Financial Hardship and Quality of Life Among Patients With Advanced Cancer Receiving Outpatient Palliative Care

Abstract

Background:

Patients with advanced cancer are increasingly experiencing financial hardship (FH) and associated negative health outcomes.

Objective:

The aims of this study were to describe FH and explore its relationship to quality of life (QOL) in patients with advanced cancer receiving outpatient palliative care (PC).

Methods:

Validated questionnaires assessed FH, QOL dimensions, symptom burden, and sociodemographic and clinical characteristics. Descriptive statistics characterized the sample and described FH. Pearson correlation and linear regression assessed relationships between FH and QOL.

Results:

The average participant (n = 78) age was 56.6 (SD, 12.2) years. Most were female (56.4%), White (50%) or Black (46.2%), and had a range of education, partner statuses, and cancer diagnoses. Median time since cancer diagnosis was 35.5 months (interquartile range, 9–57.3 months). Highest mean symptom burden scores were for pain (2.5 [SD, 1.0]) and fatigue (2.0 [SD, 1.1]), on a 0- to 3-point scale (higher score representing worse symptom burden). The median COST (COmphrehensive Score for financial Toxicity) score was 15.0 (interquartile range, 9.0–23.0). Most (70%) had some (n = 43) or extreme (n = 9) difficulty paying for basic needs. Greater than 28% (n = 21) incurred cancer-related debt. Multivariate models indicated that FH negatively affected role limitations due to physical health (P = .008), pain (P = .003), and emotional well-being (P = .017) QOL dimensions.

Conclusions:

Financial hardship, QOL, and symptom burden scores demonstrate need for continued support for and research among patients with advanced cancer. Data support links between FH and important QOL dimensions. Larger, longitudinal studies are needed to understand how FH affects QOL in patients with advanced cancer.

Implications for Practice:

Proactive financial assessment and interventions are needed to support patients with advanced cancer experiencing the cumulative effects of cancer and its treatment.

Nearly 1.9 million cancer cases will be diagnosed in 2021.¹ As cancer diagnoses increase and patients live longer following diagnoses, overall costs of cancer care also are increasing.Nationalcancercostsareprojectedtodramaticallyincrease, from $182.6 billion to $245.6 billion between 2015 and 2030 (34%).² Payers then often shift these costs to patients. In 2018, patients paid $5.6 billion out-of-pocket for cancer treatments.³ Further, as patients’ life expectancies have improved following cancer diagnoses, late and long-term effects of cancer and its treatment contribute to individuals’ annual healthcare expenditures, compromise their ability to work, and may alter individuals’ ability to maintain insurance coverage.^4,5

Financial hardship⁶ (FH) is defined as the objective financial burden, subjective distress, and behaviors used by patients to cope with costs of cancer and cancer care (eg, treatment nonadherence, skipped appointments).⁷ Previous literature^6,8 has identified factors associated with risk for FH at multiple levels, ranging from the patient- and family-level risks (eg, age, sex, race and ethnicity, educational attainment, health literacy, household income, medical insurance, social support, cancer stage, and treatment) to policy-level risks.⁸ Financial hardship has been associated with marked negative outcomes, including increased symptom burden, decreased quality of life (QOL),bankruptcy,andincreasedmortality.^9–13 Despite the known negative influence of FH on health and QOL, studies have demonstrated that adequate patient-provider communication about FH does not consistently occur in the clinical setting,^14,15 which may contribute to increased symptom burden and suffering.

The paradigm of advanced cancer treatment has evolved from noncurative, palliative treatment alone to now include advanced therapies that transform some advanced cancers into chronic illnesses. Patients living longer but facing life-limiting illness may experience cumulative FH from cancer, its chronic treatment, and its downstream effects. Recommendations from the American Society of Clinical Oncology (ASCO) emphasize that quality cancer care should include continued discussions regarding likely treatment efficacy and a realistic appraisal of subsequent patient and family costs for both early- and late-stage treatments.¹⁶ Further, a limited body of evidence suggests that patients with metastatic cancer are likely to be vulnerable to FH due to disease and population characteristics.¹⁷ Previous studies describing FH in people with cancer, however, traditionally have not been representative of patients with advanced or metastatic cancer.⁸

Early integration of palliative care (PC) into oncology care has been shown to improve QOL and mood and may prolong survival^18,19 and is recommended for all patients with advanced cancer by ASCO and the Oncology Nursing Society.^20,21 Current PC guidelines to establish excellence in care delivery²² include communication about and assessment of indicators of financial vulnerability (eg, employment change, economic security, and ability to pay bills and access medications and food). Despite these guidelines, little information has been documented regarding FH among PC recipients. Lack of knowledge about FH faced by patients with advanced cancers impedes vital patient–interdisciplinary team communication, can affect patients’ well-being and their decisions related to goals of care, and hinders the development of interventions to prevent, identify, and ameliorate FH and its downstream sequelae.

