Racial Disparities in Breast and Cervical Cancer: Can Legislative Action Work? Part 2

Stage at Enrollment by Race

As noted, earlier enrollment could mean that women enter into Medicaid and, in turn, receive treatment while still at an early stage of disease. To derive our dependent variable, we first used stage as reported in the GCCR for women entering Medicaid within 6 months of their cancer diagnosis, regardless of what stage was reported. For those who enrolled in Medicaid more than 6 months after diagnosis, we also derived a measure of stage based on Medicaid claims. To do this, we applied the ICD-9-CM codes from the Disease Staging: Coded Criteria, v. 5.24 (The MEDSTAT Group, Inc., Ann Arbor, MI) and worked in conjunction with GCCR staff to match the staging from this Disease Staging algorithm into the scheme of in situ, local, regional, or distant, as reported in the GCCR (the full list of codes for breast and our control cancers are available upon request). For those diagnosed "early" in the GCCR and enrolled in Medicaid more than 6 months after their diagnosis, we indicated a change in their stage if, using the Disease Staging algorithm and claims data, there was an indication of disease progression to regional or distant.

We only completed the stage analysis for breast cancer cases, since the registry does not include early stage or pre-invasive cervical disease cases. The sample of women enrolled pre- and post-BCCPTA for whom we had both race and stage data was 2,668 (see Table 11.3). In multivariate analysis of the full sample (Chien, Adams, & Yang, in press) we found effects of BCCPTA on preventing disease progression by the time of Medicaid enrollment. However, that analysis was limited by the relatively small sample of cancers in our control group with stage data, and the sample was further limited when divided by race. Indeed, the small sample size precluded multivariate analysis within racial groups.

Due to the importance of this issue, however, we have presented descriptive data in Table 11.3 on the percentage of breast cancer cases for which there was evidence of "disease progression" between their time of diagnosis and Medicaid enrollment in the pre- versus post-BCCPTA period, by race. For both Blacks and Whites, there was some indication that BCCPTA improved prospects for treatment at early stage. Whereas 40% of Whites were diagnosed at an early stage of breast cancer in the pre-BCCPTA period, only 36% enrolled in Medicaid without disease progression.

TABLE 11.3 Percentage of Early Stage3 at Cancer Diagnosis and at Medicaid Enrollment for Women With Treatment (Breast) and Control Cancers Pre-BCCPTA Versus Post-BCCPTA, by Race

	Pre-BCCPTA			Post-BCCPTA
	Cancer diagnosis (%)	Medicaid enrollment (%)	Tumor progression^b (%)	Cancer diagnosis (%)	Medicaid enrollment (%)	Tumor progression^b (%)	Difference (Post – Pre)	Difference- in-difference estimation
White
Breast	40.07	36.03	-4.04	49.83	49.30	-0.53	3.51
Control	25.00	22.32	-2.68	28.36	26.87	-1.49	1.19	2.32
Black
Breast	36.22	32.20	-4.02	44.46	43.61	-0.85	3.17
Control	19.59	18.56	-1.03	19.40	19.40	0.00	1.03	2.14

BCCPTA indicates the Breast and Cervical Cancer Prevention and Treatment Act.

Pre-BCCPTA, January 1999 through June 2001 ; post-BCCPTA, July 2001 through December 2004.

Sample size = 2,668.

aEarly stage included in situ and local using SEER summary-stage methodology.

bTumor progression denoted evidence of a change from early-stage cancer diagnosis in registry to stage as measured after 6 months of Medicaid enrollment.

After BCCPTA, 50% were both diagnosed and enrolled while at an early stage. Hence, there was about a 3.5 percentage point increase in the percentage of White women with breast cancer whose tumor did not progress by the time of Medicaid enrollment. For Blacks, the percentage diagnosed at an early stage pre-BCCPTA was lower than for Whites, at 36%, and only 32% enrolled without evidence of "disease progression" in this pre-BCCPTA period. In the post-BCCPTA period, the percentage of Blacks diagnosed and enrolled at an early stage was equal, almost 44%—indicating progress, but still lower than the 50% observed for Whites in the post-BCCPTA period.

