Social Psychiatry and Psychiatric Epidemiology

, Volume 47, Issue 2, pp 165–173

The DRUID study: exploring mediating pathways between racism and depressive symptoms among Indigenous Australians

Authors

    • Onemda Unit and McCaughey Centre, Melbourne School of Population HealthUniversity of Melbourne
  • Joan Cunningham
    • Menzies School of Health Research, Institute of Advanced StudiesCharles Darwin University
    • Onemda Unit and McCaughey Centre, Melbourne School of Population HealthUniversity of Melbourne
Original Paper

DOI: 10.1007/s00127-010-0332-x

Cite this article as:
Paradies, Y.C. & Cunningham, J. Soc Psychiatry Psychiatr Epidemiol (2012) 47: 165. doi:10.1007/s00127-010-0332-x

Abstract

Purpose

Racism is an important determinant of mental and physical health for minority populations. However, to date little is known about the relationship between racism and ill-health outside of the U.S. or the causal pathways between racism and poor health. This paper focuses on the relationship between racism and depression in a non-U.S. indigenous population, including examination of novel mediators and moderators.

Methods

One hundred and eighty-five adults in the Darwin Region Urban Indigenous Diabetes study responded to a validated instrument assessing multiple facets of racism. Depressive symptoms were assessed using the Centre for Epidemiologic Studies Depression Scale. Stress, optimism, lack of control, social connections, cultural identity and reactions/responses to interpersonal racism were considered as possible mediators and moderators in linear regression models.

Results

Interpersonal racism was significantly associated with depression after adjusting for socio-demographic factors (β = 0.08, p < 0.001). Lack of control, stress, negative social connections and feeling ashamed, amused or powerless as reactions to racism were each identified as significant mediators of the relationship between racism and depressive symptoms. All examined mediators together accounted for 66% of the association between interpersonal racism and depressive symptoms.

Conclusions

This study demonstrates that racism is associated with depressive symptoms in an indigenous population. The mediating factors between racism and depressive symptoms identified in this study suggest new approaches to ameliorating the detrimental effects of racism on health.

Keywords

IndigenousRacismMental healthDepressionAustralia

Introduction

Racism can be defined as avoidable and unfair phenomena that result in inequality of resources, opportunity or benefit among racial/ethnic groups. Racism can be expressed through stereotypes, prejudice or discrimination and can occur at the internalised (i.e. within an individual’s worldview), interpersonal (i.e. between individuals) and/or systemic level (i.e. policies and practices of institutions and organisations) [1]. In recent years, the study of racism and health has emerged as an area of public health research. To the end of 2007, reviews identified over 250 publications examining racism as a determinant of health and/or health behaviours, revealing strong associations between self-reported racism and ill-health among minority groups in developed countries [24]. These associations remained after adjusting for a range of confounders and occur in longitudinal as well as cross-sectional studies, suggesting that racism precedes ill-health rather than vice versa. The most consistent finding is the association between racism and mental ill-health, particularly psychological distress, depression and anxiety [24].

Depression is now considered a leading cause of morbidity worldwide [5]. Exposure to racism as a form of stress can give rise to factors that contribute to depressive symptoms (hereafter referred to as depression). These factors include negative emotional states, poor self-esteem, low self-efficacy, and reduced self-control as well as pessimism, aggression, hyper-vigilance, and rumination [24]. Racism has been linked to cortisol and HPA dysregulation [68], sleep disturbance [9], and smoking [1014] as well as alcohol and illicit drugs [1418]. These factors are, in turn, associated with depression [4, 1921].

To date, racism and health literature has focused on the U.S. and specifically African Americans [2, 3]. Currently, little is known about the causal pathways leading from racism to ill-health [22]. This paper adds to existing scholarship by examining stress, optimism, social connections, lack of control cultural identity and reactions/responses to interpersonal racism as mediators of the relationship between racism and depression among Indigenous Australians.

