aboriginal and non-Aboriginal populations of Australia,^25,60 which in part reflects 2 very different concepts of health. Australian Aboriginal peoples, like indigenous populations in many other countries, have a holistic view of health that goes beyond individual physical and mental well-being to include aspects of spirituality, connection to land, and the social, emotional, and cultural well-being of the community.^61–63 Australian Aboriginal peoples tend to ascribe their relatively poor health to broader, macrosocial factors.⁶⁴ Issues of dispossession and exclusion are key among these, and they extend to traditional land, kinship, language, and culture.^65,66 Racism is a common thread to indigenous people's history of being excluded from many aspects of social, political, and economic life in Australian society and is being cited more commonly in the literature as having adverse consequences for health.⁶⁷ Human rights contraventions are enmeshed in the postcolonial experiences of Australian Aboriginal peoples. Evidence suggests that there is a vicious cycle between human rights and health, particularly for marginalized and minority populations.^68,69 The health of Australian Aboriginals therefore may have been affected over time—directly by human rights abuses or indirectly by the systematic inequalities that they give rise to. Many of the issues discussed here—dispossession, exclusion, discrimination, marginalization, and inequality—are implicated in the unique stress profile of indigenous populations in Australia. It has been shown that chronic stress is a feature of the lives of Aboriginal people from all social classes,⁷⁰ and this may dampen the benefits that higher SES normally generates for health.

The validity of using standard SES measures in indigenous contexts has been questioned and is also central to the analysis of social inequalities in indigenous health.^71,72 Income, education, and employment can be decidedly different constructs among indigenous peoples (as are notions of health), and this reflects the different social contexts of indigenous and nonindigenous populations.⁷³ Social status in more traditional communities may be more a function of knowledge than of material resources, or it may reflect control over resources more than ownership of them.⁷² This underscores the need to reconceptualize existing notions of SES to gain a better understanding of the complexities of their relationship with indigenous health.⁶⁰ Critically, this rethink needs to be fully informed by indigenous peoples’ views on the concept of health and its determinants. These views may be varied but will most likely represent a complete paradigm shift from existing SES constructs.⁷¹

The health determinants of Australian Aboriginal peoples need to be considered in light of their unique population distribution. Although most Aboriginal people live in urban settings, they are also far more likely than are non-Aboriginal Australians to live in remote and isolated areas. Many of the factors that affect population health are unevenly distributed across areas of geographic remoteness. For example, there tend to be fewer health care services in more remote areas and a more limited range of job choices. Location is therefore likely to be an important factor that accounts for variations in health within the indigenous population.

Limitations to This Review

There are a number of limitations to this review. A focus on the published literature may have introduced publication bias that could potentially overstate the evidence supporting an association between SES and health. Our results point to a higher proportion of statistically significant associations in the academic literature (68%) than in the gray literature (47%). We did not compare social gradients between indigenous and nonindigenous populations, and this restricts the policy recommendations that can be drawn from specific review study findings.

In addition, this review focused only on studies of Indigenous Australian health, and the results may not be generalizable to indigenous populations in other countries. There is certainly evidence of positive health gradients among a number of other indigenous cultures, including Native Americans and Alaska Natives,^74–78 New Zealand Maori,^79–85 Canadian First Nations and Metis,^86–91 Inuit,^92–96 South American indigenous groups (Andean culture and Amazon Basin tribes,⁹⁷ Mapuche,⁹⁸ and Tsimane’⁹⁹), Taiwanese Aboriginals,¹⁰⁰ and Indian Adivasis.¹⁰¹ There is also, as we have found in Australia, variation in the available evidence among these indigenous populations; the US literature, for example, also features inconclusive evidence for measures of birth weight,¹⁰² mental health,¹⁰³ general health status,^77,103 health risk behaviors,^76,104 and health care actions.¹⁰⁵

Conclusions

Despite abundant evidence that SES is a critical determinant of health, there is a paucity of research that examines the relationship between SES and health among indigenous peoples. This review begins to bridge this knowledge gap by assessing evidence from the limited existing research on social gradients in indigenous health in Australia. The review reveals that, in contrast to the ubiquitous, strong associations between SES and health in the general population, there is a less universal and less consistent SES patterning in Indigenous Australian health. Notwithstanding some measurement issues in the existing studies, which may in part explain the varied findings, we believe the unique historical circumstances, social and cultural characteristics, and profound and persistent marginalization of indigenous populations in Australia are plausible explanations for a much less consistent social gradient in indigenous health. There is a critical need for future research to take into consideration these unique circumstances of indigenous populations in conceptualizing and operationalizing health and its social determinants. Future research will also need to identify and measure a range of plausible mediating factors that may help explain the social gradient or its absence. These factors include the high levels of stress that indigenous peoples are typically exposed to in daily life, loss of cultural continuity, racism, and geographic dispersion.

There is considerable potential for research on this topic to inform the development of policy and interventions that will improve the health status of indigenous peoples. Despite the mixed available evidence, our findings suggest that SES exerts an influence on a range of health outcomes and risk-taking behaviors, including mortality, kidney disease, diabetes, and smoking status. The implication here is that these aspects of health are likely to be sensitive to investments in SES. However, the extent to which such investments would translate into significant health benefits in indigenous populations will depend on the strength of the SES–health association. More rigorous research is required to assess the strength (effect size) of SES–health relationships in indigenous contexts, particularly in comparison with nonindigenous populations.

Further research is needed to provide greater insights into the gradient effect (or its absence) among indigenous population groups, with comparisons between indigenous and nonindigenous populations and between countries and regions within countries. In addition, more advanced designs (including longitudinal studies) and analytic techniques (including multilevel modeling) and alternative methods (qualitative and ethnographic studies) will enhance our understanding of the relationship between SES and indigenous health. A more robust evidence base may enable meta-analyses to be conducted on specific health factors in the future that will support the interpretation of research findings and provide more specific guidance to the application of policy interventions.