Recent Trends in Cost-Related Medication Nonadherence Among Cancer Survivors in the United States

BACKGROUND: Cancer survivors avoid necessary medications due to costs. OBJECTIVE: To estimate the prevalence of cost-related medication non-adherence (CRN) by age and insurance status over a number of years in a national sample of U.S. cancer survivors. METHODS: Using the 1999-2012 National Health Interview Survey, we examined the prevalence and correlates of self-reported CRN, that is, patient-reported inability to afford prescribed medications within the past 12 months, resulting in nonadherence among cancer survivors. Descriptive statistics and multivariate logistic regression models were used to identify time trends in CRN among cancer survivors. RESULTS: In a nationally representative sample of 20,517 cancer survivors from 1999 to 2012, 1,788 (8.7%) survivors reported CRN, representing approximately 436,498 individuals nationally. CRN increased significantly from 11.8% (1999-2005) to 16.9% (2006-2012) among younger cancer survivors (P < 0.001). Among young cancer survivors (aged 45-64 years), the uninsurance rate was higher for those reporting CRN in the years 2006-2012 (48.5%) than in the earlier period (42.5%; P = 0.043). Among older cancer survivors, insurance coverage through Medicare only was lower for individuals reporting CRN in the years 2006-2012 (5.8%) than in the earlier period (7.8%; P = 0.0210). In adjusted models, younger cancer survivors without health insurance were more likely to report CRN than those with supplemental private insurance with Medicare, and older cancer survivors with Medicare only were more likely to report CRN than those with supplemental private insurance with Medicare. CONCLUSIONS: Increasing trends in CRN were evident among younger cancer survivors.

T he cost of prescription drugs has continued to increase rapidly over the past few decades. By 2012, the aggregate expenditures on prescription drugs reached $263 billion, or 9.7%, of total U.S. health expenditures. 1 Increases in drug costs have placed an increased financial burden on patients. This increased burden has led cancer survivors to lower dosage and frequency of drugs below the recommended level and to avoid filling prescriptions to reduce medical costs. 2,3 Cost-related medication nonadherence (CRN) is associated with a wide range of negative health outcomes in many chronically ill adults, including poorer physical and mental functioning, myocardial infarction, stroke, increased use of emergency and hospital services, and death. 4,5 Increases in medical care cost is associated with the substantial increase in out-of-pocket expenses of cancer patients and their family members during diagnosis, treatment, and follow-up care. 6,7 Increased treatment cost has been shown to be a factor in medication nonadherence. 6 One study found that Medicare breast cancer survivors treated with aromatase inhibitors experienced higher levels of financial difficulties than breast cancer survivors treated with less expensive tamoxifen. 8 Financial stress is a significant concern for cancer survivors with a history of chemotherapy and radiation, and financial difficulties often lead to skipping prescribed medications. 9 Previous studies have examined CRN and its risk factors among cancer survivors. [10][11][12][13]   investigated gender-specific differences in the prevalence of CRN using the National Health Interview Survey (NHIS) and found that female cancer survivors were more likely to report CRN and that having any type of insurance reduced CRN. 10 Another study estimated the prevalence of CRN by race and ethnicity and found that older African-American and Hispanic cancer survivors were more likely to report CRN compared with • The cost of prescription drugs has continued to increase rapidly over the past few decades. • The increased cost burden of prescription drugs has led cancer survivors to lower dosages and frequency of drugs below recommended levels and to avoid filling prescriptions to reduce medical costs. • Cost-related medication nonadherence (CRN) is associated with a wide range of negative health outcomes, and previous studies have identified a number of factors associated with CRN.

What is already known about this subject
• This study examined trends in CRN over the years by age and insurance status in a national sample of U.S. cancer survivors. • A significant increasing trend in the prevalence of CRN among younger cancer survivors was found, even after adjusting for race, gender, region, household income, educational attainment, number of comorbidities, self-rated health, activity limitation, and cancer site.

