G
R p
a ’
ra
Ch ng
Kir
OB
fou
ma
the
ST
dem
vid
ges
RE
tra h so
tus e in
ono
ce/
kely
con
5%
cio
m
tion
c st
ve c
ont
Cite h K, ptio
rac yne
D isparities in health outcomes by
sta
ma
viders in contributing to these disparities ers treat patients differently depending
on patients’ race/ethnicity4-6 and SES.7,8
Fro do
Gr s-D
Me ia,
Sch ep
Sci al R
He
Pre soc
Re
Re
Re
This project was supported by the Fellowship in Family Planning and by NIH/NCRR/OD UCSF-
CT
Its
view
000
Research www.AJOG.org
ment being highly dependent on pa-
tients’ personal preferences. In addition,
the discussion of sexual behavior and
contraception use in a clinical encounter
is a culturally and socially complex area
of medicine in which providers’ subcon-
scious biases or assumptions might play
an important role.
The limited data analyzing potential
disparities in providers’ decision making
in this context suggest that providers
SI Grant no. KL2 RR024130.
contents are solely the responsibility of the authors and do not necessarily represent the official
s of the National Institutes of Health.
2-9378/free • © 2010 Mosby, Inc. All rights reserved. • doi: 10.1016/j.ajog.2010.05.009
For Editors’ Commentary, see Table of Contents
See related editorial, page 293
m the Departments of Family and Community Medicine (Drs Dehlen
umbach), Epidemiology and Biostatistics (Drs Vittinghoff and Bibbin
dicine (Drs Bibbins-Domingo and Schillinger), University of Californ
ool of Medicine; and the Department of Obstetrics, Gynecology and R
ences (Drs Dehlendorf, Ruskin, and Steinauer), Bixby Center for Glob
alth, University of California San Francisco, San Francisco, CA.
sented in preliminary form (data in article) at the Annual Meeting of the As
productive Health Professionals, Los Angeles, CA, Oct. 1-3, 2009.
ceived Nov. 10, 2009; revised Feb. 25, 2010; accepted May 5, 2010.
prints not available from the authors.
See Journal Club, page 411
ries is widespread even among those who
self-identify as nondiscriminatory.9
Previous research on the effect of pa-
tient race/ethnicity and SES on provid-
ers’ clinical behavior has focused on
provider-patient interactions around
discrete medical decisions for which
there is general consensus about appro-
priate treatment.4,5,10 Contraceptive de-
cision making, in contrast, involves the
consideration of multiple clinically ap-
propriate options, with the best treat-
rf and
omingo), and
San Francisco,
roductive
eproductive
iation of
race/ethnicity and socioeconomic
tus (SES) are well documented in
ny areas.1 The role of health care pro-
is an area of growing research,2,3 with
multiple studies suggesting that provid-
These findings are consistent with social
psychology research indicating that sub-
conscious stereotyping by social catego-
ENERAL GYNECOLOGY
ecommendations fo
randomized trial of
ce/ethnicity and soc
ristine Dehlendorf, MD, MAS; Rachel Ru
sten Bibbins-Domingo, MD, PhD; Dean
JECTIVE: Recommendations by health care
nd to vary by patient race/ethnicity and socio
y contribute to health disparities. This study inv
se factors on recommendations for contracep
UDY DESIGN: One of 18 videos depicting pati
ographic characteristics was shown to each o
ers. Providers indicated whether they would
trel intrauterine contraception to the patient sh
SULTS: Low socioeconomic status whites were
uterine contraception recommended than hig
whites (odds ratio [OR], 0.20; 95% confidenc
this article as: Dehlendorf C, Ruskin R, Grumbac
e/ethnicity and socioeconomic status. Am J Obstet G
ntrauterine contrace
e effects of patients
oeconomic status
in, MD; Kevin Grumbach, MD; Eric Vitti
hillinger, MD; Jody Steinauer, MD, MAS
viders have been
nomic status and
igated the effect of
.
s of varying socio-
4 health care pro-
ommend levonor-
n in the video.
s likely to have in-
cioeconomic sta-
terval [CI], 0.06–
0.69); whereas, socioec
Latinas and blacks. By ra
and blacks were more li
mended than low socioe
1.1–10.2 and OR, 3.1; 9
race/ethnicity for high so
CONCLUSION: Providers
ception or make assump
nicity and socioeconomi
Key words: contracepti
disparities, intrauterine c
et al. Recommendations for intrauterine contrace
col 2010;203:319.e1-8.