Indigenous Australians constitute 2.4% of the Australian population and suffer high rates of unemployment and incarceration, low income, sub-standard housing and a high burden of ill-health and mortality, including a life expectancy 9–12 years less than other Australians [23]. The poor health of Indigenous Australians has been associated with experiences of racism in several previous studies [24].

Methods

Data for this study were drawn from the DRUID study (Diabetes and Related conditions in Urban Indigenous people in the Darwin region). Eligible participants were Indigenous people aged 15 years and over who lived Darwin (capital of the Northern Territory (NT) of Australia) [25]. DRUID was a cross-sectional study in which participants were recruited through community and family networks as well as local health centre/departments [25]. As such, the study sample constitutes a non-representative sample of the eligible population. The study utilised questionnaires to assess health status, socio-demographics as well as psychosocial and behavioural factors. The study was approved by the joint Menzies School of Health Research and Northern Territory Department of Health and Community Services Human Research Ethics Committee.

Measures

The Measure of Indigenous Racism Experiences (MIRE) was used to assess interpersonal racism [26]. Respondents were asked how often they were treated unfairly because they are Indigenous across nine mutually exclusive items covering experiences in employment, domestic, educational/academic, recreational/leisure, law (enforcement), health care and public settings as well as government and non-government service provision [26]. Possible responses were: never, hardly ever, sometimes, often, or very often. Within a stress checklist, respondents were also asked if they or their family/friends had experienced discrimination (not specifically attributed to race/ethnicity) which affected them personally in the past 12 months (hereafter referred to as discrimination-related stress). Participants could respond that the event: was not experienced, had minimal/no effect, affected them somewhat, or affected them a great deal.

Age, sex, marital status and household composition were assessed along with level of education, gross weekly household equivalised income (using the modified Organisation for Economic Co-operation and Development scale) and housing tenure.

Five reactions and six responses to racism from the MIRE were also assessed as mediators. Reactions included: feeling ashamed, humiliated, anxious or fearful (‘ashamed’); feeling angry, annoyed or frustrated (‘angry’); feeling amused, contemptuous or sorry for the person who did it (‘amused’); feeling powerless, hopeless or depressed (‘powerless’); and getting a headache, an upset stomach, tensing of your muscles, or a pounding heart (‘somatic’). Responses included: ignoring, forgetting or accepting racism (‘ignoring); avoiding racism (‘avoiding’); changing self or actions to prevent racism (‘changing’); doing something about perpetrators or racist situations (‘acting’); talking to family/friends or expressing racist experiences (‘talking’); and keeping it to oneself (‘withdrawing’). These items could be answered as: never, hardly ever, sometimes, often, or very often [26].

Stress over the past 12 months was assessed using a scale constructed for the DRUID study. This included 10 items about acute stress (one-off, sudden or unexpected events including relationship breakup, death of partner/child, being fired, witness/target of violence, family member sent to jail, serious illness or accident/injury as well as other unspecified event) and 20 items about chronic stress (regular or continuing events including trouble at work, school or relationships; problems with alcohol/drug-related, gambling, money, mental/emotional, health, employment, police or access to land/country; abuse/violence, overcrowding; family in prison; pressure to fulfil cultural, family/carer responsibilities; worry about family; reliance on others for everyday needs as well as other unspecified event). Respondents were asked to include stressful events that occurred to their family/friends as well as those they experienced personally. These 30 items were assessed using a four-point response scale (event not experienced, had minimal/no effect, affected me somewhat, or affected me a great deal).

Optimism was assessed using the Life Orientation Test-Revised (LOT-R) [27]. Lack of control was assessed via four items (how much, in the past 4 weeks you felt a lack of control with life in general, finances, personal life and health) with responses: never/rarely, sometimes, often, or very often.