Study Sample
Cancer survivors were identified as those who answered "yes" to the question "Have you ever been told by a doctor that you had cancer?". Respondents who were younger than 45 years were excluded from the analysis because cancer survivors in the age group 20-39 years are often diagnosed with different types of cancers than those traditionally seen in adult oncology clinics. 19 For our analyses, we excluded respondents with a diagnosis of non-melanoma skin cancers because of their distinctly different treatment strategies and prognoses. 10 The resulting dataset included information on 22,488 cancer survivors aged 45 years or older. In the dataset, 427 participants did not report their race; 176 did not report their educational status; 1,029 did not report their household income levels; 251 did not report their health insurance status; 50 did not report their self-rated health status; and 154 participants did not report their cancer type. After excluding all cases with at least 1 missing piece of relevant data (n = 1,971), a total of 20,517 cancer survivors were included in the analysis.

Statistical Analysis
The main dependent variable of interest was CRN, which was defined as the self-reported inability to afford prescribed medications with the result of not taking medication within the previous 12 months. Inability to afford medications was identified by respondents who answered "yes" to the following question: "During the past 12 months, was there a time when you did not take your prescribed medication because of cost? Do not include over-the-counter medication." The dependent variable was a binary variable (yes/no). The main independent variable of interest was time in years. We examined CRN by age (in years) and insurance status across time in a national sample of U.S. cancer survivors, and time or year was included in the analysis as a continuous variable ranging from 1 to 14. We stratified the analyses into groups composed of individuals aged 45-64 years (younger patients) and individuals aged 65 years or older (older patients) because U.S. adults generally qualify for Medicare at age 65, and previous research has found that age modifies access to medications among cancer survivors. 20,21 A number of research studies have found that insurance type or enrollment in insurance plans modifies access to medications among cancer survivors. 13,20,22 Therefore, health insurance type was categorized into 1 of 7 groups: Medicare with supplementary private insurance, private health insurance only, Medicaid only, Medicare only, Medicare and Medicaid dual coverage, other insurance, and uninsured. To simplify health insurance type, we combined remaining health insurance types into an "other" category that included TRICARE, VA, SCHIP, and other military coverage. Gender was categorized as male or female. Race was self-reported and categorized as non-Hispanic white, black or African American, and other. non-Hispanic whites. 11 Moreover, Nekhlyudov et al. (2011) investigated CRN among elderly Medicare enrollees using the 2005 Medicare Current Beneficiary Survey and found that lower income, having nonemployer-based medication insurance, and being African American were significant risk factors. 12 Nekhlyudov et al. also found that approximately 10% of elderly Medicare beneficiaries with and without cancer reported CRN and lowered expenditures on basic needs to offset the high cost of medications. 12 Sabatino et al. (2006) examined CRN among U.S. adult cancer survivors aged 18-64 years, using the 1998 and 2000 NHIS and found that uninsured and publicly insured survivors were more likely to experience CRN. 13 They also found that greater than 20% of cancer survivors reported delaying or missing needed care within the previous year because of cost and that those unmet needs were greatest among cancer survivors with health insurance coverage; 42.8% of this group reported that they did not obtain needed prescription drugs because of cost. 13 Although a number of studies have examined the factors associated with CRN, little is known about the CRN trend over the years because of the economic downturn and health policy changes. To understand the trend in CRN, we have estimated the incidence of CRN using a national sample of cancer survivors for the years 1999-2012. Our dataset covers 7 years (1999-2005) before the introduction of Medicare Part D and 7 years after (2006-2012).
We had 2 main research questions. First, we hypothesized that, because of the rapid increase in prices of prescription drugs, CRN should show an increasing trend among younger cancer survivors (aged 45-64 years) and, in particular, those without any health insurance coverage. Given that many survivors in this age group lack health insurance coverage, 13,14 and higher out-of-pocket costs were strongly associated with nonadherence to therapy, 15 we suspected that CRN would increase among younger cancer patients. Thorpe et al. (2003) found that 11% of cancer survivors aged < 65 years lacked insurance primarily near the time of diagnosis. 14 Second, we hypothesized that CRN should show a declining trend among older cancer survivors, particularly among those with Medicare.

