Introduction

Despite a growing body of evidence supporting linkages between poverty and common mental disorders in adults living in low and middle income countries (LMICs)[1], there remains a surprising dearth of evidence examining this for adolescents within these settings[2]. Globally, mental disorders contribute significantly to Disability-Adjusted Life Years (DALYs) and may be prevalent in up to a quarter of young people[2, 3]. In South Africa, the adjusted prevalence of mental disorders in adolescence has been estimated to be between 15–17%[2, 4]. This is lower than prevalence estimates from More Economically Developed Countries [2] although higher than prevalence rates reported from other LMIC settings[5].

The presence of mental disorders in adolescence is a strong predictor for mental health problems in adulthood[3, 6] and may have a direct impact on economic and educational outcomes, as well as crime and suicide[2]. Adolescents who live in households and attend schools which are socially and economically disadvantaged are at a greater risk of mental disorders through a number of mechanisms, including increased exposure to violence, food and income insecurity, substance abuse, reduced economic protection and fewer employment or career prospects[2]. In South Africa, some of these risk factors, such as exposure to violence and poverty are more prevalent[4, 5]. However, little work has been done to assess the association of these experiences with mental disorders, within this context[5].

From 1948 until 1994 the apartheid regime in South Africa led to a racially segregated society in which access to economic resources and medical care, as well as educational and employment opportunity were severely restricted for people socially classified as ‘non-white’. Although following the end of apartheid there were dramatic changes in the constitution including the right to health for all as a fundamental principle[7, 8], investigators conducting research in the post-apartheid era continued to report widespread disparities in health and access to healthcare across ethnic groups[7, 9–11] with health inequalities continuing to disproportionately affect the black African majority population as well as coloured populations[9]. In South Africa the term ‘coloured’ is used to identify a group of people with mixed black and white ethnicity who have a relatively distinct cultural identity, particularly in the Western Cape. In the apartheid era these racial categories were used by the government to legitimise state-sponsored oppression of ‘black’ and ‘coloured’ people. As the effect of these practices on health and access to resources may still be apparent, these categories are used within the context of the study to examine persisting inequalities in the post-apartheid era. Although the same categories are employed here, they are acknowledged as socially constructed and not as essentialist representations of ‘race’[12].

South Africa has also been described as one of the most ‘unequal societies in the world’[13]. Income inequality increased in the post-apartheid years, from 1993 to 2008[14], with rising inequality fuelled by an increased share of wealth going to the country’s wealthiest decile[14, 15]. It has been suggested that racial inequality remains interwoven with economic inequality[15], with the authors of a recent OECD report suggesting that “levels of poverty and inequality continue to bear a persistent racial undertone” [15].

Thus, the interplay of racial and material inequalities within South Africa in the post-apartheid era remains complex and may be associated with detrimental consequences for adolescent mental health. Although South Africa is a middle income country, on many major health indicators its outcomes are worse than those of lower income countries[9]. In countries experiencing greater income inequality, associations between poverty and poorer mental health may be stronger than in societies which are more equitable[1, 16]. The picture is further complicated by South Africa having a high prevalence of violence, with much of this accounted for through interpersonal and gender-based violence and children experiencing or witnessing high levels of violence and neglect[16]. These are important risk factors for adverse mental health[16, 17]. The high prevalence of violence may be related to underlying income inequality[16, 18].

With this in mind, we conducted an analysis of data from the South African Health and Well-being Study (SHaW). The study was broadly based on an earlier study of risk and protective factors for adolescent mental health carried out in a socially deprived multi-ethnic area of East London, UK, the Research with East London Adolescents: Community Health Survey (RELACHS)[19]. Its purpose was to address the knowledge gap in risk and protective factors for adolescent mental health in a different multi-ethnic setting from London, UK[19]. Our primary objective was to assess the association of poverty and disadvantage with psychological morbidity. In particular we hypothesised that children living in materially disadvantaged circumstances or from historically disadvantaged ethnic/racial backgrounds would be more likely to screen positive for common mental disorders (CMD) and Post Traumatic Stress Disorders (PTSD). As a second objective, we sought to identify important experiences such as racially-motivated bullying, levels of self-esteem, social support and exposure to violence, as potential mediators of observed associations, as previous work in this area has indicated their potential importance[1, 2, 17, 20, 21].

Methods

Results

1034 (of a total of 1169 possible students) took part, resulting in a response rate of 88%. Children not completing questionnaires were absent from school on the day of data collection. Levels of missing data for each variable were in general low with most variables having valid responses for 90% or more of participants. The derived variable for household overcrowding had the greatest amount of missing observations (data available for 80% of participants on this measure), probably because this variable was derived from two other measures (no. of people in household and no. of rooms in household).

