People with colorectal cancer in SEER-Medicare

Colorectal cancer (CRC) is the second-most expensive cancer to treat, at an estimated projected annual cost in the United States of $19 to $26 billion in 2020.¹ The high costs and substantial cost-sharing associated with cancer medicines may put Medicare beneficiaries treated for CRC at risk for delaying or discontinuing their therapies, possibly increasing other health care use, such as inpatient admissions or emergency department (ED) visits.^2,3

In January 2006, with US drug prices continuing to climb ever higher, the Centers for Medicare and Medicaid Services implemented the optional Medicare prescription drug benefit (Part D) program to offer subsidized drug plans to all Medicare beneficiaries.⁴ Before Part D, Medicare beneficiaries with cancer faced substantial out-of-pocket (OOP) costs: according to data from the Medical Expenditure Panel Survey, from 1996 to 2005, 57 percent of Medicare beneficiaries over 65 years of age with cancer had annual OOP costs exceeding 10 percent of their annual income.^5,6 More-recent research suggests that since Part D’s introduction, average OOP burden among Medicare beneficiaries with cancer has declined by 43%.⁷

Published data from the National Cancer Institute’s (NCI’s) Surveillance, Epidemiology, and End Results (SEER)-Medicare program indicate that the proportion of Medicare fee-for-service (FFS) beneficiaries with CRC who were covered by the Part D program during the calendar year after diagnosis ranged from 55 percent in 2007 to 71 percent in 2013; about 57 percent to 70 percent of Medicare FFS beneficiaries without cancer were enrolled in Part D during the same period.⁸ Those who elect not to enroll in Part D initially can avoid paying a late enrollment penalty for subsequent Part D enrollment by maintaining “creditable coverage” for prescription drugs (meaning coverage from another source, such as a current or former employer or union, that is expected to pay at least as much as the standard Medicare prescription drug coverage, on average).⁹

To date, few researchers have studied the impact of Part D coverage on costs or outcomes among people with cancer.^2,10,11 Research not specific to people with cancer has found that Part D has reduced OOP costs⁴ and mortality;^12,13 decreased hospital admissions, overall resource use, and total Medicare costs;^13,14 and increased use of prescription medicines^4,14,15—all while keeping monthly premiums and program costs below projections¹⁶ and receiving high patient experience ratings.¹¹

The objective of this study was to evaluate patterns of Part D uptake and costs and to understand the effects of coverage on inpatient stays and ED use among a cohort of individuals diagnosed with CRC in the United States from 2007 through 2010 and matched beneficiaries without cancer. We evaluated the following research questions: (1) What proportion of beneficiaries with and without CRC had Part D coverage during this time period, and how did this change over time? (2) How did Part D expenditures compare between Medicare beneficiaries with and without CRC? (3) Among people with CRC, was Part D coverage associated with any differences in the occurrence or volume of inpatient admissions or ED use during the initial treatment phase?

We used the most-recent data available at the time of our study, which covers individuals diagnosed from January 2007 through December 2010 and followed through December 2013. We used cancer registry data from the SEER database linked with Medicare claims and enrollment data (SEER-Medicare). SEER-Medicare data have been used in more than 1,600 peer-reviewed studies since 1993.¹⁷ The data also include cancer stage, which is difficult to infer from claims data alone, and which is often a major driver of costs and clinical outcomes.¹⁸

The cohort of Medicare FFS beneficiaries with CRC included 12,774 people, of whom 39 percent had complete Part D coverage (Table 1).

The matched comparison cohort included 12,774 Medicare FFS beneficiaries without cancer, of whom 38 percent had complete Part D coverage (P for difference between cohorts = .119). Within both cohorts, individuals with Part D coverage differed significantly from those without complete coverage on sex, race, rurality, US Census region, and comorbidity category. Among beneficiaries with CRC, those with complete Part D coverage had a significantly different age distribution (with more enrollees in the 65–74 age category) and were less likely to be married. Among both cohorts, those with complete Part D coverage were more likely to be in the youngest age group (65–74 years), female, and white; to live in a rural location; and to live in the Midwest.

In Figure 1, we show trends in Part D uptake by cohort and time (relative to index).

Figure 1

Patterns of Part D enrollment among Medicare FFS beneficiaries with and without colorectal cancer

^a Individuals indexed/diagnosed 2008–2011.

^b Individuals indexed/diagnosed 2007–2011.

