Cost-Related Medication Nonadherence and Patient Cost Responsibility for Rural and Urban Cancer Survivors

PURPOSE:

The relationship between out-of-pocket spending and cost-related medication nonadherence among older rural- and urban-dwelling cancer survivors is not well understood.

METHODS:

This retrospective cohort study used the Surveillance, Epidemiology, and End Results Program, Medicare claims, and the Consumer Assessment of Healthcare Providers and Systems survey linked data resource linked data (2007-2015) to investigate the relationship between cancer survivors' cost responsibility in the year before and after report of delaying or not filling a prescription medication because of cost in the past 6 months (cost-related medication nonadherence). Secondary exposures and outcomes included Medicare spending and utilization. Generalized linear models assessed bidirectional relationships between cost-related medication nonadherence, spending, and utilization. Effects of residence were assessed via interaction terms.

RESULTS:

Of 6,591 older cancer survivors, 13% reported cost-related medication nonadherence. Survivors were a median 8 years (interquartile range, 4.5-12.5 years) from their cancer diagnosis, 15% were dually Medicare/Medicaid-eligible, and prostate (40%) and breast (32%) cancer survivors were most prevalent. With every $500 USD increase in patient cost responsibility, risk of cost-related medication nonadherence increased by 3% (risk ratio, 1.03; 95% CI, 1.02 to 1.04). After report of cost-related medication nonadherence, patient cost responsibility was 22% higher (95% CI, 1.11 to 1.32) compared with those not reporting nonadherence, amounting to $523 USD (95% CI, $430 USD to $630 USD). Medicare spending and utilization were also higher before and after report of cost-related nonadherence versus none. For survivors residing in rural (18%) and urban (82%) areas, residence did not modify adherence or cost outcomes.

CONCLUSION:

A bidirectional relationship exists between patient cost responsibility and cost-related medication nonadherence. Interventions reducing urban- and rural-dwelling survivor health care costs and cost-related adherence barriers are needed.

INTRODUCTION

Cancer survivors are increasingly at risk of financial hardship and its associated adverse outcomes because of out-of-pocket spending associated with initial cancer treatment and survivorship care.^1-4 Across all cancer types and stages, annualized out-of-pocket costs associated with cancer were recently estimated at $2,443 USD, $593 USD, and $4,271 USD in the initial, survivorship, and end-of-life phases of care.⁵ Cost-related medication nonadherence is a common coping behavior of financial hardship, where patients delay or skip recommended medication doses, split pills, or leave prescriptions unfilled because of their cost.^6-9 In older cancer survivors, medication nonadherence is associated with hospitalizations, lost productivity, and functional limitations, which further increase patient spending and exacerbate existing financial hardship.^10-12 Thus, decreasing both out-of-pocket spending and risks for cost-related nonadherence may optimize survivorship outcomes for cancer survivors.

Medication nonadherence and out-of-pocket health care spending may be more prevalent in areas with fewer health care resources, such as rural areas. Individuals with cancer residing in rural compared with urban areas more often report financial problems due to their cancer,¹³ potentially because of health care access issues resulting in treatment delay-related disease complications and increased severity.¹⁴ Rural survivors may also face barriers related to transportation, pharmacy proximity or density, or insurance coverage that could complicate medication adherence.^15-17 Thus, survivors residing in rural areas may have higher risk of medication nonadherence and negative sequelae of out-of-pocket spending compared with those residing in areas with more health care resources.

Directionality of the relationship between cost-related medication nonadherence and out-of-pocket health care spending is unclear. A bidirectional relationship may exist, where increased health care use and associated out-of-pocket health care spending may result in medication nonadherence, and medication nonadherence may also lead to increased out-of-pocket health care spending because of poor health and increased health care utilization.^18,19 Additionally, the relationship between cost-related medication nonadherence may differ between survivors residing in rural and urban areas because of variation in health care resources and barriers to care. Understanding these associations is important in identifying targetable areas for intervention, thus reducing out-of-pocket costs, preventing medication nonadherence, and lowering the risk of adverse outcomes. Therefore, this study seeks to understand the bidirectional relationship between cost-related medication nonadherence and patient-attributed health care expenses for older cancer survivors living in rural and urban areas.