The purpose of this pilot study was to describe FH and explore its relationship to QOL in patients with advanced cancer receiving outpatient PC services, particularly among socioeconomically diverse patients in the southern United States. We hypothesized that worse FH would be associated with poorer QOL. Specifically, we aimed to (1) describe FH and (2) examine the relationship between FH and QOL.

Methods

Design, Sample, and Setting

This was a descriptive, cross-sectional pilot study of patient-reported measures and medical record data. Eligibility criteria included age ≥18 years, advanced cancer diagnosis (ie, documented metastatic or locally advanced refractory disease), receiving outpatient PC services, ability to complete measures in English, and competency for consent. No additional exclusion criteria were applied. We recruited patients between June 2019 and February 2020 from both Emory Supportive Care Clinic and the Georgia Cancer Center for Excellence at Grady Palliative Care Clinic, which, respectively, serve a tertiary academic medical center and a vulnerable safety-net setting.

Ethical Considerations and Procedures

The institutional review boards and research committees at both sites approved the study protocol. Clinicians introduced research staff to potentially eligible patients. Research staff then obtained informed consent. Patients completed study assessments either independently or with research staff assistance in a private room at the clinic on iPads via REDCap.²³ If a participant was interested in participation but not available to complete the study assessments at the time of consent, alternative plans to collect data were made, either at a future clinic appointment, over the phone, or online via REDCap. Participants were compensated with a $10 gift card.

Framework, Variables, and Measures

This study was informed by Yabroff and colleagues’⁷ framework of FH among cancer patients, which outlines factors at hierarchical levels that are associated with FH. While this study mainly explored aspect of the framework that addresses patient, family, and provider factors, healthcare systems and policies remain important considerations for future inquiry. A battery of valid, reliable, and common measures was used to characterize the sample and assess study concepts and possible covariates that were identified in the literature (listed in measures below) as associated with key study variables.

The 11-item COmphrehensive Score for financial Toxicity (COST),^24,25 tested in samples of patients with advanced cancer, was used to assess FH. Items are evaluated using a 0- to 4-point Likert scale and include patients’ perceptions of their financial situation, costs, resources, and concerns. The instrument sum score range is 0 to 44, with lower scores indicating worse FH.

Single items from the 2018 Medical Expenditure Panel Survey Cancer Survivorship Supplement²⁶ assessed the participant’s primary caregiver, ability to cover cost of cancer care, family debt incurred due to cancer, and provider communication about out-of-pocket cancer costs.

The Short Form-36 Health Survey,²⁷ was used to evaluate QOL across 8 dimensions of physical and mental health including physical functioning, role limitations due to physical health, pain, general health, social functioning, role limitations due to emotional problems, energy/fatigue, and emotional well-being. Items are averaged to generate dimension scores (0–100), with higher scores indicating better QOL.

Items from the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE)^28,29 were used to assess symptom burden for prevalent oncology symptoms that are known to impact QOL, including fatigue, insomnia, pain, anorexia, dyspnea, cognitive problems, anxiety, nausea, depression, neuropathy, constipation, and diarrhea. Each symptom is scored using 5-point Likert scales to assess symptom attributes (ie, frequency, severity, and/or interference with daily activities). A validated and reliable composite grading algorithm was applied,³⁰ which combines 1, 2, or 3 attribute scores measured for each symptom to yield a single numerical score, a “composite grade,” for each symptom (range = 0–3, with higher grades representing worse symptom burden). A composite grade of 0 indicates no symptom burden. Each symptom composite grade score >0 indicates symptom presence.

The 3-item Brief Health Literacy Scale³¹ was used to assess health literacy. Likert scale items (1–5) provide a sum score, with higher scores indicating higher subjective health literacy. Previous literature suggests that a score >4 on all 3 items indicates adequate, versus low, health literacy.³²

The Interpersonal Support Evaluation List-12³³ assessed perceived social support. This instrument uses 4-point Likert scale items. Sum scores range from 0 to 36, with higher scores indicating better social support.

