Association of patient-reported financial barriers with healthcare utilization among Medicare beneficiaries with a history of cancer

Abstract

Purpose:

We examined characteristics associated with financial barriers to healthcare and the association of financial barriers with adverse healthcare events among US adult cancer survivors enrolled in Medicare.

Methods:

We used nationally representative Medicare Current Beneficiary Survey data (2011–2013, 2015–2017) to identify adults with a history of non-skin cancer. We defined financial barriers as cost-related trouble accessing and/or delayed care in the prior year. Using propensity-weighted multivariable logistic regression, we examined associations between financial barriers and adverse healthcare events (any ED visits, any inpatient hospitalizations).

Results:

Overall, 11.0% of adult Medicare beneficiaries with a history of cancer reported financial barriers in the prior year, with higher burden among beneficiaries <65 years of age vs. ≥65 (32.5% vs. 8.2%, p<0.0001) and with annual income <$25,000 vs. ≥$25,000 (18.1% vs. 6.9%, p<0.0001). In bivariate models, financial barriers were associated with a 7.8 percentage point (95% CI: 1.5 – 14.0) increase in the probability of ED visits. In propensity-weighted models, this association was not statistically significant. The association between financial barriers and hospitalizations was not significant in the overall population; however, financial barriers were associated with a decreased probability of hospitalization among Black/African American beneficiaries.

Conclusions:

Despite Medicare coverage, beneficiaries with a history of cancer are at risk for experiencing financial barriers to healthcare. In the overall population, financial barriers were not associated with ED visits or hospitalizations.

Implications for Cancer Survivors:

Policies limiting Medicare patient out-of-pocket spending and care models addressing health-related social needs are needed to reduce financial barriers experienced.

Introduction

As cancer care delivery costs, and thus patient out-of-pocket costs, rise [1, 2], patients and their families are increasingly at risk of financial hardship. Over 8 million cancer survivors, defined from the time of diagnosis [3], in the US experience cancer-related financial hardship and distress, termed “financial toxicity” [4–6]. Several studies suggest an association between extreme financial toxicity and heightened mortality [7–10], though the mechanisms underlying such a complex association are not well understood.

Financial barriers to care, resulting in cost-related delayed and/or forgone care, is one hypothesized mechanism between financial hardship and adverse clinical outcomes [11]. Financial barriers to care may stem from perceived and actual costs of medical care, non-medical costs associated with getting to care, and insurance-related concerns. Additionally, underlying financial vulnerability or precarity may exacerbate the potential for cancer-related financial hardship to lead to nonadherence to recommended treatment, follow-up care, and medication [12–16]. Evidence from other patient populations, including adults with cardiovascular disease and asthma, suggests that financial barriers to care can lead to adverse and high-cost healthcare utilization, such as emergency department (ED) visits and inpatient hospitalizations [17–19].

Exploration of the association between patient-reported financial barriers to care and healthcare utilization has been limited in longer term cancer survivors, who must cope with the accumulated financial burden of treatment and routine surveillance over time [20]. Additionally, cancer survivors are at heightened risk of chronic conditions (e.g., heart disease, diabetes) associated with increased out-of-pocket medical costs and lost productivity [21]. The objective of this study is to examine the association of patient-reported financial barriers to care (i.e., trouble accessing care due to cost-related reasons, delays in care due to cost) with ED visits and inpatient hospitalizations among adult cancer survivors insured by Medicare, which provides coverage to adults ages 65 years and older, as well as those with qualifying disabilities. Though Medicare is an important social protection for those who qualify, high out-of-pocket costs and non-medical costs associated with serious illness may still impose substantial financial hardship [22]. Additionally, given that 67% of cancer survivors are age 65 or older, and thus eligible for Medicare, Medicare beneficiaries are an important subsample of cancer survivors to study [23].

This study builds on a recent cross-sectional analysis documenting an association between financial hardship and ED visits among cancer survivors in the National Health Interview Survey by isolating the effects of financial barriers to care using longitudinal, patient-reported data and applying inverse probability of treatment weighting in an attempt to quantify causal effect [24, 25]. Understanding the clinical and financial consequences of financial barriers to care among Medicare beneficiaries will inform patient-centered approaches to reducing cost-related cancer outcome disparities.