METHODS

Data and Sample

This retrospective cohort study used the Surveillance, Epidemiology, and End Results Program, Medicare claims, and the Consumer Assessment of Healthcare Providers and Systems survey–linked data resource (SEER-CAHPS) from 2007 to 2015. This resource links SEER cancer registry data, Medicare enrollment and Fee-for-Service claims, and Medicare CAHPS surveys, which capture patient-reported care experience data.²⁰

Medicare fee-for-service beneficiaries with a first, primary cancer diagnosis of breast, colorectal, lung, prostate, or gynecologic (cervical, ovarian, uterine, or other female genital system) cancer between 2007 and 2014 were included. These cancer types were chosen because of clinical prevalence. Included patients had at least 18 months between their cancer diagnosis and completion of the CAHPS survey to ensure adequate data capture of cost and utilization exposures. Exclusion criteria included health maintenance organization enrollment or incomplete Medicare coverage (parts A, B, D) in the 18 months before and 12 months after the CAHPS survey date, carcinoma in situ, male breast cancers, and discordant Medicare and SEER sex designations.

Exposures and Outcomes

Cost-related medication nonadherence and out-of-pocket spending, proxied by patient cost responsibility, served as both the primary exposure and outcome of our study because of their hypothesized bidirectional relationship. Cost-related medication nonadherence was self-reported using the binary (yes or no) CAHPS survey question, “In the past 6 months, did you ever delay or not fill a prescription because you felt that you could not afford it?” A response of yes denoted cost-related medication nonadherence.

Patient cost responsibility was defined as payment to providers for which Medicare beneficiaries were responsible through deductibles, coinsurance, and copayments.^21,22 Overall and health care service–stratified costs to beneficiaries were summed from Medicare inpatient, outpatient, durable medical equipment, and prescription drug event files. Patient cost responsibility was quantified both in the year before (7-18 months before survey because of the 6-month survey lookback period) and the year after (0-12 months after survey) report of cost-related medication nonadherence (Fig 1).

FIG 1.

Study data capture timeline. CAHPS, Consumer Assessment of Healthcare Providers and Systems.

Components of health care spending and utilization potentially associated with cost-related medication nonadherence were assessed as secondary exposures and outcomes, including Medicare payment (sum of inpatient, outpatient, physician visit/carrier, durable medical equipment, home health, and hospice spending), short-stay hospitalizations, and emergency department (ED) visits. These variables were also measured before and after self-report of cost-related medication nonadherence (Fig 1).

Urban/Rural Residence

County-level patient residence at cancer diagnosis was assessed as an effect modifier and classified on the basis of USDA Rural-Urban Continuum Codes found in SEER data.^23,24 Nonmetropolitan counties were categorized as rural (Rural-Urban Continuum Codes 4‐9) and metropolitan counties as urban (Rural-Urban Continuum Codes 1-3).

Covariables

Other captured patient demographic and clinical characteristics included cancer type, age at the time of survey, sex, race/ethnicity, census tract poverty rate at cancer diagnosis, education, Medicare/Medicaid dual enrollment, National Cancer Institute comorbidity index score,^25-27 cancer stage, and the time from cancer diagnosis to CAHPS survey.

Statistical Analysis

Patient characteristics were described using medians and interquartile ranges (IQRs) for continuous variables and frequencies for categorical variables. Mean differences, or effect sizes, were calculated and assessed using Cohen's²⁸ d (0.2 small, 0.5 medium, and 0.8 large effect) or Cramer's V (0.1 small, 0.3 medium, and 0.5 large effect) to determine the magnitude of relationships in bivariate associations.

We used two models to estimate the bidirectional relationship between cost-related medication nonadherence and patient cost responsibility. The first model estimated risk of cost-related medication nonadherence by prior patient cost responsibility using modified Poisson models.²⁹ The results were reported by calculating risk ratios (RRs) and 95% CIs. The second model estimated patient cost responsibility by prior cost-related medication nonadherence using gamma generalized linear models. The results were reported using expenditure ratios (ERs), which is a ratio of costs for those reporting nonadherence versus those who did not. Using two separate models allowed assessment of causal relationships between cost-related medication nonadherence and patient cost responsibility. Total and service-specific (inpatient, outpatient, and prescription drug) patient cost responsibilities were estimated separately.

Secondary analyses estimated the bidirectional relationship between cost-related medication nonadherence and Medicare spending, ED visits, or hospitalizations using modified Poisson, gamma generalized linear, and negative binomial models as appropriate. Dollar amount differences in patient cost responsibility (β) for survivors reporting nonadherence versus those who did not were also estimated using gamma generalized and negative binomial models.

To understand the effect of residence on our exposures and outcomes, an interaction term (EXPOSURE × RESIDENCE) was added to all models described above. Interaction significance was assessed using likelihood ratio tests. All models were assessed for fit and adjusted for age, sex, race, census tract poverty rate, education, Medicare/Medicaid dual enrollment, residence, comorbidity count, cancer type, cancer stage, and time since diagnosis. Precision of estimated results were confirmed in the final analytic sample, with power for a hypothetically observed effect size Cohen's r = 0.1, 0.3, and 0.5 at approximately 100% (α = .01).

