Despite overall progress in the detection and early treatment of many cancers, disparities persist. While Black non-Hispanic women have a lower incidence of breast cancer across all ages than White non-Hispanics, their mortality rate from this cancer is higher than that for non-Hispanic Whites, and both their incidence and mortality rates for cervical cancer are higher than that experienced by non-Hispanic Whites (American Cancer Society, 2009). An important reason for disparities in mortality is that minority and other disadvantaged women are more likely to be diagnosed at advanced stages of disease, with those uninsured or Medicaid-insured consistently more likely to present with advanced stages of cancer (Chen, Schrag, Halpern, Stewart, & Ward, 2007; Elmore et al., 2005; Gwyn et al., 2004).
Breast cancer is the most common site of a new cancer among women and is second only to lung cancer in terms of deaths among women. Since many risk factors for breast cancer are not easily amenable to change, public policy has focused on early detection through screening and effective treatment of diagnosed cases (Pamuk, Makuc, Heck, Reuben, & Lochner, 1998). While studies point to continued underutilization of screening (Burns et al., 1996; Shavers & Brown, 2002; Smith-Bindman et al., 2006), the gap in mammography between non-Hispanic Blacks and Whites has narrowed, equal to about 2-3 percentage points in more recent years (Adams, Breen, & Joski, 2007; American Cancer Society, 2009). Yet, studies continue to indicate that non-Hispanic Black and Hispanic patients have an increased risk of advanced stages at diagnosis for breast and other cancers, even after controlling for insurance and other demographics (Halpern, Bian, Ward, Schrag, & Chen, 2007).
Cervical cancer is less prevalent and is preventable through early detection and treatment of precancerous conditions. With respect to invasive cervical cancer, the incidence and mortality has declined dramatically since the mid-1940s due to wide use of Papanicolaou (Pap) smear tests and detection/treatment of cervical intraepithelial neoplasia (CIN) (Casper & Clarke, 1998; Devesa et al., 1987; Schoell, Janicek, & Mirhashemi, 1999). Yet, lower income, less education, and lack of insurance (Akers, Newmann, & Smith, 2007; Bradley, Given, & Roberts, 2004) are all factors associated with higher incidence and mortality from cervical cancer, largely due to failure to detect the cancer at an early stage and the unavailability of treatment when detected (Breen, Wagener, Brown, Davis, & Ballard-Barbash, 2001; Harlan et al., 2005; Hewitt, Devesa, & Breen, 2004; Rodriguez, Ward, & Perez-Stable, 2005; Roetzheim et al., 1999; Sung, Alema-Mensah, & Blumenthal, 2002; Thorpe & Howard, 2003).
For the United States to reach the Healthy People 2020 (HP2020) goal of eliminating health disparities (U.S. Department of Health & Human Services, 2009) in breast cancer mortality and other diseases, we need more information regarding the types of interventions and/or policies that can be used to affect access to early detection and treatment for women with cancers, regardless of their racial/ethnic backgrounds.
BACKGROUND ON LEGISLATION
To help reduce disparities in breast and cervical cancer, the U.S. government began to invest in the National Breast and Cervical Cancer Early Detection Program (NBCCEDP) in 1991. Under the NBCCEDP, states received federal funding channeled through the Centers for Disease Control and Prevention (CDC) and had the option of providing supplemental funds to reach more women (Centers for Disease Control and Prevention Division of Cancer Prevention and Control, 2010). The NBCCEDP targeted racial/ethnic minorities, and during the 1991-1995 time period, the women tested were disproportionately non-Hispanic African American and Hispanic (Koh & Francis, 1990). National data have confirmed relatively greater increases in mammography use among non-Hispanic African Americans than among White women aged 40-64 in the early years of the program, from 1991 to 1994 (Makuc, Breen, & Freid, 1999). As the NBCCEDP matured, research indicated that the longevity of the program within a state was associated with significantly increased rates of both mammography and Pap tests for non-Hispanic Whites, but the data and analysis could not confirm this effect for Black, non-Hispanic women (Adams et al., 2007). However, public insurance was found to have an equal effect to that of private insurance in promoting increased testing for non-Hispanic Blacks and Hispanics, but not for White non-Hispanics, in this study. Hence, multiple public policies are likely needed to reduce disparities in screening.
