Impact of the National Breast and Cervical Cancer Early Detection Program on mammography and pap test utilization among white, Hispanic, and African American women: 1996–2000†
Presented at Exploring Models to Eliminate Cancer Disparities Among African American and Latino Populations: Research and Community Solutions, Atlanta, Georgia, April 21-22, 2005.
Prevention, including routine cancer screening, is key to meeting national goals for the elimination of death and suffering due to cancer. Since 1991, the U.S. government has invested in programs such as the National Breast and Cervical Cancer Early Detection Program (NBCCEDP) to detect breast and cervical cancer early among uninsured low-income women. A concomitant goal is reducing racial disparities in screening and early detection, and the NBCCEDP program targets low-income women who are more often racial and ethnic minorities. This paper analyzes data to test for effects of the NBCCEDP and other determinants of screening across racial/ethnic groups. We used data from the Behavioral Risk Factor Surveillance System (BRFSS) for 1996 through 2000. These data indicate that gaps in testing for breast and cervical cancers between African American and non-Hispanic white women aged 40–64 years have closed but remain for Hispanics. Multivariate findings indicate that the longevity of free screening sites through the NBCCEDP significantly increased both tests for non-Hispanic white women. The data do not confirm this effect for other racial and ethnic groups. Analysis did indicate that public insurance, or Medicaid, was equal to private insurance in promoting increased testing for African Americans and Hispanics, but not for non-Hispanic whites. Assuring that Medicaid remains available for women in this nonelderly group and increasing access to free screening sites can lead us closer to national screening goals, yet policies still need to address racial/ethnic disparities in insurance and service delivery. Cancer 2007. © 2006 American Cancer Society.
Prevention and early detection are key to reduction of the incidence and severity of many chronic diseases. Screening for breast and cervical cancer is supported by scientific evidence reflected in guidelines, and these tests are widely covered by private insurance and government programs. Although used by a large proportion of the age and gender groups for which they are recommended, poor, and especially uninsured, women are far less likely than others to receive the recommended screening.1, 2 To help reduce disparities in breast and cervical cancer test use, in 1991, the U.S. government began to invest in the National Breast and Cervical Cancer Early Detection Program (NBCCEDP). States receive funding for the NBCCEDP from the federal government, channeled through the Centers for Disease Control (CDC), and some states provide supplemental funds beyond their required match for the program.
In 2000, the release of the Healthy People 2010 (HP2010) Goal to eliminate health disparities further underscored the federal government's commitment to facilitating use of these important cancer tests.3 In this paper, we build on earlier findings showing that the longevity of the NBCCEDP in a state was associated with use of mammography to examine the effect of the availability and longevity of the NBCCEDP on non-Hispanic white, African American, and Hispanic women's cancer testing between 1996 and 2000 within a state.
Breast cancer is the most common site of new cancer among women and second to lung cancer as a leading cause of cancer deaths among women. Since the majority of known risk factors (age, family history, early menarche) for this cancer are not easily amenable to change and testing rates could be improved, the national policy focus has been on early detection and effective treatment. Although the overall HP2010 objective for mammography has been reached, the goal has not been attained among all subpopulations, and disparities remain. Moreover, the HP2010 goals for cervical cancer have not been met, even though this cancer can be prevented with Pap testing and appropriate follow-up. Disparities in use of cancer testing are related to differences in income, insurance, race, or ethnicity.4 Although we analyze all women aged 40–64 years, we are especially interested in whether women with health care access barriers obtain cancer testing.
Even though the economic growth of the later 1990s increased private insurance coverage for some low-income women, public insurance coverage slipped and racial/ethnic disparities were not eliminated.5, 6 The net result was an increase in uninsured mothers of school-aged children.7 Among working-age adults, there was a slight decline in the uninsured among non-Hispanic whites, while coverage remained stable among African Americans and Latinos during 1997–20018 and racial disparities remained in 2001.9
Other policy changes during the 1990s—such as the growth of the NBCCEDP—were especially targeted to increase cancer testing among low-income women. The NBCCEDP, established by Congress in 1990 and administered through the CDC, channels federal dollars to provide breast and cervical testing for low-income uninsured women, up to 250% of the federal poverty level in most participating states. This program has been seen as generally positive, although the fact that federal funding levels allow it to reach only 12%–15% of those eligible means that significant challenges remain.10 State participation in the NBCCEDP program grew steadily in the early and mid-1990s, so that by 1995, 70% of the states had contracted with some NBCCEDP testing sites, and as of 1998, all states had. Although a recent study reported an overall positive effect on testing from the longevity of this program,11 it did not address racial/ethnic variations in the program's effect.
