Abstract
Objective.
To determine associations between adoption of Medicaid expansion (ME) and changes in insurance status, early stage diagnosis, and cancer survival among women with endometrial carcinoma (EC).
Methods.
The National Cancer Database (NCDB) was queried for patients diagnosed with EC between the age 40–64 from 2004 to 2015. Difference-in-differences analysis quantified the impact of ME on the proportion of new EC diagnoses with insurance (vs. uninsured), the proportion diagnosed with stage I (vs. II-IV), and overall survival.
Results.
156,253 patients were included. Among 65,019 women living in ME states, ME is associated with an increase in the percent of EC cases who are insured of 1.4% (95% CI 0.9–2.0%, p < 0.0001), with strongest effects among Hispanic women, women in the lowest income quartile, and women in the second age quartile (age 53–57). There was no overall impact of ME on stage, though an increase of early stage diagnoses by 2.4% (95% CI 0.3–4.5%, p = 0.022) was observed among women age 53–57. There was a trend towards improved overall survival with ME, which was strongest in women age 53–57 (HR = 0.83, 95% CI 0.70–0.99, p = 0.037).
Conclusions.
Among women with EC, ME positively impacted insurance coverage, an important hurdle in accessing health care. In women aged 53–57, ME was associated with earlier stage at diagnosis and improved survival, suggesting that the magnitude of the improvement in insurance coverage may correlate with important clinical outcomes. Efforts should continue to understand the complexity of barriers to health care access and to develop effective strategies to surmount them.
Keywords: Medicaid expansion, Endometrial cancer, Health care disparities, Insurance coverage, Stage, Race/ethnicity
1. Introduction
Due to the heralding symptom of abnormal uterine bleeding, most women with endometrial carcinoma (EC) present with early stage disease and in these women, surgery alone often results in favorable outcomes and a high probability of cure. However, for women who present with advanced disease or suffer recurrence, curative treatment options are limited. Chemotherapy, radiation, and immunotherapy play roles in managing disease progression, but the greatest opportunity to battle EC remains in primary prevention, early disease detection, and timely surgical intervention [1-5]. Unfortunately, patients with significant barriers to health care access may not be able to take full advantage of aggressive early intervention.
In 2010, the Affordable Care Act (ACA) expanded Medicaid eligibility to patients with income up to 138% of the federal poverty line. State legislatures voted on implementation, with approximately two-thirds of states ultimately electing to expand Medicaid. Most expansion was implemented in January 2014 [6,7]. Under Medicaid Expansion (ME), an estimated 13 million previously uninsured Americans gained coverage [8,9]. ME has been linked with increased cancer screening [10,11] and improved access to surgical care for cancer patients [12,13]. In breast cancer, ME was associated with earlier stage at diagnosis and improved quality of care [14]. Few studies have evaluated the impact of ME on uterine cancers, with two showing improvement in insurance coverage but none demonstrating a shift in stage at presentation [12,15,16]. To our knowledge, no studies have evaluated the impact of ME on EC survival. Therefore, our objectives were to examine associations between adoption of ME and insurance coverage, EC stage at diagnosis, and overall survival (OS).
2. Methods
2.1. The National Cancer Database
Data were obtained from the 2016 National Cancer Database (NCDB) Participant User Files (PUF). Sociodemographic, clinical, and treatment facility characteristics were captured using standardized codes defined by the Facility Oncology Registry Data Standards (FORDS) [17]. All data are de-identified, and the study was considered exempt by the Ohio State University Institutional Review Board.
2.2. Study population
Between 2004 and 2015, 368,702 women were diagnosed with invasive EC [International Classification of Diseases (ICD)-10 morphology codes 8380-8383 8140, 8210, 8211, 8260-8263, 8560, 8570]. We excluded women younger than 40 at diagnosis, as geographic information is suppressed for this age group (n = 11,032), women aged 65 or older at diagnosis, given the likelihood of Medicare insurance (n = 160,047), and women with missing information on stage (n = 38,341), insurance status (n = 2970), or follow-up data (n = 69), leaving 156,243 women for analysis.
2.3. Medicaid expansion status
Information was available on whether and when each woman's state of residence adopted ME, in four categories: non-expansion (nineteen states), early expansion (2010–2013, five states and the District of Columbia), January 2014 expansion (nineteen states), and late expansion (after January 2014; seven states). Early expansion states had preexisting medical coverage for low-income adults which they continued and expanded using federal funding during 2010–2013 [18]. Among late expansion states, two expanded in 2014, three in 2015, and two in 2016 (after the end of follow-up) [19]. We therefore decided to divide women into two groups based on whether their state of residence expanded ME by January 2014 – early and January 2014 expansion states are considered “exposed” to ME, while non-expansion and late expansion states are considered “unexposed.”
2.4. Covariates
We included information on epidemiological, hospital, tumor, and treatment characteristics as categorized in Tables 1a and 1b. No >3% of data was missing for any variable, and missing values were included as a separate category. Overall survival (OS) was the time from diagnosis to death; women who did not die were censored at date of last contact.
Table 1a.
