Economic costs of family caregiving for persons with advanced stage cancer: a longitudinal cohort study

Abstract

Purpose:

To form a multifaceted picture of family caregiver economic costs in advanced cancer.

Methods:

A multi-site cohort study collected prospective longitudinal data from caregivers of patients with advanced solid tumor cancers. Caregiver survey and out-of-pocket (OOP) receipt data were collected biweekly in-person for up to 24 weeks. Economic cost measures attributed to caregiving were as follows: amount of OOP costs, debt accrual, perceived economic situation, and working for pay. Descriptive analysis illustrates economic outcomes over time. Generalized linear mixed effects models asses the association of objective burden and economic outcomes, controlling for subjective burden and other factors. Objective burden is number of activities and instrumental activities of daily living (ADL/IADL) tasks, all caregiving tasks, and amount of time spent caregiving over 24 h.

Results:

One hundred ninety-eight caregivers, 41% identifying as Black, were followed for a mean period of 16 weeks. Median 2-week out-of-pocket costs were $111. One-third of caregivers incurred debt to care for the patient and 24% reported being in an adverse economic situation. Whereas 49.5% reported working at study visit 1, 28.6% of caregivers at the last study visit reported working. In adjusted analysis, a higher number of caregiving tasks overall and ADL/IADL tasks specifically were associated with lower out-of-pocket expenses, a lower likelihood of working, and a higher likelihood of incurring debt and reporting an adverse economic situation.

Conclusions:

Most caregivers of cancer patients with advanced stage disease experienced direct and indirect economic costs.

Implications for Cancer Survivors:

Results support the need to find solutions to lessen economic costs for caregivers of persons with advanced cancer.

1. INTRODUCTION

Caring for a family member or friend can bring challenges for the estimated 2.8 million caregivers in the USA who provide unpaid assistance in the home for persons with cancer [1]. Over time, caregiving demands can change and the experience of being a caregiver may depend on whether the care recipient is in an active treatment phase, survivorship phase, or approaching death (Van Houtven, Ramsey [2]). Overall, we know surprisingly little about the economic costs experienced by the caregivers of persons with advanced cancer. Whereas high subjective burden, defined as stress and role strain, has been documented for caregivers of cancer patients [3], less is known about their objective burden, that is, the time and intensity of caregiving provided and its relationship to caregiver economic costs.

Even though persons with advanced cancer may not require caregiving for a long period of time given high mortality rates in advanced cancer, the caregiving provided will likely be intensive depending on phase of treatment and whether palliative and other supportive care is able to be accessed [4, 5]. Caregivers may face high expenditures for care, depletion of savings including retirement savings, and significant adjustments to their work lives. It is well known that end-of-life care is often more expensive than at any other life stage and in the case of cancer, dying in a hospital has been shown to be two times more expensive in the past 30 days of life than dying at home ($20 k versus $10 k) [6], which means caregivers of patients who do die may share some of these end-of-life costs. As such, economic costs for caregivers of persons with advanced cancer may have long-lasting effects on economic security long after caregiving activities are complete because of high costs of care and because some caregivers do not return to work after caregiving [7]. Retiring early contributes to economic insecurity in old age, and individuals who work at least through their mid-60 s are more protected from poverty in old age than those who stop working prior to their mid-60 s [8]. Working age caregivers (e.g., those under 65), therefore, are particularly at risk of economic insecurity in old age if they leave the labor force earlier than planned. Quantifying such costs can guide supportive policies for caregivers.

The purpose of this study was (1) to describe multiple dimensions of economic costs for caregivers of persons with advanced cancer and short-term changes in costs, and (2) to estimate the association between objective burden (number of up to ten activities of daily living and instrumental activities of daily living tasks over 24 h, number of up to 24 caregiving tasks over 24 h, and hours of care over 24 h) and economic costs over 24 weeks of time. Models control for caregiver subjective burden and other factors expected to be associated with caregiver economic costs. The overarching hypothesis is that higher objective burden will be associated with higher economic costs to caregivers.