ma
OCTOBER 2010 American
tion:
hoff, PhD;
mic status had no significant effect among
ethnicity, low socioeconomic status Latinas
to have intrauterine contraception recom-
omic status whites (OR, 3.4; and 95% CI,
CI, 1.0–9.6, respectively), with no effect of
economic status patients.
ay have biases about intrauterine contra-
s about its use based on patient race/eth-
atus.
ounseling, family planning, health
raception
n: a randomized trial of the effects of patients’
r i
th
i
sk
Sc
pro
eco
est
tion
ent
f 52
rec
ow
les
y be susceptible to different influ-
Journal of Obstetrics& Gynecology 319.e1
en
on
ges
lik
fec
cou
com
mi
Un
ten
tra
hig
sys
do
the
rai
no
tra
the
T
pro
ute
ula
effi
cou
tio
lik
Alt
cer
pe
cre
fec
cli
con
wh
tio
lik
co
no
use
as
na
sta
rec
SE
dif
IU
T
eth
me
we
om
da
nic
M
Sta
W
sta
ab
var
or
cla
wi
his
[ST
ina
ma
wo
mi
wa
for
rec
as
an
by
nic
tha
no
ha
Ch
tes
in
she
at
the
pro
ha
ha
thi
cei
da
or
get
for
gra
sho
I
he
var
scr
wi
the
iza
tio
act
wa
con
Stu
W
he
Pra
at
fam
ne
be
the
tra
lec
of
ab
tio
of
�3
Ag
me
ing
co
ord
od
eff
ing
car
me
the
lev
eff
Un
qu
pa
the
lik
an
gen
wo
aw
reg
nic
da
O
wh
vid
can
ba
co
IU
of
thi
can
len
of
Research General Gynecology www.AJOG.org
31
ces on their recommendations. From
e perspective, several studies have sug-
ted that providers may be especially
ely to encourage the use of highly ef-
tive contraceptive methods and dis-
rage fertility in minority and low-in-
e populations.11-13 In contrast,
nority and low-income women in the
ited States have higher rates of unin-
ded pregnancy and lower use of con-
ceptive methods than do white and
her-income women.14,15 Although
tem and patient-related factors un-
ubtedly contribute to these statistics,
presence of these disparities also
ses the possibility that clinicians may
t, in fact, be promoting effective con-
ceptive methods among patients from
se sociodemographic groups.
he effect of patient characteristics on
vider recommendations for intra-
rine contraception (IUC) is of partic-
r interest due to this method’s high
cacy,16 as any tendency toward dis-
raging the fertility of specific popula-
ns could be manifested in a greater
elihood of recommending this method.
ernatively, as many providers are con-
ned that the use of IUC could result in
lvic infections among women at in-
ased risk for sexually transmitted in-
tions,17,18 although well-designed
nical studies have indicated that these
cerns are misplaced,19,20 clinicians
o make race- and class-based assump-
ns about sexual behaviors may be less
ely to consider IUC as an appropriate
ntraceptive method for poor and mi-
rity women. As an expansion in the
of IUC is currently being advocated
a means to decrease unintended preg-
ncy,18,21,22 it is important to under-
nd whether differences in provider
ommendations by race/ethnicity and
S exist and, if so, consider how these
ferences may affect efforts to promote
C in different demographic groups.
o determine whether patient race/
nicity and SES affect provider recom-
ndations for the levonorgestrel IUC,
conducted a study of providers’ rec-
mendations using videos of stan-
rdized patients of different race/eth-
ities and SES. con
9.e2 American Journal of Obstetrics& Gynecology
ATERIALS AND METHODS
ndardized patient videos
e produced 18 videos portraying a
ndardized patient requesting advice
out contraception, with the patient
ying by race/ethnicity (white, black,
Latina), SES (low- or upper-middle
ss), and gynecologic history (a woman
th a history of a vaginal delivery and no
tory of sexually transmitted infections
Is]; a woman with a history of a vag-
l delivery and history of pelvic inflam-
tory disease [PID]; or a nulliparous
men with no history of sexually trans-
tted infections). The low SES patient
s portrayed as a housekeeper studying
her GED and the high SES patient as a
ent business school graduate working
a bank manager. Both the high SES
d the low SES patients were portrayed
the same actor within each racial/eth-
category. The providers were told
t the patient was 27 years old, had
rmal blood pressure, and had recently
d a negative test for Gonorrhea and
lamydia and a normal Papanicolaou
t. Each patient indicated that she was
a monogamous relationship and that
did not want to become pregnant for
least a few years. For the purpose of
se analyses, the primary gynecologic
file of interest was the woman who
d previously had a vaginal delivery and
d no history of STIs, as women with
s history have historically been per-
ved as ideal IUC candidates. The stan-
rdized patients who were nulliparous
had a history of PID were grouped to-
her as having perceived risk factors
complications related to IUC. Photo-
phs of the standardized patients are
wn in Figure 1, A-F.