Our study does address questions about racial/ethnic variation, specifically:
We use data for 1996–2000 from the Behavioral Risk Factor Surveillance (BRFSS) linked to state data on the NBCCEDP program to examine these questions using both descriptive and multivariate analyses. We received NBCCEDP program data on floppy disks in the form of Excel® spreadsheets from the Division of Cancer Prevention and Control within the National Center for Chronic Disease Prevention and Health Promotion. These data included 1) the year of initiation, 2) the federal funding for the NBCCEDP per year, 3) the number of NBCCEDP sites, and 4) the percentage of women covered. However, data on NBCCEDP sites and the percentage of women covered were not complete either across states or over our study time period. Hence, we used the longevity and amount of federal dollars per poor women as our key measures in our previous work11; in both instances, we linked the state data from the CDC to the BRFSS observation based on the woman's state of residence. From our earlier work, we expect the age of each state's NBCCEDP program and level of federal funding to be key predictors of mammography and Pap tests for women residents aged 40–64 years.
We also consider whether other variables amenable to policy but less directly related to delivery of Pap and mammography—insurance coverage and income level—are relatively more important to testing for some racial and ethnic groups. Because public programs are limited to low-income women and a larger percentage of non-Hispanic African American and Hispanic women have low income than do non-Hispanic white women, we anticipated that these programs would be more important to the minority groups.
A large body of literature has clearly established that economic variables such as health insurance and income affect the receipt of health care services, including cancer screening.4, 12, 13 Insurance has been found to be more directly associated with receipt of cancer testing than income.14 This is important, as the growth in the number of uninsured Americans resulted in almost 16 million women without health care coverage in 2002.9 Without coverage, women may forgo needed medical services, including screening: uninsured women are 40% less likely to have had a recent mammogram, 26% less likely to have completed a clinical breast examination, and 23% less likely to have had a Pap test than are insured women.15, 16
There are clear differences in the enabling factors related to cancer testing across racial/ethnic groups. Minorities are more likely to be in the low-income strata and either uninsured (Hispanics) or publicly insured (African Americans) than are non-Hispanic white groups.9 Non-Hispanic African Americans and Hispanics tend to have less net wealth than non-Hispanic whites,17 and this nontransitory income can be important to the utilization of health care services.18 After controlling for insurance, income, and medical costs (but not wealth), Adams and colleagues found that race and ethnicity were significant determinants of testing.11 A subsequent study found that African Americans, who are similar to non-Hispanic whites in terms of enabling factors, were just as likely to be tested as non-Hispanic whites. This did not hold for Hispanics, however, indicating that other factors explain their different screening rates.19
As noted, the NBCCEDP program especially targets racial/ethnic minorities. During 1991–1995, women tested by this program were disproportionately non-Hispanic African Americans and Hispanics,20 and national data from the early years of the program (1991–94) showed relatively greater increases in mammography use among non-Hispanic African American women than non-Hispanic white women aged 40–64 years.1 Although in 2000 non-Hispanic African American women continued to report mammography use at the same rate as did non-Hispanic white women, Hispanic women lagged behind them. There was no striking improvement in breast cancer screening rates for Hispanic women or for women without a usual source of care or insurance.2
Even when policies reduce disparities in insurance and income overall, if they affect utilization of health care services differently across racial/-ethnic groups, disparities may persist. Koh and Francis found evidence of this for infant mortality policy.21 A review by Shavers and Brown found that in most facilities, African American women were less likely to receive standard therapy for breast and other cancers within a given stage of cancer.22Unequal Treatment, a recently released IOM report, suggests racist practices shape the delivery of medical care in the U.S.23 Clearly, a broad set of factors help explain racial/ethnic disparities: cost barriers, poor services in poor communities, cultural and communication barriers, ‘fear’ or mistrust of the health care system, and general problems in the relationship and communication between patients and providers.23 There may also be systematic differences in the fear of cancer diagnosis across racial/ethnic groups. Although cultural barriers also seem important, they are often not readily measured because they require in-depth surveys. When they have been studied, researchers found that access factors and prior screening are more strongly associated with current screening.24, 25
In this paper, we limit our investigation to the economic variables of insurance and income, and to key policy variables—the longevity and federal subsidization of the state's NBCCEDP program—as we examine the likelihood that women from differing racial/ethnic backgrounds receive timely cancer screening. The effects of the NBCCEDP, controlling for income, insurance, and cost barriers, have not been examined by race/ethnicity. We contribute to the literature by examining trends in cervical and breast cancer testing from 1996 to 2000 for white non-Hispanic, African American non-Hispanic and Hispanic women, aged 40–64 years.