2. BACKGROUND AND CONCEPTUAL MODEL

Background

Cancer patients experience economic costs as a result of cancer treatment, what many researchers detail as “financial toxicity” [9]. Compared to other high-income countries, systematic reviews indicate that patients in the USA are the most exposed to the economic costs of cancer care [10, 11]. In one population-based cohort study, nearly half of patients reported financial difficulties [9] and in another, median monthly out-of-pocket (OOP) costs were nearly $600 [12]. Cancer patients have used savings, reduced costs on leisure activities, and taken on debt to meet the financial demands of cancer [13]. While being uninsured or underinsured is associated with higher economic costs [14], there is strong evidence that even among insured persons with advanced cancer, positive economic costs exist for patients.

Compared to cancer patients, we know much less about the economic costs incurred by caregivers. Often times, caregivers and the patient have shared household budgets, and as such, patient costs of cancer care also affect caregivers. In other situations, caregivers have separate finances from the patient, and yet, they too incur costs in their caregiver role. Yet most studies lack clarity on which costs are incurred by the caregiver and which are incurred by the patient. In this work, we include both caregiver costs and patient costs related to cancer care and we distinguish between direct and indirect economic costs of caregiving, as we know that costs are multidimensional. Direct costs include OOP expenditures derived from the caregiving role, such as helping with patient copays, drug costs or transportation costs for the patient, or health care expenditures resulting from caregiving, such as to address mental health consequences or physical injuries from caregiving. Indirect costs may include downstream health care costs, or work changes due to caregiving, such as stopping work (Van Houtven et al., 2010), or due to caregivers delaying their own treatments due to focusing on the patient [15]. Combined, the direct and indirect costs of caregiving can affect how a caregiver feels he/she is doing financially, called perceived financial strain. Perceived financial strain could create negative caregiver well-being even when direct economic costs are low and there is some evidence that perceived financial strain is associated with caregiver anxiety and depression [16, 17].

Past literature on economic costs of caregiving is sparse and somewhat dated, and most does not focus on caregivers of persons with advanced cancers. We briefly review other cancer caregiving literature to profile direct and indirect costs. OOP costs for caregivers of lung and colorectal cancer patients averaged $1243 over about a 6-month period, with variation by whether patients were in an active treatment phase or not [2]. In an international scoping review of 19 studies in Canada, the USA, and Europe, the mean estimated OOP cost was CA$447 per month for cancer caregivers (range: CA$25– CA$1223), depending on the type of out-of-pocket costs included (e.g., travel expenses, formal home support, medication) [18]. Regarding indirect costs to caregivers of cancer patients, one US-based study found that family members of cancer patients were more likely to delay or forego medical care [15]. In another study based in the USA, the estimated mean annual value of caregiver time was $47,710 [19, 20].

Furthermore, caregivers face a decline in their work productivity. In a cross-sectional study in the USA, about 10% of employed caregivers of patients with advanced cancer reported a loss in work hours (absenteeism) over 15 months; 15% reported impairment while at work due to caregiving (presenteeism) [21]. In a new study 30 days post-pancreatectomy, nearly 60% of working caregivers expressed having work difficulty due to caregiving tasks while working, resulting in substantial losses in work productivity over this short time period [22]. A systematic review study indicates the estimated work productivity loss as a combination of absenteeism and presenteeism due to caregiving was 21–27% [23].

Economic costs to caregivers differ across important domains, such as by race, economic status, and degree of work flexibility. In a population-based cohort study, most (73.8%) caregivers of Black cancer survivors made some employment change ranging from unpaid time off work and making schedule, hours, or duty changes [16]. In a majority White caregiver participant study in the USA, nearly a quarter of caregivers of cancer survivors made extended employment changes, such as time off, modified work duty/hours, and lost opportunities [20]. We expect differences by race to exist in the USA based on persistent policies that reduce economic stability among African Americans (e.g., red-lining, credit constraints, lower savings), and due to other consequences of systemic racism.

Total economic costs comprise both measurable indirect and direct costs. In a study of caregivers for colorectal and lung cancer patients, in the initial phase of treatment, total economic costs were $7000, $19,000 in continuing phase, and $14,000 in terminal phase [2].

