Role of Medicaid in Early Detection of Screening-Amenable Cancers

In the United States, most cases of breast, invasive cervical, and colorectal cancer and subsequent mortality can be prevented with regular screening and appropriate care following an abnormal test (1). Nonetheless, racial and ethnic disparities (2–5) and disparities due to lack of insurance persist in the diagnosis of these cancers (6, 7). Although Medicaid insurance, which covers qualifying individuals with low income, is associated with later stage disease and higher mortality, prior research indicates that many Medicaid beneficiaries enroll following their cancer diagnosis (8, 9), suggesting that they may have been uninsured prior to diagnosis and had limited access to cancer screening and treatment. The Patient Protection and Affordable Care Act (ACA) expanded Medicaid eligibility and specifically emphasized preventive care including early-stage cancer detection, which could reduce cancer disparities in low-income populations (10).

Long before the passage of the ACA, the Centers for Disease Control National Breast and Cervical Cancer Early Detection Program (NBCCEDP) provided free breast and cervical cancer screening followed by enrollment in Medicaid, under the provisions of the Breast and Cervical Cancer Prevention and Treatment Act (BCCPTA; ref. 11), for the treatment of women who screened positive (12). It is estimated that the NBCCEDP has saved thousands of lives of women who are uninsured or underinsured and who meet a family income threshold below 250% of the federal poverty line (FPL; refs. 13, 14), partially mitigating the effects of being uninsured or underinsured for a narrow, but important segment of the population (15). Studies of Medicaid expansion's impact on cancer detection suggest that it's more comprehensive coverage and emphasis on prevention is associated with early-stage detection (16, 17). Medicaid expansion under the ACA provides coverage for eligible individuals with a family income below 138% of the FPL, but many states chose not to expand Medicaid and have set adult income eligibility thresholds well below 138% of the FPL (18).

Many prior studies of cancer screening in the context of the Medicaid program are from single states or are limited to claims-based data such as the Medicaid Analytic eXtract (MAX) file that lack details of the cancer diagnosis, such as stage, and do not allow for comparisons across patient samples who are and are not covered by Medicaid. The linkage of cancer registry and Medicaid enrollment data can address some of the gaps in existing literature. Using cancer registry data from the Surveillance, Epidemiology, and End Results (SEER) program linked to Medicaid enrollment files (19), we compared the percentage of persons without Medicaid with those enrolled in Medicaid and diagnosed with distant stage disease for three screening-amenable cancers, colorectal, breast and cervical, the latter two of which are covered by the NBCCEDP. We distinguished between patients who were Medicaid insured prior to diagnosis and those who enrolled in Medicaid following diagnosis and among these patients, we identified those enrolled in Medicaid through the BCCPTA. We also provided a comparison of traditional Medicaid enrollees to NBCCEDP enrollees, who may exceed standard Medicaid income thresholds in their state. However, those enrolled in Medicaid through the NBCCEDP were also uninsured or underinsured prior to enrollment and may not have received regular screening or other health care services. To our knowledge, this study is the first to directly compare traditional Medicaid with the NBCCEDP for early cancer detection. The inclusion of colorectal cancer offers an additional comparison that includes men and a cancer site without a national coverage program. Our study adds to the existing evidence about the importance of Medicaid enrollment for preventive care.

We analyzed data from the SEER program matched with the MAX Personal Summary (PS) file based on social security number, date of birth, sex and Medicare number, if available, as described in Warren and colleagues (19). Records with a match score less than 5 were not included (8). Data were from 14 SEER population-based cancer registries (Northern California, Greater California, Los Angeles, Connecticut, Georgia, Hawaii, Iowa, Kentucky, Louisiana, New Jersey, New Mexico, Utah, metropolitan Detroit, and Seattle). The PS files contain information about Medicaid enrollment from all 50 states and the District of Columbia. We excluded patients enrolled in Medicaid in multiple states in the same month, enrolled in states where they were not diagnosed (based on registry records), not male or female, age younger than 18 years or older than 64 years, enrolled in Medicaid prior to diagnosis but identified as an NBCCEDP participant, and categorized as American Pacific Islander. Sample sizes in this last category were very small when cross-tabulated with Medicaid enrollment. We further restricted our sample to SEER records associated with first and only primary tumors (i.e., sequence number = 00) of female breast, cervical, or colorectal cancer diagnosed between 2006 and 2013 (N = 392,538), excluding those with missing area level socioeconomic data (N = 5,726). Figure 1 shows how the sample was derived.