Data and Methods
The BRFSS is a federally funded survey designed by the CDC implemented in collaboration with state health departments. Random-digit dial methods are used to derive a probability sample of households with telephones to collect data on health-related behaviors and risk factors for respondents 18 years and older. Annual core survey data included sociodemographics, risk factors, and until 2001, the receipt of breast and cervical cancer tests. Using the definition of response rates issued by the Council of American Survey Research Organization, and consistent with response rates for other telephone surveys, median BRFSS response rates were higher in 1996 (63.2%) than in 2000 (48.9%).26 However, bias in 2000 was found to be lower because it indicated a random variation not associated with the response rate.27 State samples range from 1179 to 5908 across the study period.
The self-reported data in the BRFSS are used to derive annual mammography and Pap smear testing rates for women aged 40–64 years. Despite early concerns about the reliability of self-reported data, in some studies it has been shown to correlate highly with chart audits,28 although other studies have shown some amount of overreporting with self-reported data relative to chart audits.29, 30 Further, telephone surveys are one of the least expensive methods of obtaining relatively accurate information on health services use. The BRFSS survey questions for breast and cervical cancer testing use are as follows:
“A mammogram is a radiograph of each breast to look for breast cancer. Have you ever had a mammogram?” “How long has it been since you had your last mammogram?”
“A pap smear is a test for cancer of the cervix. Have you ever had a Pap smear?” “How long has it been since you had your last Pap smear?”
Using responses to these questions for women respondents aged 40–64 years, we developed 2 outcome variables: 1) mammography within the past year; and 2) Pap test within the past year based on the American Cancer Society (ACS) guidelines (Table 1). The BRFSS also asks whether the test is part of a routine examination or related to a cancer problem; we use this additional information to define routine tests in parts of our analysis. We limit our analysis to women aged 40–64 years because a) early screening is most important for this age group, b) they represent an age group especially vulnerable to becoming uninsured in the U.S., and c) this is the core age group covered by the NBCCEDP.
Table 1. Screening Guidelines from American Cancer Society (ACS)
| Aged ≥40 y, yearly|
| Sexually active or aged ≥18 y, yearly Pap test and pelvic examination|
| ≥3 consecutive normal annual exams, physician discretion|
We pooled 1996–2000 BRFSS data to estimate multivariate logit models. Drawing on previous research on health care utilization and our earlier models on screening,11 we used the following individual controls for nonelderly women aged 40–64 years: 1) household income, 2) insurance status, 3) race/ethnicity, 4) age, 5) education, 6) whether pregnant, 7) child younger than 5 years in household, 8) reported medical cost barriers, 9) employment status, 10) urban location, and 11) reported health status. The BRFSS asks about race and whether the respondent is of Hispanic origin. We used these responses to create the following categories: non-Hispanic white, non-Hispanic African American, Hispanic, and non-Hispanic other. Following convention, we refer to these groups as white, African American, and Hispanic in the remainder of the paper. Because the numbers in the non-Hispanic ‘other’ category are very small and heterogeneous, we did not include them in this analysis.
In the BRFSS, insurance status at the time of interview is asked. We examine 3 major categories: private (including CHAMPUS), public (Medicaid and/or Medicare), and uninsured. Privately insured is the omitted or reference category. The BRFSS measures income as a categorical variable; these have been collapsed into <$25,000, from $25,000 to $49,999, and equal to $50,000 or greater. Dummy variables for the above-mentioned income categories were created; a dummy for missing income data was also included in the equation. The reference category is ‘less than $25,000.’
