Background Despite a long history of discrimination and persisting racial disparities in smoking prevalence, little research exists on the relationship between discrimination and smoking in South Africa.
Methods This analysis examined chronic (day-to-day) and acute (lifetime) experiences of racial and non-racial (eg, age, gender or physical appearance) discrimination and smoking status among respondents to the South Africa Stress and Health study. Logistic regression models were constructed using SAS-Callable SUDAAN.
Results Both chronic racial discrimination (RR=1.45, 95% CI 1.14 to 1.85) and chronic non-racial discrimination (RR=1.69, 95% CI 1.37 to 2.08) predicted a higher risk of smoking, but neither type of acute discrimination did. Total (sum of racial and non-racial) chronic discrimination (RR=1.46, 95% CI 1.20 to 1.78) and total acute discrimination (RR=1.28, 95% CI 1.01 to 1.60) predicted a higher risk of current smoking.
Conclusions Racial and non-racial discrimination may be related to South African adults’ smoking behaviour, but this relationship likely varies by the timing and frequency of these experiences. Future research should use longitudinal data to identify the temporal ordering of the relationships studied, include areas outside of South Africa to increase generalisability and consider the implications of these findings for smoking cessation approaches in South Africa.
- Priority/special populations
- Global health
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Although apartheid, which includes the period of legalised institutional racial discrimination in South African history, ended about 20 years ago (1950–1994), few articles exist on discrimination and health in South Africa. This lack of literature stands in contrast to the large number of publications produced by some other countries with a history of legalised discrimination, such as the USA.1–6 Race relations remain strained in South Africa today.7 Historically, the racial status structure of South Africa consisted of Whites at the top, Black Africans at the bottom, and Coloured (a term that generalised South Africans of mixed race into a single category) and Indians or Asians in the middle.8 According to mid-year 2011 population estimates, 79.5% of South Africans were Black Africans, 9.0% were White, 9.0% were Coloured, and 2.5% were Indian or Asian.9
In 2003, approximately 35% of South African men and 10% of South African women smoked daily or occasionally.10 Among men, prevalence of daily or occasional smoking was 32.8% for Black South Africans, 52.1% for Coloured, 35.7% for Whites and 55.5% for Indians.10 Among women, prevalence was 5.2% for Black Africans, 41.8% for Coloured, 27.3% for Whites and 13.1% for Indians.10 Racial differences in smoking prevalence are also evidenced by racial disparities in smoking-attributable mortality rates in South Africa.11 Smoking is associated with increased odds of tuberculosis, chronic obstructive pulmonary disease, lung cancer, upper digestive tract cancer, ischaemic heart disease, cirrhosis of the liver and other alcohol-related conditions among South Africans.11 ,12
We were able to locate only one article on racial discrimination and smoking in South Africa. Brook et al13 found a positive association between racial discrimination and smoking status among South African adolescents (OR=1.46, 95% CI 1.21 to 1.75).
Existing research suggests a relationship between discrimination and smoking.3–5 14–23 Several studies have found a significant relationship between experiences of racial and ethnic discrimination and current smoking for African American1 ,3 ,5 ,16 ,17 and Hispanic/Latino and Asian American samples in the USA.4 Using a multi-state sample of adults ages 45–84, Borrell et al3 found that African Americans who reported one or more lifetime experiences of racial or ethnic discrimination had 34% greater odds of current smoking than those who reported none. Among Asian Americans, Chae et al4 found a positive association between both racial and ethnic discrimination and unfair treatment with smoking. Similar findings exist for research conducted in New Zealand.14 ,24 ,25 Among adults in New Zealand, racial discrimination was significantly associated with smoking status, and the authors found a positive dose–response relationship between number of types of discrimination reported and odds of current smoking.25
In order to fully understand the relationship between discrimination and smoking in South Africa, both racial and non-racial discrimination should be assessed. South Africans tend to report lower levels of racial discrimination and higher levels of non-racial discrimination than US samples.26 However, no English language publication has examined non-racial discrimination and smoking among South Africans.26 However, several studies conducted outside of South Africa found positive associations between specific types of non-racial discrimination, such as sexual orientation,27 sexism,28 ,29 immigration-based discrimination30 and smoking.