Disenrollment Patterns by Race

We completed multivariate analysis of those enrolled in Medicaid using the same statistical methods as described above for the probability of enrollment. We noted that once women were enrolled with breast or cervical cancers in Georgia’s WHMP, they were more likely to stay enrolled post-BCCPTA (see Table 11.4). In the overall sample, we found that relative to women with one of the control cancers, women with breast or cervical cancer were 50% less likely to disenroll post- versus pre-BCCPTA (Chien & Adams, 2010). When this analysis was repeated by race (see Table 11.4), there was a comparable gain for both Blacks and Whites with breast cancer; the rate of disenrollment declined due to the BCCPTA by 55-56% for both races. There was, however, an important differential seen for Blacks with cervical cancer; the BCCPTA was associated with a reduced disenrollment rate of 48% for Whites but a 65% reduction for Blacks with this cancer. For both races with these cancers, these changes translated into almost six more months of enrollment—although, as noted, there was a slightly greater increase for Blacks versus Whites with cervical cancer. This longer enrollment period should increase the probability that women have continuity of care and complete their cancer treatment regimen while under Medicaid coverage.

Treatment Patterns by Race

In Table 11.5, we have presented data on the patterns of cancer treatment that women received once they were enrolled in Medicaid under Georgia’s WHMP. In this analysis, we limited our sample to those who enrolled in Medicaid under the BCCPTA eligibility category. In this analysis, we additionally included those who were diagnosed with pre-invasive cervical disease through Medicaid enrollment files and tested for differences by race in the receipt of treatments appropriate to this disease. We estimated logistic regression models on the odds of receipt of specific forms of cancer treatment and used White patients as the reference group. We have presented the relative odds of Black patients receiving cancer treatment as identified from Medicaid claims (the treatment codes are available upon request from the authors).

TABLE 11.4 The Unadjusted and Adjusted Disenrollment Rates (per 100 Person-Months) of Treatment (Breast and Cervical) Pre- Versus Post-BCCPTA by Race

	Pre-BCCPTA				Post-BCCPTA				Percent change (Post – Pre)/pre
	Unadjusted		Adjusted³		Unadjusted		Adjusted³		Unadjusted	Adjusted³
	Mean	95% Cl	Mean	95% Cl	Mean	95% Cl	Mean	95% Cl	(%)	(%)
White
Breast	2.72	2.38-3.10	3.61	3.50-3.72	1.25	1.09-1.44	1.61	1.56-1.66	-53.94	-55.46
Cervical	5.22	4.19-6.50	6.47	6.24-6.69	2.35	1.87-2.96	3.30	3.19-3.42	-55.00	-48.94
Control	5.81	4.89-6.90	-		6.49	5.43-7.76	-		11.76
Black
Breast	2.91	2.59-3.28	3.73	3.64-3.82	1.52	1.33-1.74	1.64	1.60-1.68	-47.71	-56.03
Cervical	5.05	3.85-6.63	6.12	5.92-6.32	2.49	1.89-3.28	2.13	2.06-2.20	-50.63	-65.22
Control	4.17	3.41-5.09	-		6.82	5.65-8.22	-		63.50

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 disenrollment rate is predicted from the difference-in-difference of the parametric hazard model with Weibull distribution. Covariates included age at diagnosis, marital status, cancer stage at diagnosis, urban/rural, resident county’s resident county’s health resources, and socioeconomic status (county has a teaching hospital, percentage of service employment in county, percentage of county employment in small firms [<10 employees], percentage Medicaid recipients in county).

TABLE 11.5 Multiple Regression Analysis of the Receipt of Treatment Among (Breast and Cervical) Cancer Cases, Odds Ratios for Blacks Versus Whites by Type or Treatment

	Adjusted^a
	OR	95% CI	p-Value
Breast Cancer Cases
Any treatment	0.718	0.296-1.741	0.463
Any drug regimen	1.035	0.662-1.617	0.880
Any radiation	1.179	0.879-1.581	0.272
Any surgery	0.942	0.655-1.356	0.749
Last/definitive surgery (either lumpectomy	0.928	0.650-1.326	0.683
or mastectomy)
Lumpectomy vs. mastectomy	0.991	0.711-1.382	0.957
Lumpectomy with vs. without radiation	1 .091	0.621-1.919	0.762
Mastectomy and follow-up One vs. none	2.681	1.220-5.894	0.014
Two and more vs. none	1.726	0.807-3.692	0.160
Cervical cancer cases
Pre-invasive
Cancer work-up	0.832	0.624-1.110	0.211
Pre-invasive treatment	1.885	1.259-2.822	0.002
Simple hysterectomy	0.494	0.311-0.783	0.003
Invasive
Cancer work-up	0.700	0.231-2.118	0.528
Invasive surgical treatment	0.417	0.203-0.857	0.017
Radiation	2.145	0.941-4.893	0.070
Chemotherapy	1.035	0.530-2.022	0.919

Sample sizes: N = 979 for breast cancer cases; N = 1,580 for cervical cancer cases.