Multivariable-adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for associations between Medicaid expansion status and patient, tumor, and treatment characteristics, National Cancer Database, 2004–2013.
| Non/late expansion states |
Expansion states |
p1 | ||
|---|---|---|---|---|
| n = 62,168 |
n = 65,019 |
|||
| n (%) | n (%) | OR (95% CI)a | ||
| Age | 0.001 | |||
| 40–52 | 16,598 (26.7) | 17,307 (26.6) | 1.00 | |
| 53–57 | 17,030 (27.4) | 17,951 (27.6) | 1.04 (1.01, 1.07) | |
| 58–61 | 16,472 (26.5) | 17,501 (26.9) | 1.06 (1.03, 1.10) | |
| 62–64 | 12,068 (19.4) | 12,260 (18.9) | 1.02 (0.99, 1.06) | |
| Race | <0.0001 | |||
| NHW | 46,937 (75.5) | 48,899 (75.2) | 1.00 | |
| NHB | 6195 (10.0) | 4243 (6.5) | 0.66 (0.63, 0.68) | |
| Hispanic | 2840 (4.6) | 4005 (6.2) | 1.36 (1.29, 1.43) | |
| API | 811 (1.3) | 3065 (4.7) | 3.65 (3.37, 3.95) | |
| Other | 5385 (8.7) | 4807 (7.4) | 0.86 (0.82, 0.89) | |
| Insurance | <0.0001 | |||
| None | 4517 (7.3) | 2967 (4.6) | 1.00 | |
| Private | 47,561 (76.5) | 51,501 (79.2) | 1.70 (1.62, 1.79) | |
| Medicaid | 3619 (5.8) | 4654 (7.2) | 1.99 (1.87, 2.13) | |
| Medicare | 5375 (8.7) | 5293 (8.1) | 1.60 (1.51, 1.70) | |
| Other government | 1096 (1.8) | 604 (0.9) | 0.85 (0.76, 0.95) | |
| Charlson comorbidity score | <0.0001 | |||
| 0 | 46,655 (75.1) | 50,437 (77.6) | 1.00 | |
| 1 | 12,790 (20.6) | 11,956 (18.4) | 0.88 (0.85, 0.90) | |
| ≥2 | 2723 (4.4) | 2626 (4.0) | 0.92 (0.87, 0.98) | |
| Year of diagnosis | 0.0004 | |||
| 2004–2005 | 10,154 (16.3) | 10,669 (16.4) | 1.00 | |
| 2006–2007 | 12,029 (19.4) | 12,699 (19.5) | 1.00 (0.96, 1.04) | |
| 2008–2009 | 12,642 (20.3) | 13,025 (20.0) | 0.97 (0.94, 1.01) | |
| 2010–2011 | 13,944 (22.4) | 14,078 (21.7) | 0.94 (0.91, 0.98) | |
| 2012–2013 | 13,399 (21.6) | 14,548 (22.4) | 1.01 (0.97, 1.04) | |
| Histology | 0.005 | |||
| Low-grade endometrioid | 45,255 (72.8) | 47,464 (73.0) | 1.00 | |
| High-grade endometrioid | 7379 (11.9) | 7504 (11.5) | 0.98 (0.95, 1.02) | |
| Serous | 2976 (4.8) | 3118 (4.8) | 1.07 (1.01, 1.13) | |
| Carcinosarcoma | 2359 (3.8) | 2478 (3.8) | 1.07 (1.01, 1.14) | |
| Mixed epithelial | 3484 (5.6) | 3800 (5.8) | 1.05 (1.00, 1.10) | |
| Clear cell | 715 (1.2) | 655 (1.0) | 0.90 (0.81, 1.00) | |
| Stage | <0.0001 | |||
| I | 46,949 (75.5) | 49,501 (76.1) | 1.00 | |
| II | 4168 (6.7) | 4254 (6.5) | 0.97 (0.93, 1.02) | |
| III | 8109 (13.0) | 8431 (13.0) | 0.99 (0.96, 1.02) | |
| IV | 2942 (4.7) | 2833 (4.4) | 0.94 (0.89, 0.99) | |
| Adjuvant treatment | <0.0001 | |||
| None | 43,017 (69.2) | 42,672 (65.6) | 1.00 | |
| Chemotherapy only | 3414 (5.5) | 3568 (5.5) | 1.08 (1.03, 1.14) | |
| Radiation only | 11,583 (18.6) | 13,485 (20.7) | 1.18 (1.15, 1.21) | |
| Chemotherapy plus radiation | 3145 (5.1) | 3916 (6.0) | 1.28 (1.21, 1.34) | |
Women with unknown values for a particular variable are included in the model but not presented in the table.
p-value from multivariable-adjusted model.
Odds ratios (ORs) and 95% confidence intervals (CIs) adjusted for age quartile and race with non/late expansion states as the reference group.
Table 1b.
Multivariable-adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for associations between Medicaid expansion status and zip-code level income and education quartiles, and facility type and location, National Cancer Database, 2004–2013.
