Understanding contributors to racial disparities in blood pressure control.

Abstract

Racial disparities in blood pressure (BP) control are well documented but poorly understood; prior studies have only included a limited range of potential explanatory factors. We examined a comprehensive set of putative factors related to blood pressure control, including patient clinical and sociodemographic characteristics, beliefs about BP and BP medications, medication adherence, and experiences of discrimination, to determine if the impact of race on BP control remains after accounting for such factors.
We recruited 806 white and black patients with hypertension from an urban safety-net hospital. From a questionnaire administered to patients after their clinic visits, electronic medical record and BP data, we assessed an array of patient factors. We then examined the association of patient factors with BP control by modeling it as a function of the covariates using random-effects logistic regression. Blacks indicated worse medication adherence, more discrimination, and more concerns about high BP and BP medications, compared with whites. After accounting for all factors, race was no longer a significant predictor of BP control.
Results suggest that equalizing patients' health beliefs, medication adherence, and experiences with care could ameliorate disparities in BP control. Additional attention must focus on the factors associated with race to identify, and ultimately intervene on, the causes of racial disparities in BP outcomes.

Full text

Page 1

1941-7713
American Heart Association. All rights reserved. Print ISSN: 1941-7705. Online ISSN: 2010 Copyright ©
Greenville Avenue, Dallas, TX 72514
Circulation: Cardiovascular Quality and Outcomes is published by the American Heart Association. 7272
DOI: 10.1161/CIRCOUTCOMES.109.860841
2010;
2010;3;173-180; originally published online January 19,
Circ Cardiovasc Qual Outcomes
Berlowitz
Nancy R. Kressin, Michelle B. Orner, Meredith Manze, Mark E. Glickman and Dan
Understanding Contributors to Racial Disparities in Blood Pressure Control



http://circoutcomes.ahajournals.org/content/3/2/173.full
on the World Wide Web at:
The online version of this article, along with updated information and services, is located
http://www.lww.com/reprints
Reprints: Information about reprints can be found online at

journalpermissions@lww.com
410-528-8550. E-mail:
Health, 351 West Camden Street, Baltimore, MD 21201-2436. Phone: 410-528-4050. Fax:
Permissions: Permissions & Rights Desk, Lippincott Williams & Wilkins, a division of Wolters Kluwer

http://circoutcomes.ahajournals.org/site/subscriptions/
online at
Subscriptions: Information about subscribing to Circulation: Cardiovascular Quality and Outcomes is
by guest on May 18, 2011circoutcomes.ahajournals.orgDownloaded from