Positive, negative and instrumental social connections were assessed using scales adapted from the Strong Heart Study [28]. Positive social connections included six items assessing the extent to which friends or relatives cared about, understood, or appreciated the respondent as well as to what degree he/she could rely upon, talk to, and relax around friends/relatives (not much, some or a lot). Negative social connections included five items assessing the degree to which friends/relatives made too many demands, argued with, criticised, let down or got on the nerves of the respondent (rarely/never, sometimes or often). Instrumental social connections were assessed with five items asking if there was someone the respondent could: go with to community/social events; borrow money from in an emergency; lend them a car or drive them somewhere if needed; ask to provide bail; and count on to check in on the respondent regularly (no; yes, one person or a few people; or yes, lots of people).

As measures of cultural identity, participants were asked if they recognise a homeland/traditional country and whether they identify with a clan, tribe or language group.

Depression was assessed using the Centre for Epidemiologic Studies Depression Scale (CES-D) [29]. The CES-D is a 20-item self-report scale that assesses depression over the past week. Each item was scored from zero to three.

Analytical approach

Racism was included as the independent and depression as the dependent variable in hierarchical linear regression models. As the internal reliability of the racism items was good (α = 0.83), responses were given values from zero to four (never to very often) and summed across the nine items, giving a potential range 0–36. Depression was measured in this study as a continuous variable and was subject to a square-root transform to approximate normality. Residuals in relation to racism approximated normality following this transform. Although residuals in relation to discrimination-related stress were somewhat non-normal, this variable was retained in analysis as a contrast to the interpersonal racism measure.

Socio-demographic variables (age, sex, household composition, martial status, equivalised household income, housing tenure and level of education) were added in the first step with racism added in the second step. As the internal reliabilities of the acute stress items (α = 0.66), chronic stress items (α = 0.82) and total stress items combined (α = 0.85) were adequate to good, weighted summary scores were calculated. The four responses were given numerical values from zero to three (no experience to affected me a great deal) and summed across items giving a potential range 0–30, 0–60 and 0–90 for acute, chronic and total stress, respectively. Positive, negative and instrumental social connections were coded as a summary score of the items in each scale. The cultural identity items were coded as dichotomous variables.

Psychosocial variables (acute, chronic and total stress, optimism, lack of control, the three measures of social connections and the two cultural identity variables) were entered in the third step. The Wald test was used to assess each step in the analysis [30]. Variance inflation factors of less than three across all models indicated against multicollinearity. Variables with significant main effects were examined as potential moderators in a model that included depression, racism, the variable in question and its interaction with racism.

Acute, chronic and total stress, optimism, lack of control, the three measures of social connections, two cultural identity variables and responses/reactions to racism were assessed separately as potential mediators of the association between racism and depression.

In order to provide increased statistical power without requiring the mediated effect to be normally distributed, bias-corrected non-parametric percentile confidence intervals were estimated from bootstrapping with 5,000 replications [31]. As variables to be examined in mediation analysis were non-normal (even under various transformations), they were dichotomised at the median and binary mediation approaches with probit models [32] were utilised [33]. Mediators found to be significant in models including only a single mediating variable between racism and depression (adjusted for age, sex, household composition, martial status, equivalised household income, housing tenure and level of education) were then entered simultaneously in a multiple mediation model that adjusted for this same set of confounders [33].

All analyses were then repeated using discrimination-related stress in place of interpersonal racism as the independent variable. All analyses were conducted using Stata 9.0 for Windows (Stata Corporation, College Station, TX, USA: 2007).

Results

Of the 363 DRUID participants, 312 completed at least some of the MIRE. Missing data for racism (n = 52), depression (n = 52) and income (n = 65) resulted in a final sample of 185 respondents. The MIRE was interviewer-administered for 137 respondents while the remaining 48 self-administered. Excluded participants were similar to those included across socio-demographic variables, with the exception that level of education was higher among included participants.