| Non/late expansion states |
Expansion states |
p1 | ||
|---|---|---|---|---|
| n = 62,168 |
n = 65,019 |
|||
| n (%) | n (%) | OR (95% CI)a | ||
| Incomeb | <0.0001 | |||
| ≤$38,000 | 12,850 (20.7) | 7254 (11.2) | 1.00 | |
| $38,000–$47,999 | 16,771 (27.0) | 11,692 (18.0) | 1.23 (1.19, 1.28) | |
| $48,000–$62,999 | 16,959 (27.3) | 17,173 (26.4) | 1.77 (1.71, 1.84) | |
| ≥$63,000 | 15,275 (24.6) | 28,590 (44.0) | 3.25 (3.14, 3.37) | |
| Education (% without high school diploma)c | <0.0001 | |||
| ≥21% | 10,588 (17.0) | 9763 (15.0) | 1.00 | |
| 13%–20.9% | 17,225 (27.7) | 14,458 (22.2) | 0.97 (0.93, 1.00) | |
| 7%–12.9% | 20,564 (33.1) | 22,061 (33.9) | 1.22 (1.18, 1.27) | |
| <7% | 13,501 (21.7) | 18,459 (28.4) | 1.55 (1.49, 1.60) | |
| Urban/rural | <0.0001 | |||
| Metro | 48,956 (78.8) | 55,150 (84.8) | 1.00 | |
| Urban | 10,257 (16.5) | 7213 (11.1) | 0.64 (0.62, 0.66) | |
| Rural | 1476 (2.4) | 719 (1.1) | 0.45 (0.41, 0.49) | |
| Hospital type | <0.0001 | |||
| Community Cancer Program | 2546 (4.1) | 3357 (5.2) | 1.00 | |
| Comprehensive Community Cancer Program | 25,343 (40.8) | 21,602 (33.2) | 0.66 (0.63, 0.70) | |
| Academic/research program | 22,678 (36.5) | 32,433 (49.9) | 1.13 (1.07, 1.20) | |
| Integrated Network Cancer Program | 11,601 (18.7) | 7627 (11.7) | 0.52 (0.49. 0.55) | |
| Hospital location | <0.0001 | |||
| New England | 1868 (2.5) | 7804 (9.8) | 1.00 | |
| Middle Atlantic | 8691 (11.4) | 17,248 (21.6) | 0.48 (0.45, 0.51) | |
| South Atlantic | 25,586 (33,5) | 6275 (7.9) | 0.06 (0.06, 0.06) | |
| East North Central | 13,492 (17.7) | 16,568 (20.7) | 0.32 (0.30, 0.34) | |
| East South Central | 7341 (9.6) | 3141 (3.9) | 0.11 (0.10, 0.12) | |
| West North Central | 6282 (8.2) | 6702 (8.4) | 0.28 (0.26, 0.30) | |
| West South Central | 10,521 (13.8) | 200 (0.3) | 0.00 (0.00, 0.01) | |
| Mountain | 2249 (3.0) | 4374 (5.5) | 0.49 (0.46, 0.53) | |
| Pacific | 299 (0.4) | 17,602 (22.0) | 14.53 (12.59, 16.77) | |
p-value from multivariable-adjusted model.
Odds ratios (ORs) and 95% confidence intervals (CIs) adjusted for age quartile and race with non/late expansion states as the reference group.
Income: quartiles based on equally proportioned income ranges among all U.S. zip codes.
Education: quartiles based on proportion of adults in the patient’s zip code who did not graduate from high school.
2.5. Statistical analysis
We compared epidemiological, hospital, tumor, and treatment characteristics of women according to dichotomized ME status using logistic regression. Odds ratios (ORs) and 95% confidence intervals (CIs) for associations between predictors and ME status were estimated in models adjusted for age quartile (40–52, 53–57, 58–61, 62–64) and race and ethnicity (Non-Hispanic White [NHW], Non-Hispanic Black [NHB], Hispanic, Asian/Pacific Islander [API], Other). These models were restricted to women diagnosed before 2014 to quantify how ME states differ from the non/late-ME states in characteristics of interest in our analysis, prior to ME enactment. We visualized the distributions of patient insurance status separately by age, race/ethnicity, and income, and the proportions of women in different insurance categories through time by ME status.
We performed difference-in-differences analyses to quantify the effect of ME on insurance status, stage at diagnosis, and OS [20]. For binary outcomes (insured vs. uninsured and early vs. late stage), we fit linear probability models [20], and for OS we fit Cox proportional hazards models. The unadjusted models included terms for time effects (year of diagnosis), geographic patterns (facility location and ME category), and an indicator that the patient lived in an ME state after ME. The coefficient on this indicator quantifies the effect of ME on the outcome of interest beyond what is expected based on geography and time. The adjusted models included terms for age (in quartiles), race, income, education, rurality, and facility type. We repeated these analyses for subgroups based on age quartiles; race/ethnicity (NHW, NHB, and Hispanic only, due to limited sample size of other groups); and income quartiles. For OS, we fit a fully adjusted model, which includes additional characteristics relevant to patient survival (histology, stage, and adjuvant treatment). In analyses of stage and survival, we do not include insurance status, as we hypothesized that insurance status would mediate associations between ME and these outcomes.
3. Results
A total of 156,243 patients were included, of whom 76,329 (48.8%) resided in states that did not adopt ME by January 2014, while 79,914 (51.2%) lived in ME states. Mean age at EC diagnosis was 55.9 years (standard deviation = 5.9).
Table 1a shows age- and race-adjusted associations between demographic, tumor, and treatment characteristics with ME status before ME implementation (2004–2013) to quantify baseline differences comparing ME to non/late-ME states. Women diagnosed at an older age were more likely to live in a ME state. Compared with NHW women, we observed lower odds of living in an ME state among NHB women (OR 0.66, 95% CI 0.63–0.68) and higher odds among Hispanic (OR 1.36, 95% CI 1.29–1.43) and API women (OR 3.65, 95% CI 3.37–3.96). ME states had higher proportions of cases covered by Medicaid, Medicare, or private insurance, and lower proportions covered by other government insurance. Having a comorbid condition vs. none (OR 0.88, 95% CI 0.85–0.90) was also related to lower odds of ME.
In terms of tumor characteristics, compared to women with low-grade endometrioid disease we observed higher odds of living in ME states among women with serous (OR 1.07,95% CI 1.01–1.13), carcinosarcoma (OR 1.07, 95% CI 1.01–1.14), or mixed epithelial tumors (OR 95% CI 1.00–1.10), but inverse associations among women with clear cell cancer (OR 0.90,95% CI 0.81–1.00). Compared to stage I, diagnosis with stages II-IV tumors was associated with lower odds of ME while receipt of any adjuvant therapy modality was positively associated with ME status (Table 1a).
Table 1b displays associations between zip-code level income and education quartiles, and facility information with ME status prior to ME. Zip codes with higher income and education were associated with ME, while urban or rural areas, compared with metro areas, were associated with late/non-ME. Compared with care received at a community cancer program, we noted inverse associations between care received at a Comprehensive Community Cancer Program (OR 0.66, 95% CI 0.63–0.70) or at an Integrated Network Cancer Program (OR 0.52, 95% CI 0.49–0.55) with ME status, while women receiving care in an academic/research program had a higher odds of living in an ME state (OR 1.13, 95% CI 1.07–1.20).