This analysis examined the role of chronic and acute racial and non-racial discrimination in smoking among respondents to the South Africa Stress and Health (SASH) study, a representative sample of South African adults. Prior research on discrimination indicates that like other stressful experiences, discrimination is multidimensional, consisting of acute experiences that are discrete and time-limited and chronic experiences that are ongoing.31
The first aim was to examine six types of discrimination—total chronic, chronic racial, chronic non-racial, total acute, acute racial and acute non-racial—in association with smoking status. The second aim was to explore the association between the number of experiences of each type of discrimination and smoking intensity. We hypothesised that all types of discrimination would be associated with a higher risk of smoking. We also hypothesised a significant positive relationship between number of experiences of discrimination and smoking intensity among smokers.
Population and sample
The SASH is a nationally representative sample of South African adults collected between 2002 and 2004 to investigate postapartheid life in South Africa.32 Human Subjects Committees from the University of Michigan, Harvard Medical School and Medical University of South Africa provided ethics approval before data collection began.33 Multi-stage stratified random sampling was used to select individuals from enumerated provinces, groupings of homes and households. Inperson interviews were conducted with one individual per household in one of seven languages based on the respondent's preference. Informed consent was obtained from all participants before interviewing began.33 Further information on sampling techniques can be found in publications by Williams and colleagues.32 ,33 After all data were collected, researchers at The University of Michigan and Harvard University created weights that adjusted for sampling techniques and non-response.33
All discrimination predictor variables were dichotomised (zero vs one or more) because of limited variability in the number of experiences of discrimination, as well as to create consistent coding across acute and chronic discrimination despite differences in the formats of these scales.
Chronic discrimination was measured using the 10-item version of the Everyday Discrimination Scale,34 which asks about day-to-day experiences of discrimination and has a Cronbach's α of 0.84. This scale is reliable across the four major racial groups in South Africa.35 The scale assesses the frequency of (1) receiving less courtesy or (2) respect than others; (3) receiving poorer service; people acting as if they are (4) afraid of you, (5) think you are dishonest, (6) think they are superior to you or (7) think that you are not smart; (8) being called names or insulted, (9) threatened or harassed; and (10) ‘followed around’ in stores. Response options were a 6-point Likert scale that ranged from ‘almost everyday’ (1) to ‘never’ (6). The number of items endorsed as occurring ‘ever’ or more frequently were added up and then dichotomised (zero vs one or more experiences) to create the total chronic discrimination variable.
An eleventh item in the scale asked, ‘What do you think was the main reason for this/these experience(s)?’ Experiences attributed to race or tribe were coded as chronic racial discrimination. Experiences attributed to age, gender, height or weight, other aspect of physical appearance, or ‘other’ were coded as chronic non-racial discrimination. Both variables were dichotomised as zero versus one or more experiences. We also examined the chronic non-racial discrimination subtypes of age, gender, physical appearance or other whenever the small number of respondents endorsing these items did not prevent us from doing so. These types of discrimination were also dichotomised as zero experiences versus one or more.
Acute discrimination was measured using nine questions about lifetime experiences of discrimination within multiple domains.32 Respondents were asked whether they had ever… (1) been unfairly fired; (2) for unfair reasons, not been hired for a job; (3) unfairly been denied a promotion; (4) unfairly been stopped, searched, questioned, physically threatened or abused by the police; (5) unfairly been discouraged by a teacher or advisor from continuing their education; (6) unfairly been prevented from moving into a neighbourhood or (7) moved into a neighbourhood where neighbours made it difficult for them or their family; (8) been unfairly denied a bank loan; and (9) received service that was worse than what other people get. Response options were ‘yes’ and ‘no’. The number of items was counted up and dichotomised (zero vs one or more experiences).
Respondents were asked to hypothesise a reason for each experience they endorsed, choosing from the same list of personal characteristics included in the Everyday Discrimination Scale. Attributions were coded as racial or non-racial discrimination based on the same criteria used for chronic discrimination. Both variables were dichotomised as zero versus one or more experiences. We also examined the acute non-racial discrimination subtypes of age, gender, physical appearance or other whenever the small number of respondents endorsing these items did not prevent us from doing so. These types of discrimination were also dichotomised as zero experiences versus one or more.