Methods

Data and Sample

We used 2011 – 2013 and 2015 – 2017 Medicare Current Beneficiary Survey (MCBS) data. MCBS data from 2014 was never released. MCBS is a nationally representative, longitudinal survey of Medicare beneficiaries administered by the Centers for Medicare and Medicaid Services (CMS). MCBS survey weights account for both differential sampling and non-response (see Supplemental Table S1 for complete survey response rates). We used MCBS’s rotating, multistage survey design to identify eight panels of participants who completed the Access to Care survey and had complete Cost and Use data in the following year (i.e., survived for the entire following year and remained in the MCBS study) (Supplemental Figure S1) [26]. The analytic sample was comprised of adults living in community settings who self-reported a prior cancer diagnosis, excluding skin cancer (due to the high prevalence of easily curable non-melanoma skin cancers).

We excluded participants who were missing data for both measures of financial barriers to care (N=22, 0.5%) or for covariates with less than 30 missing responses (N=33, 0.7%). Excluded participants did not differ from the final analytic sample across variables with complete data, with the exception of excluded participants being more likely to have household income <$25,000 and be of Hispanic or Latino ethnicity (Supplemental Table S2). The institutional review board approved this study (UNC-CH IRB#22–0467).

Measurement of Healthcare Utilization

Healthcare utilization data were captured in the MCBS Cost and Use files, which are compiled by reconciling Medicare fee-for-service (FFS) claims with self-reported healthcare events, thus providing cost and utilization data for beneficiaries enrolled in both Medicare FFS and Medicare Advantage, with data for Medicare Advantage beneficiaries being based on self-report alone [27]. We defined our primary outcomes as binary indicators of any ED visits, and any inpatient hospitalizations, in the year following survey completion. ED visits included visits that did and did not result in an inpatient admission.

Measurement of Financial Barriers to Healthcare

Patient-reported measures of financial barriers came from the Access to Care questionnaire. Participants answering affirmatively to at least one of two survey questions assessing trouble accessing care and delayed care, respectively, were considered to be experiencing financial barriers to care. Trouble accessing care was measured through the question, “In the last year, have you had any trouble getting healthcare that you wanted or needed?” Participants answering affirmatively were then asked why this was and given a range of reasons, to which they could respond yes or no. Participants responding yes to one or more cost-related reasons (i.e., do not have money, cost is too high, services/supplies not covered, not eligible for public coverage, needed transportation to doctor/hospital) were considered to be experiencing financial barriers to care. The second question asked, “In the last year, have you ever delayed seeking medical care because of cost concerns?” Participants answering affirmatively were considered to be experiencing financial barriers to care.

Measurement of Control Variables

We identified covariates hypothesized to be associated with both financial barriers and healthcare utilization based on a conceptual model of financial barriers to healthcare [28] and an integrated framework for understanding social determinants to advance cancer health equity [29]. Covariates associated with healthcare utilization were prioritized for inclusion in the propensity model. Covariates included those related to the ability to pay for healthcare (education; FFS vs. Medicare Advantage), need for healthcare (age; sex; count of self-reported comorbidities associated with financial barriers; self-reported difficulty performing one or more activities of daily living; cancer site; self-rated general health), baseline healthcare seeking behavior (self-reported usual source of care; binary indicator of whether participants reported usually going to the doctor as soon as they feel bad), living environment (marital status, metropolitan status based on the Office of Management and Budget statistical area designations), and social identity (race, ethnicity). All covariates were self-reported by participants, with the exception of age, sex, FFS vs. Medicare Advantage, and metropolitan status, which were taken from CMS administrative records. Though income was not included in the primary propensity model due to its association with healthcare utilization hypothesized to be mediated through financial barriers, we ran sensitivity analyses with income included, and results were not substantially different.

Analytic Strategy

We first assessed bivariate associations between each categorical covariate and financial barriers using chi-squared tests, accounting for the MCBS complex survey design. For each of the two primary outcomes (binary indicators of any ED visits and any inpatient hospitalizations in the year following survey completion), we then developed multivariable logistic regression models with inverse probability of treatment weights (IPTWs) to account for observable, baseline differences between cancer survivors reporting and not reporting financial barriers to care that may independently affect healthcare utilization patterns [30]. Standardized differences <10% in the weighted covariates suggested appropriate balance (Supplemental Table S3) [31].