Table 1 summarises participant characteristics. Participants ranged in age from 16 to 81 years (mean 40.2) and were predominantly female. The interpersonal racism score of 5.3 equates to a quarter of respondents reporting no racism, half reporting ‘hardly ever’ experiencing racism and the remaining quarter reporting racism ‘sometimes’, ‘often’ or ‘very often’ on average across nine items. About a quarter of respondents reported racism in only one setting, 30% in two settings, 25% in three settings and 20% in four or more settings. To our knowledge, the CESD-20 has not previously been used to assess depression among Indigenous Australians. Over 29% of the study sample scored above the CES-D cut-off score of 16 compared to about 10% of the total Australian population who report having depression [33].
Table 1

Characteristics of DRUID participants (n = 185)

Variable

Numbera

%a

Sex

 Male

62

34

 Female

123

66

Age group (years)

 16–24

30

16

 25–34

41

22

 35–44

43

23

 45–54

41

22

 55–64

20

11

 65+

10

5

Marital status

 Never married

43

23

 Married/de facto

99

54

 Separated/divorced/widowed

43

23

Gross weekly household equivalised income ($AU)

 1–199

49

26

 200–499

73

39

 500+

63

34

Household tenure

 Owned/being purchased

88

48

 Rented/other

97

52

Household composition

 All Indigenous members

110

59

 Indigenous and non-Indigenous members

75

41

Level of education

 No formal education

46

25

 Year 10/12

40

22

 Non-degree

75

41

 Degree

24

13

Identifies with a clan, tribe or language group

 Yes

130

71

 No

53

29

Recognises homeland/traditional country

 Yes

142

77

 No

42

23

Discrimination-related stress

 Yes

133

72

 No

52

28

Scored variables

Number (range)

Mean (SD)

 Interpersonal racism

185 (0–24)

5.3 (5.4)

 10-item acute stress

178 (0–18)

4.6 (4.5)

 20-item chronic stress

177 (0–52)

11.1 (8.9)

 30-item total stress

172 (0–67)

15.6 (12.1)

 Optimism

180 (11–30)

20.7 (4.1)

 Lack of control

184 (0–12)

3.3 (2.9)

 Positive social connections

179 (6–18)

15.3 (2.8)

 Negative social connections

183 (5–15)

8.2 (2.3)

 Instrumental social connections

180 (1–10)

6.1 (2.0)

 Depression(CES-D)

185 (0–52)

12.9 (10.8)

aN may not add to 185 due to missing values; percentages may not add to 100 due to rounding

Regression models are shown in Table 2. Age, sex, household composition, level of education, equivalised household income, housing tenure, and martial status accounted for 11% of the variance in depression. After adjustment for these variables, racism remained significantly associated with depression, explaining a further 5% of the variance. The psychosocial variables added in step three explained another 24% of the variance. The Wald test was significant at each step of the analysis. None of the examined interaction terms were significant. Comparable results were produced from hierarchical linear regression with discrimination-related distress (analyses not shown).
Table 2

Hierarchical regression models with depression (CES-D) as outcome

 

Model 1 β (SE) N = 185

Model 2 β (SE) N = 185

Model 3 β (SE) N = 160

Age

−0.008 (0.01)

0.002 (0.01)

0.008 (0.01)

Sexa

0.438* (0.22)

0.423 (0.22)

0.277 (0.22)

Mixed household

−0.412 (0.24)

−0.284 (0.25)

−0.434 (0.24)

Year 10/12b

−0.279 (0.32)

−0.111 (0.33)

0.010 (0.31)

Non-degreeb

0.011 (0.29)

0.161 (0.30)

0.071 (0.31)

Degreeb

−0.400 (0.39)

−0.436 (0.41)

−0.330 (0.38)

$200–499c

−0.191 (0.26)

−0.193 (0.27)

0.201 (0.26)

$500+c

0.593* (0.30)

0.701* (0.31)

−0.192 (0.31)

Own/purchasing home

−0.298(0.22)

−0.195 (0.23)

−0.164 (0.21)

Married/Defactod

−0.360 (0.30)

−0.512 (0.30)

−0.436 (0.29)

Separated/divorced/widowed

−0.162 (0.37)

−0.304 (0.39)

−0.233 (0.35)

Interpersonal racism

 

0.080*** (0.02)

0.020 (0.02)

Acute stress

  

0.016 (0.04)

Chronic stress

  

0.000 (.)

