Content uploaded by Luke J Peppone
Author content
All content in this area was uploaded by Luke J Peppone on Mar 20, 2017
Content may be subject to copyright.
Perceived Discrimination, Psychological Distress, and Current
Smoking Status: Results From the Behavioral Risk Factor
Surveillance System Reactions to Race Module, 2004–2008
Jason Q. Purnell, PhD, MPH, Luke J. Peppone, PhD, MPH, Kassandra Alcaraz, MPH, Amy McQueen, PhD, Joseph J. Guido, MS,
Jennifer K. Carroll, MD, MPH, Enbal Shacham, PhD, MPE, and Gary R. Morrow, PhD, MS
The public health burden of cigarette smoking
is well established. Smoking is the leading
preventable cause of mortality in the United
States, accounting for 1 of every 5 deaths,
1
and
is associated with cardiovascular and respira-
tory diseases as well as several cancers and
adverse reproductive effects.
2
Although rates
of smoking have declined over the past several
decades, a substantial proportion of the pop-
ulation still smokes.
Data from the National Health Interview
Survey show that 21% of adults aged 18
years or older are current smokers.
3
Smoking
prevalence is substantially higher among
those of lower socioeconomic status (SES),
and because of the disproportionate repre-
sentation of minoritiesamongthepoor,low-
SES members of racial/ethnic minority
groups are particularly affected by smoking-
related health disparities.
4,5
When consid-
ered by race, the highest rate of adult smok-
ing is found among American Indians and
Alaska Natives (23%), followed by African
Americans and Whites (both 21%); the low-
est rate is that among Asian Americans
(10%).
3
The rate of smoking in the adult
Hispanic population is 15%.
3
In light of high
levels of smoking-related morbidity and
mortality and demonstrated health dispar-
ities, it is important to understand the factors
that promote smoking among members of
different racial/ethnic groups.
Several investigators have proposed that
perceived discrimination or unfair treatment
may account in part for disparities in health
behaviors and outcomes, particularly among
members of racial and ethnic minority
groups.
6-8
Much of this work has been de-
veloped within the framework of stress and
coping models, with perceived discrimination
treated as a chronic stressor that has potentially
negative effects on health.
8,9
Investigations linking discrimination to
smoking are included in this growing body of
research. For example, studies involving ado-
lescent,
10
young adult,
11, 12
and adult
13,14
African
Americans have shown that experiences of
racial discrimination, and the stress caused by
such experiences, are positively associated with
cigarette smoking. Similar findings were
reported in a study of Asian Americans expe-
riencing high levels of unfair treatment and
racial/ethnic discrimination, who were 2 to
almost 3 times more likely to be current
smokers than those who experienced no unfair
treatment or discrimination.
15
Landrine et al. reported that racial/ethnic
discrimination is associated with psychiatric
symptoms as well as smoking behavior, and
these relationships were stronger for racial and
ethnic minority groups.
14
Krieger et al. found
a trend toward an association between dis-
crimination and smoking among Blacks and
Latinos but not among Whites.
16
Guthrie et al.
demonstrated that a daily hassles measure
of stress mediated the relationship between
reported discrimination and smoking among
African American adolescent girls, offering
support for stress and coping explanations of
the relationship between discrimination and
smoking.
10
According to Williams et al., in their review
of the literature on discrimination and health,
more research exploring mechanisms underly-
ing the association between discrimination
and health is needed.
8
Much of the research
on discrimination and smoking has been
framed within stress and coping models, but
relatively little work has been done to test
whether psychological distress mediates this
relationship.
The theoretical framework that guided our
investigation is the Clark et al. biopsychosocial
model of the effects of perceived racism,
9
which proposes that psychological distress may
be one of the psychological responses to per-
ceived discrimination. Consistent with this
model, several studies have shown an associa-
tion between perceived discrimination and
psychological distress.
17---19
However, whereas
the Clark et al. model considers the relationship
between discrimination and health outcomes,
Objectives. We examined the association between perceived discrimination
and smoking status and whether psychological distress mediated this relation-
ship in a large, multiethnic sample.
Methods. We used 2004 through 2008 data from the Behavioral Risk Factor
Surveillance System Reactions to Race module to conduct multivariate logistic
regression analyses and tests of mediation examining associations between
perceived discrimination in health care and workplace settings, psychological
distress, and current smoking status.
Results. Regardless of race/ethnicity, perceived discrimination was associated
with increased odds of current smoking. Psychological distress was also
a significant mediator of the discrimination–smoking association.
Conclusions. Our results indicate that individuals who report discriminatory
treatment in multiple domains may be more likely to smoke, in part, because of
the psychological distress associated with such treatment. (Am J Public Health.
2012;102:844–851. doi:10.2105/AJPH.2012.300694)
RESEARCH AND PRACTICE
844 |Research and Practice |Peer Reviewed |Purnell et al. American Journal of Public Health |May 2012, Vol 102, No. 5
we were interested in the health behavior of
smoking, which is commonly thought of as
a means of coping with stress
20
and is also
associated with certain forms of psychological
distress and psychopathology.
21
Smoking may
be used as a form of self-medication to alleviate
stress and associated psychological distress.
Reports on the association between dis-
crimination and smoking have also tended to
focus on single racial or ethnic groups (e.g.,
African Americans, Asian Americans) or have
compared members of a single racial/ethnic
minority group with Whites, limiting our ability
to compare the relative effects of perceived
discrimination on smoking behavior between
groups. In addition, most previous research has
involved convenience samples and, in some
cases, relatively small sample sizes, limiting
generalizability.