Trained study staff assessed participants’ performance status using the single-item Eastern Cooperative Oncology Group (ECOG) Performance Status scale.³⁴ The ECOG grade range is 0 (fully active and able to carry on predisease unrestricted performance) to 4 (completely disabled and bedridden or chair-ridden).

The 10-item self-report brief Charlson Comorbidity Index³⁵ was used to measure chronic disease burden. This index of comorbidities scale asks participants whether they have experienced or are dealing with a list of health conditions. Responses (yes/no) are weighted and range from 0 to 25, with higher score indicating worse comorbidity burden.

Additional sociodemographic and clinical variables were collected using items from standardized, institutionally developed self-report forms (ie, age; sex at birth; ethnicity; race; partner status; educational attainment; informal primary caregiver; annual household income, with categories based on household size; difficulty paying for basic needs [not at all, somewhat, extremely]³⁶) or extracted from medical records within 1 week of data collection (ie, current cancer diagnosis; time since most recent cancer diagnosis and treatment; number of cancer treatment type[s] received, body mass index, and current health insurance).

Sample Size

Because of the exploratory nature of this pilot study, a traditional sample size calculation was not performed. Instead, we focused on determining minimum detectable effect sizes. Previously reported COST parameters in patients with cancer include mean scores of approximately 22 (SD, 12).^24,25 To describe FH (aim 1), a sample size of 44 participants would be required to determine the mean COST score among oncology patients receiving PC with a 90% confidence interval of ±3 points. For aim 2, a sample size of 44 achieved 80% power to detect an effect size (Cohen f²) of 0.19 (a medium-high effect size),³⁷ attributable to the primary predictor (FH) using an F test with a significance level of .05 after adjusting for 4 additional independent variables. To account for potential loss to follow-up or missing data, we initially aimed to recruit 50 participants for the study. Because of early robust enrollment, we increased our target sample size to 100. However, recruitment was halted in March 2020 because of clinical site restrictions related to the COVID-19 pandemic.

Data Analyses

Descriptive and exploratory analyses were performed to examine data distributions, characterize the sample, and compute study variables. Scales were scored per developer guidelines; when no information was available regarding the number of items required to scale a score, scores were imputed using mean imputation if at least 80% of scale items were completed. For aim 1, descriptive statistics (mean, median, range, and SD for continuous variables; frequencies and percentages for categorical variables) were computed for all study variables including the COST measure of FH. We ran Pearson product-moment correlations to determine relationships between FH and QOL dimensions.

For aim 2, after verifying statistical assumptions, separate hierarchical linear regression models were conducted to examine the unique contribution of FH on individual QOL dimensions that were found to be statistically associated with FH in bivariate analyses. All models were adjusted for covariates identified as statistically associated with at least one QOL dimension (Pearson product-moment correlations for continuous and independent t tests or one-way analysis of variance for categorical variables). For each model, in step 1, we examined the effects of covariates on each QOL dimension. In step 2, the main predictor of interest, FH, as measured by the COST instrument, was added to the model to determine how much variability was additionally explained by the main predictor, after adjusting for the covariates for each outcome variable. Multicollinearity in regression analysis was measured by the variance inflation factor (VIF). Small VIF values indicate low correlation among predictors, and any VIF value >10 is considered to have severe multicollinearity issues.³⁸

IBM SPSS version 27 (IBM Corp, Armonk, New York) was used for data processing and analyses. The PRO-CTCAE composite grade algorithm macro was ran using SAS^® Studio, a web application for SAS (SAS Institute Inc., Cary, North Carolina). Statistical significance was set at .05.

Results

After screening charts for potential eligibility, we approached 130 patients in clinic; 12 were ineligible, and 35 declined, with patient time constraints being the most common patient-reported reason for declining. Of the 83 patients we consented, 5 patients who requested to complete surveys at home were lost to follow-up and did not complete any survey data; 5 patients partially completed the battery of surveys, and 73 participants completed the full battery of instruments (Figure 1).

Figure 1.

CONSORT (Consolidated Standards of Reporting Trials) diagram depicting screening, recruitment, and data collection.