Since the NBCCEDP did not cover all needed diagnostic services, nor any treatment costs, it inadvertently created a "treatment gap" for low-income and minority women (Lantz et al., 1999). Providers reported limiting the number of women screened because of the burden of finding treatment resources in this situation (Lantz et al., 2000). Such delays in treatment can be critical, as the literature suggests that initiation of treatment more than 90 days after an abnormal mammogram might be associated with decreased survival rates (Richards, Westcombe, Love, Littlejohns, & Ramirez, 1999). Subsequent legislation tried to address these concerns, and the outlook for early/appropriate treatment among poor (generally <250% Federal Poverty Level [FPL]) minority women diagnosed with breast and/or cervical cancer changed significantly on October 24, 2000 with the passage of the Breast and Cervical Cancer Prevention and Treatment Act (BCCPTA). The BCCPTA established a new Medicaid coverage option that permitted states to extend Medicaid to any uninsured woman under 65 diagnosed with breast or cervical cancer or pre-invasive cervical disease when screened through the NBCCEDP. States could also designate women as having been "screened under the Program" even if they were screened by other providers; 12 states adopted this option (Centers for Medicare & Medicaid Services, 2010). States also differed with respect to the recertification process used for BCCPTA women, provider payment rates, and delivery systems.
The BCCPTA legislation has the potential to reduce long-standing disparities across income groups and race. Even when policies reduce such gaps overall, however, disparities may continue if policies affect utilization of health care services differently across racial/ethnic groups (Koh & Francis, 1990). It is important, then, to gauge not only the overall effect of legislation such as BCCPTA but also its effects by race/ethnicity and, if possible, by geographic area. Due to its relative newness, there have been few national or state-specific studies. We discuss in this topic the findings of available studies of BCCPTA and our earlier analysis of Georgia’s BCCPTA program, called the Women’s Health Medicaid Program (WHMP). We add a new analysis of the effects of BCCPTA by race in Georgia. We also discuss the general challenges that researchers face when designing and completing analysis of legislation, so as to inform policy makers in a timely manner.
ESTIMATED EFFECTS OF BCCPTA LEGISLATION
The only national study of the effects of BCCPTA used a pre-post analysis of time to definitive diagnosis and treatment over all women and by race/ethnicity (Lantz & Soliman, 2009). The results indicated both a positive effect—a decrease in time to definitive diagnosis for cervical cancer— and a negative effect—a delay in time to definitive diagnosis for breast cancer cases. Moreover, both the positive and negative effects of BCCPTA held only for White non-Hispanics. While White women were more likely to have a delay in the time to definitive breast cancer diagnosis, this delay was not sufficient in length to indicate that BCCPTA changed the proportion of women who initiated treatment within a 30- or 60-day time period after diagnosis. Another negative effect of BCCPTA—an increase in time to treatment initiation (7-15 days) for cervical cancer—found in this study, held only for non-Hispanic Blacks and Hispanics. This delay was sufficient in length for there to be a significant reduction in the probability of treatment initiation within 60 days of diagnosis. These types of delays reduce the ability of the system to achieve quality benchmarks for these racial/ ethnic groups. The authors noted that these delays might be due to a lack of access to Medicaid-participating providers or, as noted in another study, delays in making and keeping appointments among disadvantaged and immigrant women (Ogilvie, Shaw, Lusk, Zazulak, & Kaczorowski, 2004). Lantz and Soliman (2009) also noted that women of color had longer mean times (days) to: (1) definitive diagnosis after an abnormal screening test, and (2) treatment initiation in both the pre- and post-BCCPTA periods, which is a cause for concern.