Conceptual Framework

A guiding conceptual framework (Fig. 1) helps frame how caregiving might affect economic costs and comes from a previously published adapted version of Pearlin’s Stress Process Coping model [24]. This adapted model posits that economic cost outcomes will be worse for high objective burden caregivers compared to low objective burden caregivers, supported by the evidence that intensive caregiving affects work and mental health more than for non-intensive caregivers (Jacobs et al., 2014; Kolodziej, Coe and Van Houtven, 2022). Furthermore, these negative consequences can arise from multiple pathways associated with intensive caregiving, such as by experiencing greater work productivity losses, and/or due to having disadvantageous baseline factors which exacerbate pathways by which caregiving has negative economic consequences, such as poor initial economic security or exposure to structural racism.

Figure 1.

Caregiver Stress Process Model and Caregiver Economic Cost Outcomes.

Note: Model adapted from Kolodziej, Coe, Van Houtven 2022, which was adapted based on Robinson et al, 2020 version of the Pearlin Stress Process Coping Model, Economic cost outcomes in italics are not measured in this study.

3. MATERIALS AND METHODS

Overview

This multi-site cohort study employed a mixed-methods, prospective, longitudinal design and followed 223 caregiver-patient pairs (dyads) for up to 6 months or until 1 month after patient death, whichever happened first. Data was collected biweekly and in-person from caregivers in their homes using qualitative interviews, quantitative surveys, and monthly structured observation. Caregivers completed more detailed interviews and patients a brief survey. All dyads participated in baseline (T1) interview, i.e., an initial observation, and up to 11 subsequent waves of data collection. OOP cost data collection began at the second visit (T2). Figure 2 illustrates the data collection time points and content for the longitudinal economic outcomes.

Figure 2.

Study data collection time points and economic outcomes content

NOTES: We excluded final interview of bereaved caregivers of patients who died during the study period as the final interview was conducted during varying amounts of time since the last biweekly interview completed while the final interview of caregivers of patients who survived the study period was 2 weeks after the second to last biweekly interview.

3.1. Participants and Recruitment

Eligible cases were the primary caregivers 25 years of age or more, of adult home-dwelling patients with metastatic (stage IV disease) solid tumors for which the median overall survival was < 12 months and who had failed first-line therapy. A primary caregiver was defined as an individual who takes on the responsibility for assisting and meeting the patient’s daily needs and who is not paid for performing this role. Participation by both caregiver and patient was an eligibility criterion for the study. All participants provided written informed consent. The study was approved by the institutional review boards for human subjects (Temple IRB #22776; VCU IRB: HM20003319).

Patients were identified through the electronic medical record and case conferences from 3 oncology clinic sites in Virginia and Pennsylvania. Patient oncologists were contacted to confirm eligibility and to request permission to contact the patient. Patients were screened by telephone to confirm eligibility including for the presence and willingness of a primary caregiver. Additionally, caregivers of eligible patients in hospice care were screened directly working with the Visiting Nurses’ Association of Philadelphia. A HIPAA waiver of authorization was obtained for eligibility identification purposes. Of the 875 patients contacted, approximately 23.9% declined participation, 50.3% were ineligible, and 25.4% provided informed consent (223 dyads). This paper includes 198 individuals who completed the OOP cost data (at least T1 and T2 interviews). In the analytic sample, the mean and median number of biweekly interviews were 7.5 and 8, respectively.

3.2. Data Collection

Patient and caregiver sociodemographic data were obtained directly from the participants. If a patient became too sick to complete their survey, caregivers did as proxies. At each time point, quantitative measures were verbally administered by the research assistants (RAs) to both patient and caregiver who each were interviewed privately. Patients received honoraria in the form of cash or visa gift card at each visit.

Caregiver Virtual Data Collection Protocol

In March 2020, IRB modifications were submitted and approved to continue data collection entirely by phone due to the COVID-19 pandemic. As such, initial and all biweekly surveys and interviews were conducted by phone after verbal consent was obtained from each participant.