We multiplied IPTWs by MCBS survey sampling weights to generate the final logistic regression weights. We also added key covariates hypothesized to influence necessary utilization (i.e., age, sex, cancer site, and comorbidity count) to the final outcome models for a doubly robust estimation of the independent effect of financial barriers to care on utilization. Variance inflation factors less than five for all covariates identified a priori suggested the absence of multicollinearity. Due to the potential for cancer site and sex causing model convergence issues, we ran sensitivity analyses without sex, and the results were not affected. Using weighted regression results, we calculated the average marginal effect of financial barriers on each utilization outcome, which can be interpreted as the average difference in the predicted probability of each outcome, holding all covariates constant, across all observations in the sample. We calculated standard errors and 95% confidence intervals using the Delta method, accounting for the MCBS complex survey design through Fay’s method of balanced repeated replication [32, 33].

We also conducted several exploratory analyses to better understand the relationship between financial barriers and healthcare utilization. In order to reduce the potential for confounding due to lack of cancer diagnostic and treatment data, we conducted multivariable analyses stratified by cancer type (small sample sizes precluded the use of inverse probability of treatment weighting). Additionally, using an inverse probability of treatment weighted Hurdle model combining a logistic regression with a zero-inflated Poisson model, we modeled the association of financial barriers with total count of ED visits in the following year. We also assessed the association of financial barriers with the likelihood of having a 30-day hospital readmission using an inverse probability of treatment weighted logistic regression. Finally, we assessed whether our primary effect measures were modified by either age at time of the survey (<65, 65 – 74, 75+) or racial identity (white vs. Black or African American). To do so, we re-estimated propensity weights within each subgroup and used these weights (multiplied by the survey weights) in outcome models which included an interaction term between the effect modifier of interest and financial barriers.

Results

Sample characteristics

The final analytic sample included 4,465 adult Medicare beneficiaries with a history of cancer, weighted to represent 23,780,784 Medicare beneficiaries. The majority of beneficiaries (88.6%) were age 65 or older, and thus age-eligible for Medicare. Beneficiaries were also majority female (56.4%), white (86.6%), not Hispanic or Latino (93.0%), married (55.7%), and from a metropolitan area (77.0%). Only 32.3% of beneficiaries reported an annual income of $50,000 or more, and 32.7% of beneficiaries reported having at least some college education. Just over 70% of beneficiaries were enrolled in a FFS (vs. Medicare Advantage) plan; 45.6% had Part D coverage; and 13.7% and 16.6% were eligible for Medicaid and a Part D low-income subsidy, respectively. The most commonly reported cancer sites were breast (21.5%), prostate (18.6%), and gynecologic (10.6%), with an additional 10.7% reporting history of cancer in multiple sites. In addition to cancer, the majority (88.8%) of beneficiaries reported one or more other chronic conditions (Table 1).

Table 1.

Weighted sample socio-demographic characteristics by self-reported financial barriers (N=4465, weighted to represent 23,780,784 Medicare beneficiaries with a history of cancer)