We used 2004 through 2008 data from the
Behavioral Risk Factor Surveillance System
(BRFSS) to examine the association between
perceived discrimination and smoking status
and whether psychological distress mediates
this association in a large multistate, multieth-
nic random sample. Beginning in 2002, an
optional module, Reactions to Race, was added
to the BRFSS and adopted by several states
in an attempt tocapture data on perceived racial
discrimination and its effects in a population-
based sample.
On the basis of findings from previous
studies examining perceived discrimination
and smoking, we hypothesized that the odds of
smoking would be higher among those who
perceived that they were treated differently
because of their race (hypothesis 1). We also
hypothesized that the association between dis-
crimination and smoking would be stronger for
racial and ethnic minority groups (hypothesis
2) and that psychological distress, as measured
by reported mental health problems, would
mediate the association between discrimination
and smoking status (hypothesis 3).
METHODS
The BRFSS is administered by the Centers
for Disease Control and Prevention along with
US states and territories. It involves a random-
digit-dialing, multistage-cluster sample survey
designed to collect data on preventive health
practices and risk behaviors linked to chronic
diseases, injuries, and preventable infectious
diseases in the adult population (those aged 18
years or older).
22
The survey has 3 parts: the
core component, optional modules, and state-
added questions.
Although the Reactions to Race module was
added in 2002, we used BRFSS data from
2004 through 2008 because the 2002 and
2003 modules were not available from the
Centers for Disease Control and Prevention for
analysis. The 2008 data were the most recent
data available at the time of the study. The
module was administered by Arkansas, Colo-
rado, Delaware, the District of Columbia, Mis-
sissippi, Rhode Island, South Carolina, and
Wisconsin in 2004; Delaware and Ohio in
2005; Michigan and Wisconsin in 2006;
Rhode Island in 2007; and Nebraska and
Virginia in 2008. We pooled 2004 through
2008 data for these 11 states and the District
of Columbia. A total of 90 363 respondents
were administered the BRFSS survey in the
states and years selected, and 85 130 of these
individuals (94.2%) completed the Reactions
to Race module.
Measures
Perceived discrimination. Perceived racial
discrimination was assessed in 2 domains (i.e.,
while seeking health care and in the workplace)
as measured by the Reactions to Race module.
Perceived racial discrimination while seeking
health care was assessed with the following
item: “Within the past 12 months when seek-
ing health care, do you feel your experiences
were worse than, the same as, or better than
people of other races?”Perceived racial dis-
crimination at work was assessed with the item
“Within the past 12 months at work, do you
feel you were treated worse than, the same as,
or better than people of other races?”
In addition to better than, the same as, and
worse than people of other races, individuals
could respond that they felt they were treated
worse than some races and better than others.
For both discrimination items, these 4 re-
sponses were coded 1 to 4, respectively, with
lower scores indicating not being the perceived
target of discrimination and higher scores in-
dicating some degree of perceived discrimina-
tion. Individuals could also respond that they
encountered only people of the same race.
These respondents were excluded from
analyses related to the 2 perceived discrimi-
nation items.
Two yes---no items were used to assess the
effects associated with perceived discrimination.
The emotional effect of perceived discrimination
was assessed with the following item:
During the past 30 days, have you felt emotion-
ally upset, for example angry, sad, or frustrated,
as a result of how you were treated based on
your race?
The physical effect of perceived discrimina-
tion was assessed with the item:
Within the past 30 days, have you experienced
any physical symptoms, for example headache,
an upset stomach, tensing of your muscles, or
a pounding heart, as a result of how you were
treated based on your race?
Psychological distress. Psychological distress
was examined as a potential mediator of the
association between perceived racial discrimi-
nation and cigarette smoking. Psychological
distress was assessed with a single item: “Now
thinking about your mental health, which in-
cludes stress, depression, and problems with
emotions, for how many days during the past
30 days was your mental health not good?”
Potential scores on the item ranged from
0 (none of the past 30 days) to 30 (all of the
past 30 days).
Current smoking. The primary outcome
variable was current smoking status, which
was assessed with the following items:
“Have you smoked at least 100 cigarettes in
your entire life?”and “Do you now smoke
cigarettes every day, some days, or not at
all?”Respondents who reported having
smoked at least 100 cigarettes and smoked
every day and some days were coded as
current smokers. Respondents who had not
smoked at least 100 cigarettes and those
who responded not at all to the question on
current smoking were coded as non-
smokers.
Potential confounders. We identified a num-
ber of variables from the literature that could
potentially confound the association between
perceived discrimination and cigarette smok-
ing. Potential confounders used in our analyses
included age, gender, self-identified race, mar-
ital status, income, education, health insurance
coverage, self-rated general health status, and
state of residence.
RESEARCH AND PRACTICE
May 2012, Vol 102, No. 5 |American Journal of Public Health Purnell et al. |Peer Reviewed |Research and Practice |845
Age in years was categorized into 6 groups:
18 to 24, 25 to 34, 35 to 44, 45 to 54, 55
to 64, and 65 or older. Self-identified racial
categories included non-Hispanic White, non-
Hispanic Black, Hispanic or Latino, and other
(Asian, Native Hawaiian, other PacificIs-
lander, American Indian, Alaska Native, or
another group). Marital status was categorized
as married or coupled; divorced, widowed, or
separated; and never married. Yearly income
was categorized into 4 groups: less than $20
000, $20 000 to $34 999, $35 000 to $74
999, and $75 000 or more. Educational
attainment was categorized as less than high
school, high school or equivalent, some col-
lege, and college.