The primary outcome variable was dichotomous smoking status (current smokers vs all others). Current smokers had smoked 100 cigarettes or more in their lifetimes and self-identified as ‘current smoker’ when given the options of current, past or never smoker. All other respondents (former and never smokers) were classified as non-smokers. For respondents with missing values, responses to a third question, ‘Are you a current smoker?’ (yes/no) were used to aid in the determination of smoking status. The mean number of cigarettes smoked per day (smoking intensity) among smokers was also examined as an outcome variable.
Covariates included age, gender, education, race, household income (adjusted by number of household members) and marital status. Because of potential differences in racial segregation across urban and rural settings in South Africa, urbanicity (urban or rural residence) was examined as a potential confounder.
Ethnic identification, which can be defined as one's sense of belonging with those of a similar ethnic group that has a similar cultural background and origin, was tested as an effect modifier and confounder of discrimination and smoking.4 ,17 Questions interpreted as assessments of ethnic identification included (1) ‘How closely do you identify with other people… of the same racial descent...?’ (1=‘very closely’ to 4=‘not at all’); (2) ‘How close do you feel in your ideas and feelings … to other people of the same racial descent?’ (1=‘very closely’ to 4=‘not at all’); (3) ‘How much time would you like to spend with … people… of your same racial group?’ (1=‘a lot’ to 4=‘none’); and (4) ‘How important … for people … from your racial group to marry … people… from this group?’ (1=‘very important’ to 4=‘not at all important’). Scores were summed and divided at the median (≤8=low ethnic identification, >8=high ethnic identification).4 ,17
χ2 Analyses were used for bivariate analyses of each type of discrimination by race and smoking status. McNemar's tests were used for bivariate statistics for the number of experiences of discrimination and smoking intensity. SAS-Callable SUDAAN 11 was used for all modelling because the clustered sampling design could contribute to correlated values within sampling tracts. SUDAAN enabled us to adjust for selection based on household size and non-response, and to make the sample comparable with the 2001 South African census data.
Because of the high prevalence of smoking among respondents in SASH, which likely violated the rare disease assumption, SUDAAN was used to construct log-binomial models of discrimination and the relative risk of smoking. In addition to models using dichotomous discrimination predictor variables, we also tested for a dose–response relationship between discrimination and smoking. We used log-binomial models to examine the relationship between the number of items endorsed for each type of discrimination and smoking risk. PROC LOGLINK was used to create multivariable log-linear regression models of the sum of the number of experiences of each type of discrimination and smoking intensity. In order to create consistency with the majority of the tobacco literature, all multivariable models were adjusted by covariates that tend to correlate with smoking behaviour: race, gender, age, education and income.36 For each type of discrimination, respondents who had one or more missing questionnaire responses were excluded from multivariable analyses.
First, total chronic discrimination and total acute discrimination were modelled in two separate models. Then, chronic racial and non-racial discrimination were modelled in one model and acute racial and non-racial discrimination in another model. In order to lessen concerns about collinearity, we excluded respondents who endorsed both racial and non-racial discrimination (n=99, 2.3% of sample), thereby creating mutually exclusive exposure categories for acute discrimination. Interaction terms were tested for each type of discrimination by demographic covariates.
The majority of the sample was Black African (76.4%), female (54.0%), between the ages of 18 and 34 (51.1%), and reported 12 years of education or less (84.7%; table 1).
Chronic racial discrimination varied significantly by race (p<0.0001; table 2). Indians/Asians and Black Africans reported the highest levels of chronic racial discrimination (17.6% and 15.6%, respectively) and non-racial discrimination (29.6% and 31.1%, respectively).
Acute racial (p=0.0033) discrimination varied significantly by race (table 2), with Indian/Asians and Black Africans reporting the highest levels of acute racial discrimination (11.2% and 7.8%, respectively). Indians/Asians and Whites had highest levels of acute non-racial discrimination (20.0% and 17.3%, respectively).
Analyses of subtypes of non-racial chronic discrimination revealed that, compared with smokers, non-smokers reported a higher prevalence of discrimination related to gender (7.8% vs 6.1%, p=0.2628) and other (13.7% vs 2.7%, p=0.0047), and smokers reported a higher prevalence of discrimination related to age (8.7% vs 6.2%, p=0.2269) and physical appearance (7.8% vs 5.9%, p=0.3966; see online supplementary table S1). Blacks reported higher levels of discrimination based on age (10.5%, p=0.0062) and physical appearance (10.5%, p<0.0001) than other racial groups (see online supplementary table S2). Women reported higher levels of sexism than men (17.7% vs 9.7%, p=0.0034; see online supplementary table S3).