n each video, the patient presented
r history as a monologue, with the only
iation being the study factors. The
ipts used in the videos were pretested
th a sample of 15 providers to ensure
maximal level of realism. Standard-
tion of verbal factors, such as inflec-
n and tone, were practiced with the 3
ors. Five health care providers
tched all 18 videos to verify overall
sistency of nonverbal and verbal
tent. dif
OCTOBER 2010
dy design
e recruited a convenience sample of
alth care providers (MDs, DOs, Nurse
ctitioners, and Physician Assistants)
meetings of professional societies of
ily medicine and obstetrics and gy-
cology. Eligibility criteria consisted of
ing a practicing health care provider in
United States who had completed
ining. After observing 1 video, se-
ted using randomly permuted blocks
18, the providers completed a survey
out their contraceptive recommenda-
ns for the patient shown, ranking each
6 methods on a scale of�3 to�3, with
indicating “Strongly Recommend
ainst”, 0 indicating “Neither Recom-
nd for nor Against” and �3 indicat-
“Strongly Recommend For.” The
mputerized survey randomized the
er in which the contraceptive meth-
s were displayed to avoid any sequence
ect. The subjects were informed dur-
the survey that the patient’s health
e insurance covered all contraceptive
thods. Our outcome of interest was
recommendation regarding the
onorgestrel IUC, as this is the more
ective of the 2 IUCs offered in the
ited States.23 Providers also answered
estions about their perceptions of the
tient in the video, indicating whether
y felt the patient was more or less
ely to experience specific outcomes
d whether she was more or less intelli-
t and knowledgeable than an average
man her age. The providers were not
are of the primary study hypothesis
arding the effect of patient race/eth-
ity and SES on provider recommen-
tions for IUC.
ur primary research question was
ether the recommendations of pro-
ers for IUC differ for African Ameri-
, Latina and White patients. We
sed our sample size on a binary out-
me of willingness to recommend an
C. We hypothesized that a difference
15 percentage points in prevalence of
s outcome would be clinically signifi-
t in populations where overall preva-
ce of the outcome is 30%. Our sample
524 provided 84% power to detect a
ference of this magnitude in separate
com
La
Sta
W
ou
for
Fis
pri
wa
spo
the
fac
an
be
rac
jec
pa
RE
Th
tw
reg
Am
Gy
of
Ph
he
stu
gyn
pa
vid
rac
ble
sta
act
we
ma
pa
mo
tie
La
ers
Re
lev
Sig
rac
(P
log
wi
lev
no
int
str
nic
wo
ha
SE
Bla
lik
pa
P�
rec
wh
A
pa
wa
cen
str
A s
F
S
A,
pat
SES
Deh
www.AJOG.org General Gynecology Research
parisons of African American and
tina women to white women.
tistical analysis
e performed bivariate analysis of the
tcome variable of recommendation
the levonorgestrel IUC using �2 tests,
her’s exact tests, and t tests as appro-
ate. Multivariate logistic regression
s performed using a dichotomized re-
nse variable of�1 or�0. Because of
complicated interplay between social
tors in other studies,24 prespecified
alyses included analysis of interactions
tween the patient characteristics of
IGURE 1
tandardized patients
Low SES white patient. B, High SES white patie
ient. E, Low SES Latina patient. F, High SES Lat
, socioeconomic status.
lendorf. Recommendation for intrauterine contraception. Am
e/ethnicity, SES, and gynecologic his- cis
y. For the multivariate model, we pre-
cified the following provider-level
iables to be included in the model:
, sex, race/ethnicity, specialty, and
vision of IUC. For all other variables,
used backward selection and in-
ded any variables that changed any of
coefficients of interest by�10%. The
jects’ perceptions of the patients were
alyzed in the same manner. All analy-
were performed using Stata Version
(Stata Corp, College Station, TX).
he Committee of Human Research at
University of California, San Fran-
, Low SES black patient. D, High SES black
patient.
bstet Gynecol 2010.
co approved this study, and all sub- hig
OCTOBER 2010 American
ts provided informed consent before
rticipation.