A binary variable created for health insur-
ance coverage was based on whether partici-
pants were enrolled in a health care plan
(including Medicare and Medicaid). Self-rated
health status was coded into 5 categories:
excellent, very good, good, fair, or poor. In the
case of all variables, don’t know and not sure
responses were categorized as missing data, as
were responses indicating refusal.
Statistical Analysis
We used SAS version 9.2 SURVEY proce-
duresinallofouranalysestoaccountforthe
complex sampling design of the BRFSS and to
obtain unbiased standard errors.
23
Sample sizes
presented in the tables are not weighted; how-
ever, all proportions, means, odds ratios (ORs),
and 95% confidence intervals (CIs) are weighted
via the final weight provided in the data set to
represent the corresponding population. We
used v
2
statistics to examine sociodemographic
characteristics by current smoking status. The
statistical significance level was set at P<.05.
We used multivariate logistic regression models
to determine associations (ORs) between per-
ceived discrimination and current smoking status
after control for all potential confounders (hy-
pothesis 1). We created separate models for health
care discrimination, workplace discrimination,
emotional reactions, and physical reactions from
the Reaction to Race module; models were
selected in a stepwise, backward manner. Initially,
all model terms (predictor variable and covariates)
were added to the logistic model. Covariates
that were not significant (P> .05) in the model
were removed 1 at a time, beginning with the
covariate with the greatest Pvalue. We report ORs
TABLE 1—Sociodemographic Characteristics, by Smoking Status, Among Adults Who
Completed the Behavioral Risk Factor Surveillance System Reactions to Race Module:
United States, 2004–2008
Nonsmokers, No. (Weighted %) Smokers, No. (Weighted %)
Total 68 626 (79.0) 16 504 (21.0)
Age, y
18–24 3092 (12.4) 1106 (16.2)
25–34 8046 (16.2) 2674 (21.0)
35–44 10 989 (18.6) 3590 (22.6)
45–54 13 500 (18.7) 4255 (21.5)
55–64 12 935 (14.2) 2884 (11.7)
‡65 19 545 (19.5) 1930 (6.7)
Missing 519 (0.3) 584 (0.2)
Gender
Male 25 364 (47.3) 6886 (52.1)
Female 43 262 (52.7) 9618 (47.9)
Race/ethnicity
Non-Hispanic White 56 442 (81.6) 12 973 (79.4)
Non-Hispanic Black 6974 (8.8) 2062 (10.7)
Hispanic 2412 (4.1) 583 (3.9)
Other 2208 (4.7) 763 (5.4)
Missing 590 (0.8) 123 (0.6)
Marital status
Married/coupled 40 912 (67.2) 7782 (54.6)
Divorced/separated/widowed 18 542 (15.3) 5591 (21.3)
Never married 8974 (17.2) 3086 (23.9)
Missing 225 (0.3) 45 (0.2)
Annual income, $
< 20 000 9977 (11.0) 3802 (18.8)
20 000–34 999 13 162 (18.2) 4130 (25.6)
35 000–74 999 20 991 (31.4) 4866 (31.0)
‡75 000 15 024 (26.3) 2037 (13.9)
Missing 9472 (13.1) 1669 (10.7)
Educational attainment
< high school 5630 (7.1) 2298 (14.4)
High school or equivalent 20 646 (29.2) 6517 (42.1)
Some college 17 416 (26.5) 4718 (28.1)
College 24 819 (37.1) 2953 (15.2)
Missing 115 (0.2) 18 (0.1)
Health insurance coverage
Yes 62 809 (90.0) 13 268 (76.5)
No 5645 (9.6) 3192 (23.1)
Missing 170 (0.5) 44 (0.4)
Continued
RESEARCH AND PRACTICE
846 |Research and Practice |Peer Reviewed |Purnell et al. American Journal of Public Health |May 2012, Vol 102, No. 5
and 95% CIs adjusted for the covariates found to
be significant in the final model.
We also tested models stratified by race/
ethnicity (hypothesis 2). Full models are
presented in Appendix A (available as
a supplement to the online version of this
article at http://www.ajph.org). We examined
similar models with the same covariates and
occasional and everyday smoking as the out-
comes.
Although the data were cross sectional, we
chose to conduct mediation analyses to pro-
vide conceptual support for psychological
distress as a mediator of the association
between perceived discrimination and smok-
ing behavior (hypothesis 3). As suggested by
Preacher and Hayes,
24
we initially tested 2
regression models to determine the coeffi-
cients and standard errors for the association
between the independent variable (i.e., per-
ceived discrimination) and the mediator (i.e.,
psychological distress) and the association of
the mediator and dependent variable (i.e.,
smoking) with the independent variable in-
cluded in the model.
We then calculated bias-corrected and ac-
celerated bootstrap confidence intervals of the
sampling distribution of the indirect effect to
assess statistically significant mediation.
25
The
bootstrapping procedure allows for a 95% CI
to be constructed around a point estimate of
the indirect effect and is not constrained by the
assumption that the sampling distribution of
the indirect effect is normal. A 95% CI that
does not include zero indicates statistically
significant mediation.