Baseline distributions of patient insurance in the 2004–2013 era is visualized separately by age, race/ethnicity, and income (Fig. 1). The majority of women have private insurance. Younger women are more frequently uninsured or covered by Medicaid than older women. Larger proportions of NHB, Hispanic, API, and Other are uninsured or covered by ME compared with NHW. Smaller proportions of women are uninsured or covered by Medicaid as zip code income quartile increased.
Fig. 1.
Baseline distributions of patient insurance in the 2004–2013 era is visualized separately by age, race/ethnicity, and income. The majority of women have private insurance. Younger women are more frequently uninsured or covered by Medicaid than older women. Larger proportions of NHB, Hispanic, API, and Other are uninsured or covered by ME compared with NHW. Smaller proportions of women are uninsured or covered by Medicaid as zip code income quartile increased.
Time trends in insurance coverage are presented in Fig. 2. We observed relatively parallel temporal trends in ME and non-ME groups between 2004 and 2013. In 2014–15, the proportion of EC cases with Medicaid in ME states increased from 8.2% to 14.1%, and the proportion of uninsured cases declined. In non/late ME states, the uninsured rate also declined, with an increase in private insurance coverage.
Fig. 2.
Insurance coverage over time. We observed relatively parallel temporal trends in ME and non-ME groups between 2004 and 2013. In 2014–15, the proportion of EC cases with Medicaid in ME states increased from 8.2% to 14.1%, and the proportion of uninsured cases declined. In non/late ME states, the uninsured rate also declined, with an increase in private insurance coverage.
3.1. Difference-in-differences analyses
Table 2 shows changes in percent insured associated with ME implementation in ME states. We quote unadjusted results; adjusted results are similar. ME was associated with an increase of 1.2% in the percent of EC cases who are insured (95% CI 0.6–1.8%, p < 0.0001). A significant improvement was seen in age quartiles Q1, Q2, and Q4, with the strongest effect in Q2 (ages 53–57; 1.7%, 95% CI 0.07–2.8%). There was a statistically significant improvement among NHW but not among NHB. Among Hispanic women, the effect was substantial (7.6%, 95% CI 4.1–11.1%). We observed significant increases in percent coverage of 4.0% and 2.1% in the lowest two income quartiles.
Table 2.
Changes in health insurance status associated with Medicaid expansion in the overall study population and stratified by age, race, zip code level income, National Cancer Database, 2004–2015.
| Model 1 |
Model 2 |
||||
|---|---|---|---|---|---|
| Percent change (95% CI) |
P | Percent change (95% CI) |
P | ||
| Overall | 1.2 (0.6, 1.8) | 5.40E-05 | 1.4 (0.9, 2) | 8.30E-07 | |
| Subgroup analysis | |||||
| Age group | 40–52 | 1.3 (0, 2.6) | 0.043 | 1.6 (0.4, 2.9) | 0.011 |
| 53–57 | 1.7 (0.7, 2.8) | 0.0017 | 2 (0.9, 3.1) | 0.00029 | |
| 58–61 | 0.7 (−0.4, 1.8) | 0.21 | 0.9 (−0.1, 2) | 0.087 | |
| 62–64 | 1.3 (0, 2.5) | 0.043 | 1.4 (0.2, 2.6) | 0.021 | |
| Race | NHW | 0.8 (0.2, 1.4) | 0.013 | 0.9 (0.3, 1.5) | 0.0055 |
| NHB | 1.2 (−1.1, 3.5) | 0.32 | 1.4 (−0.9, 3.7) | 0.23 | |
| Hispanic | 7.6 (4.1, 11.1) | 2.10E-05 | 7.3 (3.8, 10.7) | 3.30E-05 | |
| Income | Q1 | 4 (2.2, 5.8) | 1.90E-05 | 4.1 (2.2, 5.9) | 1.20E-05 |
| Q2 | 2.1 (0.7, 3.5) | 0.0029 | 2.4 (1.1, 3.8) | 0.00053 | |
| Q3 | 1 (−0.1, 2.1) | 0.073 | 1.1 (0, 2.2) | 0.041 | |
| Q4 | 0.2 (−0.5, 1) | 0.54 | 0.4 (−0.4, 1.2) | 0.29 | |
Model 1: Unadjusted model, which includes terms for geographic location (Facility Location and ME Category ofPatient Zip Code) and year (grouped into pairs). The Percent Change quantifies the change in percent of women diagnosed with early stage disease among states that adopted ME, relative to what would be expected based on non-ME states.
Model 2: Adjusted model, which includes terms in the unadjusted model plus age quartile (40–52, 53–57, 58–61, 62–64), race (NHW, NHB, Hispanic, API, Other, unknown), Charlson comorbidities (0, 1, ≥2), hospital type (community cancer, comprehensive community cancer, academic/research, integrated network cancer), hospital location (Northeast, South, Midwest, Mountain, Pacific), rurality (metro, urban, rural, unknown), educational attainment (quartiles based on proportion of adults in the patient’s zip code who did not graduate from high school: ≥21%, 13%–20.9%, 7%–12.9%, <7%), household income (quartiles based on equally proportioned income ranges among all U.S. zip codes: ≤$38,000, $38,000–$47,999, $48,000–$62,999, ≥$63,000), year ofdiagnosis (2004–2005, 2006–2007, 2008–2009).
Q: quartile.
Table 3 shows changes in the percent of early stage ECs associated with ME implementation. The overall effect of ME on the percent diagnosed at early stage was not significant (0.8%, 95% CI −0.3–1.9%, p = 0.17). The only statistically significant subgroup effect was in the second quartile of age (53–57) where ME was associated with an absolute increase of 2.2% in the percentage of cases diagnosed at early stage (95% CI 0.1–4.3%). The point estimates were positive in income Q2-Q4 – though not distinguishable from 0 statistically – the point estimate in Q1 was essentially 0. There were no significant associations by race/ethnicity subgroup.
