Comparison of the Financial Burden of Survivors of Head and Neck Cancer With Other Cancer Survivors

Key Points

Questions  How do the financial demographics of patients with head and neck cancer differ from patients with other cancers, and how are these differences associated with their medical expenses?

Findings  In this review of data from 16 771 patients with cancer in the Medical Expenditure Panel Survey, patients with head and neck cancer were more often members of a racial/ethnic minority group, poor, and less educated and had lower health status than patients with other cancers. Patients with head and neck cancer also experienced higher total medical expenses and higher out-of-pocket costs relative to their income, with the highest relative out-of-pocket costs occurring among the poor and publicly insured.

Meaning  Patients with head and neck cancer are uniquely disenfranchised and have higher medical expenses, which cumulatively increases their risk for financial burdens incurred by medical treatments.

Importance  Head and neck cancer (HNC) is more common among socioeconomically disenfranchised individuals, making financial burden particularly relevant.

Objective  To assess the financial burdens of HNC compared with other cancers.

Design, Setting, and Participants  In this retrospective review of nationally representative, publicly available survey, data from the Medical Expenditure Panel Survey were extracted from January 1, 1998, to December 31, 2015. A total of 444 867 adults were surveyed, which extrapolates to a population of 221 503 108 based on the weighted survey design. Data analysis was performed from April 18, 2018, to August 20, 2018.

Exposures  Of 16 771 patients with cancer surveyed (weighted count of 10 083 586 patients), 489 reported HNC (weighted count of 261 631).

Main Outcomes and Measures  Patients with HNC were compared with patients with other cancers on demographics, income, employment, and health. Within the HNC group, risk factors for total medical expenses and relative out-of-pocket expenses were assessed with regression modeling. Complex sampling methods were accounted for with weighting using balanced repeated replication.

Results  A total of 16 771 patients (mean [SD] age, 62.3 [18.9] years; 9006 [53.7%] female) with cancer were studied. Compared with patients with other cancers, patients with HNC were more often members of a minority race/ethnicity, male, poor, publicly insured, and less educated, with lower general and mental health status. Median annual medical expenses ($8384 vs $5978; difference, $2406; 95% CI, $795-$4017) and relative out-of-pocket expenses (3.93% vs 3.07%; difference, 0.86%; 95% CI, 0.06%-1.66%) were higher for patients with HNC than for patients with other cancers. Among patients with HNC, median expenses were lower for Asian individuals compared with white individuals ($5359 vs $10 078; difference, $4719; 95% CI, $1481-$7956]), Westerners ($8094) and Midwesterners ($5656) compared with Northwesterners ($10 549), and those with better health status ($16 990 for those with poor health vs $6714 for those with excellent health). Higher relative out-of-pocket expenses were associated with unemployment (5.13% for employed patients vs 2.35% for unemployed patients; difference, 2.78%; 95% CI, 0.6%-4.95%), public insurance (5.35% for those with public insurance vs 2.87% for those with private insurance; difference, 2.48%; 95% CI, −0.6% to 5.55%), poverty (13.07% for poor patients vs 2.06% for high-income patients), and lower health status (10.2% for those with poor health vs 1.58% for those with excellent health).

Conclusions and Relevance  According to this study, HNC adds a substantial, additional burden to an already financially strained population in the form of higher total and relative expenses. The financial strain on individuals, assessed as relative out-of-pocket expenses, appears to be driven more by income than by health factors, and health insurance does not appear to be protective.