3.3. Outcome Measures.

There are up to 11 observations for each of the economic outcomes because out-of-pocket (OOP) costs started at the second visit. Study visits occurred over a period of approximately 24 weeks or every 2 weeks. We excluded the economic outcomes from the final interview of caregivers of patients who died during the study period because of the imbalanced recall period by group. Specifically, the final interview for bereaved caregivers was conducted a minimum of 1 month after a patient’s death whereas the final interview for caregivers of patients who survived the study period was 2 weeks after the last biweekly interview. Economic outcomes are measured at the caregiver-visit level. There are two measures of direct economic costs: biweekly OOP costs attributed to caregiving and whether a caregiver took on debt ever and how much [25]. There are two measures of indirect economic costs: working for pay or not and experiencing an adverse economic situation or not.

Out-of-Pocket Expenses.

Most studies ask participants to recall expenditures, which is burdensome and results in underestimation of expenses [26]. The approach for this study was to give a medium-sized box to caregivers and patients to place receipts and bills related to the patient’s care, the caregiver’s own health care utilization, and supplementary services that assist the caregiver (e.g., house cleaning, child care) with instructions provided on allowable expenses. During the biweekly meetings, an RA recorded each receipt in a spreadsheet and asked the caregiver to detail the reason for the expense (e.g., prescriptions, equipment, house cleaning, respite care). OOP expenses had to be attributed to the caregiving role and included events, such as caregivers buying take-out foods because of long duration of patient’s medical appointment or getting a pedicure as an act of self-care and decompression from caregiving. Participants also were asked about any additional receipts/bills that were not in the box and documented. If a participant reported items without documentation, such as physician visits but no corresponding co-pays, participants were queried, their explanation documented, and, if appropriate, estimates were made based on previous records (e.g., standard co-pay from previous visits). Mileage for any driving related to caregiving, such as for patient appointments, was recorded, and cost calculated using the IRS-determined rate at the time ($0.535–$0.58/mile). OOP costs per visit reflected about 2 weeks of expenditures and total OOP expenditures for each 2-week period were calculated.

Debt due to Caregiving.

In baseline and biweekly surveys, caregivers were asked: “In the past two weeks, have you taken out any of the following types of loans to help you pay for the cost of care for [patient name]? If so, please provide your best estimate of how much you borrowed for each category.” Caregivers then detailed whether and how much (when yes) they charged on their credit card; took a home equity loan, mortgage, or cash-out refinance; took a loan from family or friends; took a bank loan, payday loan, or loan from employer; and finally, incurred any other loans/debt (and were asked to specify what kind) [27]. In this paper, we coded as “1” caregivers who reported accruing any debt and “0” otherwise.

Adverse Economic Situation.

At the baseline and biweekly surveys, caregivers were asked: “Which one of the following statements best describes your own personal economic situation?” Caregivers chose from the following statements: (1) I am in good shape, financially. I am able to save; (2) I am okay. I am saving a little and I am able to provide for my needs; (3) I am barely getting by. I have to budget carefully and I am not able to plan for the future; (4) I am falling behind. I have to use savings or go further into debt to pay my bills; or (5) I am in serious financial trouble and can’t quite see how I am going to make it. We dichotomized caregiver responses at each visit as “1” if they reported being in an adverse financial situation, that is, they reported barely getting by, falling behind, or in serious financial trouble. Those in good or okay shape were coded as “0.”

Work for Pay.

Participants were asked at the baseline survey visit “Did you do any work for pay?” If a caregiver reported working full-time (35 or more hours a week for pay) or part-time (less than 35 h a week for pay), they were considered to be working for pay. Subsequently, in the biweekly surveys, caregivers were also asked, “Since we last talked, did you do any work for pay?” If they answered no, they were coded as “0” for that study visit for “not working” and “1” for yes “working.”