Table 3.
Changes in the percent of early stage disease associated with Medicaid expansion in the overall study population and stratified by age, race, zip code level income, National Cancer Database, 2004–2015.
| Model 1 |
Model 2 |
||||
|---|---|---|---|---|---|
| Percent change (95% CI) |
P | Percent change (95% CI) |
P | ||
| Overall | 0.8 (−0.3, 1.9) | 0.17 | 0.8 (−0.3, 1.9) | 0.15 | |
| Subgroup analysis | |||||
| Age group | 40–52 | 1.6 (−0.6, 3.8) | 0.15 | 1.7 (−0.5, 3.9) | 0.13 |
| 53–57 | 2.2 (0.1, 4.3) | 0.037 | 2.4 (0.3, 4.5) | 0.022 | |
| 58–61 | −0.4 (−2.5, 1.7) | 0.72 | −0.5 (−2.6, 1.6) | 0.63 | |
| 62–64 | −0.6 (−3, 1.8) | 0.65 | −0.5 (−2.9, 1.9) | 0.67 | |
| Race | NHW | 0.3 (−1, 1.5) | 0.66 | 0.3 (−0.9, 1.6) | 0.58 |
| NHB | 0.5 (−3.3, 4.3) | 0.79 | 0.4 (−3.4, 4.2) | 0.83 | |
| Hispanic | 1.3 (−3.2, 5.7) | 0.57 | 1.1 (−3.4, 5.5) | 0.64 | |
| Income | Q1 | 0.1 (−2.7, 3) | 0.94 | −0.1 (−3, 2.7) | 0.92 |
| Q2 | 0.9 (−1.5, 3.2) | 0.47 | 0.9 (−1.4, 3.3) | 0.42 | |
| Q3 | 0.6 (−1.5, 2.7) | 0.55 | 0.6 (−1.5, 2.7) | 0.55 | |
| Q4 | 0.9 (−1, 2.8) | 0.36 | 0.9 (−1, 2.8) | 0.36 | |
Model 1: Unadjusted model, which includes terms for geographic location (Facility Location and ME Category of Patient Zip Code) and year (grouped into pairs). The Percent Change quantifies the change in percent of women diagnosed with early stage disease among states that adopted ME, relative to what would be expected based on non-ME states.
Model 2: Adjusted model, which includes terms in the unadjusted model plus age quartile (40–52, 53–57, 58–61, 62–64), race (NHW, NHB, Hispanic, API, Other, unknown), Charlson comorbidities (0, 1, ≥2), hospital type (community cancer, comprehensive community cancer, academic/research, integrated network cancer), hospital location (Northeast, South, Midwest, Mountain, Pacific), rurality (metro, urban, rural, unknown), educational attainment (quartiles based on proportion of adults in the patient's zip code who did not graduate from high school: ≥21%, 13%–20.9%, 7%–12.9%, <7%), household income (quartiles based on equally proportioned income ranges among all U.S. zip codes: ≤$38,000, $38,000–$47,999, $48,000–$62,999, ≥$63,000), year of diagnosis (2004–2005, 2006–2007, 2008–2009).
Q: quartile.
Table 4 shows hazard ratios (HRs) associated with ME implementation in ME states. Overall, the HR for ME suggested a protective effect on OS, though the association did not meet the threshold for statistical significance (HR = 0.92, 95% CI 0.85–1.01, p = 0.066). By age, we again observed the most substantial effect within the 53–57 age group (HR = 0.78, 95% CI 0.66–0.93). The HR CIs in other age categories contain 1, as do the CIs for the HRs in the race/ethnicity subgroups. Within income quartiles, the effect of ME appears strongest in the third quartile, while the point estimates in quartiles 1 and 2 are below 1 but are not statistically significant.
Table 4.
Hazard ratios (HRs) and 95% confidence intervals (CIs) for associations between Medicaid expansion in 2014–2015 and overall survival, National Cancer Database.
| HR (95% CI)a | P | HR (95% CI)b | P | HR (95% CI)c | P | ||
|---|---|---|---|---|---|---|---|
| Overall | 0.92 (0.85, 1.01) | 0.066 | 0.93 (0.86, 1.01) | 0.11 | 0.93 (0.85, 1.01) | 0.095 | |
| Subgroup analysis | |||||||
| Age group | 40–52 | 0.97 (0.78, 1.20) | 0.77 | 0.97 (0.79, 1.2) | 0.8 | 0.93 (0.75, 1.15) | 0.52 |
| 53–57 | 0.78 (0.66, 0.93) | 0.0056 | 0.77 (0.65, 0.91) | 0.0027 | 0.83 (0.70, 0.99) | 0.037 | |
| 58–61 | 0.93 (0.80, 1.08) | 0.35 | 0.94 (0.8, 1.09) | 0.39 | 0.94 (0.81, 1.10) | 0.44 | |
| 62–64 | 1.04 (0.89, 1.22) | 0.59 | 1.05 (0.9, 1.23) | 0.56 | 1.01 (0.86, 1.18) | 0.94 | |
| Race | NHW | 0.92 (0.83, 1.03) | 0.14 | 0.92 (0.83, 1.02) | 0.13 | 0.91 (0.82, 1.01) | 0.067 |
| NHB | 0.92 (0.77, 1.11) | 0.4 | 0.92 (0.77, 1.1) | 0.37 | 0.93 (0.77, 1.12) | 0.44 | |
| Hispanic | 1.07 (0.74, 1.55) | 0.71 | 1.08 (0.75, 1.56) | 0.68 | 1.03 (0.71, 1.50) | 0.87 | |
| Income | Q1 | 0.92 (0.76, 1.12) | 0.41 | 0.93 (0.77, 1.12) | 0.44 | 0.92 (0.76, 1.12) | 0.4 |
| Q2 | 0.96 (0.80, 1.14) | 0.62 | 0.97 (0.81, 1.15) | 0.71 | 0.95 (0.79, 1.13) | 0.54 | |
| Q3 | 0.84 (0.71,0.99) | 0.036 | 0.84 (0.71, 1) | 0.047 | 0.79 (0.67, 0.94) | 0.0065 | |
| Q4 | 0.93 (0.79, 1.10) | 0.4 | 0.95 (0.8, 1.12) | 0.54 | 1.00 (0.84, 1.18) | 0.97 | |
HRs from the unadjusted model, which includes terms for geographic location (Facility Location and ME Category of Patient Zip Code) and year (grouped into pairs). The HR quantifies the change in hazard associated with ME among states that adopted ME, relative to what would be expected based on non-ME states.