■■ Methods Data Source
We used data from the NHIS, a cross-sectional householdbased, nationally representative survey that has been conducted annually by the National Center for Health Statistics since 1957. 16,17 We combined data from annual waves of the survey from 1999 through 2012 to obtain a sample of cancer survivors in order to observe long-term trends in CRN. The NHIS uses a complex survey design that includes clustering, stratification, and multistage sampling to collect self-reported health data from a representative sample of the civilian, noninstitutionalized U.S. population. 18

Characteristics of Cancer Survivors by Age Group and Time: National Health Interview Survey, 1999-2012
U.S. region of residence at the time of the interview was categorized as Northeast, Midwest, South, and West. Respondents self-reported the highest grade of school completed, which was categorized as no high school diploma, high school diploma, and greater than a high school diploma. Three categories of household income were defined: lower than $35,000, $35,000-$75,000, and higher than $75,000. From the dataset, it was not possible to identify Medicare beneficiaries with Medicare Part D insurance for the same time frame related to CRN.
The number of comorbidities was categorized into 3 groups (0-1, 2-3, and 4 or more). To measure comorbidities, we accounted for the number of self-reported medical conditions, including coronary heart disease, hypertension, emphysema, stroke, asthma, heart condition, and arthritis. Self-rated health was categorized as good or better (excellent, very good, and good) and fair or poor. Activity of daily living limitations (ADLL) was categorized as "yes/no," with "yes" indicating the presence of at least 1 ADLL.
We stratified the analyses into groups composed of individuals aged 45-64 years and those aged 65 years or older. Chi-square statistics were used for testing differences in the values of different variables, including insurance type, gender, race, region, household income, education, number of comorbidities, self-rated health, activity of daily living limitations, and cancer site between years 1999-2005 and years 2006-2012. In addition, the weighted percentage of individuals reporting CRN was calculated for insurance type, gender, race, region, household income, education, number of comorbidities, selfrated health, activity of daily living limitations, and cancer site.
Multivariate logistic regression models, stratified by the 2 age groups, were used to calculate the adjusted associations between insurance type, time, and the outcome measure. Time was examined as a count variable in Table 4 and as a group variable (1999-2005 vs. 2006-2012) in Tables 1, 2, and 3. The odds ratios (ORs) with 95% confidence intervals (CI) were calculated to measure the strength of association between factors affecting CRN and incidence of CRN.
Sampling weights were derived by comparing the number of individuals in the dataset with national estimates of cancer survivors for each of the years. This comparison ensured that the survey weights could be used to obtain national values and proportions when a subset of NHIS data was included in the analysis. After applying the weights to the sample, the unweighted count of 20,517 cancer survivors represented 5.2 million cancer survivors, the aggregate number of U.S. cancer survivors aged 45 years or older for the years 1999-2012.    Table 4 shows the multivariate logit model of variables associated with CRN among cancer survivors. In addition to the variables listed in Table 3, we included year of survey as a count variable to estimate the time trend. Introduction of this variable marginally decreased ORs relative to the model in Table 3. Among younger cancer survivors, there was an increasing trend in CRN from 1999-2012, and this trend persisted among younger cancer survivors even after adjusting for all relevant covariates. Moreover, those without health All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

■■ Results
The characteristics of adults with a history of cancer, stratified by age group and survey year, are summarized in Table 1. The final sample had 20,517 cancer survivors aged 45 years or older, which represented an estimated 5.2 million U.S. cancer survivors. Younger cancer survivors were slightly more likely than older cancer survivors to be female and black or African American. Younger cancer survivors reported higher CRN, higher household income, higher educational attainment, and less activity limitations than older cancer survivors. From 1999 to 2005, 96% of older cancer survivors had Medicare alone or Medicare with private insurance or Medicaid. Medicare coverage of elderly cancer survivors increased to about 98% during 2006-2012. For younger cancer survivors, private insurance coverage was 70% during 1999-2005 and 64% during 2006-2012. Lower private insurance coverage during 2006-2012 compared with 1999-2005 was offset by increased coverage through public insurance programs, keeping the uninsurance rate more or less similar over these 2 time frames for the younger group. Therefore, these 2 age groups differed significantly in terms of insurance status and type of insurance coverage. Figure 1 presents the percentage of U.S. cancer survivors who did not take needed prescription drugs because of cost. During the period from 1999 to 2012, 1,788 cancer survivors reported CRN (8.7%). Among younger cancer survivors, the prevalence of CRN increased from 9.9% in 1999 to 16.9% in 2012 (P < 0.0001). The prevalence of CRN increased from 2.4% to 3.5% in 2012 for older cancer survivors as well (P = 0.039).   insurance coverage were 5.40 times more likely to report CRN than those with supplemental private insurance with Medicare (95% CI = 3.57-8.17). Note that insurance coverage is also time dependent, but the coefficient of the year variable should be able to indicate potential effects of increasing cost and price of medicines after controlling for insurance status and other individual level factors.