Sociodemographic and clinical characteristics of the participants who partially or fully completed study instruments are displayed in Tables 1 and 2. The average participant age was 56.6 (SD, 12.2) years. Most participants were female (56.4%), non-Hispanic (88.2%), and either White (50.0%) or Black or African American (46.2%). A broad range of partner statuses and educational backgrounds were represented. Nearly half (49.4%) had low health literacy. More than half (53.8%) identified themselves as their own caregiver. Perceived median social support was 30.0 (interquartile range [IQR], 27.3–32.0). Most participants (96%) were insured. Numerous cancer diagnoses were represented, the most common being breast (21.8%), colorectal (14.1%), and lung (11.5%). The median time since diagnosis was 35.5 months (IQR, 9.0–57.3 months). This heavily treated sample represented varying combination therapies, and most (78.2%) were receiving cancer treatment. Most participants’ (92.3%) ECOG performance status was 0 to 2. The mean Charlson Comorbidity Index was 6.6 (SD, 3.5). The PRO-CTCAE composite grade scores (see Figure 2 for distributions; additional descriptive statistics are available online (Table S1, Supplemental Digital Content, http://links.lww.com/CN/A45) indicated that the most burdensome symptoms and the most frequently reported symptoms were pain (mean, 2.5 [SD, 1.0]; reported by 94.4%), fatigue (mean, 2.0 [SD, 1.1]; reported by 93.2%), and anxiety (mean, 1.4 [SD, 1.0]; reported by 78.8%).

Table 1 •.

Sample Sociodemographic Characteristics

Characteristics	n (%)
Age (n = 78), mean (SD), range, y	56.6 (12.2), 24–78
Sex at birth (n = 78)
Female	44 (56.4%)
Ethnicity (n = 76)
Not Hispanic or Latino	67 (88.2%)
Race, self-report (n = 78)
White	39 (50.0%)
Black or African American	36 (46.2%)
Partner status (n = 78)
Married or domestic partnership	30 (38.5%)
Never married	22 (28.2%)
Divorced or separated	20 (25.6%)
Education, highest level attained (n = 78)
≤ High school	12 (15.4%)
High school graduate or GED	14 (17.9%)
Some college or associate degree	26 (33.3%)
Bachelor’s degree	16 (20.5%)
≥ Master’s degree	10 (12.8%)
Health literacy (n = 77), mean (SD), range	11.7 (3.0), 3–15
Low	38 (49.4%)
Adequate	39 (50.6%)
Informal primary caregiver (n = 78)
Self	42 (53.8%)
Spouse or partner	20 (25.6%)
Parent or legal guardian	6 (7.7%)
Child	4 (5.1%)
Home health aide	2 (2.6%)
Case manager	1 (1.3%)
Roommate or friend	1 (1.3%)
Social support (n = 72), median (IQR)	30.0 (27.3–32.0)
Currently documented health insurance (n = 78)
Medicare	32 (41.0%)
HMO	28 (35.9%)
Medicaid	12 (15.4%)
Self-pay (no insurance)	2 (2.6%)
No means to pay (no insurance)	2 (2.6%)
Group health insurance (purchased by employer and offered to employees)	1 (1.3%)
Individual health insurance (purchased by individual for self)	1 (1.3%)

Abbreviations: GED, tests of General Educational Development; HMO, Health Maintenance Organization; IQR, interquartile range.

Table 2 •.

Sample Clinical Characteristics

Characteristics	n (%)	Mean (SD), Range
Current cancer diagnosis (n = 78)
Breast	17 (21.8)
Colorectal	11 (14.1)
Lung	9 (11.5)
Liver	7 (9.0)
Prostate	6 (7.7)
Melanoma	5 (6.4)
Pancreatic	4 (5.1)
Head and neck	4 (5.1)
Gynecological	3 (3.8)
Kidney	3 (3.8)
Basal cell	3 (3.8)
Hodgkin lymphoma	2 (2.6)
Other	4 (5.1)
Time since most recent cancer diagnosis, months (n = 78), median (IQR)		35.5 (9–57.3), 1–264
Time since most recent cancer treatment, y (n = 78)
Currently receiving treatment for cancer	61 (78.2)
<1 y	15 (19.2)
1–5 y	2 (2.6)
No. of treatment types received (n = 78)
1	10 (12.8)
2	17 (21.8)
3	29 (37.2)
4	16 (20.5)
5	6 (7.7)
BMI, most recently documented (n = 78)		25.5 (6.9), 14.5–47.8
ECOG Performance Status (n = 78)
0	27 (34.6)
1	31 (39.7)
2	14 (17.9)
3	5 (6.4)
4	1 (1.3)
Charlson Comorbidity Index (n = 71)		6.6 (3.5), 0–17

Abbreviations: BMI, body mass index; ECOG, Eastern Cooperative Oncology Group; IQR, interquartile range.