We test 2 state-level measures of the NBCCEDP program. To compute the age of the state's program in years, the initiation year of the state's NBCCEDP initiation was subtracted from 2000. All states are given a value for this variable, even if they did not have a program in 1996 (age = 0); the age range over the full study period is from birth to 9 years. To calculate federal NBCCEDP dollars/poor woman for each state and year, we used CDC data on annual federal payments with data from the Current Population Survey on the number of women with household income <300% of the federal poverty level. ‘Real’ dollar measures were found by adjusting the federal NBCCEDP dollars for medical care inflation, using the change in the Consumer Price Index (all urban consumers) over the 1996–2000 time period. We also included a year-specific dummy variable (1996 is the omitted category) to capture the effects of any independent secular trends affecting screening rates. For example, increased public education/promotion that is done at a national level may increase the awareness of women in all states and hence increase testing.
Finally, we used fixed-effects models to capture state-specific characteristics that are invariant over time, such as sparsely populated rural states where travel times may be longer but do not change over time, and adjusted standard errors for repeated within-state observations and complex survey design. We tested Huber-White corrections for repeated observations of the same state overtime and found no differences in the results. We note that the time trend analyses presented here will differ from those of cross-sectional analyses, since they measure the effect of increases in federal aid to states in addition to differences in federal funding across states, whereas cross-sectional analysis measures only the latter.
Table 2 indicates that African Americans had the greatest increase (over 10 percentage points) in mammography rates and ended up with the highest average rates of annual mammography in 2000. The rates for African Americans, almost 65%, were comparable to rates for whites (63%). Even with increases during 1996–2000, Hispanics continued to have lower rates than white non-Hispanics or African Americans for both cancer tests. Yet the differences between these groups narrowed between 1996 and 2000, as gains for Hispanics relative to non-Hispanic whites were greater for both types of tests over this period.
Table 2. Annual Papanicolaou and Mammography Rates for Women Aged >40 Years by Income, Insurance, Education, and Race or Ethnicity 1996–2000, All Screens
| >$25,000 but <$50,000||64.2||66.3†||54.2||60.4*|
| Low income <$25,000, insured||57.5||64.1*||51.0||61.5*|
| Low income <$25,000, uninsured||38.4||42.6||27.0||31.7‡|
| <High school||50.0||58.8*||46.6||53.2*|
| High school graduate||61.2||66.0*||52.8||61.3*|
| Some college||65.2||68.1*||54.4||62.4*|
| College graduate||71.7||72.8||59.0||67.2*|
|Race or ethnicity|
| Non-Hispanic white||63.9||68.0*||54.5||62.7*|
| Non-Hispanic African American||66.5||69.9||53.1||64.8*|
Despite reduction in disparities by racial/ethnic groups, large disparities remained between insured versus uninsured as well as between lower versus higher income groups. The data in Table 2 show that 59% of low-income women (<$25,000) were tested for cervical cancer in 2000, compared with almost 75% of those with income more than $50,000. Although disparities remained in 2000, the differences across income groups narrowed between 1996 and 2000 for the Pap test. Disparities across income groups for mammography, however, remained virtually the same in 2000 as in 1996, equal to about 15 percentage points. This appears to be due in part to differential changes in rates across insurance groups over time. That is, there was a marked increase in cervical cancer testing among publicly insured (increasing almost 10 percentage points) relative to privately insured (increasing around three percentage points) among women aged 40–64 years over this period. Meanwhile, increases in mammography use within the publicly insured and the private insured groups were similar in magnitude (around 8 percentage points).
The uninsured are clearly disadvantaged with respect to timely cancer screening. Their rates were far below those for the insured in 1996 and in 2000 even within the same low-income strata (see Table 2). In 2000, around 42% of the low-income uninsured received Pap tests, whereas 64% of the low-income insured did. That same year, the insurance-related gap in mammography rates among low-income women was almost 30 percentage points; 32% of low-income women without insurance received a mammogram versus 62% of low-income women with insurance. The data in Table 2 show no significant increases for low-income uninsured women from 1996 to 2000 for the 2 tests examined here. We did find significance for routine testing among low-income uninsured (data not shown), as noted in a footnote to Table 2; we further test this finding in our multivariate analyses.