Distant stage cancer at diagnosis was the outcome of interest. Distant stage disease was defined using the “distant” SEER summary stage. We include unstaged tumors as part of the reference category along with “in situ,” “local,” “regional,” and “other.” Tumors may not be staged if they are assumed to be metastatic. Therefore, our decision to include unknown stage cases in the reference category may underestimate distant stage disease if the patient did not receive appropriate pathologic diagnostics.

We categorized patients’ Medicaid enrollment as: not insured by Medicaid, Medicaid insured within the 12 months prior to diagnosis, and Medicaid insured only in the month of or within the 12 months following diagnosis, depending on dates of Medicaid eligibility in the enrollment file. Those enrolled in Medicaid both before and after the month of diagnosis were categorized as “Medicaid insured prior to diagnosis.” We identified women who were diagnosed with breast or cervical cancer through the NBCCEDP and subsequently enrolled in Medicaid and classified their insurance as Medicaid after NBCCEDP. The distinction between NBCCEDP enrollment and Medicaid enrollment through other qualifying conditions is important because states vary in whether they provide full Medicaid benefits to women enrolled through NBCCEDP or a subset of benefits related to their cancer treatment; states also vary in whether they reimburse providers at the same level as for other Medicaid beneficiaries. As a result, these women may differ from other Medicaid enrolled women both in the care they receive and by having higher family incomes that exceed the Medicaid threshold in their state.

We controlled for patient characteristics (age at diagnosis and race/ethnicity) by Medicaid insurance and sex. Age groups were 18 to 40, 41 to 54, and 55 to 64 years. Racial and ethnic categories were non-Hispanic white, non-Hispanic Black, and other/unknown. The Yost index comprises census tract median household income, median rent, percent of residents who live below 150% of the FPL, an education index, percent of residents considered working class, and the percent of adult unemployed residents (20). We categorized the Yost index in quintiles with the first quintile representing patients who reside in census tracts with the lowest socioeconomic status and the fifth quintile representing patients who reside in census tracts with the highest socioeconomic status. Prior studies indicate that patients who live in rural communities have lower levels of screening and worse cancer outcomes (21). Therefore, we included a variable for rural residency based on the Rural Urban Commuting Area (RUCA). We also controlled for reporting registry and year of diagnosis.

Statistically significant differences in the distribution of patient, residency, and stage were assessed using a χ² test. Tests were two sided, and a P value of 0.01 or less was considered statistically significant for these descriptive analyses because sample sizes were generally large. In Supplementary analysis, we restricted the sample to Medicaid enrolled patients and compared the distribution of covariates, using a P value of 0.05 or less as statistically significant. We predicted the likelihood of distant stage cancer at diagnosis using adjusted logistic regression and report ORs and 95% confidence intervals (CI). We also compared whether estimates for Medicaid enrollment categories were statistically significantly different from each other using the Wald test, which produces a χ² statistic; a P value <0.05 was considered statistically significant. All analyses were stratified by cancer site and performed with SAS version, 9.4.

Table 1 reports descriptive statistics for the sample stratified by cancer site and Medicaid enrollment status. Medicaid insured patients differ from the non-Medicaid sample along most dimensions. A higher percentage of patients insured by Medicaid were non-Hispanic Black or Hispanic and resided in areas of lower socioeconomic status than patients who were not enrolled in Medicaid. Less than 1% of patients were enrolled in Medicaid prior to diagnosis and of women who were insured by Medicaid the month of or within 12 months of diagnosis, 47% and 36% were enrolled through the NBCCEDP and diagnosed with breast and cervical cancer, respectively. Supplementary Table S1 reports statistically significant differences between patients enrolled in Medicaid one or more months prior to diagnosis and patients enrolled in Medicaid 12 or fewer months following diagnosis.