For subtypes of non-racial acute discrimination, smokers had a higher prevalence of discrimination attributed to other (13.1% vs 7.9%, p=0.0125) than non-smokers (see online supplementary table S1). Whites had a higher prevalence of discrimination attributed to physical appearance (14.5%, p=0.0001) and other (86.9%, p<0.0001) than other racial groups (see online supplementary table S2). The prevalence of sexism was higher for women than men (8.3% vs 3.5%, p=0.0393; see online supplementary table S3).
In unadjusted regression analyses, chronic racial discrimination (RR=1.32, 95% CI 1.08 to 1.62) and chronic non-racial discrimination (RR=1.40, 95% CI 1.12 to 1.74) were significantly associated with smoking status (table 3). Total chronic discrimination was also significantly associated with smoking status (RR=1.33, 95% CI 1.12 to 1.59). Adjusting for covariates increased the relative risk of smoking associated with chronic racial discrimination (RR=1.45, 95% CI 1.14 to 1.85), non-racial discrimination (RR=1.69, 95% CI 1.37 to 2.08; table 3) and total discrimination (RR=1.46, 95% CI 1.20 to 1.78).
Tests for interactions revealed a significant positive interaction of chronic racial discrimination by gender (p=0.0370) and chronic non-racial discrimination by race (p=0.0009). Among those who experienced racial discrimination, the relative risk of smoking was significantly higher for men (RR=1.41, 95% CI 1.15 to 1.72) than for women (RR=1.01, 95% CI 0.61 to 1.68). Among respondents who endorsed non-racial discrimination, Black Africans (RR=1.67, 95% CI 1.24 to 2.25) and Whites (RR=2.62, 95% CI 1.69 to 4.06) had a higher risk of smoking than South Africans of Coloured (RR=1.09, 95% CI 0.67 to 1.77) or Indian, Asian or other descent (RR=1.00, 95% CI 0.57 to 1.74). Urbanicity was not included in multivariable models because it was not a significant confounder.
In models of number of chronic discrimination items endorsed and smoking risk, adjusting for weights, stratification variables and covariates, each 1-item increase in discrimination was associated with a 5% increase in risk of smoking (RR=1.05, 95% CI 1.02 to 1.08) for total chronic discrimination, 3% for chronic racial discrimination (RR=1.03, 95% CI 0.99 to 1.08) and 5% for chronic non-racial discrimination (RR=1.05, 95% CI 1.01 to 1.08). In multivariable models restricted to smokers, the number of experiences of chronic racial discrimination (Wald χ2=19.97, p=0.0295) and non-racial discrimination (Wald χ2=26.89, p=0.0027) were positively associated with mean cigarettes smoked per day.
In unadjusted analyses, neither acute racial discrimination (RR=1.22, 95% CI 0.84 to 1.77) nor acute non-racial discrimination (RR=1.25, 95% CI 0.92 to 1.68) was significant (table 3). However, total acute discrimination was significant (RR=1.37, 95% CI 1.14 to 1.64). Adjusting for covariates, racial discrimination (RR=1.07, 95% CI 0.67 to 1.70) and non-racial discrimination (RR=1.18, 95% CI 0.85 to 1.66) remained non-significant (table 3). However, total acute discrimination remained significant (RR=1.28, 95% CI 1.01 to 1.60). For acute discrimination, the only significant interaction was non-racial discrimination by gender (p=0.0391). Among respondents who endorsed acute non-racial discrimination, women had a significantly higher risk of smoking (RR=1.68, 95% CI 1.06 to 2.67) than men (RR=1.06, 95% CI 0.49 to 1.42). Urbanicity was not a significant confounder in any of the models. The number of experiences of each type of acute discrimination was not associated with cigarettes smoked per day.
For acute total discrimination, there was a non-linear relationship between the number of scale items endorsed and smoking risk. Risk of smoking was greater (RR=1.17, 95% CI 1.05 to 1.29) for one experience of discrimination versus zero than for seven experiences versus six (the maximum, RR=1.12, 95% CI 1.07 to 1.17). Results were non-significant for number of items of acute racial and non-racial discrimination endorsed.