Time period of study: July 2001 through July 2006.

aWhite is reference group, covariates included age at Medicaid enrollment, cancer stage at diagnosis, Charlson comorbidity index, enrolled in Medicaid prior to month of diagnosis, continuous enrollment of over 24 months, urban/rural percentage of resident county population with income less than $15,000, resident county’s health resources (county has hospital with oncology services, has at least one Commission on Cancer approved hospital, number of obstetricians/gynecologists per 1,000 women)

Bold data denote that the outcomes were significantly different between blacks and whites at p value <= .05.

We noted that BCCPTA women differed from other Medicaid enrollees in that they were of relatively higher income and were linked more closely to their providers due to the structure of the WHMP and its recertification process. In this analysis, then, we examined differences in treatment patterns for Black versus White women all of whom were enrolled in BCCPTA. Their BCCPTA eligibility category denoted that, while of different racial backgrounds, they were: (1) similar in terms of socioeconomic status; and (2) were being treated largely within the same health care delivery system.

In the analysis presented in Table 11.5, we looked for significant differences by race after controlling for following variables: age at Medicaid enrollment; cancer stage at diagnosis; the Charlson comorbidity index; whether enrolled in Medicaid prior to cancer diagnosis; whether enrolled over 24 months; county of residence (urban/rural); percentage of resident county population with income less than $15,000; resident county’s health resources (the county has at least one hospital with oncology services and at least one Commission on Cancer approved hospital, and the number of obstetricians/gynecologists per 1,000 women). We tested for a wide array of treatments over the 24 months after enrollment, including any drug, radiation, or surgical treatment. For breast cancer, we also tested for the probability of lumpectomy versus mastectomy as the "definitive" surgery among those with any surgery. If multiple surgeries were found in the claims data, the "definitive" surgery was assigned as the last one observed. As these analyses showed, there were no significant differences for Blacks versus Whites, with one exception. Black women receiving a mastectomy were more likely to receive at least one follow-up (chemotherapy, hormonal, or other drug regimen) than their White counterparts within the Medicaid WHMP.

In Table 11.5, we also showed the patterns of treatment in WHMP for pre-invasive cervical disease or invasive cervical cancer. Here, there were differences by race. Blacks were almost twice as likely as Whites within the WHMP to receive some treatment for pre-invasive (CIN II or CIN III) cervical disease. There was also a stark difference in the rate of hysterectomy among the races. Black women were less than half as likely to have a simple hysterectomy if their disease was pre-invasive, and to have any invasive surgery (simple or radical hysterectomy, pelvic lymph node dissection, paraaortic lymph node sampling, radical trachelectomy, removal of cervical stump, and pelvic exenteration [or] pelvic evisceration) if their cervical cancer was invasive. These differences were significant even after controlling for the covariates as noted earlier.

DISCUSSION

The strength of legislative action is that it can compel entities such as states, employers, or other geopolitical jurisdictions to provide or enhance programs that are available to all those identified as eligible within the legislative language. While the BCCPTA language left it up to state option to provide the new program, a majority of states had adopted the BCCPTA and Georgia was among the first to do so, increasing its Medicaid cancer caseload by almost one third in 2003 alone (Adams et al., 2007). Although one of the mandates under the BCCPTA was that women were eligible if they were screened for breast or cervical cancer by another legislated program, the NBCCEDP, states also had "Screening Options" to designate women as having been "screened under the Program" even if their screening provider was not funded through the NBCCEDP (Centers for Medicare & Medicaid Services, 2010). This flexibility essentially allowed states to determine availability/accessibility to the new eligibility category. States that selected only to include NBCCEDP-funded providers may have inadvertently limited women’s access to the BCCPTA (Carreyrou, 2007). However, Georgia was one of 12 states that expanded its provider network beyond the NBCCEDP to any provider in the state, thereby providing greater potential access. The data indicated that in 2003, nearly 75% of the women in Georgia coming into the BCCPTA entered through non-NBCCEDP-funded providers (Adams et al., 2007).