RESULTS

Sample Characteristics

We included 6,591 cancer survivors in our analysis (Appendix Fig A1, online only), with a median of 8 years (IQR, 4.5-12.5 years) from their cancer diagnosis when surveyed (Table 1). Survivors' median age was 76 years (IQR, 71-82 years), 82% were non-Hispanic White, 56% had more than a high school degree, and 15% were Medicare/Medicaid dually enrolled. Prostate and breast cancer survivors were most prevalent (40% and 32%, respectively) and most were diagnosed with a localized or regional cancer (92%). Median total patient cost responsibility was $1,407 USD (IQR, $741-$2,637 USD) in the year before the survey lookback period and $1,266 USD (IQR, $653-$2,493 USD) in the year after survey. Median Medicare spending was $3,480 USD (IQR, $1,525-$8 USD, 810) in the year before the survey lookback period and $3,335 (IQR, $1,407-$8,511) in the year after survey. Almost a third of survivors had at least one ED visit in the year before and after the CAHPS survey, whereas 17% had at least one hospitalization in the year before and after the CAHPS survey (Table 1).

TABLE 1.

Sample Sociodemographic and Clinical Characteristics (N = 6,591)

Cost-Related Medication Nonadherence and Patient Cost Responsibility

In our sample of cancer survivors, 13% reported cost-related medication nonadherence in the previous 6 months. Those more often reporting nonadherence had ≥ 3 comorbidities compared with none (19% v 0%, V = 0.1) and were non-Hispanic Black compared with non-Hispanic White (18% v 11%, V = 0.08).

In adjusted models, with every $500 USD increase in patient cost responsibility, risk of cost-related medication nonadherence increased by 3% (RR, 1.03; 95% CI, 1.02 to 1.04; Appendix Table A1, online only). In service-specific models, risk of cost-related medication nonadherence was highest with increases in prescription drug cost responsibility (RR, 1.07; 95% CI, 1.04 to 1.10). Survivors reporting cost-related medication nonadherence had $571 USD higher patient cost responsibility in the previous year compared with those who did not (95% CI, $482 USD to $673 USD; Fig 2A), with most of the cost difference found in outpatient spending (β = $336 USD; 95% CI, $277 USD to $404 USD).

FIG 2.

(A) Overall and service-specific patient cost responsibility before report of cost-related medication nonadherence, estimated using generalized linear models with a log link and gamma distribution (N = 6,591). (B) Overall and service-specific patient cost responsibility after report of cost-related medication nonadherence, estimated using negative binomial models (N = 6,591).

After report of cost-related medication nonadherence, patient cost responsibility was 22% higher compared with those who did not (ER, 1.22; 95% CI, 1.14 to 1.30; Appendix Table A1). In service-specific models, the ratio of patient cost responsibilities for those reporting cost-related medication nonadherence compared with those not reporting nonadherence was highest in prescription drug spending (ER, 1.22; 95% CI, 1.11 to 1.32). Survivors reporting cost-related medication nonadherence had $523 USD higher patient cost responsibilities in the subsequent year compared with those who did not (95% CI, $430 USD to $630 USD; Fig 2B).

Cost-Related Medication Nonadherence, Medicare Spending, and Health Care Utilization

Small increases in risk of cost-related medication nonadherence were seen with every $1,000 USD increase in Medicare spending (RR, 1.01; 95% CI, 1.00 to 1.01; Appendix Table A1). Risk of cost-related medication nonadherence also increased by 8% with every ED visit (RR, 1.08; 95% CI, 1.03 to 1.13) and by 10% with every hospitalization (RR, 1.10; 95% CI, 1.01 to 1.18).

After report of cost-related medication nonadherence, Medicare spending increased by 21% compared with those not reporting nonadherence (ER, 1.21; 95% CI, 1.11 to 1.33; Appendix Table A1). Survivors reporting cost-related medication nonadherence also had 22% increased risk of ED visits (95% CI, 1.05 to 1.42) and 23% increased risk of hospitalizations (95% CI, 1.02 to 1.49) compared with those who did not.

Effects of Rural and Urban Residence

Of cancer survivors in our sample, 18% resided in a rural area when diagnosed with cancer. Rural-dwelling survivors had lower education levels, with 52% reporting a high school education or lower, compared with 40% for urban-dwelling survivors (V = 0.1). Rural-dwelling survivors more often lived in high poverty areas (22% rural-dwelling v 9% urban-dwelling survivors, V = 0.16). Cost-related medication nonadherence was similar comparing rural and urban survivors, with 10% and 12% reporting nonadherence, respectively.