Total stress

  

0.027 (0.02)

Optimism

  

−0.047 (0.03)

Lack of control

  

0.058 (0.04)

Negative social connections

  

0.132** (0.05)

Positive social connections

  

0.106* (0.04)

Instrumental social connections

  

−0.006 (0.05)

Identifies with clan/tribal/language group

  

−0.115 (0.23)

Recognises homeland/traditional country

  

−0.165 (0.26)

Constant

3.765 (0.56)

2.950 (0.59)

3.992 (1.04)

Adjusted R2

0.108

0.161

0.401

p < 0.05; **p < 0.01; ***p < 0.001

aReference category is ‘male’

bReference category is ‘no formal education’

cReference category is ‘$1–$199’

dReference category is ‘never married’

Mediation models examining a single variable revealed several significant mediators (see Table 3). Three reactions to racism (‘ashamed,’ ‘powerless,’ and ‘amused’) were statistically significant mediators along with the three composite stress measures, lack of control and negative social connections. In each separate model, 22% (negative social connections) to 59% (total stress) of the association between racism and depression was accounted for by a single mediator. Models with discrimination-related stress as the independent identified a subset of these mediators as significant, namely feeling ‘ashamed,’ ‘powerless’ and the three composite stress measures (see Table 4).
Table 3

Factors mediating the association between interpersonal racism and depression (single mediation models)

Mediator

β coefficient

Bias-corrected lower CIa

Bias-corrected upper CIa

% of effect mediatedb

Ashamed

0.069

0.013

0.148

24.4

Amused

−0.063

−0.151

−0.001

−25.7

Powerless

0.094

0.027

0.173

32.2

Acute stress

0.077

0.021

0.166

27.4

Chronic stress

0.144

0.055

0.252

46.4

Total stress

0.181

0.085

0.284

58.7

Lack of control

0.124

0.049

0.207

41.7

Negative social connections

0.062

0.010

0.133

21.9

For each model, the dependent variable was depression score. Each mediator was included in a separate model along with racism score, age group, sex, household composition, marital status, equivalised household income, housing tenure and level of education

aBecause the forced symmetry of ordinary CIs can result in estimation inaccuracies, bias-correct CIs are adjusted to the percentile values of the sorted distribution of bootstrap estimates used for determining the bounds of the interval [31]

bThe proportion of the association between interpersonal racism and depression that is accounted for by this mediator (expressed as a percentage)

Table 4

Factors mediating the association between discrimination-related stress and depression (single mediation models)

Mediator

β coefficient

Bias-corrected lower CIa

Bias-corrected upper CIa

% of effect mediatedb

Ashamed

0.158

0.014

0.124

21.7

Powerless

0.072

0.026

0.137

26.3

Acute stress

0.049

0.009

0.114

17.3

Chronic stress

0.096

0.040

0.170

40.0

Total stress

0.100

0.041

0.174

33.3

For each model, the dependent variable was depression score. Each potential mediator was included in a separate model along with racism score, age group, sex, household composition, marital status, equivalised household income, housing tenure and level of education

aBecause the forced symmetry of ordinary CIs can result in estimation inaccuracies, bias-correct CIs are adjusted to the percentile values of the sorted distribution of bootstrap estimates used for determining the bounds of the interval [31]

bThe proportion of the association between discrimination-related stress and depression that is accounted for by this mediator (expressed as a percentage)

In the multiple mediation model with interpersonal racism as the independent variable, feeling ‘amused’ as a reaction to racism, acute stress, lack of control, and negative social connections emerged as significant mediators of the association between interpersonal racism and depression (see Table 5). All the mediating variables in this model (ashamed, amused, powerless, acute and chronic stress, lack of control and negative social connections) together accounted for 66% of the association between racism and depression. Feeling ‘ashamed,’ and ‘powerless’ as well as acute and chronic stress emerged as significant mediators in the multiple mediation model with discrimination-related stress as the independent variable (see Table 6). All the mediating variables in this model (ashamed, powerless, acute and chronic stress) together accounted for 53% of the association between discrimination-related stress and depression.
Table 5

