Out‐of‐pocket costs associated with head and neck cancer treatment

Abstract Background Out‐of‐pocket costs (OOPC) associated with treatment have significant implications on quality of life and survival in cancer patients. Head and neck cancer patients face unique treatment‐related challenges, but to date OOPC have been understudied in this population. Aims This study aims to identify and measure OOPC for patients with head and neck cancer (HNC) in Ontario. Methods HNC patients between 2015 and 2018 at Princess Margaret Cancer Centre in Toronto were recruited. Participants completed OOPC questionnaires and lost income questions during radiation, post‐surgery, and 3, 6, 12, and 24 months after completion of treatment. Associations between OOPC and treatment modality and disease site were tested with multivariable hurdle regression. Results A total of 1545 questionnaires were completed by 657 patients. Median estimated OOPC for the total duration of treatment for participants undergoing chemoradiation was $1452 [$0–14 616], for surgery with adjuvant radiation or chemoradiation (C/RT) was $1626, for radiation therapy alone was $635, and for surgery alone was $360. The major expenses for participants at the mid‐treatment time‐point was travel (mean $424, standard error of the mean [SEM] $34) and meals, parking, and accommodations (mean $617, SEM $67). In multivariable analysis, chemoradiation, surgery with C/RT, and radiation were associated with significantly higher OOPC than surgery alone during treatment (791% higher, p < .001; 539% higher, p < .001; 370% higher, p < .001 respectively) among patients with non‐zero OOPC. Participants with non‐zero OOPC in the laryngeal cancer group paid 49% lower OOPC than those with oropharyngeal cancers in adjusted analysis (p = .025). Conclusions Patients undergoing treatment for HNC pay significant OOPC. These costs are highest during treatment and gradually decrease over time. OOPC vary by patient demographics, clinical factors, and, in particular, treatment modality.


| INTRODUCTION
The costs associated with cancer care may be significant and are often overlooked by healthcare providers. 1,2 Out-of-pocket costs (OOPC) for medical care may cause distress among cancer patients and their families and impact their social, psychological, and spiritual wellbeing. 3 The OOPC borne by cancer patients may impact their ability to participate in treatment and surveillance, which may in turn affect their prognosis. 4 New treatment and surveillance modalities may lead to increased OOPC. 5,6 In Canada and in the province of Ontario, healthcare coverage is universal. However, many costs are still incurred by patients including travel and accommodation costs, prescription medications, and many allied health care services. Some of these costs may be offset for patients who have private insurance coverage. Despite having universal healthcare coverage, patients may suffer a significant financial burden for cancer care. 7,8 Recent studies have shown that cancer patients are more likely to forego medical care or medications due to financial concerns than other patients. 9,10 Moreover, as many as 13.4% of cancer patients spend more than 20% of their income on health care. 11 Annual medical expenditure and productivity loss have been investigated in specific types of cancer, such as breast, colorectal, and prostate cancer. [12][13][14] Although there have been some studies describing the magnitude of OOPC in head and neck cancer (HNC) patients in countries such as China and India, [15][16][17][18][19] there is little work to date that has investigated the impact of OOPC in HNC patients in North America. 20,21 Prior North American studies have utilized national databases, such as the Surveillance Epidemiology and End Results (SEER), to estimate OOPC. [22][23][24] This approach to estimating OOPC may not fully capture all relevant costs or the full financial impact on patients. The Patient Self-Administered Financial Effects (P-SAFE) instrument was previously designed to examine the financial effects of a cancer diagnosis. 7,25 However, we felt the OOPC associated with HNCs are unique and as such we developed and validated a new tool to examine the costs in this specific population.
To our knowledge, this is the first study aiming to comprehensively measure different types of OOPC incurred by HNC patients in North America.

| Study population and data sources
From 2015 to 2018, adult (>18) patients with HNCs (oral cavity, pharynx, larynx, nasal cavity, and unknown primaries) treated with curative intent were approached to participate in a prospective study at the Princess Margaret Cancer Centre in Toronto, Ontario, Canada. Ethics approval was obtained for this study from the University Health Network (UHN) Research Ethics Board (REB), REB#07-0521. Informed consent was obtained from each participant. Participants' OOPC were assessed immediately after surgery (in surgically resected patients); at mid-radiation treatment (in patients receiving radiation treatment); and at 3, 6, 12, and 24 months after completion of treatment. The questionnaire was administered prospectively to all potentially interested patients. Patients were enrolled from all timepoints simultaneously. As such, not all patients were followed from treatment initiation to the last follow-up; some were enrolled well after their treatment. Demographic data and household income were collected around the time of diagnosis; patient-reported lost wages due to cancer was collected at 12 months post-treatment.
The OOPC questionnaire used in this study ( Figure S1) was devel-