Page 2

Understanding Contributors to Racial Disparities in Blood
Pressure Control
Nancy R. Kressin, PhD; Michelle B. Orner, MPH; Meredith Manze, MPH;
Mark E. Glickman, PhD; Dan Berlowitz, MD, MPH
Background—Racial disparities in blood pressure (BP) control are well documented but poorly understood; prior studies
have only included a limited range of potential explanatory factors. We examined a comprehensive set of putative
factors related to blood pressure control, including patient clinical and sociodemographic characteristics, beliefs about
BP and BP medications, medication adherence, and experiences of discrimination, to determine if the impact of race on
BP control remains after accounting for such factors.
Methods and Results—We recruited 806 white and black patients with hypertension from an urban safety-net hospital.
From a questionnaire administered to patients after their clinic visits, electronic medical record and BP data, we assessed
an array of patient factors. We then examined the association of patient factors with BP control by modeling it as a
function of the covariates using random-effects logistic regression. Blacks indicated worse medication adherence, more
discrimination, and more concerns about high BP and BP medications, compared with whites. After accounting for all
factors, race was no longer a significant predictor of BP control.
Conclusions—Results suggest that equalizing patients’ health beliefs, medication adherence, and experiences with care
could ameliorate disparities in BP control. Additional attention must focus on the factors associated with race to identify,
and ultimately intervene on, the causes of racial disparities in BP outcomes. (Circ Cardiovasc Qual Outcomes. 2010;
3:173-180.)
Key Words: blood pressure ? hypertension ? race ? disparities
H
vascular, and renal disease.1It is more frequent among
African Americans2and accounts for a significant portion of
racial differences in mortality, through excess cardiovascular
morbidity.3Many patients with hypertension have poorly
controlled blood pressure (BP), and African Americans are
disproportionately represented among this group,4even after
controlling for comorbidities such as diabetes and renal
disease.4–7
The reason for this racial disparity is not well under-
stood. Many prior studies of BP control have only exam-
ined a narrow range of potential etiologic factors—usually
clinical characteristics and sometimes including selected
sociodemographic factors.8,9Recently, authors have sug-
gested that patient self-management attitudes and behav-
iors10and other attitudes, beliefs, and experiences that
might affect medication nonadherence11might be potential
causal pathways to disparities in chronic disease outcomes
such as BP control.
ypertension, which affects more than 73 million Amer-
icans, is a major risk factor for cardiovascular, cerebro-
Bosworth et al12proposed an organizing framework for
the psychosocial and cultural domains they theorized
would affect BP control, incorporating patient character-
istics, including age, education, health literacy, and psy-
chological factors such as beliefs and attitudes about health
and illness, social/cultural environmental factors, includ-
ing culturally linked perceptions of hypertension and
therapies for it, and the medical environment, including the
provider-patient relationship and interactions. The model
does not specify causal associations, and, perhaps as a
result, it has not been tested empirically, so the presence or
strength of the hypothesized associations is not known.
The model also did not include comorbid conditions. Thus,
we sought to extend this work and the model itself by
examining the contribution of the previously proposed and
additional putative causal factors, in a more diverse sample
in a different setting. We hypothesized that after adjusting
for a more extensive set of potential confounders, race
would no longer be significantly associated with BP
control.
Received March 6, 2009; accepted December 3, 2009.
From the Section of General Internal Medicine (N.R.K., M.M.), Boston University School of Medicine, Boston, Mass; VA Boston Healthcare System
(N.R.K.), Boston, Mass; Center for Health Quality, Outcomes, and Economic Research (N.R.K., M.B.O., M.E.G., D.B.), Bedford VAMC, Bedford, Mass;
and the Health Policy and Management Department (M.E.G., D.B.), Boston University School of Public Health, Boston, Mass.
The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Correspondence to Nancy Kressin, PhD, General Internal Medicine, Boston University School of Medicine, 801 Massachusetts Ave, No. 2090, Boston,
MA 02118. E-mail nkressin@bu.edu
© 2010 American Heart Association, Inc.
Circ Cardiovasc Qual Outcomes is available at http://circoutcomes.ahajournals.orgDOI: 10.1161/CIRCOUTCOMES.109.860841
173
by guest on May 18, 2011circoutcomes.ahajournals.orgDownloaded from