Figure 2 compares the percentage of patients diagnosed with distant stage disease by cancer site and Medicaid enrollment category. Fewer than 5% of women who were not Medicaid insured were diagnosed with distant stage breast cancer (A). Only about 5% of women who were Medicaid enrolled prior to diagnosis were diagnosed with distant stage disease. Among women enrolled in Medicaid following diagnosis, about 7% of those who enrolled through the NBCCEDP were diagnosed with distant stage disease and just over 10% of women without the NBCCEDP were diagnosed with distant stage disease. Women diagnosed with distant stage cervical cancer follow a similar pattern however, women enrolled in Medicaid prior to diagnosis had the lowest percentage of distant stage cases (B). Patients diagnosed with colorectal cancer had the highest percentage of distant stage cases with patients not enrolled in Medicaid and those enrolled prior to diagnosis having a similar percentage whereas those who enrolled in Medicaid following diagnosis having the greatest percentage of distant stage disease (C).

Table 2 reports ORs associated with likelihood of distant stage disease for the three cancers, adjusting for covariates. Women enrolled in Medicaid prior to a breast cancer diagnosis were more likely to be diagnosed with distant stage disease than women who were not enrolled in Medicaid (OR = 1.45; 95% CI = 1.10–1.92). However, women who enrolled in Medicaid through NBCCEDP (OR = 2.10; 95% CI = 1.98–2.23) and women who enrolled in Medicaid following diagnosis without the NBCCEDP (OR = 2.89; 95% CI = 2.74–3.04) had an odds of distant stage diagnosis that was more than two and three times that of women who were not Medicaid insured, respectively. A Wald test confirmed that these ORs are statistically significant from each other (P < 0.001).

A similar pattern is observed for women diagnosed with cervical cancer, although Medicaid enrollment prior to diagnosis was not associated with distant stage disease at diagnosis. Medicaid enrollment following diagnosis, regardless of whether the woman was diagnosed through the NBCCCEDP, was associated with advanced stage cervical cancer (OR = 1.38; 95% CI = 1.20–1.57 and OR = 2.02; 95% CI = 1.83–2.25). These ORs were not statistically significantly different (P = 0.18). Patients who enrolled in Medicaid following a colorectal cancer diagnosis had a substantially higher likelihood of distant stage relative to patients who were not insured by Medicaid (OR = 2.35; 95% CI = 2.27–2.44).

Despite controls for Medicaid enrollment, some disparities in stage at diagnosis remained. Non-Hispanic Blacks were more likely to be diagnosed with distant stage breast (OR = 1.28; 95% CI = 1.22–1.36) and colorectal cancer (OR = 1.06; 95% CI = 1.02–1.11) relative to their non-Hispanic White counterparts. Hispanics, however, were less likely to be diagnosed with distant stage cancer than non-Hispanic Whites across all three sites. Older patients were less likely to be diagnosed with distant stage disease. Residing in higher socioeconomic status areas was associated with a lower likelihood of distant stage breast cancer. Socioeconomic disparities were not observed for women diagnosed with cervical and colorectal cancer in estimates controlling for Medicaid enrollment.

We conducted an analysis using population-based, national data to explore how individual characteristics and Medicaid enrollment, including enrollment through the NBCCEDP, influenced the likelihood of distant stage diagnosis in three screen amenable cancers. Stage at diagnosis is the best indicator of cancer prognosis and the driver of cancer treatment plans. The incidence of distant stage disease, the most severe stage and most challenging to treat, can be greatly reduced for screen amenable cancers by following widely accepted guidelines for screening (22). We found that patients diagnosed with cervical and colorectal cancer insured through Medicaid prior to diagnosis had nearly the same odds of distant stage disease as patients who were not insured by Medicaid. In this SEER sample, women enrolled in Medicaid prior to diagnosis had the lowest percentage of cervical cancer. A prior study specific to Ohio reported comparable percentages of distant stage cervical cancer in women who were not enrolled in Medicaid (5.8%) to women enrolled in Medicaid prior to cervical cancer diagnosis (ref. 23; 6.7%). Patients diagnosed with breast and cervical cancer who enrolled in Medicaid prior to diagnosis also had a lower odds of distant stage disease than patients enrolled in Medicaid through the NBCCEDP, despite having met lower income thresholds than those imposed by the NBCCEDP. This finding suggests that women who have low incomes but not sufficiently low to meet Medicaid's income threshold may be more vulnerable to distant stage diagnosis.