Following the advice of Preacher (written
communication, February 2010), we used 2
million bootstrap resamples. Although formal
TABLE 1—Continued
General health status
Excellent 14 050 (22.0) 2042 (13.9)
Very good 23 567 (37.1) 5008 (31.5)
Good 19 840 (27.7) 5670 (35.2)
Fair 7793 (9.4) 2530 (13.8)
Poor 3127 (3.4) 1197 (5.3)
Missing 248 (0.3) 56 (0.3)
State or city of residence
Arkansas 2997 (3.9) 880 (4.8)
Colorado 4701 (7.1) 1077 (6.6)
Delaware 6372 (2.6) 1745 (2.8)
District of Columbia 2242 (0.9) 514 (0.9)
Michigan 4451 (15.7) 1093 (16.9)
Mississippi 3942 (4.1) 1119 (5.0)
Nebraska 13 305 (2.8) 2460 (2.4)
Ohio 5392 (16.9) 1535 (18.2)
Rhode Island 6174 (3.2) 1359 (2.8)
South Carolina 4230 (4.9) 1189 (5.9)
Virginia 4346 (12.9) 814 (9.5)
Wisconsin 10 474 (25.0) 2719 (24.2)
Note. The sample size was n = 85 130. All group differences were significant at the P< .001 le vel.
TABLE 2—Experiences of Discrimination, by Race/Ethnicity: Behavioral Risk Factor Surveillance System
Reactions to Race Module: United States, 2004–2008
Non-Hispanic White, No. (Weighted %) Non-Hispanic Black, No. (Weighted %) Hispanic, No. (Weighted %) Other, No. (Weighted %)
Health care
Treated better 9228 (13.6) 611 (7.4) 310 (10.4) 313 (10.1)
Treated the same 48 624 (83.8) 6322 (78.8) 2189 (81.5) 2066 (82.7)
Treated worse than some but better than others 338 (1.1) 280 (4.9) 35 (2.2) 61 (3.4)
Treated worse 1011 (1.5) 746 (8.9) 153 (5.9) 118 (3.8)
Workplace
Treated better 1830 (4.7) 161 (4.4) 95 (6.1) 92 (4.7)
Treated the same 35 657 (91.5) 3758 (72.7) 1483 (81.8) 1367 (83.9)
treated worse than some but better than others 294 (1.2) 221 (6.0) 42 (2.8) 50 (3.3)
Treated worse 1073 (2.7) 855 (16.9) 200 (9.3) 129 (8.1)
Emotional
Yes 2945 (4.5) 2197 (16.3) 509 (12.8) 496 (10.1)
No 65 899 (95.5) 6652 (83.7) 2462 (87.2) 2417 (89.9)
Physical
Yes 2132 (5.7) 1369 (27.8) 338 (19.3) 291 (21.6)
No 66 653 (94.3) 7485 (72.2) 2630 (80.7) 2618 (78.4)
Note. All group differences were significant at the P< .001 level.
RESEARCH AND PRACTICE
May 2012, Vol 102, No. 5 |American Journal of Public Health Purnell et al. |Peer Reviewed |Research and Practice |847
effect size estimates for mediation involving
dichotomous outcomes have not been devel-
oped, we also followed the advice of Preacher
(written communication, October 2011) in
attempting to provide at least some information
on the practical significance of mediation re-
sults. We report the ratio of the indirect effect
to the total effect, or the mediation ratio (P
M
),
ð1ÞPM¼cc9
c
where cis the total effect of the independent
variable (perceived discrimination) on the de-
pendent variable (current smoking) and c9is
the effect of the independent variable on the
dependent variable after control for the medi-
ator (psychological distress).
26
RESULTS
Table 1 provides a sociodemographic de-
scription of the study sample by smoking status.
In comparison with nonsmokers, smokers were
more likely to be younger, male, unmarried, and
uninsured; to have lower income and educa-
tional attainment; and to have relatively poorer
self-reported health status. Overall smoking rates
were 24.5% among non-Hispanic Blacks,
23.3% among those in the “other”racial/ethnic
category, and 20.5% among both non-Hispanic
Whites and Hispanics.
Table 2 presents data on perceived dis-
crimination experiences by race/ethnicity. The
majority of respondents across racial/ethnic
groups reported that they were treated the
same as others. Non-Hispanic Blacks were most
likely to report perceived discrimination in
both the health care and workplace domains,
followed by Hispanics, those in the “other”
racial/ethnic category, and, finally, non-
Hispanic Whites. The same pattern was
observed for adverse physical responses to
perceived discrimination, with the exception
that those in the “other”racial/ethnic category
reported higher levels of negative emotional
responses to discrimination than did Hispanics
and non-Hispanic Whites.
Table 3 presents results from the logistic
regression models testing the association of
experiences of discrimination with current
smoking status. Because all covariates were
significant, all were retained in the final models.
Consistent with our first hypothesis, the odds of
current smoking were higher among individ-
uals who perceived that they were treated
differently because of their race. In comparison
with those who received treatment similar to
people of other races, those who reported being
treated worse than people of other races in
health care settings were 18% more likely to be
current smokers (adjusted OR = 1.18; 95%
CI = 1.09, 1.26), and those who reported
worse treatment in the workplace were 13%
more likely to smoke (adjusted OR = 1.13;
95% CI = 1.03, 1.23). Emotional and physical
responses to discrimination were not associ-
ated with current smoking after adjustment for
relevant covariates. Contrary to our second
hypothesis, the patterns of association re-
mained the same when interactions by race/
ethnicity were considered.