With increasing survivability of cancer, the number of cancer survivors in the United States is expected to increase to 18 million by 2020.¹ The cost of cancer care has also doubled during the past decade, contributing 4.9% of all medical expenditures and continuing to increase.^2,3 The financial burden to patients, especially those with cancer, is increasingly recognized. Out-of-pocket cancer expenses often consume up to 27% of income in low-income households,⁴ and it is the most common cause of medical bankruptcy.⁵

This financial burden is acutely relevant in head and neck cancer (HNC), which occurs disproportionately in the socioeconomically disenfranchised, is costly to treat, and creates long-term health needs.^6-9 Head and neck cancer encompasses mucosal and salivary gland malignant tumors of the upper aerodigestive tract,¹⁰ thus directly affecting vital functions, including voice, swallowing, and cosmesis. Head and neck cancer constitutes 3% of new cancer diagnoses, with 436 060 current survivors,^8,11 and its incidence is increasing.¹² Survival has slowly improved¹³ to 64.5% at 5 years,¹⁴ largely driven by shifting causes from carcinogen exposures to human papillomavirus–mediated disease.^15,16

Most existing literature on HNC expenses is limited to subpopulations defined by the treating institution, payer, age, site, or treatment modality.^17-20 A systematic review²¹ in 2014 concluded that no studies up to that time had assessed the societal burden of HNC medical cost. The largest study⁹ used a national sample and established a median treatment cost of $79 151, with variability attributed to payer, treatment modality, and health status. However, that study was restricted to insured patients and did not include income data to contextualize out-of-pocket expenses. Only 2 studies^22,23 analyzed a nationally representative sample, but the scopes were limited to defining expenses attributable to HNC. No prior study, to our knowledge, directly compares HNC finances and expenses with those of other patient populations.

The present study used the Medical Expenditure Panel Survey (MEPS) database to assess the societal and individual burden of HNC. Comparisons were made with other cancer survivors, trends were assessed for total and relative expenses, and factors associated with increased total and personal expenses were identified.

Data were collected from the MEPS database using household and condition survey files from January 1, 1997, through December 31, 2015.²⁴ Data analysis was performed from April 18, 2018, to August 20, 2018. The survey methods have been previously described in detail.²⁵ Briefly, the survey provides a nationally representative, annual assessment of the nation’s medical expenses and includes data on individual demographics, employment, income, insurance, and medical conditions. Individuals younger than 18 years were excluded. For the years when age at diagnosis was available (2007-2012), length of follow-up was calculated based on age at the time the survey was completed. This study was considered exempt from review and patient consent by the Saint Louis University Institutional Review Board based on use of deidentified, publicly available data. The need for participant informed consent was also waived by the Institutional Review Board for the same reason.

Demographic characteristics included age, sex, race/ethnicity (white, black, Asian, Hispanic, or other), and marital status (married, single, or widowed or separated). Race/ethnicity was self-identified by the survey participants. In addition, educational level was available since 2011 (less than high school, high school graduate, or beyond high school). Economic variables included insurance (any private, public only, or uninsured), employment, total family income, and poverty level. Poverty level was divided into 5 categories based on the percentage of the local poverty level: poor (<100%), near poor (100%-125%), low income (>125%-200%), middle income (>200%-400%), and high income (>400%).

Several metrics of patient health were investigated. Individual self-report of general and mental health was assessed on a 5-point semicontinuous scale, ranging from poor (score of 1) to excellent (score of 5). In addition, 2 comorbidity indexes were calculated from International Classification of Disease, Ninth Revision (ICD-9) codes listed in the condition data files. First, the Charlson Comorbidity Index with Quan modification²⁶ was used to allow comparisons with other oncologic literature. Second, the Elixhauser index with van Walraven modification²⁷ was used because of its stronger correlation with inpatient mortality and medical expenses.²⁸

Total expenses incorporate expenses reported by the household and practitioner survey components. Expenses and income were corrected for inflation to 2014 US dollars using the gross domestic product for total expenses and price index for individual costs and income based on MEPS guidelines. Relative out-of-pocket expenses were calculated as the percentage of total income spent on out-of-pocket medical expenses.

Individuals were categorized as having HNC, other cancer, or no cancer based on the MEPS clinical classification system.²⁹ Individual characteristics were compared between patients with HNC, other cancers, and no cancer. Differences between means, medians, and proportions for each variable are reported together with 95% CIs.

Annual estimates were calculated for HNC and other cancer groups. Total and relative out-of-pocket expenses were assessed over time, and trends were assessed by linear regression models.