Figure 2.

Distribution of PRO-CTCAE Symptom Composite Grades. Abbreviation: PRO-CTCAE, Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events.

See Table 3 for descriptive data for FH and related constructs. Although some data were missing, 26 (34.7%) reported annual household income below or near poverty levels, and 15 (20%) had low income. More than 70% had some (n = 43) or extreme (n = 9) difficulty paying for basic needs. Only 36.5% (n = 27) had been able to cover their share of cancer care costs. More than 28% (n = 21) incurred debt due to cancer; of those, most had incurred <$10 000 in debt. The median COST score of FH was 15.0 (IQR, 9.0–23.0). Forty-nine participants (68.1%) said that, since first being diagnosed with cancer, neither their doctor nor any healthcare provider had discussed their out-of-pocket cancer care costs with them.

Table 3 •.

Sample Financial Hardship Data

Characteristics	n (%)
Household income, yearly, before taxes (n = 75)a
<$20 000 (below or near poverty)	26 (34.7)
$20 000–$39 999 (low income)	15 (20.0)
$40 000–$99 999 (middle class)	12 (16.0)
$100 000–$149 999 (upper middle class)	5 (6.7)
>$150 000 (high income)	5 (6.7)
Unknown	2 (2.7)
Prefer not to answer	10 (13.3)
How difficult is it to pay for your basic needs? (n = 74)
Not at all difficult	22 (28.2)
Somewhat difficult	43 (55.1)
Extremely difficult	9(11.5)
Have you been unable to cover your share of cost of cancer care? Yes (n = 74)	27 (36.5)
Have you or anyone in your family had to borrow money or go into debt because of cancer? (n = 74)
Yes	21 (28.4)
No	53 (71.6)
If yes to cancer debt, how much? (n = 21)
<$10 000	11 (52.4)
$10 000–$24 999	6 (28.6)
$25 000–$49 999	2 (9.5)
$50 000–$74 999	2 (9.5)
Financial hardship, COST measure (n = 75), median (IQR), range	15.0 (9.0–23.0), 0–44
	Mean (SD) = 16.8 (10.1)
Since first cancer diagnosis, has any doctor or healthcare provider discussed with you your cost of OOP cancer care? (n = 72)
Discussed it with me in detail	3 (4.2)
Briefly discussed it with me	15 (20.8)
Did not discuss it at all	49 (68.1)
I don’t remember	5 (6.9)

Abbreviations: COST, Comprehensive Score for Financial Toxicity; IQR, interquartile range; OOP, out-of-pocket.

^aIncome categorization is based on the rounded average US household size of 3, which was consistent with our sample mean household size of 2.6.

See Table 4 for QOL dimension scores and correlations with the COST. Quality-of-life scores were generally low, the lowest of which were for role limitations due to physical health (20.6 [SD, 34.7]) and role limitations due to emotional problems (34.2 [SD, 41.0]). Correlations between FH and QOL dimension scores identified statistically significant, weak, positive associations between FH and the following QOL dimensions: emotional well-being (r = 0.393, P < .001), pain (r = 0.320, P = .005), role limitations due to physical health (r = 0.282, P = .015), and role limitations due to emotional problems (r = 0.276, P = .017).

Table 4 •.

QOL Dimension Scores and Bivariate Correlations With the COST Measure of Financial Hardship

QOL Dimensions	Mean (SD), Range	Correlation With FH
Physical functioning (n = 73)	40.3 (23.8), 0–90	r = 0.062, P = .599
Role limitations due to physical health (n = 74)	20.6 (34.7), 0–100	r = 0.282, P = .015
Pain (n = 75)	40.7 (25.4), 0–100	r = 0.320, P = .005
General health (n = 74)	38.4 (19.4), 0–85	r = 0.025, P = .832
Social functioning (n = 74)	51.8 (26.1), 0–100	r =0.183, P = .119
Role limitations due to emotional problems (n = 75)	34.2 (41.0), 0–100	r = 0.276, P = .017
Energy/fatigue (n = 75)	34.8 (19.9), 0–85	r =0.014, P = .236
Emotional well-being (n = 75)	64.2 (20.4), 12–100	r = 0.393, P < .001

Abbreviations: COST, COmphrehensive Score for financial Toxicity; FH, financial hardship; QOL, quality of life.

