Accuracy of Diagnostic
M
Mammography
h att F
Facilities
iliti S
Serving
i
Vulnerable Populations
p
L. Elizabeth Goldman, UCSF
Rod Walker, Group Health
Diana Miglioretti, Group Health
Rebecca Smith
Smith--Bindman, UCSF
Karla Kerlikowske, UCSF
Breast cancer disparities
Mortality
M t lit rates
t for
f breast
b
t cancer are higher
hi h in
i vulnerable
l
bl
populations
Biologic differences in breast cancer by race
Documented disparities in treatment and treatment
delays by race/ethnicity, insurance status, & income
Facility-level differences in diagnostic accuracy of
Facilitymammography
h could
ld contribute
t ib t tto di
diagnostic
ti d
delays
l
or
unnecessary diagnostic procedures
Variability in mammography interpretive
performance
Diagnostic interpretive performance differs
across mammography
h ffacilities
iliti
Due to patient mix, skill of radiologists,
equipment, and practice patterns
Research Question
Does the interpretive performance of diagnostic
mammography differ between facilities serving
vulnerable women and those serving nonnonvulnerable women?
Breast Cancer Surveillance Consortium
Methods
Study period: 19991999-2005
Subjects:
j
– Women age 4040-80
– Diagnostic mammography
Additional evaluation of abnormal screening
mammogram
Evaluation of a symptomatic breast problem
Outcomes
Sensitivityy
False positive rate (1 – specificity)
Cancer detection rate = number cancers
detected per 1,000 diagnostic mammograms
*Cancer: Invasive carcinoma or DCIS within 1 year of
*Cancer:
diagnostic mammogram
Definition of a facility serving vulnerable
women
< 83% have at least a high school
education
> 30% AfricanAfrican-American, Hispanic,
Pacific--Islander/Hawaiian,
Pacific
Islander/Hawaiian Native
American
< $45,000 average median household
income
> 52% rural residents
*composite score: sum of 4 binary vulnerability measures
Goldman, Medical Care, 2008
Results: Study Population
Additional evaluation of a recent
abnormal screening mammogram
Evaluation of a symptomatic breast
problem
NonNonvulnerable
(score = 0)
Non-vulnerable
Non(score = 0)
Moderately
vulnerable
(score = 1 or 2)
Highly
vulnerable
(score = 3 or 4)
Moderately
vulnerable
(score = 1 or 2)
Highly
vulnerable
(score = 3 or 4)
# of facilities
62
63
28
77
70
29
# of women
43,305
25,904
6,990
44,538
21,195
8,501
# of
diagnostic
mammos
47,609
28,372
7,483
52,050
23,193
9,544
Results: Study Population
Facilities serving vulnerable women
– Women tended to be older
– Without prior mammography or very distant
mammography (> 43 months)
– More positive final interpretations
Additional evaluation of abnormal screening mammogram,
adjusted*
Sensitivity
Non--vulnerable
Non
Vulnerable
Education
Race/ethnicity
/
87.8%
87.8%
%
84.7%
83.5%
%
0.77
0.70
(0.44, 1.34)
(
(0.37,
1.31))
0.346
0.263
Rural/urban residences
Income
87.7%
88 6%
88.6%
87.2%
83 4%
83.4%
0.96 (0.61, 1.50)
0 64 (0
0.64
(0.40,
40 1
1.02)
02)
0.850
0.060
0 060
86.0%
82.7%
84.2%
93.2%
Reference
0.72 (0.45, 1.15)
0.56 (0.32, 0.98)
0.62 (0.33, 1.19)
1.64 (0.34, 7.97)
TREND
0.164
0.042
0.150
0.537
0.610
Composite vulnerability score
0
1
2
3
4
89.4%
OR (95% CI)
p-value
* Adjusted
Adj t d for
f age, breast
b
t density,
d
it time
ti
since
i
last
l t screening
i mammogram,
BCSC site
Additional evaluation of abnormal screening mammogram,
adjusted*
Non--vulnerable
Non
Vulnerable
OR (95% CI)
p-value
Education
Race/ethnicity
10.3%
10
3%
10.3%
10.1%
10
1%
10.4%
0.97
0 97 (0
(0.76,
76 1
1.24)
24)
1.01 (0.76, 1.35)
0.827
0 827
0.928
Rural/urban residences
Income
9.8%
9.6%
11.0%
12.8%
1.14 ((0.94, 1.38))
1.39 (1.13, 1.70)
0.175
0.002
10.4%
12.0%
11.9%
11 9%
9.6%
Reference
1.11 (0.91, 1.37)
1.32 (1.02, 1.70)
1 30 (0
1.30
(0.96,
96 1
1.76)
76)
1.02 (0.63, 1.66)
TREND
0.309
0.038
0.092
0 092
0.943
0.723
False Positive Rate
Composite vulnerability score
0
1
2
3
4
9.4%
* Adjusted for age, breast density, time since last screening mammogram, BCSC site
Additional evaluation of abnormal screening mammogram,
adjusted*
Non-vulnerable
Non(per 1000
exams)
Vulnerable
(per 1000
exams)
Education
Race/ethnicity
30.3
30.4
27.1
26.0
0.89
0.85
(0.69, 1.15)
(0.63, 1.15)
0.360
0.285
Rural/urban residences