Among patients with HNC, factors associated with increased total and relative out-of-pocket expenses were assessed using generalized linear models. To select the optimal model, several models were explored to best reflect the data, as summarized by Mihaylova et al.³⁰ Given that less than 1% of individuals within each cancer group had nonzero expenses, a 2-stage model was not required. Instead, to account for the skewed distribution, a log-link was used. To allow inclusion of individuals with no expenses, these values were set nominally to $1. For those with out-of-pocket expenses but no income, the relative out-of-pocket expense would be infinite, so it was set to the 97.5th percentile of all finite values. These transformations were confirmed to normalize the distributions of total and relative expenses and accommodate a gaussian distribution without substantially affecting median values.

Bivariable models were created using demographics, economic factors, health status, age of diagnosis, and time since diagnosis with the outcome variables. Ordinal variables were analyzed as linear and semicontinuous. Variables were eligible for inclusion in the initial multivariable model if the 90% CI excluded 1 and data were available for all study years. A stepwise analysis was then used to maximize the predictive ability of the model. To assess the association of variables available for limited years, a sensitivity analysis was performed using only the data from 2007 to 2012 with all variables available. Results of the linear models are reported as β with 95% CI after inverse log transformation so that β = 1 represents no association between the independent and dependent variables.

Data were downloaded from MEPS publicly available files, then imported into R statistical computing software, version 3.2.3 (R Foundation for Statistical Computing) for analysis using the survey package. Complex sampling techniques were accounted for using sample weights with balanced repeated replication techniques. Comorbidity indexes were calculated with the icd package. Results are presented as weighted estimates with 95% CIs.

From January 1, 1997, to December 31, 2015, a total of 16 771 surveyed individuals reported a history of cancer (mean [SD] age, 62.3 [18.9] years; 9006 [53.7%] female), including 489 (3.0%) with HNC. This number extrapolates to a national representative weighted sample of 10 083 586 patients with cancer and 261 631 patients (2.7%) with HNC (Figure 1). Of the 131 patients with data from 2007 to 2012 (which includes age of the patient at time of HNC diagnosis), 35 (26.7%) had been diagnosed within the year before the survey.

Patient characteristics are summarized in Table 1, including individuals without cancer for comparison. Compared with patients with other cancers, patients with HNC were less likely to be female, white, and highly educated and more likely to be poor and publicly insured. The general and mental health status of patients with HNC was rated worse. Patients with HNC also had more comorbidities based on the Charlson comorbidity index and Elixhauser index.

Annual medical expenditure varied widely among patients with cancer (range, $0-$797 300). Median expenses were higher for patients with HNC compared with patients with other cancers. The out-of-pocket expenses were similar between groups; however, the out-of-pocket expense constituted a larger percentage of income for patients with HNC.

Total medical expenditures have increased over time (Figure 2A) for patients with other cancers ($273 per year; 95% CI, $170-$374) and patients with HNC ($252 per year; 95% CI, −$256 to $860); however, the 95% CI for HNC included $0 per year. Relative out-of-pocket expenses were generally stable over time (Figure 2B) for HNC (1.45% per year; 95% CI, −1.73% to 4.65%) and other cancer groups (0.25% per year; 95% CI, 4.67%-4.98%), with the notable exception of 2013, during which there was a substantial increase in relative out-of-pocket expense (13.1%; 95% CI, 1.87%-24.4%) before returning to baseline levels in 2014.

In the bivariable models, total medical expenses were associated with race/ethnicity, region, educational level, general health status, and mental health status (Table 2). Specifically, Asian individuals had lower expenses than white individuals, and Midwesterners and Westerners had lower expenses than Northeasterners. Better general and mental health status was associated with lower expenses. Higher educational level was associated with higher expenses.

