Volume 8: No. 3, May 2011
Rashid Njai, PhD, MPH; Paul Z. Siegel, MD, MPH; Jacqueline W. Miller,
MD; Youlian Liao, MD
Suggested citation for this article: Njai R,
Siegel PZ, Miller JW, Liao Y. Misclassification of survey responses and
black-white disparity in mammography use, Behavioral Risk Factor
Surveillance System, 1995-2006. Prev
Chronic Dis 2011;8(3):A59.
http://www.cdc.gov/pcd/issues/2011/may/10_0109.htm. Accessed [date].
PEER REVIEWED
Abstract
Introduction
The validity of self-reported data for mammography differ by race. We assessed
the effect of racial differences in the validity of age-adjusted, self-reported
mammography use estimates from the Behavioral Risk Factor Surveillance System
(BRFSS) from 1995 through 2006 to determine whether misclassification
(inaccurate survey question response) may have obscured actual racial
disparities.
Methods
We adjusted BRFSS mammography use data for age by using 2000 census estimates
and for misclassification by using the following formula:
(estimated prevalence − 1 + specificity) / (sensitivity +
specificity − 1). We used values reported in the literature for the formula
(sensitivity = 0.97 for both black and white women, specificity = 0.49 and 0.62,
respectively, for black and white women).
Results
After adjustment for misclassification, the percentage of women aged 40 years or
older in 1995 who reported receiving a mammogram during the previous 2 years was
54% among white women and 41% among black women, compared with 70% among both
white and black women after adjustment for age only. In 2006, the percentage
after adjustment for misclassification was 65% among white women and 59%
among black women compared with 77% among white women and 78% among black women
after adjustment for age only.
Conclusion
Self-reported data overestimate mammography use — more so for black women than
for white women. After adjustment for respondent misclassification, neither
white women nor black women had attained the Healthy People 2010
objective (≥70%) by 2006, and a disparity between white and black women
emerged.
Back to top
Introduction
Behavioral Risk Factor Surveillance System (BRFSS) data show that in 1995 the
percentage of both black and white women aged 40 years or older who had received
a mammogram during the past 2 years was near the Healthy People 2010
objective of 70% (objective 3-13) (1). By 1997, the BRFSS race-specific
prevalence estimates had increased and remained above the Healthy People 2010
objective through 2006 (2). Racial differences in the validity of these
self-reported data can, however, result in misclassification (defined as an
inaccurate answer) of survey responses that may obscure true black-white
disparities in mammography use rates (3-8).
A recent meta-analysis of studies that measured the race-specific validity of
survey questions about self-reported mammography use against documented sources,
such as medical and billing records, found that the specificity of survey
questions that measure mammography use is lower among black women than white
women (4). In this article, specificity refers to the probability that a woman
who does not have a documented mammogram actually reports she did not
have a mammogram, whereas sensitivity refers to the probability that a woman who
does have a documented mammogram actually reports she did have a mammogram.
After adjusting race-specific mammography use rates from the 2000 National
Health Interview Survey for sensitivity and specificity, Rauscher and colleagues
found increases of up to 15 percentage points in the disparity between black and
white women; estimates for the percentage of women who reported having had a
mammogram in the past 2 years fell from 68% to 37% for black women and from
72% to 58% for white women (4).
Death rates from breast cancer remain higher for black women than for white
women, even though breast cancer incidence is higher among white women (9). This
observation has been referred to as the breast cancer incidence-mortality
paradox (10). Mammograms are a key tool for detecting breast cancer at an early,
treatable stage and thereby reducing death rates from the disease; yet black
women are more likely than white women to have advanced-stage cancer at
diagnosis. BRFSS data, however, consistently indicate that black women receive
mammograms as frequently as white women (11).
The objectives of this study were to adjust BRFSS mammography data for
misclassification and measure the extent to which this misclassification
resulted in overestimates of mammography use rates and might have obscured a
disparity between black and white women. We included data from a 12-year period,
1995 through 2006, and compared adjusted rates with the Healthy People 2010
objective that at least 70% of women aged 40 years or older should have had a
mammogram within the past 2 years. We also discuss how adjusting mammography use
data for misclassification may help to explain part of the breast cancer
incidence-mortality paradox.