One of the few state-specific studies of BCCPTA used Massachusetts (MA) data to analyze two legislative policy changes. The first was the initiation of targeted funding for case managers through the NBCCEDP (Public Law 105-340) in 1998 and, subsequently, the implementation of BCCPTA in 2004 (it was delayed in MA). Their results indicated that low-income women (<250% FPL) in Massachusetts experienced a decrease in the probability of a delayed diagnosis (>60 days from an abnormal mammography to a diagnostic resolution) after the introduction of NBCCEDP case management, but that there was no additional improvement in this outcome from the implementation of BCCPTA (Lobb, Allen, Emmons, & Ayanian, 2010). Important to our focus on disparities, the reduction in the percentage of women with a delayed diagnosis, from 33% to 23% of those with an abnormal mammogram, did not differ by race or ethnicity. However, neither the case management nor BCCPTA policy implementation was associated with decreases in treatment delay (>90 days after abnormal mammogram) for any racial/ethnic group.
BCCPTA in Georgia
Given the latitude that states had to implement BCCPTA, as well as the wide variation observed in states’ sociodemographics and Medicaid eligibility rules that affected women with these cancers prior to BCCPTA, it is likely that the impact of this legislation on the timing of women’s enrollment, treatment, and patterns of treatment varied from state to state. To date, we know of no state-specific studies of BCCPTA other than the Massachusetts and earlier Georgia studies. The Georgia analysis used a quasi-experimental analytic design, with treatment and control groups, to better identify the effect of BCCPTA separate from other changes occurring over time that could affect access to cancer care for all low-income women. The specific statistical analysis used to examine outcomes of BCCPTA (time to enrollment, probability of enrollment while at an early stage, and disenrollment and treatment patterns) are described in later sections of this topic, along with the results on each outcome.
The Georgia analysis was possible due to the creation of a unique database. Specifically, the Georgia Comprehensive Cancer Registry (GCCR) incident cases from January 1, 1999 through December 31, 2004 were linked to Medicaid enrollment files from January 1998 through December 2005. The GCCR has been a statewide, population-based cancer registry in Georgia since 1999. After linking with patients’ encrypted social security numbers, female Medicaid enrollees diagnosed with breast, cervical, and one of five control cancers (bladder, colorectal, melanomas of the skin, non-Hodgkin lymphoma, and thyroid), and enrolled in Medicaid either prior to or after their cancer diagnosis, were identified using both the GCCR and diagnosis codes on claims. Women ever enrolled under the BCCPTA, were also identified using just enrollment data, since women with pre-invasive cervical disease or in situ stage cervical cancer were omitted from the GCCR and, yet, newly eligible for Medicaid under BCCPTA. Georgia’s Medicaid program provided monthly enrollment records as well as all inpatient, outpatient, and drug claims detail through June 2006. Thus, we could observe treatments through June 2006 for incident cases enrolled by December 2005. Using this linked file, we were able to: (1) document the timing of both Medicaid enrollment and disenrollment in relation to a woman’s cancer diagnosis; (2) identify the progression of the disease by the time of enrollment; and (3) examine the receipt of alternative treatments over the period of time she was enrolled in Medicaid.
Analysis of the effects of BCCPTA on enrollment indicated that out of 1,000 women diagnosed with breast or cervical cancer in Georgia from 1999 through 2004, the implementation of BCCPTA, or the WHMP in Georgia, led to an additional two or three more cases enrolling in Medicaid per month (Adams, Chien, Florence, & Raskind-Hood, 2009). It also resulted in these women getting enrolled in Medicaid more quickly; analysis indicated that the time between initial cancer diagnosis as reported in the GCCR and eventual Medicaid enrollment was shortened by between 7 and 8 months. Once enrolled, Georgia’s BCCPTA recertification process was easier than for other Medicaid eligibility groups (e.g., self-report, no income verification) and, hence, created a situation in which coverage was more stable. Relative to women with one of the control cancers, the rates of disenrollment from Georgia Medicaid declined over 50% for women with either breast or cervical cancer from the pre- to the post-BCCPTA period (Chien & Adams, 2010).