On the basis of statistical significance in the bivariable models, race/ethnicity and region were incorporated into a multivariable model. Additional models were explored by individually adding variables with potential confounding effects, but each of these led to lack of convergence. In the sensitivity analysis, age at diagnosis and educational level of patients with HNC were included but did not produce different findings or improve the fitting of the model. Therefore, the final model was limited to race, region, and health status. Of these, only Asian race remained independently associated with total expenses. In addition, several categorical variables showed substantial variability of expenses between groups but with wide 95% CIs that included 1. For example, median expenses were less for black individuals ($5313) than white individuals ($10 078), and individuals within 1 year of their HNC diagnosis also had higher median expenses ($18 113 vs $8476).

Relative out-of-pocket expenses varied widely (median, 4.03%; range, 0%-916%). As expected, higher total income and higher income rating on the poverty index were associated with lower relative expenses (Table 3). Other variables associated with higher relative expenses were region, unemployment, public insurance, and lower health status. Relative expenses also varied by education; however, the wide CI reduced the ability to make conclusions regarding the association between educational level on out-of-pocket expenses. A multivariable model was constructed using important variables identified based on our study design; however, insurance and poverty level were removed sequentially because of a lack of convergence. Ultimately, income, sex, employment, and health status were included in the final model, but none were independently associated with relative out-of-pocket expenses.

It is increasingly important for practitioners and the health care system to understand the financial burden on patients and society. Individuals with HNC are expected to be particularly vulnerable to financial strains given the established association with lower socioeconomic status.^6,7 This study, which used national data from nearly 2 decades, found that patients with HNC in the United States are uniquely disadvantaged compared with patients with other cancer in terms of poverty, educational level, and overall health. In addition, socioeconomic variables and geographic region were identified as risk factors associated with higher total medical expenditures and individual burden.

Patients with HNC are demographically distinct from patients with other cancers. Compared with the general population, patients with HNC are classically described as male, white, and middle-aged.^31,32 This study provides perspective by comparing these patients with patients with other cancers and revealing that patients with HNC are more often of a minority race, less educated, poorer, sicker, and lacking private insurance, all of which can exacerbate existing socioeconomic disenfranchisement.^33-35 On the basis of the MEPS data, patients with non-HNCs benefit from 23% higher median income compared with patients with HNC. Compounding this discrepancy, patients with HNC have higher medical expenses (median, $8101 vs $5930). This finding expands on prior literature that demonstrates higher HNC treatment costs⁹ by showing their overall medical expenses remain persistently elevated after the initial treatment period. Compared with the general population, patients with cancer have not only higher medical expenditures but also higher out-of-pocket costs.^5,36 For patients with HNC, limited resources and higher expenses manifest as higher mean out-of-pocket expenses as a percentage of their income. In some cases, out-of-pocket medical expenses consume all a household’s income for a given year. Despite substantial changes during the past 2 decades in the treatment of HNC, the economy, and the health care system, these excess financial burdens have remained largely stable. The spike seen in relative out-of-pocket expenses in 2013 represents relatively high expenses and relatively low income among patients with HNC. This outlier year occurred during the initiation of Medicaid expansion in many states in 2014, which led to volatility in some public insurance markets and may explain the observed fluctuation in relative out-of-pocket expenses.³⁷

Within patients with HNC, several demographic factors were associated with higher absolute and relative medical expenses. Some of these findings were expected: medical expenses were higher for those with lower health status and within the first year of HNC diagnosis, and relative expenses were greater among those with lower incomes. However, the observed variation by race, sex, and educational level highlight important disparities and are less readily explainable. Socioeconomic factors, especially race and insurance status, have been correlated with expenses for several disease categories.^38-41 Reduced access to medical services among lower socioeconomic classes may contribute to these disparities.⁴² In particular, racial differences in access persist even after controlling for income and insurance.⁴³ However, total expenses represent a complex summation of access, disease burden, and care setting, which all vary with demographics.^42,44 For cancer populations, expenses may be driven partly by stage, which correlates with treatment costs and long-term rehabilitative needs.^19,20 Unfortunately, demographics continue to serve as markers of stage of presentation, treatment, and survival in HNC^31,45-48 and likely affect medical expenses through this mechanism.