Sample Characteristics	Overall (N = 4,465)	No Financial Barriers (N = 4029)	Financial Barriersa (N = 436)	p-valueb
	Weighted column %
Age				<0.0001
<65	11.45%	8.68%	33.81%
65 – 74	43.21%	43.75%	38.80%
75 – 84	32.8%	34.08%	22.41%
85+	12.55%	13.48%	12.55%
Sex				0.0001
Male	43.63%	44.84%	33.87%
Female	56.37%	55.16%	66.13%
Race				0.0034
White	86.55%	87.28%	80.68%
Black or AA	7.859%	7.53%	10.56%
Otherc	5.586%	5.20%	8.76%
Ethnicity				0.2101
Not Hispanic/Latino	92.67%	92.91%	90.72%
Hispanic/Latino	7.334%	7.09%	9.28%
Annual Household Income				<0.0001
<$25,000	36.51%	33.60%	60.1%
$25,000 - $49,999	29.03%	29.92%	21.89%
$50,000+	32.32%	34.31%	16.19%
Not reported	2.139%	2.179%	1.82%
Education				0.0057
Less than HS	16.35%	15.93%	19.78%
HS graduation	50.99%	50.52%	54.87%
Some college +	32.65%	33.55%	25.36%
Marital status				0.0013
Married	55.7%	56.86%	46.31%
Not married	44.3%	43.14%	53.69%
Rurality				0.2482
Non-metropolitan	22.8%	22.48%	25.41%
Metropolitan	77.2%	77.52%	74.59%
Medicare plan type d				0.1146
Fee-for-service	70.21%	70.65%	66.62%
Medicare Advantage	29.79%	29.35%	33.38%
Low-income Subsidy Eligibility				<0.0001
Not eligible	83.42%	85.4%	67.42%
Full subsidy	15.75%	14.00%	29.89%
Partial subsidy	0.8266%	0.60%	2.69%
Part D Coverage				0.0088
No	54.38%	55.25%	47.37%
Yes	45.62%	44.75%	52.63%
Medicaid Dual Eligibility				<0.0001
Not eligible	86.35%	87.66%	75.73%
Dual eligible	13.65%	12.34%	24.27%
Employer-sponsored Insurance Coverage				0.0001
No	70.26%	68.96%	80.78%
Yes	29.74%	31.04%	19.22%
# of Comorbidities e				<0.0001
Cancer only	11.21%	11.89%	5.70%
1 or 2	48.51%	50.12%	35.49%
3 or 4	28.6%	27.61%	36.60%
5+	11.68%	10.38%	22.21%
Cancer Type				0.0004
Breast	21.52%	21.53%	21.45%
Prostate	18.61%	19.41%	12.16%
Gynecologic	10.59%	9.73%	17.54%
Colorectal	8.737%	9.09%	5.89%
Bladder	3.4%	3.56%	2.09%
Lung	3.214%	3.12%	3.99%
Multiple	10.73%	22.93%	25.44%
Other	23.2%	10.64%	11.43%
Activities of Daily Living (ADL) Limitations				<0.0001
None	64.11%	66.93%	41.29%
1 or more ADLs	35.89%	33.07%	58.71%
Self-reported health				<0.0001
Excellent or good	70.67%	73.48%	47.92%
Fair or poor	28.08%	25.22%	51.27%
Unknown	1.249%	1.30%	0.81%
Usual source of care				0.0738
Yes	96.26%	96.52%	94.24%
No	3.74%	3.48%	5.76%
Do you seek a doctor’s care as soon as you feel badly?				<0.0001
Yes	36.70%	38.20%	24.51%
No	62.58%	61.02%	75.19%
Unknown	0.7264%	0.78%	0.30%
Survey Year f				0.0130
2011	32.61%	33.14%	28.32%
2012	13.87%	13.26%	18.83%
2015	35.64%	36.10%	31.92%
2016	17.87%	17.49%	20.93%

Abbreviations: High school (HS); Activities of daily living (ADL); African American (AA)

^a.Financial barriers were self-reported by participants and defined as trouble accessing needed or wanted healthcare for cost-related reasons (i.e., do not have money, cost is too high, services/supplies not covered, not eligible for public coverage, needed transportation to doctor/hospital) and/or delays in seeking medical care because of cost concerns in the last year.

^b.P-values reported are from the results of Chi-squared tests.

^c.Includes Asian, Native Hawaiian or Pacific Islander, American Indian or Alaska Native, Other race, More than one race

^d.Based on coverage beneficiary had for the majority of the year.

^e.Includes self-reported conditions associated with financial barriers in bivariate analysis (p<0.20): hypertension/high blood pressure, myocardial infarction/heart attack, heart failure, other heart condition (e.g., valve, rhythm), stroke/brain hemorrhage, diabetes, rheumatoid arthritis, intellectual disability, depression, other mental disorder, emphysema/asthma/COPD, complete/partial paralysis.

^f.Year of baseline survey (healthcare utilization measured in the year following).

Self-reported financial barriers to care

Overall, 11.0% (95% CI: 10.0% - 12.1%) of adult Medicare beneficiaries with a history of cancer reported financial barriers in the prior year, with higher burden among beneficiaries <65 years of age vs. ≥65 (32.5% vs. 8.2%, p<0.0001), with annual income <$25,000 vs. ≥$25,000 (18.1% vs. 6.9%, p<0.0001), and identifying as persons of color vs. white (15.8% vs. 10.3%, p=0.001). Beneficiaries eligible for a Part D low-income subsidy were more likely to report financial barriers compared to those who were not eligible, with those with eligibility for only a partial (vs. full) subsidy reporting the highest prevalence of financial barriers (35.8% vs. 20.9%, p<0.0001). Across cancer sites, financial barriers were most commonly reported among beneficiaries with gynecologic cancer (18.2%) and lung cancer (13.7%). Financial barriers were also highly associated with number of comorbidities (p<0.0001), limitations in activities of daily living (p<0.0001), and self-reported fair or poor health (p<0.0001) (Table 1, Supplemental Figure S2).