In this topic, we have extended the Georgia analysis by first examining whether the enrollment effects seen for the overall sample differed by race for either breast or cervical cancer cases and, in turn, whether there were differences in the rate at which Blacks versus Whites disenrolled from Medicaid. A key reason to enroll women more quickly was to allow them to start treatment while at an early stage of their cancer and, in turn, potentially reduce morbidity and mortality. An important outcome would be that women diagnosed at an early stage enrolled in Medicaid while still at an early stage of disease. We completed analysis of disease progression by time of enrollment for breast cancer cases using the stage at diagnosis as found in the GCCR data. We note that, throughout this analysis, we analyzed differences only between Blacks and Whites, and not between other races. To categorize women by race as completely as possible, we used our two sources of information, the GCCR and Medicaid enrollment files, with the registry being the "gold" standard. Both data sources contained a Hispanic category and where this was denoted, we identified these women and omitted them from the analysis. Hence, our groups are largely non-Hispanic Blacks and Whites; hereafter, we refer to them as Blacks and Whites.
BCCPTA by Race in Georgia
The sample for our analyses of the enrollment outcomes by race began with the 31,825 breast, cervical (local or later stage), and control cancer incident cases as found in the GCCR from 1999 through 2004. This population was predominately White at 74%, with 22% Black, and 4% of other racial/ethnic groups. After we excluded those aged under 19 at the time of diagnosis (n = 44), those who had enrolled after December 2005 (n = 54)—since only part of the sample could be followed after 2005—and those for whom this cancer was not their first (n = 3,431), the sample for the enrollment analysis by race was 28,296. Of these, a total of 3,404 enrolled in Georgia Medicaid in the month of, or sometime after, their diagnosis. Among these women, 49% were White and 46% were Black; the remaining 5% of other or unknown race/ethnicity were omitted from the analysis. The final sample sizes were 27,103 for analysis of enrollment rates and 3,227 for analysis of the duration between cancer diagnosis and Medicaid enrollment. We discuss later the samples used for analysis of stage at time of Medicaid disenrollment and treatment patterns for those with breast and cervical cancer and enrolled in Medicaid.
Enrollment in BCCPTA in Georgia by Race
To examine the effects of BCCPTA on enrollment among women with breast and cervical cancers in Georgia, we performed a time-to-event (here, enrollment) model. This model estimated the likelihood of the event at a given time for the woman who was not enrolled before her cancer diagnosis. The likelihood function of enrollment by time (t) is as follows:
where the hazard (h) is enrollment in Medicaid in month i after their cancer diagnosis; Cancer is a dummy variable denoting the primary cancer site; BCCPTA is a dummy indicating the pre-/post-policy implementation periods; Cancer X BCCPTA is a dummy variable denoting the interaction of the pre-/ post-BCCPTA period with the primary cancer site; X. is a vector of individual covariates; Tt is a year dummy; and Ct is a vector of county covariates that could affect the probability that a woman was uninsured. The effect of BCCPTA was derived by using difference-in-difference analysis. That is, we estimated the effect of BCCPTA by using our control cancers to "difference" out the effects of other factors that changed pre- and post-BCCPTA that could affect the probability that any woman with cancer, not just those affected by BCCPTA, enrolled in Medicaid. This was estimated using the interaction term in Equation 1.
While the widely used Cox proportional hazard rate model (nonpara-metric model) is more flexible for big datasets, our data failed to meet the assumption of proportional hazards for several subgroups. Hence, we tested and reported results of the parametric model with a Weibull distribution. The Weibull had the highest likelihood ratio across alternative functional forms and the results were robust across functions.