In follow-up analyses (Table 4) designed to
further characterize the association between
race-based discrimination and current smok-
ing, we examined ORs and CIs for each of the
responses independently, with the reference
category being responses indicating that the
treatment received was the same as that for
people of other races. Findings differed
according to the domain in which discrimina-
tion occurred. Individuals who perceived that
they were treated better than other groups in
health care settings were 21% less likely to be
current smokers (adjusted OR = 0.79; 95%
CI = 0.70, 0.90), whereas those who perceived
that they were treated worse than other groups
in the workplace were 42% more likely to
smoke (adjusted OR = 1.42; 95% CI = 1.17,
1.73).
As expected, everyday smokers were more
likely than occasional smokers, and occasional
smokers were in turn more likely than non-
smokers, to report being the targets of per-
ceived discrimination in both health care set-
tings (4.3%, 3.3%, and 2.1%, respectively;
weighted) and the workplace (6.5%, 5.8%, and
4.0%, respectively; weighted). Smokers were
more likely than nonsmokers to report emo-
tional and physical symptoms in response to
perceived discrimination, although occasional
smokers were more likely than everyday
smokers to report both emotional (13.5% vs
11.3%) and physical (8.5% vs 8.2%) symptoms.
The results of logistic regression models with
occasional and everyday smoking as the out-
comes are presented in Appendix B (available
as a supplement to the online version of this
article at http://www.ajph.org). In a logistic
regression analysis with occasional smoking as
the outcome, neither perceived discrimination
nor its effects were associated with occasional
smoking. In a similar analysis with everyday
smoking as the outcome, odds of everyday
smoking were increased among individuals
who reported being treated worse than mem-
bers of other races in health care settings
(adjusted OR = 1.21; 95% CI = 1.12, 1.31)
as well as the workplace (adjusted OR = 1.15;
95% CI = 1.04, 1.27).
Follow-up analyses by item response
showed that individuals who perceived better
TABLE 3—Crude and Adjusted Odds Ratios for Current Smoking, by Experiences of
Discrimination: Behavioral Risk Factor Surveillance System Reactions to Race Module:
United States, 2004–2008
Measure Crude OR (95% CI) SE Adjusted OR (95% CI) SE
Perceived discrimination
Health care 1.52** (1.43, 1.62) 0.03 1.18** (1.09, 1.26) 0.03
Workplace 1.23** (1.13, 1.33) 0.04 1.13** (1.03, 1.23) 0.05
Emotional effects of discrimination
No emotional effects (Ref) 1.00 1.00
Emotional effects 1.49 (1.34, 1.67) 0.06 1.09 (0.96, 1.24) 0.07
Physical effects of discrimination
No physical effects (Ref) 1.00 1.00
Physical effects 1.51 (1.33, 1.71) 0.06 0.99 (0.85, 1.15) 0.08
Note. CI = confidence interval; OR = odds ratio. All analyses were adjusted for age, gender, race/ethnicity, marital status,
annual income, education, health insurance, general health status, and state of residence.
**P< .001.
RESEARCH AND PRACTICE
848 |Research and Practice |Peer Reviewed |Purnell et al. American Journal of Public Health |May 2012, Vol 102, No. 5
treatment than people of other races in health
care settings were 28% less likely to be
everyday smokers (adjusted OR = 0.72; 95%
CI = 0.62, 0.82), and those who perceived
worse treatment in this setting were 24% more
likely to be everyday smokers (adjusted
OR = 1.24; 95% CI = 1.00, 1.53). In the
workplace, only individuals who perceived
worse treatment than people of other races were
significantly more likely to be everyday smokers
(adjusted OR = 1.46; 95% CI = 1.17, 1.82).
As hypothesized, the association between
perceived discrimination and smoking status
was mediated by psychological distress.
Bias-corrected and accelerated bootstrap (BCa)
point estimates were 0.05 (95% CI = 0.04,
0.06; P
M
= 0.08) for the indirect effect of
perceived discrimination in health care settings
and 0.04 (95% CI = 0.02, 0.05; P
M
= 0.21) for
perceived discrimination in the workplace.
Because the confidence intervals for these
point estimates did not include zero, we con-
clude that there is statistically significant me-
diation of the relationship between perceived
discrimination in both domains and current
smoking status through the effect of psycho-
logical distress.
Our crude measure of effect size suggests
that 8% of the total effect of perceived dis-
crimination in health care settings on smoking
status was accounted for by its indirect effect
through psychological distress and that 21% of
the total effect of perceived discrimination in
the workplace on smoking status was accounted
for by its indirect effect through psychological
distress. We found a similar pattern in mediation
models for occasional smoking with respect to
health care settings (BCa point estimate = 0.03;
95% CI = 0.02, 0.03; P
M
=0.17)and the
workplace (BCa point estimate = 0.02; 95%
CI = 0.01, 0.03; P
M
= 0.08), as well as a similar
pattern for everyday smoking with respect to
health care settings (BCa point estimate = 0.05;
95% CI = 0.04, 0.06; P
M
=0.11)and the
workplace (BCa point estimate = 0.04; 95%
CI = 0.03, 0.05; P
M
=0.25).