Relative out-of-pocket expenses were used to assess the burden of medical expenses on the individual patient and family. Similar to total expenses, variability in this metric was associated with demographic, health, regional, and insurance differences. However, the variability in relative expenses appears to be explained primarily by the differences in income. In other words, the individual’s financial burden from out-of-pocket expenses depends more on income than health, demographics, or even insurance status. The paradoxical finding that insurance status does not protect against out-of-pocket costs may be attributable to less intensive treatments among those with inadequate insurance coverage. This would, unfortunately, also result in worse oncologic outcomes in these groups.^49-51 These findings are important for practitioners discussing financial burden with patients, yet most oncologists are uncomfortable having these discussions⁵² despite the recommendations to discuss the financial implications of cancer with patients⁵³ and patients’ desire to hear this information.⁵⁴

For HNC, regional variation exists for absolute and relative medical expenses. Regional variation in medical costs was brought to public attention in 1973 by the Dartmouth Atlas study.⁵⁵ Since then, regional differences are often cited as an example of unwarranted variability and are attributed to waste or disparities in care.⁵⁶ For patients with HNC, MEPS data revealed higher costs in the Northeast followed by the South, West, and Midwest. Meanwhile, the highest relative out-of-pocket expenses were in the South. However, none of this variation remained when controlling for other factors. This finding suggests that variability in this case is attributable to the individual level rather than regional differences in the health care system based on the data available.

This study uses nationally representative medical expenditures from the MEPS database and benefits from the rigorous methods and quality controls used by the survey. The prevalence and demographic profile in the weighted estimates of the HNC population from the MEPS data are consistent with other national representative cancer databases, which supports the accuracy of the MEPS estimates.^32,57 The scope of this survey facilitates comparisons to other cancer sites and could be used in future research to compare with noncancer cohorts. However, these data have limitations inherent to survey studies, such as nonresponse and recall bias. In addition, the breadth of the MEPS survey across disease conditions necessitates practical limitations on disease-specific details. As a result, HNC-specific details are lacking, such as stage, site, and treatment modality, which can affect costs.⁹ The patients identified in the HNC cohort were limited to survivors available to survey. Therefore, patients at higher risk for HNC mortality may be underrepresented, and the financial burden on their families would not be captured.

Many gaps remain in our understanding of the financial burden of HNC on individual families and society. Previous studies^58,59 have estimated expenses directly attributable to an HNC cancer diagnosis, whereas this study focused on overall medical expenses. Further research is needed to assess the indirect financial costs to families and society as a whole. These costs may include rehabilitation, loss of personal and family income, and the inability to obtain future insurance coverage. An improved understanding is needed regarding the association of payer mix with expenses, outcomes, and disparities. This study sets a foundation for future research by establishing burdens unique to HNC and demonstrating variation within the HNC population that may not be justifiable.

Accepted for Publication: December 11, 2018.

Corresponding Author: Sean T. Massa, MD, Department of Otolaryngology–Head and Neck Surgery, Washington University in St Louis, 660 South Euclid Ave, PO Box 8115, St Louis, MO 63110 (seanmassa@wustl.edu).

Published Online: February 21, 2019. doi:10.1001/jamaoto.2018.3982

Author Contributions: Drs Massa and Adjei Boakye had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Osazuwa-Peters, Ward.

Study concept and design: Massa.

Acquisition, analysis, or interpretation of data: Massa, Osazuwa-Peters, Adjei Boakye, Walker.

Drafting of the manuscript: Massa, Osazuwa-Peters, Adjei Boakye, Ward.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Massa, Adjei Boakye.

Administrative, technical, or material support: Massa, Osazuwa-Peters.

Study supervision: Massa.

Supervision: Walker, Ward.

Conflict of Interest Disclosures: None reported.

Meeting Presentation: Part of this work was presented at the Multidisciplinary Head and Neck Cancer Symposium; February 16, 2018; Scottsdale, Arizona.