| Analysis
The distribution of OOPC across time-points was reported descriptively.
Baseline household income and lost income due to cancer are reported using descriptive statistics. Spearman correlations were computed between OOPC and baseline income, as well as OOPC and lost income due to cancer. Kruskal-Wallis tests were used to determine associations between treatment modality and lost income due to cancer. The association between lost income and OOPC was assessed by Spearman correlation. Univariable and multivariable regression analyses were used to test associations between OOPC and clinico-demographic patient characteristics. Regression analysis was conducted for both mid-treatment costs (using total estimated OOPC during treatment) and follow-up costs at 3 months post-treatment. As is often the case with studies measuring cost data, the distribution of total OOPC was right-skewed and zero-inflated (a significant proportion of patients reported zero costs). We therefore applied a hurdle model, a two-part mixture model that accounts for two separate stochastic processes occurring within the data. First, the probability of patients reporting non-zero OOPC was modeled using logistic regression. Then, among patients reporting non-zero costs, OOPC were log-transformed and modeled using linear regression. This log-linear regression was used to find the percent increase in OOPC attributable to group membership for each clinico-demographic factor. Variable selection for multivariable models was conducted by keeping any variable with p < .1 in the logistic or linear regression steps.

| Relationships between OOPC and clinicodemographic factors
Results of univariable and multivariable regression analysis on estimated OOPC during treatment are presented in Table 3. 91% of patients reported OOPC greater than $0 overall. Treatment modality, stage at diagnosis, ECOG status, disease site, employment, education, income, and age were retained in the multivariable model after variable selection.
Among those with greater than $0 OOPC, treatment modality remained significant in adjusted analyses; participants receiving radiation paid 370% more than those with surgery alone during treatment (p < .001).
The group with surgery plus C/RT paid 539% more (p < .001), and the chemoradiation group paid 791% more (P < .001). Participants with nonzero OOPC in the laryngeal cancer group paid 49% lower OOPC than those with oropharyngeal cancers in adjusted analysis (p = .025).
Three hundred and eighty-four patients contributed OOPC at 3-months post-treatment; of those, 85% of patients reported costs greater than zero. Univariable and multivariable analysis is presented in Table 4. For the 3-month post-treatment multivariable model, treatment modality, disease site, and education were retained after variable selection. In multivariable analysis, treatment modality was not significantly associated with the odds of paying any OOPC, or amount paid above $0. Participants who paid greater than $0 with laryngeal cancers also had significantly lower odds of reporting non-zero costs compared to oropharynx (p = .001).
Due to the high proportion of participants who did not report household income, we performed a sensitivity analysis while including those who did not indicate an income. In one analysis, this group had a higher likelihood of non-zero OOPC than those with income <$40 000, however this was not statistically signifi-  lower household income, 11,36,37 and lower education level 11,37 have previously been associated with higher costs in our jurisdictions, these were not independently significant in our study. This is in keeping with other published literature from Canada that has shown income and education level to be poor predictors of OOPC. 27 This likely reflects the nature of head and neck cancer care in Ontario, which is universal and centralized. Fifth, 48.5% of patients preferred not to report their income, which limited ability to interpret the overall burden of OOPC on this sample. We attempted to address this by performing a sensitivity analysis that included this group, and were able to confirm our initial findings.
Lastly, collecting patients' perceptions of their OOPC was beyond the scope of this study. This is important to consider as objective costs may not fully reflect the impact of treatment. The financial toxicity of cancer treatment is being increasingly recognized and there are specific instruments to evaluate for it including the COmprehensive Score for financial Toxicity (COST) and Financial Index of Toxicity (FIT) instruments, the latter which was developed specifically for HNC. 38,39 Additional research on the perceived impact of OOPC on patients' finances, willingness to pay additional medical costs, and patients' level of regret about making medical decisions due to costs, would be an important component in assessing the true burden of cancer-related OOPC.
In conclusion, this study found that Canadian HNC patients pay significant OOPC toward their cancer care, which tend to be highest during treatment and decrease over time. These costs vary by patient demographics, clinical factors, and, in particular, treatment modality.