3.4. Primary Explanatory Variables.

We examine three measures of objective caregiver burden measured at baseline looking back over a 24-h period: (1) the number of up to ten ADL/IADL tasks provided (see Table 1) [28, 29]; (2) the number of all caregiving tasks provided, using a list developed for this project; and (3) the time spent caregiving. The number of ADLs ranged from 0 to 4, where a higher score indicates a higher objective burden of care. ADLs included helping the patient with dressing or bathing; walking; toileting/incontinence/diaper care; or moving in or from a bed or chair. The number of IADLs ranged from 0 to 6, where a higher score indicates a higher objective burden of care. IADLs included preparing and/or cleaning up meals; doing housework; doing laundry; working on yard or house maintenance; shopping for groceries or other necessities; and providing transportation. For (2), ADLs, IADLS, and all other tasks were included in the “all caregiving tasks” measure of objective burden (Table 1). Specifically, caregivers were asked about 24 items from a list of other common tasks that cancer caregivers perform, such as arranging medical care visits for the patient, time spent on billing problems, or administering medication by injection, IV, or pain medication. Finally, caregivers could fill in other caregiving tasks not named. Time spent providing care encompassed time spent on activities detailed from this longer list of tasks (Table 1).

Table 1.

Activities Related to Caregiving “in an average day in your life as a caregiver,” as reported at the baseline survey.

Prepare and/or clean up meals**	Take care of wounds or bed sores
Do housework**	Sit with [PATIENT] alone (watching TV, reading, talking)
Do laundry**	Sit with [PATIENT] and other family/friends
Provide transportation**	Spend time on the telephone with insurance company
Go with [PATIENT] to an appointment	Arrange appointments, medical care, visitors, or supplies, etc.
Pay medical bills	Spend time on the phone with billing department (hospital, doctor’s office)
Take care of a feeding machine, catheter, colostomy	Help [PATIENT] with dressing or bathing**
Hook up or unhooked an IV	Help [PATIENT] with walking**
Give oral medications (except pain medication)	Help [PATIENT] with toileting/incontinence/diaper care**
Give medicine by injection	Help [PATIENT] with moving in or from bed or chair**
Give pain medication	Help [PATIENT] with problem symptoms or side effects
	Any other caregiving tasks or activities (please describe)

Note: The most common “other” caregiving tasks (included in the time measure) described included reminding the patient of appointments and tasks, regular “check-ins”, phone-calls and visits to assess their needs, and assisting the patient with their non-medical finances.

We also examined caregiver subjective burden using the Zarit Caregiver Subjective Burden 12-item measure [30]. The Zarit Caregiver Subjective Burden scale yields a score ranging from 0 to 48, where a higher score indicates higher burden (15 or higher is considered significant burden).

3.5. Demographics included caregiver demographics (age [in years], sex [male, female], race [Black, White, other], education [high school diploma/GED compared to greater than high school]), caregiver self-reported health, whether the caregiver lives with the patient, caregiver work status prior to becoming a caregiver, patient demographics (age, sex, race), and patient’s insurance type (non-mutually exclusive categories: employer based, Marketplace, Medicare, Medicaid, other public insurance, any other insurance). All demographics were measured at the baseline visit (T1).

4. STATISTICAL ANALYSIS

4.1. Primary Analysis

We used generalized linear mixed effects models (GLMM) to examine associations between caregiver objective burden and the economic cost outcomes over time, controlling for subjective burden [31, 32]. The analysis was based on the first 11 visits of quantitative data in the models in which the participants reported outcomes over the prior 2 weeks.

Each model adjusts for relevant caregiver and patient characteristics and includes caregiver-level random effects to account for the correlation between a caregiver’s repeated measures over time. Equation 1 shows the general mean structure of the GLMM used to examine each outcome. We used general linear mixed models for all outcomes except to model out-of-pocket expenditures and incurring any debt, for which we used a Gaussian distribution with log link and a binomial distribution with logit link, respectively.

$$Y_{ij}={\beta }_0+{\beta }_1{W}_i+{\beta }_2{Z}_i+\tau T^{\mathrm{\text{'}}}+\phi C^{\mathrm{\text{'}}}$$ Equation 1.