Ethnic identification was not a significant effect modifier or confounder for any type of discrimination.
This analysis used a nationally representative probability sample of South African adults to examine the association between multiple types of discrimination and smoking status. For racial and non-racial discrimination, the significance of these associations varied by whether discrimination was acute or chronic. However, both total chronic and acute discrimination were significantly associated with smoking status. Analyses of the number of experiences of discrimination and both smoking status and smoking intensity suggest that a dose–response relationship may exist for some types of discrimination and smoking.
Interaction terms suggested variation in the relationship between discrimination and smoking by gender and race. Chronic racial discrimination may impact smoking more for men than women, and acute non-racial discrimination may impact smoking more for women than men. A number of potential explanations exist for this finding. Gender-based differences in the prevalence and reaction to experiences of sexism, as well as differences in psychological and physiological responses to chronic and acute stress, may explain these relationships. Unmeasured mediators or confounders, differences in stress responses to discrimination,37 ,38 gender differences in socialisation patterns and social support networks,39 and gender norms about smoking in South Africa may have also influenced these findings.28 ,37 Analyses of subtypes of non-racial discrimination did not explain the interactions we found as women were more likely than men to report sexism for both chronic (p=0.0034) and acute (p=0.0393) discrimination.
The association between chronic non-racial discrimination and smoking was significant for Black Africans and Whites but not Coloured or Indians, Asians or other race. This finding is consistent with studies that found racial/ethnic variation in the relationship between discrimination and smoking in US samples.17 ,19 ,23 Interestingly, the Coloured and White groups had low levels of discrimination but a high prevalence of smoking. These findings are likely due to the fact that these groups have a higher prevalence of demographic characteristics that are associated with a higher prevalence in the sample (eg, male gender, ages 35–49, low or high income, and separated, widowed, or divorced marital status; see online supplementary table S4). Unmeasured confounders, variation in successful smoking cessation by race in South Africa38 ,40 and racial differences in influences on smoking behaviour may have also contributed to these findings.38
Surprisingly low levels of racial discrimination were reported given South Africa's history of racial conflict, particularly when compared with levels of racial discrimination in the USA.26 In addition to a positive view of race relations postapartheid and marked racial differences in socioeconomic status (SES), current segregation and isolation may limit inter-racial interaction, especially as equals, contributing to artificially low levels of perceived interpersonal racial discrimination.35 ,41
Our results were similar to those of the only other study on racial discrimination and smoking in South Africa. Among South African adolescents, Brook et al13 found a significant positive association between racial discrimination and being an experimental (OR=1.41, 95% CI 1.14 to 1.75) or current smoker (OR=1.46, 95% CI 1.21 to 1.75) versus a non-smoker.
Our results were also consistent with findings from non-South African samples. We were able to locate one such study that examined both chronic and acute discrimination. However, the analysis did not differentiate between racial and non-racial discrimination. High (PR=1.41, 95% CI 1.15 to 1.74) and moderate (PR=1.19, 95% CI 1.05 to 1.36) scores on the Everyday Discrimination Scale were associated with a higher prevalence of current smoking among pregnant women in the USA.1 There was also a significant relationship between scores from a 2-item acute discrimination scale and current smoking, but the relationship was not significant in adjusted models. Borrell et al3 found a positive association between racial discrimination and current (vs never) smoking for both Black (OR=1.34, 95% CI 1.00 to 1.81) and White Americans (OR=1.88, 95% CI 1.02 to 3.49) enrolled in the Multi-Ethnic Study of Atherosclerosis. In a study of African American college students, respondents who experienced racial harassment were twice as likely to smoke on a daily basis as those who reported none (OR=2.10, 95% CI 1.23 to 3.58).15 Chae et al4 found a significant relationship between high levels of racial discrimination and current smoking among Asian Americans (OR=3.06, 95% CI 1.07 to 9.72). Using a New Zealand sample, Harris et al25 found a significant relationship between overall discrimination (OR=1.36, 95% CI 1.17 to 1.58), as well as discrimination-related personal attacks in general (OR=1.46, 95% CI 1.25 to 1.71) and physical (OR=2.29, 95% CI 1.64 to 3.21) and verbal attacks specifically (OR=1.34, 95% CI 1.15 to 1.57), and current smoking status.