Bivariate analyses among patient sociodemographic and clinical variables and these four QOL dimensions are presented as supplementary online material (available online in Tables S2–S4: Supplemental Digital Content 2, http://links.lww.com/CN/A46; Supplemental Digital Content 3, http://links.lww.com/CN/A47; Supplemental Digital Content 4, http://links.lww.com/CN/A48), revealing significant associations between QOL dimensions and PRO-CTCAE composite grade symptom burden scores. The top 3 scores (ie, pain, fatigue, and anxiety) were significantly correlated with many of the other symptoms and QOL dimensions. To avoid multicollinearity among predictors, symptom burden scores were not included in the regression models and are instead reported descriptively (Figure 2; additional data available online in Table S1, Supplemental Digital Content, http://links.lww.com/CN/A45) and as QOL correlations (available online in Table S2, Supplemental Digital Content, http://links.lww.com/CN/A46). Associations among QOL dimension scores and both categorical (ie, sex, race, and education) and continuous (ie, age, performance status, health literacy, FH, social support, and body mass index) predictor variables were also examined (available online in Tables S3 and S4: Supplemental Digital Content 3,http://links.lww.com/CN/A47;Supplement Digital Content 4, http://links.lww.com/CN/A48). Age, race, and performance status, which were statistically associated with at least one QOL dimension in these data explorations, were included as covariates in multivariate analyses.

Separate hierarchical linear regression tests were conducted to examine the unique contribution of FH on the four QOL dimensions identified as correlated with the COST measure (Table 5). Only participants who completed the full survey battery (n = 73) were included in multivariate analyses because of missing data. No multicollinearity was detected in the regression analysis (all VIFs <2). Adjusting for age, race, and performance status, lower FH (higher scores) was significantly associated with better QOL (higher scores) in the following QOL dimensions: role limitations due to physical health (B = 1.31, P = .008), pain (B = 1.03, P = .003), and emotional well-being (B = 0.65, P = .017). Financial hardship explained a significant amount of variance in these 3 QOL dimensions: 10% of role limitations due to physical health, 11% of pain, and 7% of emotional well-being. Although FH was associated with the role limitations due to emotional problem QOL dimension in bivariate analysis, it was no longer significant (P = .104) when adjusting for age, race, and performance status.

Table 5 •.