Association of financial barriers with ED utilization

The prevalence of ED visits among Medicare beneficiaries with a history of cancer was 26.1% (95% CI: 24.6 – 27.6). Table 2 displays the average marginal effects of financial barriers on the probability of having an ED visit in the following year. In bivariate estimation, before inverse probability of treatment weighting, Medicare beneficiaries experiencing financial barriers to care had an expected probability of having an ED visit that was 7.78 percentage points (95% CI: 1.53 – 14.02) higher than beneficiaries who did not report financial barriers. In the propensity weighted multivariable model, this association was insignificant (0.71; 95% CI: −4.99 – 6.41). Further, the association was not modified by age (<65 vs. 65 – 74 vs. 75+) or race (white vs. Black or African American) (Figure 1, Supplemental Table S4). The multivariable model controlling for all covariates, instead of using the IPTWs, shows that the number of chronic conditions was the strongest predictor of ED utilization, with having 5+ conditions (vs. cancer only) associated with a 17.9 (95% CI: 10.7 – 25.0) percentage point increase in the probability of having one or more ED visits in the following year (Table 2).

Table 2.

Association of financial barriers to healthcare with emergency department visits in the following year (N=4465)

	Average Marginal Effects (95% Confidence Interval)
	Without IPTWs		With IPTWs
	Bivariate	Multivariate a	IPTWs only	Doubly Robust
Financial Barriers (ref: No)
Yes	0.078** (0.02 – 0.14)	0.014 (−0.05 – 0.08)	0.005 (−0.05 – 0.06)	0.007 (−0.05 – 0.06)
Age (ref: <65)
65 – 74		−0.078** (−0.14 – −0.01)		−0.142*** (−0.21 – −0.08)
75 – 84		−0.022 (−0.08 – 0.03)		−0.076** (−0.14 – −0.02)
85+		0.050 (−0.02 – 0.12)		0.000 (−0.07 – 0.07)
Sex (ref: Male)
Female		−0.015 (−0.06 – 0.03)		−0.036 (−0.08 – 0.01)
Cancer Type (ref: Prostate)
Breast		0.050 (−0.02 – 0.12)		0.082** (0.01 – 0.15)
Gynecologic		0.060 (−0.02 – 0.14)		0.111*** (0.03 – 0.19)
Colorectal		0.072* (−0.01 – 0.15)		0.096** (0.02 – 0.18)
Bladder		0.048 (−0.05 – 0.14)		0.217*** (0.11 – 0.32)
Lung		0.132*** (0.04 – 0.23)		0.207*** (0.10 – 0.31)
Multiple		0.018 (−0.03 – 0.07)		0.057** (0.01 – 0.11)
Other		0.037 (−0.03 – 0.10)		0.117*** (0.06 – 0.18)
# of Comorbidities (ref: Cancer only)
1 or 2		0.039 (−0.01 – 0.09)		0.032 (−0.02 – 0.08)
3 or 4		0.073*** (0.02 – 0.13)		0.112*** (0.06 – 0.17)
5+		0.179*** (0.11 – 0.25)		0.220*** (0.15 – 0.29)
Race (ref: White)
Black or AA		0.056** (0.00 – 0.11)
Other1		−0.066* (−0.13 – 0.00)
Ethnicity (ref: Not Hispanic of Latino)
Hispanic/Latino		0.034 (−0.03 – 0.09)
Education (ref: Less than HS)
HS graduation		−0.018 (−0.06 – 0.03)
Some college +		−0.024 (−0.07 – 0.02)
Rurality (ref: Metropolitan)
Non-metropolitan		0.042*** (0.01 – 0.07)
Activities of Daily Living (ADL) Limitations (ref: None)
1 or more ADLs		0.051** (0.01 – 0.09)
Self-reported health (ref: Excellent or good)
Fair or poor		0.069*** (0.03 – 0.11)
Unknown		−0.020 (−0.21 – 0.17)
Medicare plan type (ref: Fee-for-service)
Medicare Advantage		−0.036** (−0.07 – −0.01)

^***p<0.01,

^**p<0.05,

^*p<0.1

Abbreviations: Inverse Probability of Treatment Weights (IPTWs), Activities of Daily Living (ADLs)

^a.Model also controls for marital status, usual source of care, and whether beneficiary seeks a doctor’s care when feeling badly.

Figure 1.