DISCUSSION
Our findings from this large multistate,
multiethnic study showed that regardless of
racial/ethnic background, individuals who
reported being treated poorly because of
their race were modestly more likely to be
current smokers. Workplace discrimination
also emerged as potentially more relevant to
current smoking than discrimination (i.e.,
worse treatment) in health care settings. In
addition, our results revealed that the asso-
ciation between perceived discrimination and
smoking status was mediated by psychologi-
cal distress, accounting for between 8% and
21% of this association. The positive rela-
tionship between perceived discrimination
and smoking status is consistent with pre-
vious studies.
10---15
To our knowledge, however, this study is the
first to report that the association between
discrimination and smoking holds equally
across diverse racial and ethnic groups. Our
findings contrast with at least 2 studies showing
that the association between discrimination and
smoking was stronger for members of racial/
ethnic minority groups than for Whites
12,14
and
another reporting a trend in this direction.
16
In the Coronary Artery Risk Development in
Adults (CARDIA) Study, discrimination was
positively, although not significantly, associated
with tobacco use among Whites; it was also
associated with use of other substances (i.e.,
marijuana and cocaine).
12
The use of multidi-
mensional, multi-item scales to measure dis-
crimination in the CARDIA Study
12
and the
study by Landrine et al.
14
may account for
differences between those studies and ours in
the strength of the association between dis-
crimination and smoking in different racial and
ethnic groups. It is also possible that, regardless
of background, individuals who perceive that
they are being treated unfairly because of their
race in multiple domains of life are more likely
to be smokers.
Still, it should be noted that members of
racial and ethnic minority groups were more
likely to report perceived discrimination both
in our sample and in previous studies,
12,16
suggesting that discrimination may be a more
important contributor to the occurrence of
smoking in these populations. Although per-
ceptions of race-related discrimination are rel-
atively rare in the non-Hispanic White popu-
lation, assessing whether other types of
discrimination (e.g., discrimination based on
age, gender, socioeconomic position, disability,
or sexual orientation) are also associated with
an increased risk of cigarette smoking may be
warranted in light of our findings. The litera-
ture’s focus on differences in the impact of
discrimination on health as a function of SES,
gender, and other sociodemographic charac-
teristics
7,8
also suggests that the interactions
between these factors and discrimination
should be examined in future research on
smoking and other health behaviors.
Our study also adds to the existing literature
by formally examining the role of psychological
distress as an important mediator linking dis-
crimination with smoking. Our results suggest
that the adverse psychological impact of
TABLE 4—Crude and Adjusted Odds Ratios for Current Smoking, by Perceived
Discrimination: Behavioral Risk Factor Surveillance System Reactions to Race Module:
United States, 2004–2008
Measure Crude OR (95% CI) SE Adjusted OR (95% CI) SE
Health care discrimination
Treated the same (Ref) 1.00 1.00
Treated better 0.57*** (0.51, 0.63) 0.06 0.79** (0.70, 0.90) 0.06
Treated worse than some but better than others 1.62** (1.22, 2.17) 0.15 1.37 (0.98, 1.91) 0.17
Treated worse 1.86*** (1.56, 2.21) 0.09 1.20 (0.98, 1.46) 0.10
Workplace discrimination
Treated the same (Ref) 1.00 1.00
Treated better 1.11 (0.93, 1.33) 0.19 1.16 (0.96, 1.41) 0.21
Treated worse than some but better than others 1.57* (1.07, 1.33) 0.09 1.30 (0.86, 1.96) 0.10
Treated worse 1.66 (1.41, 1.96) 0.08 1.42*** (1.17, 1.73) 0.10
Note. CI = confidence interval; OR = odds ratio. All analyses were adjusted for age, gender, race/ethnicity, marital status,
annual income, education, health insurance, general health status, and state or residence.
*P< .05; **P< .01; ***P< .001.
RESEARCH AND PRACTICE
May 2012, Vol 102, No. 5 |American Journal of Public Health Purnell et al. |Peer Reviewed |Research and Practice |849
discrimination in various domains may lead
individuals to smoke. To our knowledge, only
1 previous study examined mediation of the
association between discrimination and smok-
ing in a relatively small sample of African
American adolescents.
10
Guthrie et al. found
that daily hassles mediated the relationship
between racial discrimination and cigarette
smoking in their sample.
The findings from our study, which involved
a large, diverse sample of adults representing
several states, provide further evidence for
the role that psychological distress plays in
smoking among those who experience discrim-
ination. Our findings are also consistent with
a large body of research that links perceived
discrimination to psychological distress
7,17,18
and suggest that at least 1 of the deleterious
effects of discrimination on health could be
through the psychological distress it engenders,
along with the negative health behaviors that
may be recruited to cope with such distress.
Another important implication of our find-
ings is the different results observed in follow-
up analyses of the discrimination---smoking
association. When item responses were con-
sidered individually instead of as an index of
discrimination (i.e., 1 = better treatment, 4 =
worse treatment), perceptions of worse treat-
ment than people of other races were associ-
ated after adjustment with current smoking in
the workplace but not in health care settings.
Better treatment in health care settings was
protective in terms of current smoking. This
finding makes intuitive sense because work-
place discrimination is likely more chronic and
more salient in that it is closely tied to one’s
economic well-being. This result should be
interpreted with caution, however, given the
nearly identical adjusted odds ratios for health
care and workplace models when perceived
discrimination was treated as continuous as
well as the overlap in confidence intervals of
the coefficient for worse treatment in both
continuous and categorical treatments of the
perceived discrimination variable. The protec-
tive nature of preferential treatment in health
care settings is an intriguing finding worthy of
further investigation.