The data in Table 11.1 provided summary statistics on the estimated effects of BCCPTA on the numbers of women enrolling in Medicaid by race. This type of analysis was reported earlier, for our full sample (Adams et al., 2009). As these data showed, in the pre-BCCPTA period, Black women in Georgia with either breast or invasive cervical cancers were more likely to enroll in Medicaid than White women. For example, the rates of enrollment among Blacks were 5-8 per 1,000 person-months for breast and invasive cervical cancer cases, respectively, versus only 2-4 for White women with these same cancers prior to BCCPTA. The effect of the implementation of BCCPTA, or Georgia’s WHMP, was to increase the rate of enrollment for women with both cancers and for the racial groups within them. Although the rates per 1,000 person-months increased more (increases of 2-4 women per month) for Blacks than for Whites (increases of 1-2 women per month), the percentage increase was actually higher for Whites (almost 65%) than for Blacks (37% for breast, and 44% for cervical cancer cases) due to the relatively lower baseline rates of enrollment among Whites.
TABLE 11.1 Unadjusted and Adjusted Enrollment Rates (per 1,000 Person-Months) for Treatment (Breast and Cervical) and Control Cancers Pre- Versus Post-BCCPTA by Race
|
|
Pre-BCCPTA |
Post-BCCPTA |
Percentage Change (Post – Pre)/Pre |
|||||||
|
|
Unadjusted |
Adjusted3 |
Unadjusted |
Adjusted3 |
Unadjusted |
Adjusted3 |
||||
|
|
Mean |
95% Cl |
Mean |
95% Cl |
Mean |
95% Cl |
Mean |
95% Cl |
(%) |
(%) |
|
White |
||||||||||
|
Breast |
1.20 |
1.09-1.33 |
2.04 |
1.94-2.13 |
2.77 |
2.57-2.99 |
3.37 |
3.22-3.53 |
129.95 |
65.66 |
|
Cervical |
4.63 |
3.78-5.66 |
3.87 |
3.58-4.16 |
11.90 |
10.15-13.95 |
6.39 |
5.93-6.84 |
157.22 |
64.93 |
|
Control |
1.24 |
1.07-1.45 |
1.65 |
1.43-1.89 |
32.65 |
|||||
|
Black |
||||||||||
|
Breast |
5.29 |
4.81-5.83 |
5.43 |
5.23-5.62 |
10.11 |
9.37-10.90 |
7.43 |
7.17-7.68 |
90.93 |
36.86 |
|
Cervical |
10.28 |
8.03-13.15 |
8.36 |
7.79-8.94 |
23.07 |
19.07-27.91 |
12.06 |
11.26-12.85 |
124.53 |
44.18 |
|
Control |
5.16 |
4.33-6.14 |
7.74 |
6.67-8.98 |
49.96 |
|||||
BCCPTA indicates the Breast and Cervical Cancer Prevention and Treatment Act; Cl = confidence interval.
Pre-BCCPTA, January 1999 through June 2001 ; post-BCCPTA, July 2001 through December 2004.
Sample size = 27,103.
aThe adjusted enrollment rate is predicted from difference-in-difference estimation of the parametric hazard model with Weibull distribution. Covariates included age at diagnosis, marital status, cancer stage at diagnosis, urban/rural, resident county’s health resources, and socioeconomic status in resident county (county has a teaching hospital, percentage of service employment, percentage of small-firm [<10 employees] employment, unemployment rate).
We also used the data to examine time to enrollment in Medicaid pre-and post-BCCPTA, as shown in Table 11.2 . Since we had data on the actual date of diagnosis but we only had the month of enrollment, we assigned each date of diagnosis to the appropriate month and then counted the months from diagnosis to Medicaid entry or date of death. Women enrolled in the month of diagnosis were assigned a time of 0.5 months. To estimate our models on time to enrollment, we used ordinary least square (OLS) models with log transformation:
Again, the difference-in-differences was derived from the interaction term in the estimated equation.