HRs from the partially adjusted model, which includes terms in the unadjusted model plus age quartile (40–52, 53–57, 58–61, 62–64), race (NHW, NHB, Hispanic, API, Other, unknown), Charlson comorbidities (0, 1, ≥2), hospital type (community cancer, comprehensive community cancer, academic/research, integrated network cancer), hospital location (Northeast, South, Midwest, Mountain, Pacific), rurality (metro, urban, rural, unknown), educational attainment (quartiles based on proportion of adults in the patient’s zip code who did not graduate from high school: ≥21%, 13%–20.9%, 7%–12.9%, <7%), household income (quartiles based on equally proportioned income ranges among all U.S. zip codes: ≤$38,000, $38,000–$47,999, $48,000–$62,999, ≥$63,000), year of diagnosis (2004–2005, 2006–2007, 2008–2009).
HRs from the fully adjusted model, which includes all terms in the partially adjusted model plus histological subtype (low-grade endometrioid, high-grade endometrioid, serous, carcinosarcoma, clear cell, mixed epithelial), stage (I, II, III, IV), and treatment (none, chemotherapy only, radiation only, chemoherapy plus radiation, unknown).
4. Discussion
We show strong evidence for an association between ME and improved insurance coverage, and a suggestion of better OS among women diagnosed with EC; however, an overall association with improvements in early stage diagnosis was not detected. We observed improvements in insurance coverage with decreasing quartiles of zip code level income and among Hispanic and NHW women. Additionally, ME was associated with improved OS among women in certain age and income categories. Our focus on the impact of health care policy among women with EC, the most common gynecologic malignancy for which mortality is increasing, represents an important addition to the growing body of literature related to ME effects among cancer patients.
Our findings related to insurance coverage are in line with prior studies. For example, using data from the Surveillance, Epidemiology, and End Results (SEER) program, Moss et al. reported improved insurance coverage for women newly diagnosed with cervical, uterine, or ovarian malignancies in states that expanded ME [12]. Specifically, the uninsured rate decreased by approximately 50% (8% uninsured to 4% uninsured) between 2011 and 2013 and 2014 for uterine cancer patients living in expansion states. Our subgroup results diverge from those reported by Moss and colleagues [12]; specifically in regards to race/ethnicity, reductions in uninsured proportions among White, Black, Hispanic, and Asian patients were identified in the SEER population, while we noted significant reductions for White and Hispanic women only. Previous data regarding new gynecologic oncology cancer patients of all disease sites show that low-income, uninsured patients are disproportionately comprised of racial minorities [21,22]. Other data in the general population similarly demonstrated increased coverage associated with ME for all racial subgroups including NHB and NHW, with the largest increase in coverage observed among Hispanic patients, which is consistent with our findings [23].
Not surprisingly, our study demonstrated that the percent change for insurance coverage increased most for women with lower levels of income, consistent with other large database studies. [15,24]. The relationship between broadened insurance coverage under ME as it relates to age in our study is less clear. Given that younger women tended to have higher rates of being uninsured and using Medicaid prior to 2014, we expected the strongest impact of ME would be among women in the youngest quartile (age 40–52), with lessening impact as age at diagnosis increased. That hypothesis is consistent with a previous study that suggested that the benefit of increased coverage from ME is most pronounced in younger patients [23]. However, in our study, insurance coverage improved more in the 53–57 age group than in the 40–52 age group. One possible explanation may be that in our patient population, younger healthier individuals may have been be less likely to pursue coverage even if eligible, while older patients, potentially with other health problems, may be more likely to take advantage of new opportunities for coverage.
The overall effect of ME on the rate of early stage diagnosis was lower than expected. Only the subgroup of women aged 53–57, who displayed one of the larger increases in insurance coverage, had a statistically significant improvement in early stage diagnosis associated with ME. In an NCDB analysis including multiple cancer types, Jemal and colleagues reported small but significant shifts in the proportion of early stage diagnoses among individuals diagnosed with colorectal, lung, breast, melanoma, and pancreatic cancers; however, no stage shift was noted from women with a uterine corpus diagnosis, though follow-up was limited to 1 year after ME [15]. Indolent tumor biology of low-grade endometriod disease may mute the impact of improved access to care with regard to stage at presentation. Alternatively, it is possible that our population included patients with postmenopausal bleeding prior to ME who sought medical care only after insurance coverage was obtained, and that a stage shift might be observed with more follow-up time. A recent study by Tsui et al. demonstrated that 53.2% of newly enrolled Medicaid patients (cancer diagnosed <6 months from the time of enrollment in Medicaid) had advanced stage disease compared to 39.5% of patients who were previously enrolled in Medicaid [25]. They also demonstrated an increasing proportion of late-stage disease for all Medicaid patients in all three cancer subtypes studied (breast, colorectal, and cervical). Conversely, in a 2019 SEER database analysis of the eight most prevalent cancers (including EC), Mesquita-Neto et al. reported that post-ME, significantly more cancer patients were diagnosed at an earlier stage (31.1% compared to 27.6% of cancer patients pre-ME; p < 0.001) [13].