Page 3

WHAT IS KNOWN
● Racial disparities in blood pressure (BP) control are
well documented but poorly understood, and prior
studies have only included a limited range of poten-
tial explanatory factors.
WHAT THE STUDY ADDS
● Blacks reported more experiences of discrimination,
worse medication adherence, and more concerns
about BP.
● Patient beliefs about hypertension can affect their BP
control.
● Comorbid conditions, whose prevalence varies by
race/ethnicity, can also affect BP control and thus are
important to account for.
● We suggest an enhanced model of factors leading to
disparities in BP control.
Methods
Sample
We identified all white and black patients ages 21 and older with 3
separate outpatient diagnoses of hypertension in 2004 in the primary
care practices of a northeastern academically affiliated urban safety
net tertiary care hospital. We use the term “black” to refer to patients
of African or Caribbean descent. The study was approved by the
institutional review board, and all subjects gave informed consent.
Study staff tracked clinic visits of these 10 125 patients over 19
months (October 2004 to June 2006), and, as they presented for care,
approached 3526 of them to request study participation. Those
willing were asked a series of questions and administered a 6-item
cognitive screen to determine eligibility.13However, 654 patients
(19% of 3526) overtly refused to participate and 920 patients (26%
of 3526) responded that they were unable to participate that day but
were potentially willing in the future, all before we were able to
assess their eligibility. Subsequently, 1083 patients (55% of the
remaining 1952) were excluded for reasons including cognitive
difficulties, hearing impairment, not speaking English, or not being
prescribed antihypertensive medications; we enrolled 869 patients.
We then applied this 55% exclusion rate to the 1574 nonrespondents
whose eligibility we had been unable to assess (654 refusers and 920
with no time) and estimated that 708 (45%) probably would have
been eligible. Thus, we calculated our participation rate as partici-
pants/participants?likely eligible subjects (869/869?708?55%).
We subsequently excluded 63 additional patients for whom study
staff were unable to obtain BPs, for a final sample of 806 (Figure 1).
We evaluated the representativeness of the enrolled cohort, com-
pared with those eligible to be enrolled, using the limited data
available on the nonenrolled patients. Enrolled patients were more
likely to be white (43% versus 32%, P?0.0001) and were younger
(mean age, 59 versus 65 for nonenrolled patients, P?0.0001), but
there was no difference in sex distribution from the parent population
of eligible patients (not shown).
Measures
Patient Characteristics
Sociodemographic Characteristics and Medical History
Patient sociodemographic characteristics including race (assessed
using US Census categories), education, and income were obtained
through self-report. To assess patient literacy, we used the REALM-
Short Form.14We used 5 separate dichotomous (yes/no) questions to
assess insurance status, asking if patients currently have health
insurance coverage through Medicare, Medicaid, Medigap, Free
Care, or other insurance.
Patient clinical data were extracted from the electronic medical
record, including age, sex, height, weight, and hypertension diagno-
sis. The electronic medical record was searched to obtain diagnoses
of comorbid conditions including renal insufficiency, coronary artery
disease, peripheral vascular disease, nicotine dependence, hyperlip-
idemia, diabetes mellitus, congestive heart failure, cerebrovascular
disease, and obesity, because these conditions may influence the
management of hypertension.15Obesity was considered a current
diagnosis for any patient with an electronic medical record diagnosis
or a calculated body mass index of at least 30.
Health Beliefs and Illness Perception
We examined a broad spectrum of patient beliefs and perceptions
about high BP and related medications. We used the “Beliefs about
Medicines Questionnaire”; 10 items assess patient concerns about
present and potential future adverse effects from their medications
and 8 items measure patients’ beliefs regarding the necessity of their
medications; we edited the items to relate responses specifically to
BP medicines16,17(items scored on a 5-point scale: strongly agree to
strongly disagree). Scores were summed within each scale to create
an overall scale score (range, 8 to 40 for the “necessity” scale,
Cronbach ??0.81; 10 to 50 for the “concerns” scale, ??0.80). Each
score was divided by the number of items to obtain a mean summary
score (range, 1 to 5); higher numbers indicated either greater
concerns about medications or greater beliefs in their necessity.
We used 4 additional items from our prior work18to evaluate the
degree of seriousness with which patients perceive hypertension,
asking “How serious do you think high BP is, in general?”; “How
serious do you think your high BP is, given your current use of
medication?”; “If you did not take your BP medications, how likely
do you think it would be that your BP would get worse over the next
year?”; and “If you did not take your BP medications, how likely do
you think it is that you would develop other health problems over the
10,125 eligible patients with
hypertension managed in
primary care
6599 patients not approached
-not scheduled for an appointment-not scheduled for an appointment
-study staff did not find them on the
day of an appointment
3526 3526 patients approached
for enrollment
tith d
1083 patients excluded
-257 saw medical student
-247 used medication dispenser
-199 cognitively impaired
-149 race other than White or Black
-71 did not speak English
-62 not prescribed antihypertensive
medication
-30 participating in another study of
hypertension
-16 hearing impaired
-52 all other reasons
2443 patients potentially
eligible
1574 patients declined to participate
but we were unable to assess their
eligibility
-654 overtly refused to participate
-920 did not have time to participate
th t di di t dthat day, indicated willingness to
participate in the future, but staff
were unable to enroll them at a future
visit
illit
869 patients enrolledp
63 63 patients did not have a BP reading
taken from a research assistant
ididhBP di
806 patients in the final
sample
Figure 1. Flow chart of patient recruitment.
174Circ Cardiovasc Qual Outcomes
March 2010
by guest on May 18, 2011 circoutcomes.ahajournals.orgDownloaded from