Effect modification of the association between financial barriers and healthcare utilization by age group and racial identity

In exploratory analysis, we also conducted multivariable analyses stratified by cancer type, finding that financial barriers were associated with a 13.4 percentage point (95% CI: −2.28 – 29.0) increase in the probability of having an ED visit in the following year among individuals reporting a cancer diagnosis in multiple sites. Associations within all other cancer types were insignificant (Supplemental Table S5). Lastly, we modeled the association between financial barriers and the total number of ED visits in the following year. Both the bivariate and inverse probability of treatment weighted associations were insignificant (Supplemental Table S6).

Association of financial barriers with inpatient hospitalization

The prevalence of inpatient hospitalizations among Medicare beneficiaries with a history of cancer was 18.7% (95% CI: 17.4 – 20.1). Table 3 displays the average marginal effects of financial barriers on the probability of being hospitalized in the following year. In both bivariate estimation and the propensity weighted multivariable model, the association between financial barriers and inpatient hospitalization was insignificant (−0.99; 95% CI: −4.9 – 2.9 in the propensity weighted model). As with ED visits, the multivariable model controlling for all covariates, instead of using the IPTWs, shows that the number of chronic conditions was the strongest predictor of hospitalization. Having 5+ conditions (vs. cancer only) was associated with an 8.5 (95% CI: 2.7 – 14.4) percentage point increase in the probability of having one or more inpatient hospitalizations in the following year (Table 3).

Table 3.

Association of financial barriers to healthcare with inpatient hospitalizations in the following year (N=4465)

	Average Marginal Effects (95% Confidence Interval)
	Without IPTWs		With IPTWs
	Bivariate	Multivariate a	IPTWs only	Doubly Robust
Financial Barriers (ref: No)
Yes	0.022 (−0.03 – 0.07)	−0.011 (−0.05 – 0.03)	−0.012 (−0.05 – 0.03)	−0.010 (−0.05 – 0.03)
Age (ref: <65)
65 – 74		−0.034 (−0.08 – 0.01)		−0.078*** (−0.13 - -0.03)
75 – 84		0.020 (−0.03 – 0.07)		−0.038 (−0.08 – 0.01)
85+		0.072** (0.02 – 0.13)		0.017 (−0.04 – 0.07)
Sex (ref: Male)
Female		−0.046** (−0.08 - -0.01)		−0.038** (−0.07 – 0.00)
Cancer Type (ref: Prostate)
Breast		0.026 (−0.03 – 0.08)		0.04 (−0.01 – 0.09)
Gynecologic		0.045 (−0.02 – 0.11)		0.035 (−0.03 – 0.10)
Colorectal		0.033 (−0.02 – 0.09)		0.133*** (0.07 – 0.20)
Bladder		0.085** (0.00 – 0.17)		0.295*** (0.19 – 0.40)
Lung		0.094** (0.02 – 0.17)		0.073** (0.00 – 0.14)
Multiple		0.023 (−0.02 – 0.06)		−0.01 (−0.04 – 0.03)
Other		0.070** (0.01 – 0.13)		0.069** (0.01 – 0.12)
# of Comorbidities (ref: Cancer only)
1 or 2		0.014 (−0.03 – 0.06)		0.034* (0.00 – 0.07)
3 or 4		0.072*** (0.03 – 0.12)		0.116*** (0.08 – 0.16)
5+		0.085*** (0.03 – 0.14)		0.163*** (0.11 – 0.22)
Race (ref: White)
Black or AA		−0.022 (−0.06 – 0.02)
Other1		−0.076*** (−0.13 - −0.03)
Ethnicity (ref: Not Hispanic/Latino)
Hispanic/Latino		0.006 (−0.05 – 0.06)
Education (ref: Less than HS)
HS graduation		−0.035* (−0.07 – 0.00)
Some college +		−0.047** (−0.09 – 0.00)
Rurality (ref: Metropolitan)
Non-metropolitan		0.006 (−0.03 – 0.03)
Activities of Daily Living (ADL) Limitations (ref: None)
1 or more ADLs		0.067*** (0.03 – 0.09)
Self-reported health (ref: Excellent or good)
Fair or poor		0.070*** (0.03 – 0.11)
Unknown		−0.079 (−0.18 – 0.03)
Medicare plan type (ref: Fee-for-service)
Medicare Advantage		−0.035** (−0.06 - −0.01)

^***p<0.01,

^**p<0.05,

^*p<0.1

Abbreviations: Inverse Probability of Treatment Weights (IPTWs)

^a.Model also controls for marital status, usual source of care, and whether beneficiary seeks a doctor’s care when feeling badly.