There were also noteworthy differences
when we considered occasional and everyday
smoking as subcategories of current smoking.
Individuals who perceived worse treatment
than people of other races in health care
settings were also more likely to be everyday
smokers, suggesting an interesting interaction
between perceptions of discrimination, the
setting in which discrimination is perceived,
and frequency of smoking. The stronger me-
diation by psychological distress in health care
settings among occasional smokers may be
related to the episodic nature of perceived
discrimination in these settings, which may lead
to less regular smoking, whereas everyday
smokers who report discrimination in the
workplace may be more likely to experience
chronic psychological distress. This speculative
hypothesis needs to be tested in studies de-
signed for this purpose.
Limitations
Of course, our use of cross-sectional data pre-
cludes us from being able to draw any conclu-
sions about causality in this study. As a check on
our mediation models, we did run analyses with
the predictor and mediator reversed (i.e., per-
ceived discrimination mediating psychological
distress and smoking) and found near-zero esti-
mates of indirect effects. This leads us to believe
that our hypothesized model is at least preferable
to a model in which perceived discrimination is
the mediator. Additional longitudinal studies are
needed to confirm our findings.
The limitations inherent in our estimates of
the effect size of mediation should also be
noted, given that methods of effect size esti-
mation for dichotomous outcomes in mediation
analyses have not been developed. The use
of the mediation ratio is limited in that it does
not perform strictly as a proportion in all
instances (e.g., P
M
may be negative or > 1
depending on the relation of cand c9, although
neither was the case for our data) and it is
unstable in small samples (also not the case in
the present study).
26
Therefore, we advise
caution in interpreting effect sizes within the
context of our mediation analyses.
Our study was also limited by the restricted
availability of measures of perceived discrim-
ination and psychological distress within the
BRFSS and the Reactions to Race module. Our
results may underestimate the association
between perceived discrimination and smok-
ing status, and multidimensional, multi-item
measures of the constructs under study may
have yielded different results with respect
to the strength of the association by race/
ethnicity.
Finally, the inconsistent adoption of the op-
tional Reactions to Race module among the states
during the study period limits the generalizability
of our findings. Characteristics of the states in
which the module was used may differ in
significant ways from those of states that elected
not to use it. Therefore, our results cannot be
viewed as representing the nation as a whole.
Conclusions
Despite this study’s limitations, it involved
the largest and most diverse sample among
the studies published to date reporting associ-
ations between perceived discrimination and
cigarette smoking. It highlights a potentially
high-risk group of individuals who report feel-
ing unfairly treated because of their race and
who may be smoking as a means of coping with
the psychological distress associated with dis-
crimination. Identifying these individuals for
targeted smoking cessation interventions may
improve cessation rates. Our findings also
suggest that alternative forms of coping with
discrimination may be a fruitful area of dis-
cussion in counseling interventions designed
to help individuals quit smoking.
In future research, longitudinal designs
should be used to confirm that psychological
distress mediates the association between per-
ceived discrimination and smoking, and mea-
sures of these constructs should represent their
complex, multidimensional nature. Other forms
of discrimination and additional health behav-
iors should also be considered. j
About the Authors
Jason Q. Purnell and Kassandra Alcaraz are with the
George Warren Brown School of Social Work, Washington
University in St. Louis, St. Louis, MO. Luke J. Peppone and
Gary R. Morrow are with the Department of Radiation
Oncology, University of Rochester School of Medicine &
Dentistry, Rochester, NY. Amy McQueen is with the De-
partment of Medicine, School of Medicine, Washington
University in St. Louis. Joseph J. Guido is with the De-
partment of Community & Preventive Medicine, University
of Rochester School of Medicine & Dentistry. Jennifer K.
Carroll is with the Department of Family Medicine, Uni-
versity of Rochester School of Medicine & Dentistry. Enbal
Shacham is with the Department of Behavioral Science &
Health Education, Saint Louis University School of Public
Health.
Correspondence should be sent to Jason Q. Purnell, PhD,
MPH, Washington University in St. Louis, One Brookings
Drive, Campus Box 1196, St. Louis, MO 63130 (e-mail:
RESEARCH AND PRACTICE
850 |Research and Practice |Peer Reviewed |Purnell et al. American Journal of Public Health |May 2012, Vol 102, No. 5
jpurnell@gwbmail.wustl.edu). Reprints can be ordered at
http://www.ajph.org by clicking the “Reprints”link.
This article was accepted January 6, 2012.
Contributors
J. Q. Purnell led the conceptualization, design, and
drafting of the article. L. J. Peppone conducted analyses
and contributed to drafting the article. K. Alcaraz led the
analyses and contributed to drafting the article. A.
McQueen assisted with and reviewed mediation analyses
and contributed to interpretation. J. J. Guido extracted
and managed the data, conducted analyses, and con-
tributed to drafting the article. J. K. Carroll and G. R.
Morrow reviewed analyses and contributed to the de-
velopment of the article.
Acknowledgments
J. Q. Purnelland L. J. Peppone wishto acknowledge support
from a National Cancer Institute cancer prevention and
control research training grant (R25CA102618).