Although the BCCPTA legislation should have had a strong effect across all states, the one national study of the BCCPTA presented troubling results in the form of increased time to treatment initiation (7-15 days) for Blacks and Hispanics with cervical cancer. This effect resulted in a significant reduction in the probability of treatment initiation within 60 days of treatment. If such effects held in other evaluations of BCCPTA nationally and within each state, this would indicate that the potential of this legislation to improve the system’s ability to achieve quality benchmarks for all racial/ ethnic groups was less than optimal. State-specific studies can shed insight on the particular effects of BCCPTA as it was implemented and managed within each state’s provider and public health systems.

The findings presented here for the state of Georgia largely showed an improvement in access to Medicaid and its providers for both Blacks and Whites. While differences were somewhat nuanced, the Georgia BCCPTA, WHMP, was associated with effects that may address long-standing disparities in breast and cervical cancer across the races. For example, the number of Black women with breast or cervical cancer enrolling in Medicaid was greater in both the pre- and post-BCCPTA periods, and the increases associated with BCCPTA meant that even more Black women enrolled per 1,000 person-months observed in the post-BCCPTA period. Given the lower incidence of breast cancer among Black women, the higher enrollment rates for Blacks might just reflect a higher likelihood that they were otherwise uninsured or that they fell under the NBCCEDP income eligibility criteria. In Georgia, it appears that the WHMP helped Black women experience a 1-month gain in time to enrollment relative to Whites and, in turn, an increase in the percentage entering Medicaid while still at an early stage of disease. Once in treatment for breast cancer, the data indicated that Blacks received similar treatment—and if there were any differences, it was more follow-up for Black women who received mastectomies.

With respect to cervical cancer cases, the effects of BCCPTA in Georgia also appeared positive. There was again an increase in enrollment, with the number of Black incident cases enrolling in the post-BCCPTA period being twice that for Whites. While the drop in time to enrollment with WHMP was not as large for Blacks as for Whites, there was a greater reduction in the rate at which Blacks subsequently disen-rolled from Medicaid. This might be associated with the significantly higher probability of treatment for Blacks versus White with pre-invasive cervical cancer in the WHMP. The only remaining significant difference in treatment patterns seen for Blacks versus Whites with cervical cancer was the markedly lower probability of surgical treatment. This difference held for both pre-invasive and invasive cervical cases and was consistent with the small literature on disparities in cervical cancer treatment. One study indicated that African Americans above age 35 were less likely to receive a hysterectomy (del Carmen et al., 1999) and another indicated that among women with stage IB cervical cancer, Black women were less likely than White women to be treated with surgery or combined therapy (Thoms et al., 1998). Both of these differentials could be related to the higher mortality rates observed for Black women with cervical cancer, and could diminish the ability of the WHMP or other insurance programs to lower the mortality rate of Blacks with this cancer.

CHALLENGES AND LIMITATIONS

Analysis of the effects of legislation poses challenges with respect to the study design and data needs as well as administrative, and even political, issues. When legislation is passed, it is of great importance for both the political leaders and the parties affected by the legislation to discern whether it has its intended effects. Yet, researchers cannot rely on randomization of persons into groups affected and not affected by the legislation in order to detect true "effects." The second-best solution is to use a quasi-experimental design, or one in which the researcher identifies the "targeted" group—those affected by the change inherent in the legislation—and a "control" group—similar to the targeted group, but not affected by the legislative change in policy. Once this is done, the researcher then needs data on outcomes and confounding factors for both of these groups and for periods before and after the change in policy.

Such a design creates extensive data needs, especially in the area of cancer control, where stage of cancer is a critical variable for analysis, and registry data is the "gold standard" for this measure. Since registry data provide only a part of the picture, many have stressed the importance of linking cancer registry data with other data, such as Medicaid and Medicare claims. These data are seen as providing valuable enhancements in terms of insurance coverage, treatment (e.g., chemotherapy), and outcomes over longer periods of time after a cancer diagnosis. To analyze the effects of legislative policy such as the BCCPTA, for example, we needed to know the timing of enrollment, details of the disease at that time, and other confound-ers or outcomes (comorbidity, length of time to treatment, disenrollment, etc.), which required data observed over a longer follow-up period than that covered by the cancer registry.