In adjusted models, no effect modification by survivor residence was seen when comparing risk of cost-related medication nonadherence by prior health care expenditures or expenditures after report of cost-related medication nonadherence (Appendix Table A2, online only). Rural- and urban-dwelling survivors reporting cost-related medication nonadherence had similar differences in patient cost responsibility in the previous year compared with those who did not (rural β = $570 USD, 95% CI, $326 USD to $877 USD; urban β = $566 USD, 95% CI, $471 USD to $675 USD; Appendix Fig A2A, online only). Similar differences in patient cost responsibility in the subsequent year were also found for rural- and urban-dwelling survivors reporting cost-related medication nonadherence compared with those not reporting nonadherence (rural β = $$477 USD, 95% CI, $237 USD to $785 USD; urban β = $572 USD, 95% CI, $458 USD to $704 USD; Appendix Fig A2B, online only).

DISCUSSION

In this study of older cancer survivors, 13% reported cost-related medication nonadherence. These results were similar to those reported by Nekhlyudov et al,³⁰ who found 10% of older cancer survivors reported cost-related medication nonadherence using the Medicare Current Beneficiary Survey and Medicare claims. Furthermore, our results suggest a bidirectional relationship exists between patient cost responsibility and cost-related medication nonadherence. Small increases in risk of cost-related medication nonadherence were found with every $500 USD increase in patient cost responsibility, whereas larger increases in patient cost responsibility were found for those reporting previous cost-related medication nonadherence compared with those who did not. This study adds to the current findings on cost-related nonadherence^31-37 by quantifying patient cost responsibility, as well as Medicare spending and health care utilization, and examining the bidirectional nature of these patterns within a robust sample of cancer survivors. Our results suggest a need for interventions, such as improved patient-provider communication, education-based drug information sharing, or patient monitoring and follow-up tested in survivors closer to diagnosis,^38-40 to both decrease patient cost responsibility and to monitor and support medication adherence when patients are faced with high patient cost liability many years into survivorship, even when insured.

Risk of cost-related medication nonadherence was highest by prior prescription drug spending, with a 7% increased risk of cost-related medication nonadherence with every $500 USD spent on prescription drugs. Because all survivors in our sample were enrolled in Medicare Part D prescription drug coverage, addressing drug spending even among those insured may be an ideal area for policy-level interventions to decrease patient spending and reduce risk of cost-related medication nonadherence. No out-of-pocket spending limit exists for Medicare Part D enrollees; thus, recent policy efforts have sought to cap annual out-of-pocket costs at $2,000 USD.⁴¹ This could result in substantial out-of-pocket savings, since almost 1.5 million enrollees spent above the catastrophic coverage threshold in 2019.⁴² However, cancer survivors in our study spent well under the catastrophic coverage threshold, yet reported cost-related medication nonadherence. Because survivors in our sample reporting nonadherence spent about $70 USD more on prescription drugs annually than those who did not report nonadherence, it is possible that small offsets in costs could increase medication adherence. However, prevalent issues surrounding prescription drug spending and patient financial burden remain.

Our results support the hypothesized mechanism from medication nonadherence to poorer health, thus driving up health care spending because of increased utilization. Our findings are consistent with a study of Medicare beneficiaries without cancer, where worse self-reported health was found in those reporting cost-related medication nonadherence compared with those adherent.⁴³ The increased risk for ED visits and hospitalizations after report of cost-related medication nonadherence may reflect difficulties in comorbidity management, since 69% of survivors in our study reporting cost-related medication nonadherence had at least one other noncancer comorbidity compared with 56% who did not report nonadherence. Tools such as those included in geriatric assessment could be an ideal provider-level intervention to guide comorbidity management in older cancer survivors and potentially improve medication adherence, improve health, and decrease costs.^44-46 The 2018 American Society of Clinical Oncology guidelines recommend geriatric assessments for adults age ≥ 65 years with a history of cancer.⁴⁷ However, only half of cancer providers are aware of this guideline, with less than half using the comorbidity-specific tool with their patients.⁴⁸ Because cancer survivors receive follow-up care not only from oncologists, but also from primary care providers, geriatricians, or many other specialists, research aimed at improving geriatric assessment dissemination, implementation, and utilization by all providers caring for older cancer survivors is warranted to potentially prevent cost-related medication nonadherence and associated patient spending.