Human Participant Protection
Because this study’s analyses involved secondary data, no
protocol approval was needed.
References
1. Danaei G, Ding EL, Mozaffarian D, et al. The
preventable causes of death in the United States: com-
parative risk assessment of dietary, lifestyle, and meta-
bolic risk factors. PLoS Med. 2009;6(4):e1000058.
2. The Health Consequences of Smoking: A Report of the
Surgeon General. Atlanta, GA: Centers for Disease Con-
trol and Prevention; 2004.
3. Pleis JR, Lucas JW, Ward BW. Summary Health
Statistics for U.S. Adults: National Health Interview Survey,
2008. Hyattsville, MD: National Center for Health
Statistics; 2009.
4. Centers for Disease Control and Prevention. Vital
signs: current cigarette smoking among adults aged
>or=18 years—United States, 2009. MMWR Morb
Mortal Wkly Rep. 2010;59(35):1135---1140.
5. Tobacco Use Among U.S. Racial/Ethnic Minority
Groups. Atlanta, GA: Centers for Disease Control and
Prevention; 1998.
6. Krieger N. Embodying inequality: a review of
concepts, measures, and methods for studying health
consequences of discrimination. Int J Health Serv.
1999;29(2):295---352.
7. Williams DR, Mohammed SA. Discrimination and
racial disparities in health: evidence and needed research.
J Behav Med. 2009;32(1):20---47.
8. Williams DR, Neighbors HW, Jackson JS. Racial/
ethnic discrimination and health: findings from commu-
nity studies. Am J Public Health. 2003;93(2):200---208.
9. Clark R, Anderson NB, Clark VR, Williams DR.
Racism as a stressor for African Americans: a biopsy-
chosocial model. Am Psychol. 1999;54(10):805---816.
10. Guthrie BJ, Young AM, Williams DR, Boyd CJ,
Kintner EK. African American girls’smoking habits and
day-to-day experiences with racial discrimination. Nurs
Res. 2002;51(3):183---190.
11. Bennett GG, Wolin KY, Robinson EL, Fowler S,
Edwards CL. Perceived racial/ethnic harassment and
tobacco use among African American young adults. Am
J Public Health. 2005;95(2):238---240.
12. Borrell LN, Jacobs DR Jr, Williams DR, Pletcher MJ,
Houston TK, Kiefe CI. Self-reported racial discrimination
and substance use in the Coronary Artery Risk De-
velopment in Adults Study. Am J Epidemiol. 2007;166
(9):1068---1079.
13. Landrine H, Klonoff EA. Racial discrimination and
cigarette smoking among blacks: findings from two
studies. Ethn Dis. 2000;10(2):195---202.
14. Landrine H, Klonoff EA, Corral I, Fernandez S,
Roesch S. Conceptualizing and measuring ethnic dis-
crimination in health research. J Behav Med. 2006;29
(1):79---94.
15. Chae DH, Takeuchi DT, Barbeau EM, Bennett GG,
Lindsey J, Krieger N. Unfair treatment, racial/ethnic dis-
crimination, ethnic identification, and smoking among
Asian Americans in the National Latino and Asian Amer-
ican Study. Am J Public Health. 2008;98(3):485---492.
16. Krieger N, Smith K, Naishadham D, Hartman C,
Barbeau EM. Experiences of discrimination: validity and
reliability of a self-report measure for population health
research on racism and health. Soc Sci Med. 2005;61
(7):1576---1596.
17. Kessler RC, Mickelson KD, Williams DR. The
prevalence, distribution, and mental health correlates of
perceived discrimination in the United States. J Health Soc
Behav. 1999;40(3):208---230.
18. Williams DR, Williams DR, Yu Y, Jackson JS,
Anderson NB. Racial differences in physical and mental
health: socio-economic status, stress and discrimination.
J Health Psychol. 1997;2(3):335---351.
19. Williams DR, Williams-Morris R. Racism and mental
health: the African American experience. Ethn Health.
2000;5(3---4):243---268.
20. Parrott AC. Does cigarette smoking cause stress? Am
Psychol. 1999;54(10):817---820.
21. Kalman D, Morissette SB, George TP. Co-morbidity
of smoking in patients with psychiatric and substance use
disorders. Am J Addict. 2005;14(2):106---123.
22. Centers for Disease Control and Prevention. Behav-
ioral Risk Factor Surveillance System annual survey data.
Available at: http://www.cdc.gov/brfss/technical_infodata/
surveydata.htm. Accessed March 1, 2012.
23. SAS/AF 9.2 Procedure Guide. Cary, NC: SAS In-
stitute Inc; 2008.
24. Preacher KJ, Hayes AF. SPSS and SAS procedures
for estimating indirect effects in simple mediation models.
Behav Res Methods Instrum Comput. 2004;36(4):717---
731.
25. Selig JP, Preacher KJ. Monte Carlo method for
assessing mediation: an interactive tool for creating
confidence intervals for indirect effects. Available at:
http://www.quantpsy.org/medmc/medmc.htm.
Accessed March 1, 2012.
26. Preacher KJ, Kelley K. Effect size measures for
mediation models: quantitative strategies for communi-
cating indirect effects. Psychol Methods. 2011;16(2):
93---115.
RESEARCH AND PRACTICE
May 2012, Vol 102, No. 5 |American Journal of Public Health Purnell et al. |Peer Reviewed |Research and Practice |851