Back to top
Methods
BRFSS is the largest state-based telephone survey of the civilian,
noninstitutionalized adult population. Using SAS-callable SUDAAN version 9.2
(RTI International, Research Triangle Park, North Carolina) we weighted BRFSS
data for probability of selection and to match the age-, race-, and sex-specific
populations from annually adjusted intercensal estimates. The survey, which uses
a complex sampling design, collects self-reports of health risk behaviors
monthly in all 50 states, the District of Columbia, and 3 US territories: Guam,
Puerto Rico, and the US Virgin Islands. Detailed descriptions of the methods,
questionnaires, and other technical survey details are available from the BRFSS
website (www.cdc.gov/brfss/index.htm). We used the Women’s Health Module of
BRFSS, which includes mammography use questions and was administered as part of
the core survey each year from 1995 through 1999 and in even-numbered years
since 2000. The number of black and white women aged 40 years or older who
answered the mammography use questions increased from 39,025 in 1995 to 156,982
in 2006. For the years when the Women’s Health Module was not included in the
BRFSS core survey — 2001, 2003, and 2005 — we used the midpoint of the
prevalence estimates from the previous and following years.
Chi-square analysis was used to assess differences in
demographic characteristics between black and white women (P < .001).
A meta-analysis of 12 studies including more than 4,000 white women and 1,000
black women reported that the sensitivity of self-reported mammography questions
for use within the previous 2 years was 0.97 for both black and white US women
aged 40 years or older. Specificity, however, was lower for black women (0.49)
than for white women (0.62) (4). These studies compared survey data of black and
white women’s self-reported mammography use with a review of medical or billing
records to confirm whether the self-report was accurate (12-15). Women included
in these 12 studies were aged 40 years or older, generally represented
convenience samples from a range of low socioeconomic status (SES) to
middle-upper SES groups, and were selected from intervention, clinic, and
Medicare populations.
To understand the relationship between black and white estimates for
mammography use independent of age, we first age-adjusted the weighted
prevalence estimates to the 2000 US census standard population. Next, using the
values of sensitivity and specificity obtained from Rauscher (4), we adjusted
the age-adjusted prevalence estimates for misclassification, using the following
formula (15): final-adjusted prevalence = (estimated prevalence − 1 +
specificity) / (sensitivity + specificity − 1). Thus, for black
women, the formula was (estimated prevalence − 1 + 0.49) /
(0.97 + 0.49 − 1), and for white women it was (estimated prevalence − 1 +
0.62) / (0.97 + 0.62 – 1).
For these analyses, only non-Hispanic black and non-Hispanic white women aged
40 years or older who responded to the mammography questions in BRFSS were
included. Respondents’ race was based on self-report. Race is included as a
characteristic of interest because it is a proxy for potentially unequal
psychosocial and other environmental exposures (16). Race is subsequently used
to examine the existence of an absolute disparity between white and black women
following necessary statistical adjustment of race-specific prevalence estimates
for mammography use. Chi-square tests were used to examine white-black
differences in demographic data. In addition, age-adjusted and final-adjusted
(adjusted for both age and misclassification) annual BRFSS prevalence estimates
were tested by using linear regression trend analysis to determine whether time,
race, and time-race interaction statistically affect the estimated prevalence of
mammography use.
Back to top
Results
The distributions for all selected demographic characteristics differed by
race (P < .001)
(Table). Black women were
younger, had a higher prevalence of employment and a lower prevalence of health
insurance coverage, were less educated, and had lower incomes than their white
counterparts.
Among women aged 40 years or older who participated in the 1995 BRFSS, 70% of
both white and black women reported having had a mammogram during the previous 2
years (Figure). After adjustment for misclassification, the prevalence in 1995
fell to 54% among whites and 41% among blacks. In 2006, adjustment for
misclassification resulted in a drop from 77% to 65% for white women and
from 78% to 59% for black women. Based on final adjusted results, prevalence
estimates for both black and white women in 1995 were substantially below the
Healthy People 2010 objective of 70%. Estimates were higher in subsequent
years, but by 2006, they had not reached the 70% objective for either group.