Although we were unable to locate any publications that studied non-racial discrimination as one overarching category, studies on specific types of non-racial discrimination, such as sexism, support a relationship between non-racial discrimination and smoking. Among Chinese female rural-to-urban migrant restaurant and hotel workers and sex workers in China, perceived discrimination (due to migrant status) was positively associated with the risk of being a current smoker.30 Restaurant and hotel workers, and sex workers, who reported one or more experiences of discrimination had 2.52 (95% CI 1.32 to 4.79) and 1.63 (95% CI 1.28 to 2.06) times the prevalence of smoking, respectively, compared with members of those groups who reported none. Borrell et al29 found that female Spanish natives in Spain who reported any experience of sexism in the past year had higher odds of being a current or ex-smoker (vs never smoker, OR=1.45, 95% CI 1.07 to 1.96). An analysis of female American college students also found a positive significant relationship between sexism and smoking status (p<0.05).28
Several potential explanations3 ,42 have been offered for the relationship between discrimination and smoking. First, smokers may actually experience higher levels of discrimination due to smoking-related stigma, especially Black women in South Africa, for whom smoking is considered especially taboo.38 Some smokers may falsely attribute this discrimination to personal characteristics other than smoking status. It is possible that our other discrimination category may have captured discrimination related to smoking status. Reports of other discrimination did differ significantly by smoking status for acute discrimination. Second, smoking may seem less risky to an individual whose experiences of discrimination have been life-threatening. If these experiences are perceived as a greater health risk than smoking, these individuals may not be motivated to quit.43 Third, discrimination is a stressor that impacts physiological reactivity to stress, such as cortisol production, and smoking also impacts perceived stress.
Because of different physiological responses, mild chronic stressors may have a greater impact on health outcomes than infrequent (acute) stressors, even when severe.1 ,34 We hypothesise that the differences in our results between acute and chronic discrimination are due to differences in the physiological effects of acute and chronic stress.44 Acute stressors lead to an initial increase in cortisol exposure, which returns to normal or slightly below normal shortly after the stressor is removed. In contrast, chronic stressors are associated with chronically high cortisol levels,45 which may impact nicotine sensitivity31 and psychological coping with stress,3 as well as decrease self-control in resisting unhealthy behaviours.42 Smoking is often used as a coping mechanism because nicotine decreases anxiety and generates a sense of well-being.46 In an analysis of American girls ages 11–19, Guthrie et al5 found that stress explained a significant proportion (p<0.05) of the relationship between discrimination and smoking. However, additional research is needed.
The primary limitation of this analysis is the cross-sectional nature of the data, which precludes us from making causal statements based on our results. One of the limitations of this analysis is its use of self-report. However, existing research suggests that self-reported smoking is a reasonably valid measure of cigarette consumption in population-level studies.47 As always, unmeasured confounding must be considered as a potential limitation. However, careful assessment of potential confounders decreases the likelihood of unmeasured confounding. Another potential limitation is our use of 2-category smoking outcome, which may have occluded differences between former smokers and current and non-smokers. However, we conducted additional analyses that coded smoking status into three categories (current, former, never) but did not find any significant differences between current and former or former and never smokers.
Because many of the respondents did not specify the type of acute or chronic non-racial discrimination (eg, sexism, ageism), another limitation is that there was insufficient statistical power to analyse each type of non-racial discrimination separately (eg, sexism, ageism) for the entire sample. However, cell sizes were large enough to run these analyses for Black Africans. The relationship between chronic discrimination and smoking was driven by racial discrimination (RR=1.20, 95% CI 1.08 to 1.33), age (RR=1.98, 95% CI 1.14 to 3.44) and other (RR=1.56, 95% CI 1.04 to 2.32). Acute racial and non-racial discrimination were not significant for this group. Unfortunately, despite these analyses, we were still unable to make comparisons across multiple races.