Though the association between financial barriers and inpatient hospitalizations was not modified by age, we did identify a significant difference in the association among white vs. Black or African American beneficiaries (Figure 1, Supplemental Table S4). Financial barriers were associated with a 20.1 percentage point (95% CI: 10 – 31) decrease in the probability of having an inpatient hospitalization in the following year among Black or African American beneficiaries, whereas financial barriers were not associated with inpatient hospitalizations among white beneficiaries.

In exploratory analysis, we also conducted multivariable analyses stratified by cancer type, finding that financial barriers were associated with an 8.2 percentage point (95% CI: −0.02 – 16.4) decrease in the probability of inpatient hospitalization among patients with gynecologic and lung cancers, the cancer types with the largest reported burden of financial barriers. Associations within all other cancer types were insignificant (Supplemental Table S7). Lastly, we modeled the association between financial barriers and having an inpatient admission within 30 days of discharge. Only 1.16% of the sample had 30-day readmissions in the year following survey administration. In both bivariate estimation and the propensity weighted multivariable model, the association between financial barriers and 30-day readmission was insignificant (Supplemental Table S8).

Discussion

Over one in ten Medicare beneficiaries with a history of cancer reported financial barriers to medical care, with a higher prevalence among beneficiaries <65 and those with more chronic conditions. Financial barriers were not associated with healthcare utilization (ED visits and inpatient hospitalizations, specifically) after inverse probability of treatment weighting in the overall sample; however, financial barriers may still impose other negative outcomes, such as patient distress or downstream clinical outcomes not included in our analysis. Additionally, in sub-group analyses, we found a statistically significant association between financial barriers and reduced hospitalizations among Black or African American beneficiaries and among beneficiaries with gynecologic and lung cancers. We also found that, among beneficiaries reporting a cancer diagnosis in multiple sites, financial barriers were associated with a statistically significant increase in the probability of having an ED visit in the following year. Differences by racial identity and cancer type suggest the need for future research investigating sub-group differences in the consequences of patient-reported financial barriers.

The findings from our study stand in contrast to prior research demonstrating a significant association between financial hardship and ED visits among cancer survivors in an analysis of 2013 – 2017 National Health Interview Survey data [24]. Several differences in the methodological approach, measurement, and population could explain the differences in findings. First, Zheng and colleagues’ analysis used cross-sectional data in which measurements of self-reported financial hardship and ED visits were assessed over the same 12-month period. Second, the prior study used an expansive definition of financial hardship (i.e., intensity of medical and non-medical financial hardship, including material, psychological and behavioral domains) [34], whereas we focused on financial barriers to medical care (i.e., trouble accessing care and/or delayed care due to cost) to focus on one hypothesized mechanism leading to adverse health outcomes. Lastly, and perhaps most importantly, the National Health Interview Survey includes patients of all ages and insurance types, including individuals who are uninsured. Though Zheng and colleagues stratified their analysis by age (<65 vs. ≥ 65), and the association with ED visits was present in both groups (with the majority of the age ≥ 65 likely insured by Medicare), it is likely that the broader sample contributed to differences in findings. This points to the need for further research exploring the association between cancer-related financial hardship and healthcare utilization among patients with a range of insurance types to inform interventional work.

Our study adds to a growing body of literature documenting the disparate burden of cost-related care interference, with patients of color [35, 36], low-income patients [37], and uninsured or publicly insured patients [37, 38] at highest risk. Our finding that financial barriers were associated with a lower likelihood of having any hospitalizations in the following year among Black or African American Medicare beneficiaries is directionally in contrast to a prior study documenting a positive association between cost-related medication underuse and hospitalization among adults with cardiovascular disease [18]. It is possible that the inverse association documented in our analysis stems from financial barriers preventing beneficiaries from seeking out needed medical care or elective procedures, as we did not isolate unnecessary or avoidable hospitalizations. Further research is needed to investigate the differential association between financial barriers and healthcare utilization among white vs. Black or African American beneficiaries, with particular attention paid to the influence of structural inequities on institutional environments (e.g., health system policies), living environments (e.g., accessibility of care), and interpersonal interactions (e.g., between patients and providers) [29].