Figure. Age-adjusted and final-adjusted estimates for
mammography use among white and black women, Behavioral Risk Factor Surveillance
System, 1995-2006. Data refer to women who reported having a mammogram
within the past 2 years. Final-adjusted estimates were obtained by adjusting the age-adjusted estimates
for misclassification using the following formula with race-specific specificity
(sp) and sensitivity (se) (white se = .97, sp = .62; black se = .97, sp = .49):
(estimated
prevalence − 1 + sp) / (se + sp −
1). See formula in Methods. Percentages for 2001, 2003, and 2005 are
the averages of the previous and following years. Abbreviations: adj., adjusted;
HP 2010, Healthy People 2010 goal.
[A
tabular version of this figure is also available.]
The age-adjusted prevalence of mammography use among black women from 1995
through 2006 was generally equal to or slightly higher than that among white women
(Figure). After adjustment for misclassification, however, prevalence estimates
were on average 8.95 percentage points (t = −6.75, P < .001) lower
overall among black women than among white women. Linear regression trend
analysis of adjusted prevalence estimates (regressing year and race on use) also
showed a significant positive relationship (β = 1.19, standard error = .19, P
< .001) between time (years) and mammography use in the past 2 years when
controlling for race. In additional trend analyses there was no significant
interaction between year and race (t = 1.13, P = .27) over time
(ie, the slopes of the 2 regression lines were not significantly different);
thus the rate of increase in mammography use over time was the same for black
and white women from 1995 through 2006.
Back to top
Discussion
According to unadjusted and age-adjusted BRFSS data, there was little to no
disparity from 1995 through 2006 between the percentages of black and white
women aged 40 years or older who had a mammogram during the past 2 years. Also,
mammography use rates for both black and white women were consistently at or
above the Healthy People 2010 objective of 70%. However, after adjustment
for respondent misclassification, mammography use rates for neither white women
nor black women had attained the Healthy People 2010 objective by 2006,
and a disparity between white and black women emerged.
Lack of a disparity in mammography use between black and white women has been
widely reported among studies that focus on racial, ethnic, and socioeconomic
disparities (2,17,18). However, the results presented in our study
suggest that previous notions of black-white parity in
mammography use should be reexamined (13). Women of lower SES, immigrant women, black women, and
women from other minority racial or ethnic groups remain populations of concern
because of disparities in stage of breast cancer at diagnosis and breast cancer
death rates (16,19-21). The disparity in mammography use observed in our
study may help explain racial disparities in stage of breast cancer at diagnosis
and in breast cancer death rates. Other factors that might contribute to the
incidence-mortality paradox include differences in the biology of the disease
between black women and white women, differences in stage at diagnosis that are
unassociated with mammography use, and differences in treatment following
diagnosis (11,21,22).
Lack of access to care — because of high cost, not having a usual source of
care, or lack of health insurance — remains a barrier to mammography use.
Lower-income, elderly, and immigrant women may encounter barriers because of
language or health literacy problems. Additional factors that may reduce
mammography use and contribute to disparities include patient knowledge,
attitudes, and cultural beliefs (22-24). For example, Rawl and colleagues found
that white women perceived greater benefits from receiving a mammogram than did
black women (23).
This study has several limitations. Ideally, the measures of validity used to
adjust BRFSS prevalence estimates would be based on studies of nationally
representative samples of women. Although such data are not available, the
studies included in the meta-analysis by Rauscher et al (4) do include white and
black women from a spectrum of ages (40 years or older), regions, and SES
groups. Rauscher et al do not report a significant difference between the
specificity of self-reported mammography use data for black and white women
(0.49 vs 0.62); however, the 95% confidence intervals for these measures do not
overlap (0.42-0.57 vs 0.61-0.64). The number of studies that measured
sensitivity and specificity for black women was limited; thus the values
obtained by Rauscher et al may be vulnerable to sample variation, external
generalizability, and other sources of measurement error. Also, the application
of measures of validity from the random-effects meta-analysis does assume
study-to-study variability and suggests uncertainty in estimating the underlying
parameters (ie, sensitivity and specificity) (4).
Additional limitations are that the sensitivity and specificity measures that
we used did not account for SES. Although sensitivity and specificity may differ
between women with higher and lower SES, there is a dearth of literature in this
area (4,25). The sensitivity and specificity measures also did not account for
other factors such as attitudes and women’s knowledge of breast cancer and
screening mammography. Another limitation is that the BRFSS questionnaire
(similar to other comparable national surveys) does not distinguish whether a
woman received a mammogram for screening or diagnostic purposes. Finally, the
median state and territorial Council of American Survey Research Organizations
response rates for BRFSS have been low in recent years; from 2000 through 2006
they ranged from 49% to 58%. Median cooperation rates during the same period
ranged from 53% to 77% (26).