■■ Discussion
In this nationally representative sample of U.S. cancer survivors, 8.7% (an estimated 436,498 individuals nationally) reported CRN from 1999 to 2012. A significantly increasing trend in the prevalence of CRN was evident among younger cancer survivors over the years (P < 0.001). The prevalence of CRN showed no time trend for the years 1999-2012 for the older cancer survivors. Among younger cancer survivors (aged 45-64 years), uninsured status was found to be a significant risk factor of CRN. For younger and older patients, risk of CRN was found to be higher for individuals with Medicare only coverage. Therefore, having Medicare coverage alone does not protect individuals effectively against CRN. Among younger patients, prevalence of CRN was higher among female patients, although it did not show any statistical differences by race after controlling for insurance status, income, region, household income, educational attainment, number of comorbidities, self-rated health, activity limitation, and cancer site. For older patients, race was an important factor affecting prevalence of CRN; African Americans were 85% more likely to experience CRN compared with whites, after controlling for other potential covariates.
Household income was clearly an important factor affecting prevalence of CRN for both younger and older patients. The likelihood of reporting CRN declined rapidly with the increase in household income. Increasing number of comorbidities also tended to increase the prevalence of CRN for all patients, and the likelihood ratios of the variable were similar for younger and older patients. Activity limitations and fair or poor health status increased the prevalence of CRN.
Although This increase in CRN can be explained by the time trend in CRN and the changes in insurance status of the population over the years. Less than 15% of younger patients were covered by Medicare even in the 2006-2012 period. According to Yin et al. (2008), Medicare Part D coverage resulted in modest increases in average drug use among Medicare Part D beneficiaries. 22 Langa et al. (2004) showed that a previous history of cancer was associated with an additional $240 per year in outof-pocket expenditures, whereas current cancer treatment was associated with an additional $670 per year in out-of-pocket expenditures, even after adjusting for differences in sociodemographics, living situation, functional limitations, comorbid chronic conditions, and insurance coverage. 6 Increasing out-ofpocket expenses is likely one of the factors contributing to the CRN time trend.

Limitations
This study has some limitations to consider. First, patients were asked to report CRN over the previous 12 months, so the possibility of recall bias is always present. Second, although the NHIS was conducted by trained interviewers, the survey did not verify reported medical conditions and care utilization with medical records. Finally, the survey suffers from selection bias, since less healthy patients, who are likely to be institutionalized, are not in the sample. The effect of Medicare Part D coverage could not be evaluated in this study because of the difficulty in defining the Part D coverage for the time frame over which CRN was defined. Moreover, only 37% of the older patients of the study (older than 65 years), were enrolled in Part D.

■■ Conclusions
In this nationally representative study, we found that younger cancer survivors without health insurance were more likely to report CRN than those with supplemental private insurance with Medicare, and older cancer survivors with Medicare only were more likely to report CRN than those patients with supplemental private insurance with Medicare, even after controlling for variables such as race, gender, region, household income, educational attainment, number of comorbidities, self-rated health, activity limitation, and cancer site. This finding is especially true for uninsured younger cancer survivors who faced the largest cost barriers to prescribed medication. Approximately 14.6% of younger cancer survivors reported inability to afford medications in recent years.