Health insurance is an essential determinant of access to medical care, including preventive services (24). The ACA sought to expand Medicaid as a way to provide health insurance to low-income uninsured individuals. Prior to the ACA, NBCCEDP was implemented (in 1990) to provide access to breast and cervical cancer screening for low-income uninsured or underinsured women who were not otherwise eligible for Medicaid (25). A comparable program for colorectal cancer screening does not exist. Although there is room for improvement in early detection, our findings suggest that public insurance prior to diagnosis is critical to preventing distant stage disease among low-income individuals with screening-amenable cancers and that targeted programs such as the NBCCEDP are not substitutes for broad health insurance coverage. We also report the overwhelming proportion of distant stage disease among Medicaid enrollees is driven by those who are enrolled following diagnosis. These patients have low survival probabilities under the best of circumstances.

Prior studies report that patients insured by Medicaid have worse cancer survival (26). However, many of these studies do not account for timing of enrollment in Medicaid relative to diagnosis, leading them to conclude that Medicaid provides no survival advantage to patients (27, 28). Our study is in line with studies that report continuous Medicaid coverage is superior to discontinuous coverage but not at the level of other forms of health insurance (29). Our study also aligns with studies that report higher screening rates in states that expanded Medicaid (30–32). Older patients diagnosed with breast and colorectal cancer were also less likely to be diagnosed with distant stage disease, perhaps because they fell within screening guidelines. Health equity is an important national goal, and although the addition of Medicaid variables to our regression models attenuated the likelihood of distant stage disease, disparities remained for Black and low-income patients diagnosed with breast cancer, suggesting insurance alone is not sufficient to remove all barriers to care. Disparities also remained for Black patients diagnosed with breast and colorectal cancer.

Our study has limitations. The first is the absence of Medicaid claims that could provide information on type and frequency of cancer screening. Second, we cannot identify whether those who are not Medicaid insured are uninsured, privately insured, or insured by another source. Finally, the data do not reflect changes in policies following 2013 that may have affected access. Many states passed Medicaid expansions, and managed care, which emphasizes preventive care, has become more widespread in recent years. Therefore, new trends may have emerged with more people qualifying for Medicaid and receiving full coverage for care (33), suggesting that we may underestimate the impact of Medicaid enrollment prior to diagnosis.

Our study highlights the importance of health insurance coverage prior to diagnosis and demonstrates that while targeted programs such as the NBCCEDP provide critical access to screening, they are not a substitute for comprehensive insurance coverage. Except for Medicaid enrollment prior to diagnosis for patients diagnosed with cervical and colorectal cancer, the incidence of distant stage disease is higher among patients covered by Medicaid compared with other forms of insurance, leaving room for additional policy improvements in access and screening for low-income individuals. Our findings also reflect the extent to which Medicaid is the insurance of the last resort by absorbing many patients with distant stage disease when they meet state enrollment conditions based on a combination of income, medical spending, and/or health condition (8) and when treatment is less effective. Finally, we demonstrate that disparities remain, particularly for Black patients and patients residing in low-income areas where access to quality care may be substandard despite health insurance coverage through Medicaid. This, too, represents an opportunity for additional policy and care delivery measures to eliminate disparities.

C.J. Bradley reports personal fees from IMS outside the submitted work. No disclosures were reported by the other authors.

The views expressed here are those of the authors only and do not necessarily reflect the views of the NCI or NIH.

C.J. Bradley: Conceptualization, formal analysis, methodology, writing–original draft, writing–review and editing. L.M. Sabik: Conceptualization, formal analysis, methodology, writing–review and editing. J. Entwistle: Conceptualization, formal analysis, methodology, writing–review and editing. J.L. Stevens: Conceptualization, formal analysis, methodology, writing–review and editing. L. Enewold: Conceptualization, formal analysis, methodology, writing–review and editing. J.L. Warren: Conceptualization, formal analysis, methodology, writing–review and editing.

C.J. Bradley's research was financially supported by a contract with the NCI. The contents of this article have not been copyrighted or published previously and the contents of this article are not now under consideration for publication elsewhere. No financial disclosures were reported by the authors of this article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.