Most studies that have measured the validity of mammography survey questions
were conducted in the 1990s. These studies should be repeated to confirm whether
sensitivity and specificity have changed. Such studies should be conducted in
diverse populations and include an assessment of data sources such as medical
and billing records.
In addition to updating validity measures of the standard wording in surveys
of mammography use, it would also be useful to identify alternative wording that
might have higher validity. Most surveys, including BRFSS and the National
Health Information Survey, use the following introductory wording: “These next
questions are about mammograms, which are X-ray tests of the breast to look for
cancer.” In 1992, BRFSS used different introductory wording: “I would like to
ask you a few questions about a medical exam called a mammogram. A mammogram is
an X-ray of the breast and involves pressing the breast between two plastic
plates.”
The reported prevalence of mammogram use was lower when this more graphic
wording was used than when standard wording was used; this reduction in
prevalence was greater among black women than white women (27). These effects on
measured prevalence are consistent with the hypothesis that questionnaire
language that clarifies that a mammogram involves “pressing the breast between
two plastic plates” improves the specificity (ie, results in fewer
false-positive responses).
Cultural sensitivity and awareness should be applied when addressing
black-white and other racial/ethnic disparities in breast cancer
detection and treatment. To be effective, interventions designed to overcome
persistent inequalities must take into account differences in race, culture,
language, SES, and age (24). Our study reinforces that these considerations can
apply to the validity of surveillance data as well (4,27). Surveillance,
intervention, and policy must account for the unique characteristics of women
from each racial/ethnic group. Increasing mammography use — especially among
underserved populations — remains a priority as public health professionals
strive to eliminate breast cancer disparities and to decrease breast cancer
death rates.
Back to top
Author Information
Corresponding Author: Rashid Njai, PhD, MPH, Community Health and Program
Services Branch, Division of Adult and Community Health, National Center for
Chronic Disease Prevention and Health Promotion, Centers for Disease Control and
Prevention, 4770 Buford Hwy NE, MS K-30, Atlanta, GA 30341. Telephone:
770-488-5215. E-mail: RNjai@cdc.gov. At
the time of the study, Dr Njai was affiliated with the Epidemic Intelligence
Service assigned to the Division of Adult and Community Health.
Author Affiliations: Paul Z. Siegel, Jacqueline W. Miller, Youlian Liao,
Centers for Disease Control and Prevention, Atlanta, Georgia.
Back to top
References
- 3-13. Increase the proportion of women aged 40 years and older who have
received a mammogram within the preceding 2 years. Healthy People 2010.
http://www.healthypeople.gov/2010/document/html/objectives/03-13.htm. Accessed
February 24, 2011.
- Ryerson AB, Miller JW, Eheman CR, Leadbetter S, White MC.
Recent
trends in U.S. mammography use from 2000-2006: a population-based analysis.
Prev Med 2008;47(5):477-82.
- Newell SA, Girgis A, Sanson-Fisher RW, Savolainen NJ.
The
accuracy of self-reported health behaviors and risk factors relating to
cancer and cardiovascular disease in the general population: a critical
review. Am J Prev Med 1999;17(3):211-29.
- Rauscher GH, Johnson TP, Cho YI, Walk JA.
Accuracy of self-reported cancer-screening histories: a meta-analysis.
Cancer Epidemiol Biomarkers Prev 2008;17(4):748-57.
- McPhee SJ, Nguyen TT, Shema SJ, Nguyen B, Somkin C, Vo P, et al.
Validation of recall of breast and cervical cancer screening by women in an
ethnically diverse population. Prev Med 2002;35(5):463-73.
- Zapka JG, Bigelow C, Hurley T, Ford LD, Egelhofer J, Cloud WM, et al.
Mammography use among sociodemographically diverse women: the accuracy of
self-report. Am J Public Health 1996;86(7):1016-21.
- McGovern PG, Lurie N, Margolis KL, Slater JS.
Accuracy of self-report of mammography and Pap smear in a low-income urban
population. Am J Prev Med 1998;14(3):201-8.
- Craig BM, Quinn GP, Vadaparampil ST.
Sensitivity of self-report mammography use in older women. Am J Prev Med
2009;37(5):441-4.