Although these descriptive patterns suggest that differences in the rates of testing narrowed between racial/ethnic groups over this time period, these patterns could reflect changes in the underlying determinants (e.g. income, insurance) for each racial/ethnic group, effects of the NBCCEDP program, or other factors. It is also possible that some determinants are of more import for 1 racial/ethnic group versus another. In the next part of the analysis, we use multivariate analysis to examine the importance of the key determinants of testing for the 3 racial/ethnic groups, including the longevity of the NBCCEDP and its federal funding.
We present the results of selected independent determinants on mammography and Pap testing within the last year separately for African Americans, Hispanics, and white non-Hispanics in Tables 3 and 4. An adjusted Wald test indicated that the set of coefficients were significantly different across the 3 racial/ethnic groups, and hence we present separate equations for each. We denote whether the odds ratios are significantly different across the racial/ethnic groups by comparing the confidence intervals for African Americans and Hispanics for overlap with those for white non-Hispanics. The significance of the odds ratio as well as differences across racial/ethnic groups is noted in the tables.
Table 3. Adjusted* Odds Ratios for Mammography by Income and Insurance Categories. Pooled 1996–2000 State Data for Women Ages 40–64 by Race or Ethnicity†
| <$25,000, ref cat||—||—||—|
| >$ 25,000 but <$50,000||1.15§ (1.00–1.33)||1.04 (.85–1.29)||1.24‖ (1.17–1.31)|
| ≥$50,000||1.29§ (1.06–1.58)||1.36§ (1.05–1.75)||1.53‖ (1.44–1.63)|
| Private insurance, ref cat||—||—||—|
| Public||.87 (.73–1.04)||.93 (.71–1.21)||.86‖ (.79–.93)|
| Uninsured||.49‖¶ (.41–.58)||.44‖¶ (.35–.54)||.33‖¶ (.31–.36)|
|Medical costs a barrier|
| Yes||.65‖ (.56–.75)||.71‖ (.58–.87)||.62‖ (.58–.66)|
| No, ref cat||—||—||—|
| Program maturity||1.13 (.85–1.49)||1.12 (.92–1.36)||1.05‖ (1.01–1.08)|
| Federal dollars/1000 poor women||1.03 (.93–1.15)||.97 (.86–1.09)||1.00 (.97–1.02)|
| 1996, ref cat||—||—||—|
| 1997||1.14 (.82–1.57)||.83 (.61–1.14)||.98 (.91–1.04)|
| 1998||.95 (.54–1.69)||.73 (.47–1.12)||1.02 (.94–1.11)|
| 1999||.89 (.39–2.05)||.77 (.42–1.40)||1.09 (.97–1.20)|
| 2000||1.01 (.33–3.08)||.95 (.40–1.92)||1.13 (.98–1.30)|
Table 4. Adjusted* Odds Ratios for Pap Smear by Income and Insurance Categories. Pooled 1996–2000 Data for Women Ages 40–64 by Race or Ethnicity†
| <$25,000, ref cat||—||—||—|
| >$ 25,000 but <$50,000||1.14 (.98–1.32)||.99 (.80–1.23)||1.17§ (1.11–1.23)|
| ≥$50,000||1.45§ (1.16–1.81)||1.48§ (1.11–1.98)||1.48§ (1.39–1.57)|
| Private insurance, ref cat||—||—||—|
| Public||.93 (.77–1.11)||.88 (.67–1.14)||.88§ (.81–.95)|
| Uninsured||.46§ (.39–.55)||.48§ (.38–.59)||.38§ (.35–.40)|
|Medical costs a barrier|
| Yes||.68§ (.58–.78)||.70§ (.57–.85)||.70§ (.66–.74)|
| No, ref cat|| || || |
| Program maturity||1.12 (.88–1.43)||1.19 (.97–1.46)||1.04* (1.01–1.08)|
| Federal dollars/1000 poor women||1.01 (.90–1.13)||1.02 (.90–1.15)||1.02‖ (1.00–1.05)|
| 1996, ref cat||—||—||—|
| 1997||1.10 (.81–1.48)||.98 (.71–1.35)||.99 (.92–1.06)|
| 1998||.84 (.51–1.39)||.81 (.52–1.36)||.89§ (.82–.97)|
| 1999||.73 (.36–1.48)||.69 (.36–1.29)||.95 (.85–1.06)|
| 2000||.69 (.27–1.80)||.66 (.29–1.51)||.94 (.82–1.08)|
The results in Table 3 indicate that insurance and income are independent factors affecting mammography for all 3 racial/ethnic groups; however, being publicly rather than privately insured significantly lowers the relative odds of testing (.86) only for whites. For African Americans and Hispanics, there is no statistically significant difference between these odds. Moreover, while being uninsured significantly lowers the odds of testing for all racial/ethnic groups, it lowers them more for whites. For whites, the odds for the uninsured are 0.33 those of the privately insured, whereas the odds for African Americans are 0.49. The confidence interval for this odds ratio does not overlap that for whites and is denoted as significantly different.