Obtaining permission to use data such as Medicaid claims requires a state-by-state effort and is really only the first step in the process. Linkage of registry and administrative data for research purposes also requires knowledge of the Medicaid program itself, the strengths and weaknesses of these data, and significant programming expertise and funding. A recent paper on the issues entailed in such efforts (Bradley, Penberthy, Devers, & Holden, 2010) noted that this type of linkage poses the following challenges: (1) understanding who "owns" the data and what this means in terms of access to it; (2) developing collaborations with these and possibly, other organizations; (3) establishing a strong case for the value of linking the data; (4) addressing the mechanics of the linkage itself; (5) assessing alternative methods of linking and finally; (6) evaluating the quality of the linkage. Our effort in Georgia entailed all of these challenges, and working out the details across multiple agencies led to significant delays in obtaining the data for analysis and, in turn, its completion.

Due to these challenges and the subsequent demands on time and resources, linkages to Medicaid data are more likely to be state-specific than national—like the linkage of Medicare and Surveillance, Epidemiology, and End Results (SEER) data. This is appropriate since Medicaid is not a uniform, national program as is Medicare and, indeed, exhibits significant variation across states. However, such studies based on linked state data will likely be limited in number and the research, as shown here, will be limited to state-specific results. This is a key limitation of our analysis. Use of only the Georgia data means that we could not generalize the results; nor could we make comparisons with the effects of BCCPTA in states with more restrictive policies. For example, we were not able to evaluate the effects of different levels of state implementation of the NBCCEDP legislation, to determine the extent to which this led to reductions in disparities in other states. If the data were linked over a longer period, we could have tested for effects of Georgia’s change in its recertification process—but this remains another limitation at the present time.

Other limitations of our analysis might include the choice of control cancers. We began with women with colorectal cancer as our control group and found that not only was the group of women entering Medicaid with this cancer relatively small in number but, due in part to the structure of Medicaid eligibility, they were far more likely to be at a regional or distant stage than those with breast cancer. Early on in our analysis, then, we expanded to include other cancers in our control group. We worked closely with physicians and registry staff in choosing these cancers, with one criterion being that screening and early detection could possibly lead to more treatment options and/or better outcomes for the cancer. Such choices may be questioned by other researchers or clinicians and this, as well as the imprecision of measuring stage for women entering Medicaid more than 6 months after their diagnosis, can be seen as additional limitations of our effort.

Despite such limitations, more power would come from having ongoing linkages that allow researchers to measure effects of BCCPTA state by state and over longer periods. Linked data would also allow for analysis of other policy changes, adherence to guidelines or other quality metrics and, if linked to private claims data, comparison across payers. Such analyses must be left to future research efforts.

CONCLUSION

The quasi-experiment created by the BCCPTA legislation and its implementation showed the potential of our nation to expand access to insurance and treatment, in a very timely fashion, to low-income persons with clear medical needs. This legislation was implemented as the WHMP in Georgia and in such a way that women got into the program more quickly, received appropriate and perhaps more complete treatment. Further, in the case of cervical cancer, treatment was extended to those with pre-in-vasive disease in a timely manner. The latter is a highly cost-effective use of public funds, since it can prevent this disease from becoming invasive. Most important, the positive effects of BCCPTA in Georgia were shared by both Blacks and Whites, and where differences emerged, they reflected somewhat quicker enrollment, a longer time on Medicaid, and more treatment for Blacks—as seen in the case of pre-invasive cervical disease and follow-up after mammography. Another key difference seen in our data, lower hysterectomies among Blacks with cervical cancer, perhaps reflects a choice made to retain reproductive capacity. It is important, however, to assess whether these are fully informed choices and, in turn, if they are related to higher rates of mortality among Black women with this cancer. Finally, the linkage of registry and Medicaid claims data clearly enhanced our ability to look at BCCPTA, and such linkages would allow for analysis of future legislation and, indeed, measuring access for women with breast, cervical, and other cancers in the periods leading up to and following the implementation of the Patient Protection and Affordable Care Act (PPACA) in 2014.