- United States cancer statistics: 2004 incidence and mortality. National
Program of Cancer Registries; 2007.
http://www.cdc.gov/cancer/npcr/npcrpdfs/US_Cancer_Statistics_2004_
Incidence_and_Mortality.pdf. Accessed August 19, 2009.
- Howard DL, Penchansky R, Brown MB.
Disaggregating the effects of race on breast cancer survival. Fam Med
1998;30(3):228-35.
- Smith-Bindman R, Miglioretti DL, Lurie N, Abraham L, Barbash RB,
Strzelczyk J, et al.
Does
utilization of screening mammography explain racial and ethnic differences
in breast cancer? Ann Intern Med 2006;144(8):541-53.
- Tumiel-Berhalter LM, Finney MF, Jaen CR.
Self-report and primary care medical record documentation of mammography and
Pap smear utilization among low-income women. J Natl Med Assoc
2004;96(12):1632-9.
- Holt K, Franks P, Meldrum S, Fiscella K.
Mammography self-report and mammography claims: racial, ethnic, and
socioeconomic discrepancies among elderly women. Med Care
2006;44(6):513-8.
- Champion VL, Menon U, McQuillen DH, Scott C.
Validity of self-reported mammography in low-income African-American women.
Am J Prev Med 1998;14(2):111-7.
- Rothman KJ, Greenland S, Lash TL, editors. Modern epidemiology. 4th
edition. Philadelphia (PA): Lippincott Williams and Wilkins; 2008.
- Haas JS, Earle CC, Orav JE, Brawarsky P, Keohane M, Neville BA, et al.
Racial segregation and disparities in breast cancer care and mortality.
Cancer 2008;113(8):2166-72.
- Blackman DK, Bennett EM, Miller DS.
Trends
in self-reported use of mammograms (1989-1997) and Papanicolaou tests
(1991-1997) — Behavioral Risk Factor Surveillance System. MMWR Surveill
Summ 1999;48(6):1-22.
- Sabatino SA, Coates RJ, Uhler RJ, Breen N, Tangka F, Shaw KM.
Disparities in mammography use among US women aged 40-64 years, by race,
ethnicity, income, and health insurance status, 1993 and 2005. Med Care
2008;46(7):692-700.
- Curtis E, Quale C, Haggstrom D, Smith-Bindman R.
Racial
and ethnic differences in breast cancer survival: how much is explained by
screening, tumor severity, biology, treatment, comorbidities, and
demographics? Cancer 2008;112(1):171-80.
- Sassi F, Luft HS, Guadagnoli E.
Reducing racial/ethnic disparities in female breast cancer: screening rates
and stage at diagnosis. Am J Public Health 2006;96(12):2165-72.
- Vainshtein J.
Disparities in breast cancer incidence across racial/ethnic strata and
socioeconomic status: a systematic review. J Natl Med Assoc
2008;100(7):833-9.
- Lannin DR, Mathews HF, Mitchell J, Swanson MS, Swanson FH, Edwards MS.
Influence of socioeconomic and cultural factors on racial differences in
late-stage presentation of breast cancer. JAMA 1998;279(22):1801-7.
- Rawl SM, Champion VL, Menon U, Foster JL.
The
impact of age and race on mammography practices. Health Care Women Int
2000;21(7):583-97.
- Kreuter MW, Sugg-Skinner C, Holt CL, Clark EM, Haire-Joshu D, Fu Q, et al.
Cultural tailoring for mammography and fruit and vegetable intake
among low-income African-American women in urban public health centers. Prev
Med 2005;41(1):53-62.
- Lawrence VA, De Moor C, Glenn ME.
Systematic differences in validity of
self-reported mammography behavior: a problem for intergroup comparisons? Prev Med 1999;29(6 Pt 1):577-80.
- Behavioral Risk Factor Surveillance System summary data quality reports,
2000-2006. Centers for Disease Control and Prevention; 2007.
http://www.cdc.gov/brfss/technical_infodata/
quality.htm. Updated April 30,
2009. Accessed July 21, 2009.
- Siegel PZ, Qualters JR, Mowery PD, Campostrini S, Leutzinger C, McQueen DV.
Subgroup-specific effects of
questionnaire wording on population-based estimates of mammography
prevalence. Am J Public Health 2001;91(5):817-20.
Back to top