Those who report medical costs as a barrier have lower odds of using mammography even after accounting for insurance coverage, and this finding held for all racial/ethnic groups. The odds of being tested for those reporting medical cost barriers range from 0.62 (whites) to 0.71 (Hispanics) relative to those not reporting cost barriers. The time trend variable is not significant, as shown in Table 3, indicating no evidence of an independent secular trend.
A key result shown in Table 3 is the significant, positive effect of being in a state with an older NBCCEDP program on the odds of receiving a mammogram in the past year for white women. The higher odds ratio (1.05) for whites in states with older programs is similar to the overall effect reported in an earlier paper.11
Table 4 presents multivariate results for Pap tests separately for the 3 racial/ethnic groups. Again, the key economic variables of insurance and income affect the odds of being tested for all groups, but the magnitude of the effects differs. For example, having public versus private insurance (the reference category) significantly lowers the odds of testing only for whites. Moreover, the odds of white non-Hispanic uninsured are 0.38 of the privately insured, compared with those for African American uninsured (.46) and Hispanic uninsured (.48), although the odds ratios are not statistically different across the racial/ethnic groups. Medical costs are a barrier to Pap testing for all racial/ethnic groups as they were for mammography; those reporting cost barriers have lower odds of getting the test, in the range of 0.68 (African American) to 0.70 (white and Hispanic), compared with those without such barriers.
Finally, similar to mammography, the older the state's NBCCEDP program, the higher the odds that a white woman had received a Pap test for cervical cancer. Again, while the magnitude of these odds is much higher for African American (1.12) and Hispanic (1.19) than for non-Hispanic white women, the BRFSS data do not indicate statistical significance, except at the P = .09 level for Hispanics.
It is perhaps easier to interpret the results in terms of predicted probabilities rather than odds ratios. To this end, in Table 5 we present mean predicated probabilities by income, insurance for differing ages, and federal funding of the NBCCEDP program. In Tables 3 and 4, we present these results for variables showing statistical significance. As the data in Table 5 show, the predicted probability of the uninsured receiving mammography or pap tests is lower for white non-Hispanic uninsured (26% and 36.7%) than for uninsured African Americans (38.4% and 48.4% respectively) or Hispanic uninsured (35.0% and 45.5% respectively). The effects of increased longevity are illustrated with predicted probabilities for white non-Hispanics at 1-, 3-, and 5-year intervals. These results indicate that the probability of mammography testing increases from 54.2% to 58.5% and for Pap, from 64.4% to 66.2%, across these time intervals. Increases in federal funding per poor women of the magnitude simulated in Table 5 could significantly increase probabilities of screening. However, mean federal funding per (1000) poor women for whites was only $2113 in 2000 (data not shown); for non-Hispanic African Americans, the mean was $1711, and for Hispanics the value was $1595. Differences in means reflect differences in federal funding in states where these subgroups reside.