Factors mediating the association between interpersonal racism and depression (multiple mediation model)

Mediator

β coefficient

Bias-corrected lower CIa

Bias-corrected upper CIa

% of effect mediatedb

Ashamed

0.007

−0.051

0.066

2.4

Amused

0.077*

−0.167

−0.018

−25.6

Powerless

0.056

−0.009

0.137

18.5

Acute stress

0.049*

0.005

0.124

16.1

Chronic stress

0.041

−0.051

0.135

13.6

Lack of control

0.076*

0.022

0.153

25.3

Negative social connections

0.048*

0.007

0.116

16.0

Total indirect effect

0.200*

0.025

0.349

66.4

The multiple mediation model was adjusted for age group, sex, household composition, marital status, equivalised household income, housing tenure and level of education

p < 0.05

aBecause the forced symmetry of ordinary CIs can result in estimation inaccuracies, bias-correct CIs are adjusted to the percentile values of the sorted distribution of bootstrap estimates used for determining the bounds of the interval [31]

bThe proportion of the association between interpersonal racism and depression that is accounted for by this mediator (expressed as a percentage)

Table 6

Factors mediating the association between discrimination-related stress and depression (multiple mediation model)

Mediator

β coefficient

Bias-corrected lower CIa

Bias-corrected upper CIa

% of effect mediatedb

Ashamed

0.018

−0.030

0.077

5.8

Powerless

0.045

−0.002

0.111

14.6

Acute stress

0.041*

0.001

0.106

13.2

Chronic stress

0.058*

0.007

0.131

18.9

Total indirect effect

0.162*

0.083

0.257

52.5

The multiple mediation model was adjusted for age group, sex, household composition, marital status, equivalised household income, housing tenure and level of education

* p < 0.05

aBecause the forced symmetry of ordinary CIs can result in estimation inaccuracies, bias-correct CIs are adjusted to the percentile values of the sorted distribution of bootstrap estimates used for determining the bounds of the interval.[31]

bThe proportion of the association between discrimination-related stress and depression that is accounted for by this mediator (expressed as a percentage)

Discussion

To our knowledge, this is the first study to show that racism is associated with depression among an indigenous population.

Lack of control as a determinant of morbidity and mortality is now well-established [34, 35]. However, this is the first study to show that lack of control acts as a mediator of the association between racism and depression. In a study involving 108 Arab Americans, lack of control completely mediated the association between racism and self-esteem, and partially mediated the association between racism and psychological distress [36]. Similarly, mastery mediated the relationship between racism and psychological distress among 485 African-, Native- and Asian-Americans [37]. Racism may lead to lack of control by creating unfair and unpredictable demands as well as attenuating rewards resulting from effort.

While other studies examining positive social support as a mediator of the relationship between racism and ill-health have had mixed findings [3840], this study identifies for the first time the role that negative social connections may play in mediating the ill-effects of racism. It is possible that racism perpetrated by friends or relatives may take the form of demands, arguments, criticism, being let down or annoyance. Alternatively, or in addition, stress related to racism may precipitate negative social connections or reduce the capacity of individuals to tolerate social connections (hence increasing the reporting of negative connections). This latter possibility is supported by a study which found racism to be associated with increased reporting of routine social interactions as harassing, exclusionary, and unfair [41].

Two previous studies have examined social connections as moderators of the association between racism and ill-health. A study involving 323 African Americans found that quality of life for those with positive social connections was not diminished by racism [42] while a study involving 3,012 Mexican-Americans found that racism was associated with poor physical health only for those without instrumental social support [43]. Although positive social connections were significantly (negatively) associated with depression in the DRUID sample, no moderating effects were found.

Although a handful of previous studies have found that improved health outcomes are associated with active rather than passive coping responses to racism [40, 44, 45], none of the six responses to racism mediated the association between racism and depression in this study.