Page 4

next year?” (all scored on a scale of 1 to 5: extremely serious to not
at all serious for the former 2 questions; very likely to very unlikely
for the latter 2). We also included 10 individual items from the
“cause” dimension of the “Illness Perception Questionnaire” to
assess illness identity, cause, timeline, consequences, and cure
control to examine patients’ subjective beliefs about the etiology of
their high BP (responses on a 5-point scale; range, strongly agree to
strongly disagree).19
Perceived Discrimination in Health Care
We included 3 measures to assess patient perceptions of race-based
discrimination in the health care setting. We used 5 questions that
were a subset of the Commonwealth Fund 2001 Health Care Quality
survey, focusing specifically on the patient perception of his/her
provider and of encountering discrimination while receiving medical
care, in general.20We created an additional question about the
patient’s perception of his/her provider’s understanding of the
patient’s cultural background and how it affects his/her health; each
item was examined individually. We also included 7 dichotomous
items from a measure of patient perceptions of discrimination in
accessing health care,21counting the number of experiences reported
and creating a continuous variable (higher score indicates more
experiences of discrimination21; ??0.90).
Medication Adherence and Hypertension Management
To assess medication adherence, we used the Hill-Bone Compliance
to High Blood Pressure Therapy Scale, comprising items scored on
a 4-point scale (“none of the time” to “all of the time”).22We
included the 9-item adherence subscale, which has been validated
against BP control, summing the items to create a scale score (range,
9 to 36; higher scores indicate less adherence; ??0.74).
Outcome Assessment: BP Control
Research staff assessed patient BP using an automatic, portable
machine (Omron HEM-907, Bannockburn, Ill), which was validated
according to the international validation protocol and deemed an
appropriate instrument for accurate BP measurement.23We excluded
the 7% of patients without a BP reading from study staff. We
dichotomized the BP readings to indicate whether each patients’ BP
was controlled or not, for example, when their systolic BP exceeded
140 mm Hg or their diastolic BP exceeded 90 mm Hg, according to
the Joint National Committee on Hypertension 7 standards at the
time of the study, which also specified that for patients with diabetes
or renal insufficiency, BP should not exceed 130/80 mm Hg.15
Statistical Analyses
We first examined bivariate associations between race and each of
the sociodemographic, clinical, attitudinal, experiential, and medica-
tion adherence variables, using ?2or t tests, as appropriate. Next, we
performed bivariate analyses to determine whether this same set of
variables was associated with BP control (yes/no). Finally, BP
control was modeled as a function of the covariates using random-
effects logistic regression. The random effects, which were assumed
to be mean-zero gaussian additive errors on the logit scale, accounted
for 2 levels of clustering: patients within providers and providers
within clinics.
We fit 6 sets of models of increasing numbers of covariates,
including only patients with complete data. In the first, we examined
the effects of race alone on BP control. Next, we examined the
effects of race on BP control, adjusting for age and other sociode-
mographic characteristics. The third model added medication adher-
ence, the fourth added health beliefs, the fifth added experiences of
discrimination, and the sixth added comorbid conditions. Within
each model, we chose the subset of additional variables through an
ordinary logistic regression stepwise selection procedure, forcing in
provider indicators as fixed effects, keeping significant variables
from the prior model, and including race in all models. This
procedure prevents collinearity and overparameterization. We used a
probability value of 0.05 for variables to enter or be removed from
the model. c-Statistics and Hosmer-Lemeshow analyses were per-
formed on the models resulting from the stepwise procedure. Each
model was then rerun with the random effects terms. All analyses
were conducted using SAS 9.1.3 (SAS Institute, Cary, NC) with the
exception of the random-effects logistic regression, which was
conducted using the statistics package R (version 2.8.024).
Results
Sociodemographic Characteristics of the Sample
Most of the sample were black (57%) or female (65%), with
an overall mean age of 59 years. White patients were older
(61 versus 58 years), and more likely to have at least a high
school education (90% versus 71%). Black patients were
more likely to have insurance coverage through Medicaid or
Free Care (42% versus 22%, 41% versus 25%, respectively),
whereas white patients were more likely to have “other”
(private) insurance (59% versus 35%). There were race
differences in combined household family income, with a
greater percentage of blacks earning low incomes of
?$20 000 (58% versus 36% whites) and being less likely to
have a literacy score of 9th grade or higher (48% versus
84%), (all P?0.001; Table 1).
BP and Medical History by Race
A greater percentage of white patients had controlled BP
(65% versus 53%), with lower systolic and diastolic BP than
blacks (white systolic BP, 130 mm Hg versus 132 mm Hg;
white diastolic BP 76 mm Hg versus 81 mm Hg). White
patients were more likely to have hyperlipidemia (60% versus
47%), peripheral vascular disease (7% versus 4%), benign
prostatic hypertrophy (6% versus 2%), coronary artery dis-
ease (19% versus 8%), and cerebrovascular disease (7%
versus 4%). Black patients were more likely to have diabetes
(40% versus 24%), renal insufficiency (7% versus 4%),
congestive heart failure (5% versus 1%), and to be obese
(65% versus 52%, all P?0.05).
Health Beliefs and Illness Perceptions
Blacks had significantly more concerns about their BP
medications than whites (2.5 versus 2.1, P?0.0001). White
patients were significantly more likely to respond that their
BP was less serious, given their current use of medications
(3.3 versus 2.8, P?0.0001).
When asked about the causality of high BP, blacks agreed
more with the notion that it is caused by a germ or virus or
that diet, pollution, or heredity played a major role in causing
BP (Table 1). Blacks were more likely to indicate that it was
just by chance that they became ill with hypertension, that
other people played a large role in causing their BP, or that
high BP was caused by poor medical care in the past.
Perceived Discrimination
When asked if there was ever a time they would have gotten
better medical care if they belonged to a different race or
ethnic group or if they ever felt that a doctor or medical staff
judged them unfairly or treated them with disrespect because
of how well they spoke English, blacks were more likely to
respond “yes” than whites (19% versus 1% and 4% versus
0%, respectively). Although all patients generally agreed that
their provider understood their background and values, black
patients agreed less strongly (1.5 versus 1.4), and though all
patients disagreed that their provider looks down on them and
Kressin et alRacial Disparities in Blood Pressure Control
175
by guest on May 18, 2011circoutcomes.ahajournals.orgDownloaded from