In addition, because of the small number of Whites who reported chronic (n=5, 1.6%) and acute (n=5, 3.1%) racial discrimination, these percentages should be interpreted with care given high relative SEs (50.0% and 29.4%, respectively). However, these values did not appear to impact our analyses; dropping Whites from analyses yielded similar results for chronic (racial: RR=1.39, 95% CI 1.08 to 1.79; non-racial: RR=1.53, 95% CI 1.20 to 1.93) and acute discrimination (racial: RR=1.05, 95% CI 0.66 to 1.65; non-racial: RR=1.04, 95% CI 0.77 to 1.41). A final limitation is that, although we used the labels ‘acute’ and ‘chronic’ discrimination to conform to previous publications on these questionnaires,32 ,34 we acknowledge that these terms are complex and that definitions may overlap.
This analysis also has several strengths. This is the first analysis of discrimination and smoking among South African adults. Because of the inclusion of both racial and non-racial discrimination, this analysis captured experiences of discrimination that may not have been examined had analyses only focused on either racial or non-racial discrimination. South Africans may tend to attribute racial discrimination to other factors, possibly due to a reluctance to discuss race in the postapartheid era.34 ,35 This analysis is also unique because it provides insight into the role of both chronic and acute racial and non-racial discrimination in smoking behaviour. Also, the use of SUDAAN and weighting allowed researchers to adjust for any bias created by sampling techniques and made the results generalisable to South African adults. Yet another strength of this analysis is the use of relative risks, which provided more conservative effect estimates than ORs, thereby reducing the risk of Type I error. The finding of a dose–response relationship between chronic discrimination and smoking intensity suggests that our findings are unlikely to be spurious.
Despite our extensive analysis of multiple types of discrimination, additional research is needed. Future research should use longitudinal data to examine the temporal ordering of racial and non-racial discrimination and smoking initiation, maintenance, and cessation in South Africa, as well as to examine stress as a potential mediator. Additional research should also attempt to generalise our results to populations outside of South Africa. We also recommend that, despite our non-significant results, the potential role of ethnic identity be considered in future analyses of discrimination and smoking.4 ,17 We also propose that future research tests social policies that may combat discrimination. In October of 2013, South African President Jacob Zuma called for a conference to combat discrimination suffered by apartheid victims, proposing measures such as increasing opportunities for Black businesses.48 In addition to these types of policies, social policies that promote racial integration also have the potential to increase positive interracial relations by reducing prejudice and increasing tolerance.49 Because of the link between socioeconomic status and race, housing and educational policies that provide opportunities for low-income individuals to live in middle income settings may decrease racial and socioeconomic segregation. These policies have been somewhat effective in the USA, where SES and race are closely related.
We also propose additional research on smoking cessation interventions in South Africa, which could be tailored by race and consider the potential role of stress and discrimination in smoking behaviour. A study of South African adolescents found that social and cognitive predictors of being a former smoker, such as attitudes towards smoking, mood, and smoking by friends and family, varied by race.38 However, because of a lack of research, effective targets for smoking cessation remain unclear. The link between stress and smoking50–52 suggests that interventions that combine smoking cessation with stress management techniques may be effective, especially for disadvantaged populations.53 However, we are unaware of any study that has used randomised controlled trials to examine the efficacy of a stress management component. Because of a potential link between alcohol dependence and smoking cessation in South Africa,54 another opportunity for intervention is to incorporate smoking cessation into alcohol dependence programmes.
The solution to these issues, if one exists, is extremely complex. However, given the potential link between discrimination and smoking, we cannot afford to ignore it.
What this paper adds
Smoking prevalence remains high among South African adults, and South Africa has a long history of discrimination.
This is the first paper to examine the relationship between racial and non-racial discrimination and smoking status among South African adults.
This study analysed a nationally representative sample of South African adults and found a significant positive relationship between discrimination and smoking status. Our findings suggest that the relationship between discrimination and smoking may vary by duration of the discrimination (chronic or acute) and the perceived reasons for these experiences (due to race or non-racial attributes).
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online tables
Contributors LMD, IK and CO: design, data analysis and manuscript preparation. DMW: design, data collection, data analysis and manuscript preparation.
Funding This work was supported by National Cancer Institute grant numbers 3R25CA057711-18S1 and R25CA-113710, and the National Center for Chronic Disease Prevention and Health Promotion grant number 1R01DP000097-01.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The South Africa Stress and Health study data are not publicly available but available on request from the authors of the original publication on this dataset.
Correction notice This article has been corrected since it was published Online First. The grant number ‘R01CA-61021’ was removed from the Funding section.