Reactions to racism, however, emerged as significant mediators in this study. It is likely that feeling ashamed, humiliated or anxious as well as feeling powerless, hopeless and depressed mediated the association between racism and depression because these affective states are closely related to depression as a condition. Although no previous research has examined reactions as mediators of the relationship between racism and ill-health, a study involving 183 Indigenous Australians found that racism which evoked an emotional/physical response was related to poor general health [46].

Arguably, the most novel finding to emerge from this study is that feeling amusement, contempt or sorrow for a perpetrator of racism as reaction to racism attenuated the association between racism and depression. This suggests the need for further investigation into the contexts in which ‘laughing it off’ is beneficial to those targeted by racism.

Although a previous study involving 215 multiethnic college students found that optimism was associated with depression in a model that also included perceived racism [47], this construct failed to emerge as a statistically significant mediator between racism and depression in the DRUID study.

Reviews in this field suggest that stress acts as a mediator between racism and poor health [2, 3]. However, only a few studies have examined such a role, with stress shown to be a mediator for smoking [12], depression/anxiety [48, 49], and hypertension [50]. Both acute and chronic stress emerged as important mediators in this study, remaining significant in multiple mediator models (with the exception of chronic stress in the interpersonal racism model).

Although three previous studies have found that ethnic identity buffers racism-related stress [48, 51, 52], a recent review of 12 studies notes that identity is not sufficient to ameliorate the effects of racism on the development of depression [40]. It is not clear why the two aspects of cultural identity assessed in this study failed to emerge as either significant mediators or moderators of the association between racism and depression.

The robustness of study findings are supported by broadly similar findings for both interpersonal racism and discrimination-related stress. Nonetheless, there are several study limitations to note. Racism is a difficult concept to measure accurately. It is evident that racism can go unnoticed when it does occur and be perceived when it is not ‘objectively’ present (i.e. an avoidable and unfair inequality of resources, opportunity or benefit has not, in fact occurred). The measure used in this study assesses subjective or perceived experiences of racism that respondents were willing to report. However, as with experiences of stress more generally, perceived racism itself is the source of stress for an individual, regardless of its objective veracity [24]. Moreover, there is evidence that respondents are more likely to underestimate than overestimate experiences of racism [53, 54]. This may be due to the poorly understood (and largely invisible) nature of systemic racism [3], the protective effects that may accrue from not attributing experiences to racism [55] and the negative social repercussions of labelling an experience as racism.

Although 32 longitudinal studies suggest that the primary direction of causation is from racism to ill-health rather than ill-health leading to increased reporting of racism [24, 5658], this latter scenario cannot be ruled out in this cross-sectional study. Older, female higher socioeconomic status participants were over-represented in this study compared to the Darwin or Australian Indigenous populations [25]. Failure to identify any significant moderators may have been due to the recognised low power of moderation tests [59] combined with a relatively small sample size.

This study highlights the complex pathways leading from racism to mental ill-health and elucidates several new avenues for further research. Improved understanding of racism as a determinant of health will assist in efforts to counteract racism as a growing public health concern.

Acknowledgments

The authors gratefully acknowledge the support of DRUID study participants, study staff, members of the Indigenous Steering Group, and partner organisations. Ms Hannah Reich provided research assistance in the drafting of this manuscript. The DRUID Study was funded by the National Health and Medical Research Council (NHMRC Project Grant #236207), with additional support from the Australian Government Department of Employment and Workplace Relations, the Clive and Vera Ramaciotti Foundation, the Vincent Fairfax Family Foundation, the International Diabetes Institute (AusDiab Partnership), and Bayer HealthCare. The DRUID Study is an in-kind project of the Cooperative Research Centre for Aboriginal Health (CRCAH). The first author was supported by an NHMRC Population Health Capacity-Building Program (#236235). The second author was supported by an NHMRC Career Development Award (#283310) and an NHMRC Research Fellowship (#545200).

Copyright information

© Springer-Verlag 2010