Demographic features of the full sample are displayed in supporting information (S1 Table). Table 1 displays the distribution of disadvantage indicators across the sample, by race/ ethnicity. In general, children who described their race/ ethnicity as ‘black’ appeared more disadvantaged across most durable asset/household indicators. The strongest evidence of disadvantage was evident for measures indicative of ability to afford basic items. For example, black children reported that they were less likely to afford basic items such as soap to wash (15% versus 1.5% coloured children, 3.9% white children 0% other groups(x² = 69.65;p = 0.04). More than a quarter of black children reported being unable to afford more than one pair of shoes (compared to 4.3% coloured children and 4.8% white children, (x² = 112.63;p = 0.04)). 37.5% of black children were in a lower grade for their age, compared with 8.3% coloured children, 12% white children and 12% children reporting their ethnicity as ‘other’ (x² = 117.97;p = 0.01). Maternal employment was broadly equivalent across the sample, although paternal unemployment was more likely for children reporting their ethnicity as ‘black’ (46.1%) or ‘other’ (38.5%) (vs. 14.3% of white children). The presence of household electricity was fairly universally distributed across the full sample, irrespective of ethnicity. Eight children describing themselves as belonging to an ‘other’ ethnic/ racial group appeared to be relatively disadvantaged across most of the measures.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Disadvantage indicators by race/ ethnicity.

https://doi.org/10.1371/journal.pone.0154478.t001

PCA was used to derive a single asset index of durable assets and household amenities. The first component had an eigenvalue of 3.00 and accounted for 30% of variation in the original measures. Survey weighted means with 95% confidence intervals on this measure by race/ ethnicity were: white children:1.05 (-0.46,2.57); black children:-1.24 (-2.82,0.33); coloured children:0.40 (-0.49,1.28); Indian children:1.58 (1.10,2.06); ‘other’ children:-0.02 (-2.11,2.07) (lower values indicating greater disadvantage) (see S1 Fig which displays the distribution of the PCA-derived asset index by ethnicity and for the full sample).

In the full sample, the unadjusted prevalence of depression was 41% and anxiety was 16%, giving a combined prevalence for common mental disorders (anxiety or depression) of 47%. By self-ascribed ethnicity the prevalence of CMD was: 29% (white children), 58% (black children), 45% (coloured children), 24% (Indian children) and 41% (children self-classifying as ‘other’ ethnicity). The prevalence of PTSD in the full sample was 21% and by ethnicity: 14% (white children), 34% (black children) and 19% (coloured children).

Table 2 displays the unadjusted prevalence of CMD and PTSD according to race/ ethnicity and material disadvantage indicators. Broadly, across each of the racial/ ethnic groups, a gradient was evident in the association of increasing disadvantage, although both black and white children in the poorest fifth did not always follow this pattern. Put in another way, reduced access to resources in terms of ability to afford basic items or ownership of basic amenities and durable asset index was broadly associated with an increased prevalence of either CMD or PTSD. Broadly, for each level of disadvantage, the unadjusted prevalence of common mental disorders or PTSD was greater in black or coloured children than in white children (Table 2). There were too few observations to derive stable estimates for Indian children or children describing their racial/ ethnic group as ‘other’.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Prevalence of CMD and PTSD.

https://doi.org/10.1371/journal.pone.0154478.t002

There was strong evidence of an association between race/ ethnicity and exposure to violence or racially motivated bullying, with coloured children most exposed to violence, and Indian children most likely to be bullied because of religion or race. Exposure to the adverse mediators of racial bullying, poorer social support and lower self-esteem, each followed a linear trend with increasing material disadvantage/ reduced access to resources (S2 Table).

Table 3 displays the relative risk for association of independent variables with CMD and PTSD, after adjusting for gender. After adjusting for gender, both CMD and PTSD were more than twice as likely in black children, relative to white children and CMD was 1.73 times more prevalent in coloured children (95% CI:1.11, 2.70). Estimates for PTSD were not available for Indian and ‘other’ children, due to small numbers in cells. There was a dose-response relationship for the summed ‘ability to afford basic items’ measure with both CMD and PTSD. As access to healthcare may be distinct from access to financial resources we repeated analyses using the scale which assessed ‘ability to afford basic items’ with healthcare measures taken out. Associations persisted and were similar, after the question on ‘ability to afford medicines and visits to the doctor’ was removed from the ‘ability to afford basic items’ scale (Per one item increase for CMD this was RR: 1.17 (95% CI: 1.07, 1.28) p = 0.009 (trend), for PTSD this was RR: 1.23(95% CI:1.06, .42) p = 0.02 (trend)). After adjusting for gender, children who reported not being able to afford medicines or visits to their doctor were 1.17 (95% CI: 1.06, 1.29) times more likely to screen positive for CMD and 1.23 times more likely (95% CI: 1.05, 1.45) to screen positive for PTSD. Most of the putative mediators were associated with CMD and PTSD (Table 3). Associations between most of the putative mediators and CMD and PTSD persisted after adjustment of ethnicity and all socioeconomic position/ disadvantage indicators (S3 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Gender-adjusted association of individual disadvantage indicators with CMD and PTSD.