Hierarchical Linear Regression Models of Predictors of the QOL Dimensions

QOL Dimensions	Predictors	Model 1				Model 2
		B (95% CI)	SE B	β	P	B (95% CI)	SE B	β	P
Role limitations due to physical health	Intercept	14.72 (−25.58, 55.01)	20.20		.469	25.95 (−13.42, 65.31)	19.73		.193
	Age	0.05 (−0.66, 0.76)	0.35	0.02	.893	−0.54 (−1.34, 0.26)	0.40	−0.20	.181
	Race a (Ref: White)
	Black	5.01 (−12.00, 22.02)	8.53	0.07	.559	8.16 (−8.27, 24.57)	8.23	0.12	.325
	Other	−18.72 (−72.37, 34.93)	26.89	−0.09	.489	−18.75 (−70.03, 32.53)	25.70	−0.09	.468
	ECOG	1.49 (−7.63, 10.60)	4.57	0.04	.746	0.28 (−8.47, 9.03)	4.39	0.01	.949
	COST totala					1.31 (0.36, 2.26)	0.48	0.38	.008 b
	Model summary	R2 = 0.02; adjusted R2 = −0.04; F = 0.33				R2 = 0.12; adjusted R2 = 0.05; ΔR2 = 0.10 b; F = 1.80
Pain	Intercept	32.92 (4.62, 61.23)	14.19		.023	41.71 (14.40, 69.03)	13.69		.003
	Age	0.29 (−0.21, 0.79)	0.25	0.14	.247	−0.17 (−0.73, 0.38)	0.28	−0.08	.539
	Race (Ref: White)
	Black	−5.01 (−16.92, 6.89)	5.97	−0.10	.404	−2.56 (−13.90, 8.79)	5.69	−0.05	.654
	Other	−6.63 (−44.48, 31.22)	18.98	−0.04	.728	−6.64 (−42.37, 29.09)	17.91	−0.04	.712
	ECOG a	−6.38 (−12.81, 0.06)	3.22	−0.24	.052	−7.32 (−13.42, −1.22)	3.06	−0.27	.019 c
	COST total a					1.03 (0.37, 1.69)	0.33	0.41	.003b
	Model summary	R2 = 0.08; adjusted R2 = 0.03; F = 1.58				R2 = 0.20; adjusted R2 = 0.14; ΔR2 = 0.11b; F = 3.34 b
Role limitations due to emotional problems	Intercept	−7.15 (−53.49, 39.20)	23.24		.759	0.88 (−45.94, 47.69)	23.47		.970
	Age	0.81 (0.00, 1.62)	0.41	0.25	.051	0.39 (−0.56, 1.34)	0.48	0.12	.419
	Race (Ref: White)
	Black	−0.95 (−20.43, 18.53)	9.77	−0.01	.923	1.29 (−18.15, 20.73)	9.75	0.02	.895
	Other	−0.08 (−62.05, 61.89)	31.07	0.00	.998	−0.09 (−61.33, 61.15)	30.70	0.00	.998
	ECOG	−4.06 (−14.59, 6.47)	5.28	−0.09	.445	−4.92 (−15.38, 5.53)	5.24	−0.11	.351
	COST totala					0.94 (−0.20, 2.07)	0.57	0.23	.104
	Model summary	R2 = 0.06; adjusted R2 = 0.01; F = 1.11				R2 = 0.10; adjusted R2 = 0.03; ΔR2 = 0.04; F = 1.45
Emotional well-being	Intercept	38.38 (16.26, 60.51)	11.09		.001	43.92 (22.05, 65.79)	10.96		<.001
	Age a	0.53 (0.14, 0.92)	0.19	0.32	.008 b	0.24 (−0.21, 0.68)	0.22	0.15	.288
	Race (Ref: White)
	Black	−3.72 (−13.03, 5.58)	4.66	−0.09	.427	−2.18 (−11.26, 6.90)	4.55	−0.05	.634
	Other	−12.70 (−42.29, 16.88)	14.83	−0.10	.395	−12.71 (−41.32, 15.89)	14.34	−0.10	.378
	ECOG	−2.20 (−7.23, 2.83)	2.52	−0.10	.386	−2.80 (−7.68, 2.09)	2.45	−0.13	.257
	COST totala					0.65 (0.12, 1.18)	0.27	0.32	.017 c
	Model summary	R2 = 0.13; adjusted R2 = 0.08; F = 2.67 c				R2 = 0.20; adjusted R2 = 0.14; ΔR2 = 0.07c; F = 3.48 b

Abbreviations: ΔR², the change in R²; β, standardized coefficient; B, unstandardized coefficients; CI, confidence interval; COST, COmphrehensive Score for financial Toxicity; ECOG, Eastern Cooperative Oncology Group; QOL, quality of life; R², coefficient of Determination; Ref, reference value.

^aPredictor(s) had a statistically significant association with the outcome variable in bivariate analysis (P < .05).

^bP < .01.

^cP < .05.

Discussion

This study advances the science and fills a critical knowledge gap by characterizing FH in commonly underrepresented patients with advanced cancer. This sample of patients with cancer, including patients with low income, the majority of whom had received 3 or more cancer treatment types and experienced high FH, low QOL, and symptom burden, despite receiving outpatient PC services. Findings describe patients’ difficulty paying for basic needs, inability to cover cancer care costs, family debt, and lack of provider/patient conversation about costs of cancer care (since diagnosis). The COST measure of FH in this socioeconomically and racially diverse sample indicated lower (worse) FH scores than data reported in other patient samples with metastatic cancer in the United States and Hong Kong.^25,39,40 Further, risk factors for FH in cancer mirror sociodemographic characteristics that were represented in this sample, including race and income.⁷

As hypothesized and consistent with previous literature,^39,41–43 our study confirmed that higher FH contributed significantly to worse QOL in several dimensions that are critical to patients facing serious illness, including role limitations due to physical health, pain, and emotional well-being. The dimensions that were found to be significantly associated with FH are also consistent with prior literature QOL constructs such as role changes, pain, and emotional/psychological well-being.^12,44,45 New literature⁴⁰ suggests that a COST cutoff score of 17.5, which is similar to our median sample score of 16.8, may provide acceptable sensitivity and specificity to predict QOL as measured by the Functional Assessment of Cancer Therapy—General. While self-reported race was a variable included in model testing (that did not produce significant associations in multivariate testing), we recognize that race is a social construct, not a biological variable, and suggest that future studies continue to explicitly measure and examine root causes underpinning health and well-being, such as systemic racism.