Continuous Part D enrollment during the diagnosis year and 2 subsequent years, termed “complete Part D coverage” in this study, was associated with a reduction in ED visits among nondual Medicare FFS beneficiaries with CRC over 65 years of age during the 12 months postindex, after adjusting for potential confounding effects. Among people with complete Part D coverage, no significant differences were observed between beneficiaries with CRC and those without cancer in Part D costs. Although Part D uptake significantly increased in the year postindex, no significant differences were observed between cohorts in overall uptake of the program. Of note, our estimates of Part D uptake differ from those published by the SEER-Medicare program⁸ because we excluded dual enrollees from our sample, as they are automatically enrolled in Part D.²⁵

Both CRC and non-CRC cohorts showed statistically significant increases in Part D uptake over time, consistent with trends in the larger Medicare population. According to the Kaiser Family Foundation, the program had average annual increases of 6.5 percent from 2006 through 2013.²⁶ Our study found that Part D enrollment increased by 9 percent over time in this sample of people with CRC and by 10 percent in the matched cohort. The differences we observe could be because we have only nondual FFS beneficiaries in our data.

The association of Part D with slightly reduced odds of an ED visit only among people with CRC is puzzling, partly because the Part D program does not directly cover cancer treatments for this population. Although oral chemotherapy for CRC was available during the study period in the form of capecitabine (Xeloda), it was covered under Part B, and not Part D, because it is an alternative form of the infusion therapy 5-Fluorouracil.²⁷ The rules for oral antinausea drugs are much the same.²⁸

Because of these coverage policies, the observed effects of Part D among people with CRC could be related to better management of medication-sensitive conditions, such as diabetes, congestive heart failure, stroke, and myocardial infarction. As shown in Table 1, those with complete Part D coverage were significantly more likely to have at least one major comorbidity. Theoretically and empirically, improved access to medicines leads to improved adherence, which helps improve blood pressure, cholesterol, and blood glucose levels, potentially preventing or delaying ED visits.^10,29–31 Inpatient use in one period decreases overall ED use the next, suggesting that the intensive care provided during hospital stays may be effective at addressing clinical issues that could subsequently manifest in emergencies.³² Therefore, it is also possible that the effects among Part D enrollees in the CRC cohort are related to receiving more health care services in general.

Other research on the effects of Part D on outcomes among people with cancer is limited. One study used Medical Expenditure Panel Survey data for 2002 through 2010 and employed a difference-in-differences design to explore the effects of Part D enrollment on OOP costs, medicine use, hospitalizations, ED visits, and outpatient visits among Medicare beneficiaries with cancer (n = 4,729). The authors found that Part D was associated with a 43 percent decline in OOP costs and fewer outpatient visits.³³ Studies among women with breast cancer have found that Part D was associated with reduced mortality, improved adherence to medicine, and reduced racial/ethnic disparities.^34,35 More broadly, researchers have failed to find any associations between Part D enrollment and self-reported health status, limitations in activities of daily living and instrumental activities of daily living, ED visits and hospital admissions (prevalence, counts, and spending), or death among Medicare enrollees, according to nationally representative data from the Medicare Current Beneficiary Survey (n = 56,293) combined with Medicare claims.³⁶ One study that did examine the effect of Part D on ED visits among a general Medicare population found that, using a difference-in-differences approach, Part D enrollment reduced non-urgent ED visits but not urgent ED visits.³⁷ The authors’ findings support our conclusion in this study: Part D may have led to better health management and reduced unnecessary ED usage.

The current study excluded Medicare Advantage (MA) enrollees; they are not included in SEER-Medicare, although they represent a large and growing group of seniors. Part D studies using only Medicare FFS claims could be susceptible to confounding because of an increase in MA enrollment during the study period. We addressed this by including an index year indicator in the models, which adjusts for differences in MA enrollment over time. We have no reason to suspect that including MA enrollees would have changed our findings; however, this is an area deserving further exploration.

Although we have characterized groups in terms of complete vs. partial/no Part D enrollment, most of those in the partial/no Part D group probably had some kind of insurance coverage for prescription drugs, whether through a former employer or union, some other program, or self-funded; still, an estimated 12 percent of Medicare beneficiaries had no drug coverage during the study period.²⁶ Unfortunately, our data did not provide any information about alternative coverage or lack thereof.

Our outcome measures were limited to ED visits and hospitalizations during the 12-month period after diagnosis. Despite these limitations, our study’s relatively large sample of more than 25,000 Medicare beneficiaries, along with a matched-cohort design using the reliable and well-validated SEER-Medicare database, allows us to get a clearer picture of costs, enrollment patterns, and health outcomes among people with CRC and similar people without cancer. Future research is needed to understand a broader set of potential outcomes, including OOP costs, ideally with better information on prescription drug coverage among those not covered by Part D.

The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or RTI International.

Funding support for Lisa M. Lines, Sujha Subramanian, and Sonja Hoover was provided by the Centers for Disease Control and Prevention (Contract No. 200-2008-27958-0036, to RTI International). The authors declare no conflicts of interest.

RTI Press Associate Editor: Franziska Rokoske