Relative Risk (RR) and 95% Confidence Intervals (95% CIs).

https://doi.org/10.1371/journal.pone.0154478.t003

After adjusting for disadvantage indicators and ethnicity in models, most associations of disadvantage with common mental disorders were attenuated. After adding putative mediators to models, most associations of ethnicity and disadvantage with CMD were partially attenuated, supporting at least partial mediation by these variables (Table 4). Of note, however, black and coloured children continued to experience an elevated relative risk of CMD, compared to white children, even after adjusting for potential mediators (Table 4) and children who reported experiencing violence were more than twice as likely to screen positive for CMD in adjusted models (Table 4) (RR:2.39 (95% CI:1.22,4.68; p<0.001). As with the analyses concerning common mental disorders, most associations for PTSD were attenuated with the addition of potential mediators. Overall, associations with race/ ethnicity and disadvantage indicators were less marked with PTSD than with CMD, in adjusted models (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multivariable associations with Common Mental Disorders (CMD) (n = 583) & Post Traumatic Stress Disorders (PTSD) (n = 532); Relative Risk (RR) and 95% Confidence Intervals (95% CIs).

https://doi.org/10.1371/journal.pone.0154478.t004

Discussion

Strengths and limitations

Strengths of the survey include its high response rates with low levels of missing data reducing non-response bias. In addition, the sampling frame led to an ethnically and socio-economically diverse sample, representative of pupils attending schools within a large metropolitan area of Cape Town. This permitted the assessment of disadvantage across multiple domains, as well as an estimation of prevalence of CMD and PTSD. Most measures have adequate validity and reliability in this or other equivalent settings, confirmed by a prior pilot study conducted within this setting[22]. Use of materials in languages other than English has meant that assessment would have been less prone to measurement bias.

Limitations of the survey include its cross-sectional design, limiting inferences around direction of association and the possibility of reverse causality in some of the measures, for example children who are currently depressed may also have low self-esteem as a result of a depressed mental state. In addition, although the sample was representative of a defined region in Cape Town, the generalisability of findings to other parts of South Africa remains unclear, further work is needed to establish how the findings reported here differ from other parts of South Africa. It is clear that mental disorders in adolescence have multifactorial origins, associated with, but not confined to, poverty and social disadvantage [2]. It was not possible to adjust for these other indicators,(which may include stressful life events and academic stress, family history of mental illness and living with a parent with a mental disorder[2]). It is possible that these, as well as other variables, played a part in the findings which we could not account for. Our analyses also treated gender as a confounder as there was no statistical evidence in our analyses in support of effect modification by gender. Previous work has suggested that types of mental disorder experienced in adolescence also vary by gender[2]. The multifactorial nature of adolescent mental disorders within this context, including the role of gender in accounting for associations, should be investigated in future work.

Future research and policy

Future research should focus on the inter-play of poverty with racial/ ethnic disparities[9], and longitudinal analyses of the associations between material disadvantage and mental health among adolescents in low and middle-income countries such as South Africa. Our study focused on the experience of adolescent learners within multi-ethnic schools in South Africa. Future work could also take into account the role of the school context (for example ethnic composition) in accounting for mental health in schools where children form a relative minority[55]. Larger nationally representative cohort studies that follow children through adolescence into adulthood are likely to yield more specific information on the complex mechanisms underlying the associations between material advantage, mediating variables (such as violence, social support, bullying and self-esteem) and mental health, observed in this study. Further work on the development of valid and reliable mental health and socio-economic measures is integral to this task[56]. In addition, there is an urgent need for studies that evaluate the impact of interventions on two hypothesised mechanisms of the poverty-mental health association in LMIC: firstly population-level intervention studies which address the root causes of material and social inequalities[1, 54] and assess their impact on mental health disparities; and secondly mental health intervention studies that evaluate the effect of a wide range of prevention and treatment interventions on material and educational disadvantage over time[57]. From a policy perspective, this study presents compelling evidence that mental health is integrally linked with material and ethnic disadvantage among adolescents in South Africa, and should be included in poverty alleviation interventions targeting young people.

In conclusion

In conclusion, this report confirms disparities in the mental health of young people growing up in South Africa, 15 years after the demise of apartheid. These findings raise some concern given the known continuity of childhood mental health with adult mental health[6] as well as the role which childhood disadvantage and childhood psychological morbidity plays in a range of adverse downstream health and social outcomes[58–60].

Supporting Information

Acknowledgments

We would like to thank all of the learners and schools that took part in the study. We would like to thank Arlene Kleinhans for her leading role in the data collection. Our colleague and friend Professor Alan J Flisher died tragically on 18 April 2010. We would like to acknowledge his substantial contribution to adolescent mental health in South Africa and his warm, wise and generous support of this research.