One of the first studies investigating FH and QOL in advanced cancer patients, this study sheds light on potential clinician and interdisciplinary team member assessment and conversation gaps. Although literature suggests that patients desire cost communication with their providers, our findings that patients reported lack of conversation about costs of care are consistent with prior cancer literature describing a disconnect in whether these conversations actually occur.^14,15,46 Interestingly, a study of the Medical Expenditure Panel Survey data found that greater cancer pain is associated with lower likelihood of cost discussions with healthcare providers,⁴⁷ making this topic particularly important for patients receiving PC. Cost communication may also play an important role in reducing patients’ out-of-pocket costs, as was reported in a prior study that examined relationships between cost discussions and out-of-pocket costs in a sample of solid tumor patients undergoing cancer treatment.⁴⁸

Strengths and Limitations

We demonstrated ability to recruit vulnerable PC study participants with advanced cancer in 2 busy outpatient clinics. Purposeful and targeted recruitment strategies contributed to a socioeconomically diverse sample of PC patients with cancer, of which 46% self-reported being Black or African American, groups that are historically underrepresented in research studies and FH cancer literature. Results should be interpreted within the context of several limitations and within the context of the data collected just prior to the COVID-19 pandemic and economic downturn, which may have exacerbated FH. In addition, this sample was clinically heterogeneous in terms of diagnoses and included patients receiving care in a safety-net hospital, which may limit generalizability of findings to broader populations. The cross-sectional design limits our interpretation to associations among variables, whereas a prospective study would be necessary to determine temporal relationships. Analytically, our multivariate model fit suggests that there were other important variables influencing variance in QOL dimension variables that were not included in this study, and mean imputation may introduce bias to findings. However, we accomplished the pilot study goal of examining relationships between FH and QOL.

Clinical Implications

As the cost of cancer care increases and patients and families are required to manage increasing out-of-pocket costs, individuals are experiencing increasing FH and negative outcomes, even when insured, as our sample was. In addition, the paradigm for advanced cancer care continues to evolve from palliative treatment to “make patients comfortable” to include relatively aggressive and costly therapy for an expectation of life extension. To ameliorate suffering and optimize health-related QOL, discussions surrounding cancer cost and FH must begin at diagnosis, be addressed with each treatment decision,¹⁶ and span across the care continuum. Our findings demonstrate the need for improved clinical assessment for and communication about FH, which some suggest may be aided by implementing routine electronic collection of patient-reported measures of FH during clinical encounters.⁴⁹

Oncology and PC clinician discussions about treatment decisions should include FH as a component of the risk-versus-benefit portion of the discussion. Although this should be done at every level of the cancer continuum, with tailoring to the patient’s illness trajectory from time of diagnosis through progression to advanced disease, there is an opportunity to identify this stressor during evaluations in the PC clinic. Further education and training will be needed to guide clinicians in these discussions and to make clinicians aware of available resources. While findings suggest gaps in patient-provider communication, there is also opportunity for interdisciplinary collaboration with financial counselors, navigators, and social workers to address these issues and improve QOL. Published tools exist to screen patients for FH, including a single item on the National Comprehensive Cancer Network Distress Thermometer problem list.⁵⁰ The Oncology Nursing Society’s Nurse Navigator⁵¹ toolkit also includes information and resources to assist nurses and interprofessional team members when assessing for FH and supporting patients who are experiencing common financial concerns. Another resource for communication is the Avalere Health and Robert Wood Johnson Foundation’s Cost Conversation project.⁵²

Research Implications

Findings from this descriptive study advance FH science by filling a critical knowledge gap about FH experienced by patients with advanced cancer. Early studies are testing the efficacy of navigation and pharmacy programs to address FH in cancer,^53–55 but questions remain regarding efficacy, reproducibility, appropriate time of contact, and best content to deliver and who is best to have these conversations.^56,57 Larger, longitudinal studies are needed to fill research gaps about how FH affects QOL in patients with advanced cancer across the disease trajectory, to elicit vulnerable timepoints for intervention, to develop and refine meaningful measures of FH, and to guide policy development to support patients, families, providers, and the healthcare system.

Conclusions

Results indicate that FH is highly prevalent and is negatively associated with QOL in this sample of patients with advanced cancer experiencing symptom burden. Findings highlight a need for continued clinical support and longitudinal research studies to understand experiences and support patients experiencing the cumulative effects of advanced cancer and to assist clinicians and engage systems in meeting these needs.