Income
28.4
30.5
32.5
27.6
1.15 (0.94, 1.41)
0.90 (0.72, 1.12)
0.161
0.345
27.7
28.1
27.7
44.1
Reference
0.87 (0.70, 1.08)
0.88 (0.68, 1.16)
0.87 (0.64, 1.19)
1.42 (0.84, 2.40)
TREND
0.213
0.370
0.393
0.185
0.216
0 216
Cancer Detection Rate
Composite vulnerability score
0
1
2
3
4
31.6
OR (95% CI)
p-value
* Adjusted for age, breast density, time since last screening mammogram, BCSC site
Evaluation of breast symptoms
symptoms, adjusted*
adjusted
S
Sensitivity
ii i
N -vulnerable
NonNon
l
bl
V l
Vulnerable
bl
Education
Race/ethnicity
81.9%
81
9%
81.5%
79.8%
79
8%
81.1%
0.87
0
87
0.97
(0
(0.63,
63 1
1.20)
20)
(0.67, 1.42)
0.400
0 400
0.888
Rural/urban residences
Income
80.4%
81.6%
83.5%
80.9%
1.24 (0.93, 1.66)
0.95 (0.70, 1.29)
0.139
0.751
82.5%
81.3%
79 4%
79.4%
87.9%
R f
Reference
1.10 (0.78, 1.55)
1.01 (0.69, 1.48)
0 89 (0
0.89
(0.60,
60 1
1.32)
32)
1.72 (0.70, 4.20)
TREND
0.595
0.958
0.561
0 561
0.232
0.358
Composite vulnerability
score
0
1
2
3
4
81 2%
81.2%
OR (95% CI)
p-value
l
* Adjusted for age, breast density, time since last screening mammogram, site
Evaluation of breast symptoms
symptoms, adjusted*
adjusted
NonN
Nonvulnerable
V l
Vulnerable
bl
Education
Race/ethnicity
6.1%
6 1%
6.2%
8.2%
8 2%
8.0%
1.39
1
39
1.32
(1
(1.08,
08 1
1.79)
79)
(0.98, 1.76)
0.011
0 011
0.066
Rural/urban residences
Income
5.7%
6.0%
8.5%
7.8%
1.55 (1.27, 1.88)
1.34 (1.08, 1.66)
<0.001
0.008
6.9%
7.3%
8.8%
10.6%
R f
Reference
1.31 (1.06, 1.63)
1.39 (1.08, 1.78)
2.12 (1.27, 2.32)
2.12 (1.36, 3.32)
TREND
0.014
0.011
0.001
0.001
0.001
F l Positive
False
P ii R
Rate
Composite vulnerability score
0
1
2
3
4
5.4%
5 4%
OR (95% CI)
p-value
l
* Adjusted for age, breast density, time since last screening mammogram, site
Evaluation of breast symptoms (adjusted)
Non-vulnerable
Non(per 1,000
exams)
Vulnerable
(per 1,000
exams)
Education
Race/ethnicity
36.5
35.9
39.6
44.3
1.09
1.25
(0.88, 1.35)
(0.98, 1.60)
0.440
0.069
Rural/urban residences
Income
34.6
35.6
42.9
41.0
1.26 (1.05, 1.50)
1.16 (0.97, 1.40)
0.011
0.110
38.3
41 2
41.2
41.2
45.9
Reference
1.15 (0.94, 1.41)
1 25 (0
1.25
(0.99,
99 1
1.57)
57)
1.28 (0.98, 1.67)
1.40 (0.91, 2.14)
TREND
0.169
0.058
0 058
0.070
0.122
0.102
0
0
Cancer Detection Rate
Composite vulnerability score
0
1
2
3
4
33.5
OR (95% CI)
p-value
Limitations
Income based on ZIP code average, may not
represent individual woman’s true income
Income is absolute and purchasing power may
vary across the United States
Conclusions
Facilities serving low income women are more likely to
recommend biopsy after diagnostic mammography
performed to evaluate abnormal screening
mammograms
Facilities serving low income
income, rural
rural, and minorities are
more likely to recommend biopsy after diagnostic
mammography performed to evaluate breast symptoms
Cancer detection similar between facilities serving
vulnerable and nonnon-vulnerable women
Policy Implications
Interventions to improve interpretive
performance of diagnostic mammography at
facilities serving vulnerable women may reduce
invasive procedures in women without cancer
Increased referral rates for biopsy may be
appropriate in settings with both higher cancer
prevalence and higher lost to followfollow-up rates
Future Work
Determine
D t
i if iincreased
d rates
t off recommendation
d ti
for biopsy among women without cancer (false
positives) actually lead to more biopsies in
women without cancer
Evaluate facility characteristics that may
contribute to differences in referral patterns for
biopsies
Evaluate whether referrals “appropriate”
Acknowledgments
California Breast Cancer Research Project
Project, Grant #14IB
#14IB0062
AHRQ KO8,, Grant #1 K08 HS018090-01
National Cancer Institute Breast Cancer Surveillance
Consortium (U01CA63740, U01CA86076, U01CA86082,
U01CA63736 U01CA70013 U01CA69976
U01CA63736,U01CA70013,
U01CA69976,
U01CA63731, U01CA70040)
The collection of cancer data used in this study was
supported in part by several state public health
departments and cancer registries throughout the U.S.
For a full description of these sources, please see:
htt //b
http://breastscreening.cancer.gov/work/acknowledgemen
t
i
/
k/ k
l d
t.html.