Table 5. Predicted Probabilities of Mammography and Pap Screening by Income and Insurance Categories, Longevity of NBCCEDP and Levels of Federal Dollars/1000 Poor Women*
| >$ 25,000 but ≤$50,000||61.9||56.8||57.1||73.5||66.4||64.7|
| Private insurance||63.0||60.6||61.6||74.5||70.5||69.7|
| Public insurance||58.8||59.1||54.8||66.7||62.9||57.5|
| 1-y longevity||—||—||54.2||—||—||64.4|
| 3-y longevity||—||—||57.3||—||—||65.8|
| 5-y longevity||—||—||58.5||—||—||66.2|
|Federal dollars/1000 poor women|
Our results confirm that there are positive effects from the length of time the CDC BCCEDP program has been in a state. However, the BRFSS data confirm this only for white non-Hispanic women. Even though the point estimates actually indicate higher effects for African Americans and Hispanics, the confidence intervals for these groups are too large to show a significant effect. The BRFSS sample has power to test for effect differences of as little as 5%, except in 1996. Although the changes over time often exceeded these levels, the magnitude of difference in test use between whites and other racial/ethnic groups was small over much of the study period.
There is also less variation in state-based than in individual-based variables, and our measures of the NBCCEDP program are state-based. This is compounded by the concentration of minority groups in fewer states than whites. Fewer observations and less variation both contribute to less precise estimates for African Americans and Hispanics, as demonstrated by the large confidence intervals for these groups. When we tested the model on subsets of states with concentrations of African Americans and Hispanics, the results remained insignificant. Thus, although it is likely that the longevity of the NBCCEDP within a state positively impacted testing for African Americans and Hispanics, as we found for non-Hispanic whites, statistical limitations prevent us from drawing this conclusion.
The individual-level determinants of cancer testing that we analyzed vary by racial/ethnic group for both Pap and mammogram. Being uninsured or publicly insured is a greater barrier for white women than for African American and Hispanic women. Moreover, being uninsured lowers the odds of screening more for non-Hispanic whites than for African Americans or Hispanics. This may indicate more experience, knowledge, and support networks among African American and Hispanic women relative to non-Hispanic white women for using safety net providers when uninsured and for accessing public insurance or subsidized services. There is some evidence for this. A recent qualitative study of the California NBCCEDP9 found that “women seeking health care services through community health centers and public health departments may be more knowledgeable about NBCCEDP screening services than are women with a regular source of care or those seeing private doctors.” Most of the women in the focus groups were nonwhite and uninsured.
Our finding that minorities have similar odds of testing whether publicly or privately insured is of importance, given the more heavy reliance on public insurance by African Americans and Hispanics in 2000 (7.0% for whites, 9.3% for African Americans and 16.6% for Hispanics in the BRFSS data) than in 1996 (5.8% for whites, 15.6% for African Americans and 12.0% for Hispanics). Thus, the availability and ‘take-up’ of public insurance by minority women aged 40–64 years seems to have played a role in the gains in testing rates made by African Americans and perhaps in narrowing the gaps in testing for Hispanics than for whites over the 1996–2000 period.
The BRFSS data also showed that women who report medical costs as a barrier had lower odds of using mammography even after accounting for insurance coverage, and this finding held for all racial/ethnic groups. While public insurance, especially Medicaid or dual Medicaid/Medicare coverage, should remove most out-of-pocket medical costs, any remaining out-of-pocket costs for preventive care could create a cost barrier for low-income groups. In addition, some private insurance does not cover preventive services.31
There are a number of important limitations to this study. The analysis is ecologic, so that cause and effect cannot be determined. The 2 measures we evaluated, program longevity and funding per state, were the only structural measures available to us for all states and over the study time period; however, these measures may be imprecise or not sensitive enough to adequately capture the program's interventions effects on minority women.
This and issues related to sample size may underlie our inability to find a statistically significant effect of the NBCCEDP on African American and Hispanic women; that is, the lack of statistical significance on the NBCCEDP age variable for minorities is likely related to sample size issues. As noted, the BRFSS has adequate power (90% power; 95% confidence level) for detecting a relatively small difference (5 percentage points from baseline) in testing rates across the races but the actual magnitude of these differences is smaller than that during much of our study period. Smaller BRFSS sample sizes for African Americans and Hispanics are compounded by the concentration of these racial/ethnic groups in certain states, resulting in less variation in the state-based NBCCEDP program variable for them. Models tested on subsets of states with concentrations of African Americans and Hispanics still showed no significant results at P = .05 or lower for these groups.