SULTS
e videos were shown at 4 meetings be-
een September 2007 and May 2008; 2
ional and 1 national meeting of the
erican College of Obstetricians and
necologists, and the national meeting
the American Academy of Family
ysicians. Five hundred twenty-four
alth care providers completed the
dy, and the race/ethnicity, SES, and
ecologic profile of the standardized
tients were balanced between all pro-
er characteristics except provider
e/ethnicity in the overall sample (Ta-
1). Within each strata defined by the
ndardized patients’ gynecologic char-
eristics, the provider characteristics
re balanced, with the exception that
le providers assigned to standardized
tients with perceived risk factors were
re likely to be assigned the black pa-
nt and less likely to be assigned the
tina patient than were female provid-
(P� .02).
commendations for
onorgestrel IUC
nificant interactions between patient
e/ethnicity and gynecologic history
� .05) and patient SES and gyneco-
ic history (P � .04) were identified
th respect to recommendations for the
onorgestrel IUC. As the woman with
perceived risk factors was of primary
erest, we focused our analysis on this
atum (n � 173). Analyzing race/eth-
ity and SES separately, low SES
men were significantly less likely to
ve IUC recommended than were high
S women (57% vs 75%; P � .01).
ck women were significantly more
ely to have IUC recommended com-
red with white women (75% vs 57%;
.04) while there was no difference in
ommendations between Latina and
ite women (66% vs 57%; P� .31).
s an additional interaction between
tient race/ethnicity and SES (P� .02)
s identified, Figure 2 presents the per-
t of providers recommending IUC
atified by both race/ethnicity and SES.
ignificant difference between low and
tor
spe
var
age
pro
we
clu
the
sub
an
ses
9.2
T
the
nt. C
ina
J O
h SES white women (P� .01) and low
Journal of Obstetrics& Gynecology 319.e3
T
C
C
s
M
.
R
.
A
.
S
.
P
.
P
.
F
c
.
B
.
P
r
.
A
.
H
.
P
.
D
Research General Gynecology www.AJOG.org
31
ABLE 1
haracteristics of study subjects, by standardized patient characteristics
haracteristics of study
ubjects
Standardized patient characteristics
All
subjects
White
n � 179
Black
n � 172
Latina
n � 173
P
value
High SES
n � 262
Low SES
n � 262
P
value
No risk
factors
n � 173
Perceived
risk factors
n � 351
P
value
ale sex, % 53.6 52.5 59.3 49.1 .16 50.8 56.5 .19 52.0 54.4 .61
...............................................................................................................................................................................................................................................................................................................................................................................
ace/ethnicity, % .96 .38 .008
.......................................................................................................................................................................................................................................................................................................................................................................
White 76.9 79.9 75.0 75.7 78.2 75.6 84.4 73.2
.......................................................................................................................................................................................................................................................................................................................................................................
Black 7.8 7.3 7.6 8.7 6.1 9.5 2.3 10.5
.......................................................................................................................................................................................................................................................................................................................................................................
Latina 3.8 2.8 4.1 4.6 5.0 2.7 3.5 4.0
.......................................................................................................................................................................................................................................................................................................................................................................
Asian 9.2 7.8 11.1 8.7 8.4 9.9 6.9 10.3
.......................................................................................................................................................................................................................................................................................................................................................................
Other 2.3 2.2 2.3 2.3 2.3 2.3 2.9 2.0
...............................................................................................................................................................................................................................................................................................................................................................................
ge, y (mean/SD) 45.9 (10.5) 44.9 (9.7) 46.9 (11.3) 45.8 (10.4) .22 44.6 (10.0) 47.2 (10.8) .10 45.5 (10.9) 46.0 (10.3) .47
...............................................................................................................................................................................................................................................................................................................................................................................
pecialty, % .94 .80 .57
.......................................................................................................................................................................................................................................................................................................................................................................
Obstetrics/Gynecology 59.0 59.8 59.9 57.2 58.0 59.9 59.5 58.7
.......................................................................................................................................................................................................................................................................................................................................................................
Family Medicine 38.7 38.6 37.8 39.9 39.3 38.2 39.3 38.5
....................................................................................................................................................................................................................................................................................................