The prevalence of financial barriers reported in this study suggests the need for policy and programmatic intervention to reduce both the out-of-pocket costs of medical care and the costs associated with accessing that care. FFS Medicare beneficiaries without supplemental coverage must pay 20% coinsurance for Part B outpatient services, and out-of-pocket costs vary for Part D prescription drug plans. The Inflation Reduction Act passed by the US Congress in August 2022 is a promising step toward reducing out-of-pocket spending for prescription drugs. In 2024, coinsurance above the Part D catastrophic threshold will be eliminated, and in the following year, out-of-pocket drug spending will be capped at $2,000 annually [39]. Further expansion of Part D low-income subsidy benefits will make Part D coverage more accessible and affordable for beneficiaries with incomes up to 150% of the federal poverty level. Notably, we found that the prevalence of financial barriers was highest for beneficiaries with a partial low-income subsidy. As of 2024, the partial subsidy will be replaced by the full low-income subsidy, addressing this potential coverage gap. Despite these landmark changes to Medicare Part D, additional protections are needed to address out-of-pocket spending in Medicare Parts A and B.

In addition to out-of-pocket spending on medical care, Medicare beneficiaries may also face cost-related challenges accessing care, such as paying for transportation, parking, and lodging in the case of care that is distantly-located. Further, other health-related social needs stemming from financial hardship (i.e., housing instability, food insecurity) have the potential to interfere with an individual’s ability to manage and seek care for health concerns [40], particularly among cancer survivors who may be managing ongoing treatment or late and long-term effects [41]. CMS is increasingly incentivizing health systems and healthcare practitioners to screen for and address health-related social needs through programs like the Accountable Health Communities Model [42] and the forthcoming Enhancing Oncology Model [43, 44]. However, building effective models of clinical and community collaboration to address whole-person health is challenging, particularly given scarce community resources and limited social safety net programs in the United States.

Our findings must be considered in light of several limitations. Financial barriers captured in the MCBS survey may stem from a variety of sources (e.g., underlying financial hardship, other chronic conditions) and may not be related to the cost of cancer itself, particularly for patients who were no longer undergoing active treatment at the time of the survey, which is likely the majority of our sample (based on data collected in 2011, only about 13% of the 2009 and 2010 panels included in our analysis were diagnosed within the prior year). MCBS does not include data on cancer stage, treatment history, or time since diagnosis. These data are important omitted variables; however, given that the sample is more representative of long-term cancer survivors, comorbidities, self-reported health status, and activities of daily living should capture important late and long-term effects of treatment that are more relevant to the healthcare utilization captured in this study, which may or may not be directly related to cancer. Of note, by including the count of comorbidities as an indicator of overall comorbidity burden we were not able to shed light on the mechanisms through which comorbidities influence healthcare utilization outcomes, though this was not the main goal of our analysis. Linkage of nationally representative survey data to the Surveillance, Epidemiology, and End Results (SEER) program data could provide opportunities to investigate patterns of cost-related care interference in relation to cancer care trajectories. Given that financial hardship is exacerbated during active treatment, it is possible that the relationship between financial barriers and healthcare utilization may be stronger among patients undergoing active treatment period. Furthermore, we did not attempt to isolate unnecessary or avoidable ED visits and hospitalizations, though we did include doubly robust controls for medical complexities contributing to necessary use. It is possible that isolating avoidable use may lead to a stronger association with financial barriers.

Additionally, though a notable strength of this analysis is the inclusion of both Medicare Advantage and fee-for-service beneficiaries, the survey-reported utilization data for Medicare Advantage beneficiaries is subject to self-report bias and missingness. This may explain our multivariable model findings that Medicare Advantage beneficiaries had a significantly lower probability of having both ED visits and inpatient hospitalizations in the following year compared to fee-for-service beneficiaries. As of 2019, MCBS has started to use Medicare Advantage encounter data in the utilization reconciliation process, meaning that future analyses with more recent MCBS data will not be subject to this same limitation.

Despite Medicare insurance coverage, beneficiaries with a history of cancer are at risk for experiencing financial barriers to healthcare. Further research is needed to understand the consequences of patient-reported financial barriers to care on healthcare utilization, and disparate health outcomes, with particular attention to inequities in how financial barriers are experienced by race and other social identities. In the meantime, given the prevalence of financial barriers among Medicare beneficiaries with a history of cancer, policies limiting Medicare patient out-of-pocket spending and care models addressing health-related social needs are needed to reduce the financial barriers experienced.