Few prior studies have examined the impact of ME on survival. An association between ACA implementation and improved survival in colorectal carcinoma was demonstrated, along with an increase in screening and early stage diagnosis [26]. Although our overall association between ME and OS was not significant, our results certainly suggest further investigation of any such relationship would be worthwhile. ME was associated with improved survival in the 53–57 age group, where the largest improvements in insurance coverage and presentation at early stage were also observed. As the NCDB does not report cancer-specific survival, the survival benefit could be related to improvements in cancer care and in other aspects of health care. As women diagnosed with EC often carry comorbid diagnoses, this group of women may particularly benefit from improved access to the health care system. Regardless, observing any effect among EC patients, where survival is generally quite favorable, is notable, and strongly recommends evaluating the potential impact of ME on cancers with less favorable prognosis and potential benefit from early diagnosis.
Our results suggest that improved insurance coverage is a promising step towards more equitable health care. However, better insurance coverage is not synonymous with better health care. Barriers to receipt of quality care are complex and extend beyond insurance coverage. They include transportation issues, geographic distance, racial and cultural factors, education level, distrust of health care providers, and low income [27-31 ]. Some have argued that improved coverage has not resulted in improved health care access, while others demonstrated that ME was an independent predictor of access to surgery for cancer patients [9,13]. Specific for EC, disparities in mortality based on race have been well described [32-35]. Health insurance coverage alone does not address many of the multifactorial barriers to care, and other outcomes (such as stage, access to surgery, and survival) must be examined. For example, it is notable that in our results, we find a large improvement in insurance coverage among Hispanic women, but that does not seem to translate into improvements in stage at diagnosis or OS.
Strengths of this study include the large sample size, adjustment for demographic and clinical confounders, and the analysis approach accounting for baseline geographic differences in health care and legislature. Limitations include lack of information on EC-specific survival and individual-level income and education. Within county lines, income and education can vary widely, potentially impacting individuals' access to health care. Additionally, insurance status is only documented at diagnosis, but patients' insurance may change (and affect survival) after diagnosis. In order to examine the effect of ME precisely, we excluded patients 65 and older who would be eligible for Medicare, resulting in a younger population with an average age at diagnosis of 55. It is possible that this age-restricted cohort is enriched for patients with hereditary cancer syndromes which could diminish a potential effect of improved coverage on survival, as these patients may be at risk for developing other life-limiting cancers.
5. Conclusions
Early diagnosis of EC and improved access to care are paramount for the application of curative treatment. We found no significant overall improvement in early stage diagnosis or survival associated with ME. However, among women age 53–57, we observed improved insurance coverage, early stage diagnosis, and OS associated with ME. We also identified patient characteristics associated with residing in ME states, which may deserve consideration in further research relating disparities to cancer outcomes. Additional gynecologic malignancies warrant similar investigation of the impact of ME on health care access. Efforts aimed at identifying and overcoming barriers to the receipt of quality cancer care should remain a priority.
HIGHLIGHTS.
Medicaid expansion improves insurance coverage for endometrial cancer patients.
Expansion improves coverage to a larger extent for some groups.
Some groups that experience a larger increase in coverage present at earlier stage.
Acknowledgments
Funding support
This work was supported by the National Cancer Institute (K01CA21845701A1 to ASF).
Footnotes
Declaration of competing interest
The authors have no relevant conflicts of interest.
References
- [1].Nomura H, et al. , Effect of Taxane plus platinum regimens vs doxorubicin plus cisplatin as adjuvant chemotherapy for endometrial cancer at a high risk of progression: a randomized clinical trial, JAMA Oncol 5 (6) (2019) 833–840. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Matei D, et al. , Adjuvant chemotherapy plus radiation for locally advanced endometrial cancer, N. Engl. J. Med 380 (24) (2019) 2317–2326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].de Boer SM, et al. , Adjuvant chemoradiotherapy versus radiotherapy alone for women with high-risk endometrial cancer (PORTEC-3): final results of an international, open-label, multicentre, randomised, phase 3 trial, Lancet Oncol 19 (3) (2018) 295–309. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Ott PA, et al. , Safety and antitumor activity of pembrolizumab in advanced programmed death ligand 1-positive endometrial cancer: results from the KEYNOTE-028 study, J. Clin. Oncol 35 (22) (2017) 2535–2541. [DOI] [PubMed] [Google Scholar]
- [5].Le T, et al. , Histopathologic assessment of chemotherapy effects in epithelial ovarian cancer patients treated with neoadjuvant chemotherapy and delayed primary surgical debulking, Gynecol. Oncol 106 (1) (2007) 160–163. [DOI] [PubMed] [Google Scholar]
- [6].Mazurenko O, et al. , The effects of Medicaid expansion under the ACA: a systematic review, Health Aff (Millwood) 37 (6) (2018) 944–950. [DOI] [PubMed] [Google Scholar]
- [7].Sommers BD, Grabowski DC, What is Medicaid? More Than Meets the Eye. Jama 318 (8) (2017) 695–696. [DOI] [PubMed] [Google Scholar]