Another limitation is that behavioral measures are self-reported and these are subject to recall and social-desirability bias. The measure of screening used is the most recent screening event and not routine or regular screening, which is what is needed to ensure early detection (breast cancer) or prevention (cervical cancer). By using the questions regarding testing over a shorter time period, we tried to minimize recall bias. We did estimate models on measures of routine testing and found results similar to those presented here. One exception in this modeling was significance of the longevity variable for African Americans, but only at the P = .06 level.
Another limitation is that other important determinants related to physician visits and physician recommendations to obtain cancer screening were not available in the BRFSS. We also could not use the BRFSS to proxy which women were actually eligible for NBCCEDP programs. Rather, we relied on structural measures to test its effect. As described throughout this text, we tested several factors, such as insurance, income, and reported cost barriers, that could affect access. We may have overadjusted by including the latter measure. When we tested models that omitted this variable, however, results were stable.
A number of other potentially important determinants were not included as controls in our analysis. These include capacity measures such as number and proximity of providers per woman served and outreach efforts from the community screening programs. These variables also were not available to us.
Significant strides in reducing racial/ethnic disparities in mammography and Pap testing rates were made between 1996 and 2000. This analysis examined which determinants helped reduce racial/ethnic disparities in cancer testing. We found that public insurance programs, largely Medicaid for the 40- to 64-year-old age group, were especially important for African Americans and Hispanics. These groups are more likely to be publicly insured than whites and have similar odds of being tested regardless of whether they have public or private insurance coverage.
Our results indicate the NBCCEDP program is significantly associated with increased testing among low-income, uninsured white women. It likely also positively affected minorities, though this could not be ascertained with the data available to us. That the program is significantly associated with increased testing among some women is a critical finding, because the NBCCEDP is a large, nation-wide program that offers free screening for breast, cervical, and in some states, other cancers (e.g. colorectal). Since women who screen for 1 cancer are likely to be screened for others, the benefits seen here are likely understated. Moreover, the NBCCEDP provides additional services in terms of treatment referrals, case management, and social support. Yet, the program could be improved in several ways. The most important change would be to increase levels of federal and matching state funding. As noted, congressional appropriations allowed the NBCCEDP to test only about 12%–15% of low-income uninsured women nationally in fiscal year 2000.32 The nation would be more likely to meet its 2010 cancer screening goals for underserved subpopulations of women if funding were increased to cover more eligible women and if states were funded to further publicize the program. We note that some states' outreach efforts include monitoring of screening rates, examining racial disparities, and evaluating the role of their NBCCEDP in getting women screened at early stages.33–35 Despite increases in screening among all racial/ethnic groups, we found that large disparities persisted across income groups for both Pap and mammography testing. It is a strong indictment of health care delivery in the U.S. that we have not remedied these disparities. This is especially true for Pap testing, which is inexpensive and can be administered during a regular medical visit. Moreover, regular Pap testing prevents cervical cancer and has a strong evidence base for the recommendation that it should be a routine part of health care for every woman aged 40–64 years.36 The 2010 goal for 70% of all women to have obtained a mammogram was met by 2000, though disparities remained among racial/ethnic, geographic, and low-income groups.37 Although our results show African Americans with higher mammography rates than non-Hispanics whites, they are still lower than the ACS guidelines (HP2010 goals are biennial). African Americans have always had higher rates of Pap testing than non-Hispanic whites; this held as early as 198538 and remained true in both in 1996 and 2000. Pap testing will need to improve, however, if the nation is to meet the Healthy People goals for 90% of women aged 18 years and older to have obtained a Pap test within the past 3 years.
In addition to preventive services, women from all backgrounds need timely access to treatment. Now that Medicaid coverage for treatment of women diagnosed with breast or cervical cancer is immediately available through the Breast and Cervical Cancer Prevention and Treatment Act of 2000, low-income women and their providers may feel more comfortable taking the first step, age-appropriate screening. Our analysis showed that Medicaid and the NBCCEDP contribute to meeting these goals. Meeting the nation's Healthy People goals for important subpopulations requires addressing remaining disparities in service delivery. To do this, more eligible women will need to have access to these programs, and programs will need to be adequately funded. Even in a time of fiscal stress, it is critical for the nation to maintain its commitment to Medicaid, NBCCEDP, and other public programs that provide testing, diagnosis, and treatment of cancer for poor women.
The opinions expressed in this paper are those of the authors and do not necessarily represent those of the funding entity or the National Cancer Institute.