Page 5

Table 1.Sociodemographic, Clinical, Attitudinal, Belief, and Experience Variables by Race and BP Control
Sociodemographic Characteristics and
Medical HistoryOverall, %White, %Black, %
P Value for Race
Differences
Controlled
BP, %
Uncontrolled
BP, %
P Value for Difference
by BP Control
% White
Mean age, y
% Male
Education, % ?12 y
% Income ?$20 000
Insurance status
Medicare
Medigap
Medicaid
Other
Free care
Literacy categories
?3rd grade
4th to 6th grade
7th to 8th grade
?9th grade
% With controlled BP
Mean systolic BP, mm Hg
Mean diastolic BP, mm Hg
Nicotine dependence
Hyperlipidemia
Diabetes
Peripheral vascular disease
Renal insufficiency
Benign prostatic hypertrophy
Coronary artery disease
Obesity
Congestive heart failure
Cerebrovascular disease
Health Beliefs and Illness Perceptions
BMQ: Mean necessity of medications,
mean score
BMQ: Mean concerns about
medications, mean score
How serious do you think high BP is,
in general*
How serious do you think your high
BP is, given your current use of
medication?*
If you did not take BP meds,
likelihood that BP would get worse
w/in a year†
If you did not take BP meds,
likelihood that you would develop
other health problems w/in a year†
Illness Perceptions Questionnaire Items‡
A germ or virus caused my high BP
Diet played a major role in causing
my high BP
Pollution caused my high BP
43
59
35
79
48
n/a
0.0007
?0.0001
?0.0001
?0.0001
48§
59
34
81
43
36§
60
38
76
56
0.0009
0.0853
0.2919
0.0521
0.0003
61
46
90
36
58
27
71
58
39
3
33
45
34
39
3
22
59
25
40
2
42
35
41
0.6889
0.4991
?0.0001
?0.0001
?0.0001
37
2
29
51
32
43
4
40
38
37
0.0984
0.2143
0.0012
0.0003
0.1072
4
9
2
2
5
3
8
4
14
33
48
53
132
81
11
27
58
24
63
58
131
79
12
84
65
130
76
22
67
?0.0001
0.0009
0.0277
?0.0001
0.0839
0.0002
?0.0001
0.0383
0.0470
0.0003
?0.0001
0.0004
0.0091
0.0382
0.0884
n/a
n/a
n/a
0.7303
0.9302
?0.0001
0.9918
?0.0001
0.1107
0.2151
0.2373
0.0635
0.2162
n/a
n/a
n/a
7
52
25
5
3
4
14
58
2
4
n/a
n/a
n/a
8
53
45
5
10
2
11
62
5
6
759
52
33
5
6
4
13
59
3
5
60
24
7
4
6
19
52
1
7
47
40
4
7
2
8
65
5
4
3.73.73.7 0.15423.73.70.9739
2.32.12.5
?0.0001 2.32.40.0407
1.5 1.51.50.46071.51.60.1147
3.0 3.32.8
?0.00013.12.9 0.0010
1.51.4 1.50.1284 1.51.4 0.1348
1.71.71.7 0.2307 1.7 1.70.4281
3.9
2.3
4.2
2.4
3.7
2.1
?0.0001
0.0017
3.9
2.2
3.8
2.3
0.1208
0.6715
3.73.83.5 0.0002
3.7 3.70.7546
(Continued)
176Circ Cardiovasc Qual Outcomes
March 2010
by guest on May 18, 2011circoutcomes.ahajournals.org Downloaded from