- [8].Kominski GF, Nonzee NJ, Sorensen A, The Affordable Care Act’s impacts on access to insurance and health care for low-income populations, Annu. Rev. Public Health 38 (2017) 489–505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Manchikanti L, et al. , A critical analysis of Obamacare: affordable care or Insurance for many and coverage for few? Pain Physician 20 (3) (2017) 111–138. [PubMed] [Google Scholar]
- [10].Zerhouni YA, et al. , Effect of Medicaid expansion on colorectal cancer screening rates, Dis. Colon Rectum 62 (1) (2019) 97–103. [DOI] [PubMed] [Google Scholar]
- [11].Fedewa SA, et al. , Changes in breast and colorectal cancer screening after Medicaid expansion under the Affordable Care Act, Am. J. Prev. Med 57 (1) (2019) 3–12. [DOI] [PubMed] [Google Scholar]
- [12].Moss HA, Havrilesky LJ, Chino J, Insurance coverage among women diagnosed with a gynecologic malignancy before and after implementation of the Affordable Care Act, Gynecol. Oncol 146 (3) (2017) 457–464. [DOI] [PubMed] [Google Scholar]
- [13].Mesquita-Neto JWB, et al. , Disparities in access to cancer surgery after Medicaid expansion, Am. J. Surg 219 (1) (2020) 181–184 Epub 2019 Jun 26. [DOI] [PubMed] [Google Scholar]
- [14].Ajkay N, et al. , Early impact of Medicaid expansion and quality of breast cancer care in Kentucky, J. Am. Coll. Surg 226 (4) (2018) 498–504. [DOI] [PubMed] [Google Scholar]
- [15].Jemal A, et al. , Changes in insurance coverage and stage at diagnosis among nonelderly patients with cancer after the Affordable Care Act, J. Clin. Oncol 35 (35) (2017) 3906–3915. [DOI] [PubMed] [Google Scholar]
- [16].Han X, et al. , Comparison of insurance status and diagnosis stage among patients with newly diagnosed cancer before vs after implementation of the patient protection and Affordable Care Act, JAMA Oncol 4(12) (2018) 1713–1720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Bilimoria KY, et al. , The National Cancer Data Base: a powerful initiative to improve cancer care in the United States, Ann. Surg. Oncol 15 (3) (2008) 683–690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].States Getting a Jump Start on Health Reform's Medicaid expansion, [cited 2020 February 10]; Available from: https://www.kff.org/health-reform/issue-brief/states-getting-a-jump-start-on-health/ 2012. [Google Scholar]
- [19].Status of State Action on the Medicaid Expansion Decision, [cited 2020 February 10]; Available from: https://www.kff.org/health-reform/state-indicator/state-activity-around-expanding-medicaid-under-the-affordable-care-act/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Location%22,%22sort%22:%22asc%22%7D 2020. January 10. [Google Scholar]
- [20].Ryan AM, Burgess JF Jr., Dimick JB, Why we should not be indifferent to specification choices for difference-in-differences, Health Serv. Res 50 (4) (2015) 1211–1235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Courtney-Brooks M, et al. , The patient protection and Affordable Care Act: impact on the care of gynecologic oncology patients in the absence of Medicaid expansion in Central Virginia, Gynecol. Oncol 130 (2) (2013) 346–349. [DOI] [PubMed] [Google Scholar]
- [22].Duska LR, Engelhard CL, The patient protection and Affordable Care Act: what every provider of gynecologic oncology care should know, Gynecol. Oncol 129 (3) (2013) 606–612. [DOI] [PubMed] [Google Scholar]
- [23].Wehby GL, Lyu W, The impact of the ACA Medicaid expansions on health insurance coverage through 2015 and coverage disparities by age, race/ethnicity, and gender, Health Serv. Res 53 (2) (2018) 1248–1271. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Soni A, et al. , Effect of Medicaid expansions of 2014 on overall and early-stage cancer diagnoses, Am. J. Public Health 108 (2) (2018) 216–218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Tsui J, et al. , Association of Medicaid enrollee characteristics and primary care utilization with cancer outcomes for the period spanning Medicaid expansion in New Jersey, Cancer 125 (8) (2019) 1330–1340. [DOI] [PubMed] [Google Scholar]
- [26].Gan T, et al. , Impact of the Affordable Care Act on colorectal cancer screening, incidence, and survival in Kentucky, J Am Coll Surg 228 (4) (2019) 342–353.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Syed ST, Gerber BS, Sharp LK, Traveling towards disease: transportation barriers to health care access, J. Community Health 38 (5) (2013) 976–993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Lazar M, Davenport L, Barriers to health care access for low income families: a review of literature, J. Community Health Nurs 35 (1) (2018) 28–37. [DOI] [PubMed] [Google Scholar]
- [29].Devoe JE, et al. , Insurance + access not equal to health care: typology of barriers to health care access for low-income families, Ann. Fam. Med 5 (6) (2007) 511–518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Barrington DA, et al. , Distance from a Comprehensive Cancer Center: a proxy for poor cervical cancer outcomes? Gynecol. Oncol 143 (3) (2016) 617–621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Shalowitz DI, Vinograd AM, Giuntoli RL 2nd, Geographic access to gynecologic cancer care in the United States, Gynecol. Oncol 138 (1) (2015) 115–120. [DOI] [PubMed] [Google Scholar]
- [32].Mukerji B, et al. , Racial disparities in young women with endometrial cancer, Gynecol. Oncol 148 (3) (2018) 527–534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Bregar AJ, et al. , Disparities in receipt of care for high-grade endometrial cancer: a National Cancer Data Base analysis, Gynecol. Oncol 145 (1) (2017) 114–121. [DOI] [PubMed] [Google Scholar]
- [34].Randall TC, Armstrong K, Differences in treatment and outcome between African-American and white women with endometrial cancer, J. Clin. Oncol 21 (22) (2003) 4200–4206. [DOI] [PubMed] [Google Scholar]
- [35].Maxwell GL, et al. , Racial disparity in survival among patients with advanced/recurrent endometrial adenocarcinoma: a Gynecologic Oncology Group study, Cancer 107 (9) (2006) 2197–2205. [DOI] [PubMed] [Google Scholar]