The results in Table 11.2 were of special importance, as they indicated striking delays between the time of cancer diagnosis and eventual Medicaid enrollment prior to BCCPTA. Based on the descriptive (unadjusted) data in Table 11.2, women with breast cancer experienced delays in time to enrollment equal to 15 months for Blacks and up to 18 months for Whites during this period. While time to enrollment was shorter for women in both treatment and control groups after BCCPTA implementation, the delay decreased by 9, to almost 14, months for women with breast or cervical cancer (see unadjusted difference pre-BCCPTA vs. post-BCCPTA, in Table 11.2) versus a decline of only around 4 months for women with cancers in the control group.
After we controlled for other covariates that could affect the time to enrollment, the reduction in these delays was statistically significant, and greater for women in the treatment versus control group of cancers, indicating an "effect" of the BCCPTA. Moreover, the reduction in delays to enrollment was greater for Blacks with breast cancer (9 months) than for Whites (around 8 months), so that Blacks enrolled somewhat more quickly (around 4 months) than Whites (4.5 months) in the post-BCCPTA period. Patterns for women with cervical cancer (local or later stage) were different in that the reductions in time to enrollment were slightly greater for Whites (8.5 vs. 8 months for Blacks), so that White women enrolled within 3 months while Blacks took just a little over 3 months.
TABLE 11.2 Unadjusted and Adjusted Time (Months) From Initial Cancer Diagnosis to Eventual Medicaid Enrollment for Women With Treatment (Breast and Cervical) Versus Control Cancers Pre-BCCPTA Versus
|
|
Pre-BCCPTA |
Post-BCCPTA |
Difference (Post – Pre) |
||||||
|
Unadjusted |
Adjusted3 |
Unadjusted |
Adjusted3 |
Unadjusted |
Adjusted3 |
||||
|
|
Mean |
95% Cl |
Mean |
95% Cl |
Mean |
95% Cl |
Mean 95% Cl |
Mean |
Mean |
|
White |
|||||||||
|
Breast |
18.31 |
16.32-20.31 |
12.79 |
12.42-13.16 |
4.46 |
3.79-5.14 |
4.58 4.44-4.71 |
-13.85 |
-8.21 |
|
Cervical |
14.46 |
11.27-17.66 |
11.02 |
10.65-11.40 |
2.82 |
1.86-3.78 |
2.56 2.48-2.65 |
-11.64 |
-8.46 |
|
Control |
13.49 |
10.93-16.06 |
- |
8.98 |
7.47-10.49 |
- |
-4.51 |
||
|
Black |
|||||||||
|
Breast |
14.92 |
13.20-16.64 |
13.38 |
12.96-13.81 |
4.34 |
3.74-4.94 |
3.98 3.85-4.11 |
-10.58 |
-9.40 |
|
Cervical |
11.56 |
7.27-15.85 |
10.89 |
10.51-11.27 |
2.54 |
1.46-3.63 |
3.03 2.92-3.14 |
-9.02 |
-7.97 |
|
Control |
10.39 |
7.56-13.22 |
7.00 |
5.66-8.33 |
-3.39 |
||||
BCCPTA indicates the Breast and Cervical Cancer Prevention and Treatment Act; Cl = confidence interval.
Pre-BCCPTA, January 1999 through June 2001; post-BCCPTA, July 2001 through December 2004.
Sample size = 3,227.
aThe adjusted time from initial cancer diagnosis to eventual Medicaid enrollment is predicted from the difference-in-difference estimation of the regression model with log-linear distribution. Covariates included age at diagnosis, marital status, cancer stage at diagnosis, urban/rural, resident county’s health resources, and socioeconomic status in resident county (county has a teaching hospital, percentage of service employment, percentage of small-firm [<10 employees] employment, unemployment rate).
Stage, Enrollment, and Treatment Patterns by Race
We next used the Georgia data to examine stage at time of enrollment, dis-enrollment of those women who had enrolled in Medicaid with treatment and control cancers, and treatment patterns for those enrolled within the BCCPTA eligibility category and whom we could observe for up to 2 years within the Medicaid claims. These are discussed, in turn, in the following sections.