Our study found similar associations comparing cost-related medication nonadherence and cost, and utilization for rural- and urban-dwelling survivors, suggesting geographic disparities may not be as prevalent during survivorship. Varied health care access outcomes have been found comparing urban and rural cancer survivors.⁴⁹ A previous study using SEER-CAHPS data found rural cancer survivors were more likely to report getting needed care quickly when compared with urban survivors.⁵⁰ Another study found no differences in Medicare spending or health care utilization comparing older cancer survivors traveling longer than 1 hour to receive care, a travel distance common for rural-dwelling survivors, compared with < 30 minutes.⁵¹ Studies using nationally representative data found both higher and lower risk of financial hardship when comparing rural and urban cancer survivors.^52,53 As more attention is turned toward geography-related health care disparities, researchers must remember the vast heterogeneity of both urban and rural areas with respect to sociodemographics, economic structures, and availability of health care resources.^54,55 Because system-level interventions such as patient navigation have been shown to reduce survivor Medicare spending and hospitalizations,⁵⁶ future research should incorporate multiple dimensions of geography to identify the optimal intervention to decrease medication nonadherence, drive down costs, and decrease preventable health care utilization.

Our results should be considered within several limitations. Patient cost responsibility may overestimate true out-of-pocket spending, since it reflects potential costs before application of supplemental insurance coverage held by a majority of beneficiaries.⁵⁷ However, these estimates may also underestimate the true out-of-pocket burden for cancer survivors, as other health care–related costs such as transportation to clinical appointments, time spent waiting and receiving care in clinical settings, and lost household productivity are common.⁵⁸ Although patient cost responsibility captured financial outlay, no information was available on resources to pay for care to assess financial hardship. Because our sample primarily comprised long-term survivors, nonadherence estimates may not be generalizable to a recently diagnosed sample. We were unable to capture the specific medications to which survivors reported nonadherence or delineate between cancer-specific and other health care costs. Unmeasured confounding may exist because of the inability to capture survivor frailty, functional status, comorbidity severity, or quality of life. Our study included Medicare fee-for-service beneficiaries only, and the results may not be generalizable to Medicare Advantage enrollees who often face different cost-sharing structures. Our study included primarily prostate or breast cancer survivors, which may also limit generalizability. Patient residence was assessed at cancer diagnosis, rather at CAHPS survey completion. Finally, cost-related medication nonadherence may be a persistent issue, yet was assessed at one time point in our study via self-report on a single-item question, which may introduce recall or reporting bias.

In conclusion, this study found that 13% of older, long-term cancer survivors reported cost-related medication nonadherence in the previous 6 months. Cost-related medication nonadherence was preceded by increases in patient cost responsibility, ED visits, and hospitalizations, and followed by higher patient cost responsibility, Medicare spending, and ED visits when compared with survivors not reporting nonadherence. No differences in spending or utilization were seen comparing rural- and urban-dwelling survivors reporting cost-related medication nonadherence and those who did not. Interventions, including lowering prescription drug costs, comorbidity management, and individualized assessment of survivor needs, could aid in breaking the bidirectional relationship between patient cost responsibility and cost-related medication nonadherence, ultimately improving the physical and financial health of older cancer survivors.

APPENDIX

TABLE A1.

Model-Estimated Risk of Cost-Related Medication Nonadherence by Prior Annual Health Care Expenditures and Utilization, and Expenditure Ratio of Annual Health Care Expenditures and Risk of Health Care Utilization After Report of Cost-Related Medication Nonadherence (N = 6,591)

TABLE A2.

Model-Estimated Risk of Cost-Related Medication Nonadherence by Prior Annual Health Care Expenditures and Utilization, and Expenditure Ratio of Annual Health Care Expenditures and Risk of Health Care Utilization After Report of Cost-Related Medication Nonadherence by Residence (N = 6,591)

FIG A1.

Exclusion cascade. HMO, health maintenance organization; SEER-CAHPS, Surveillance, Epidemiology, and End Results Program, Medicare claims, and the Consumer Assessment of Healthcare Providers and Systems.

FIG A2.

(A) Overall patient cost responsibility by residence before report of cost-related medication nonadherence, estimated using generalized linear models with a log link and gamma distribution (N = 6,591). (B) Overall patient cost responsibility by residence after report of cost-related medication nonadherence, estimated using negative binomial models (N = 6,591).

AUTHOR CONTRIBUTIONS

Conception and design: Courtney P. Williams, Amy Davidoff, Michelle Mollica, Janet S. de Moor

Collection and assembly of data: Courtney P. Williams, Michelle Mollica

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Cost-Related Medication Nonadherence and Patient Cost Responsibility for Rural and Urban Cancer Survivors

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).