Y_ij represents each of the outcomes for caregiver I at time j; W_ij represents the three measures of objective burden at baseline (each modeled separately) for caregiver I ; Z_i represents the Zarit Subjective Burden score at baseline; T represents a vector of indicators for each of the measurement collection time points, as well as baseline caregiver demographics (age, sex, race, education), baseline self-reported health, whether the caregiver coresides with the patient, whether the caregiver worked prior to becoming a caregiver (with the exception of the not working outcome), baseline patient demographics (age, sex, race) and patient’s baseline insurance status. Mixed effects model parameters were estimated and tested using MEGLM (Stata). We calculated the marginal effects of each of the measures of burden. When calculating the marginal effects from the generalized linear mixed models, the average marginal effects were derived from marginal probabilities and conditional on the random effect. With four economic outcomes and 3 objective burden measures, we ran 12 models total. For the out-of-pocket expenditures, we excluded one extreme outlier observation from the descriptive statistics and all analysis but present the results including the outlier in the Appendix. We also ran a sensitivity analysis including the final study visit for those caregivers whose patient had died, with a longer recall period between the second to last and last study visit to check consistency of the main results to this exclusion (results in Appendix).

5. RESULTS

Table 2 shows descriptive statistics of the 198 caregivers overall and then separated by higher compared to lower caregiving intensity (for each objective burden measure). The caregiver mean age is 56 years old, 74% female, 41% Black or African American, 66% had completed some education beyond high school, 25% report baseline health status of fair/poor, and 65% live with the patient. Nearly 86% of caregivers report feeling pretty well/very well prepared to be a caregiver. The average Zarit Subjective Burden score is 11.8. Prior to caregiving, 53% of caregivers were working full-time and 20% were working part-time. Since caregiving, percent of caregivers working full-time and percent working part-time decreased (from 53 to 33% and 20 to 17%, respectively). On average, the patients are 61 years old, 51% male, 41% Black or African American, 41% with Medicare, 32% with Medicaid, 30% with employer-based health insurance, and 6% without any health insurance.

Table 2.

Baseline Characteristics of Caregiver and Patient Participants, Overall and by Median Baseline Objective Burden

	Overall Sample	Number of ADL/IADL Tasks			Number of All Caregiving Tasks			Time Spent Helping
Factor		Below Median	At or Above Median	p-value	Below Median	At or Above Median	p-value	Below Median	At or Above Median	p-value
N	N=198	N=81	N=117		N=87	N=111		N=98	N=100
Caregiver Demographics, N (%)
Age, mean (SD)	55.88 (14.01)	54.50 (14.60)	56.82 (13.57)	0.25	54.67 (15.18)	56.81 (13.03)	0.29	54.09 (15.15)	57.61 (12.64)	0.078
Gender				0.96			0.89			0.47
Male	51 (25.76%)	21 (25.93%)	30 (25.64%)		22 (25.29%)	29 (26.13%)		23 (23.47%)	28 (28.00%)
Female	147 (74.24%)	60 (74.07%)	87 (74.36%)		65 (74.71%)	82 (73.87%)		75 (76.53%)	72 (72.00%)
Race				0.60			0.27			0.37
White	107 (54.04%)	42 (51.85%)	65 (55.56%)		47 (54.02%)	60 (54.05%)		52 (53.06%)	55 (55.00%)
Black	81 (40.91%)	36 (44.44%)	45 (38.46%)		38 (43.68%)	43 (38.74%)		43 (43.88%)	38 (38.00%)
Other	10 (5.05%)	3 (3.70%)	7 (5.98%)		2 (2.30%)	8 (7.21%)		3 (3.06%)	7 (7.00%)
Education				0.54			0.76			0.62
High School Diploma/GED	66 (33.33%)	25 (30.86%)	41 (35.04%)		30 (34.48%)	36 (32.43%)		31 (31.63%)	35 (35.00%)
Greater than High School Diploma/GED	132 (66.67%)	56 (69.14%)	76 (64.96%)		57 (65.52%)	75 (67.57%)		67 (68.37%)	65 (65.00%)
Health Status: Fair/Poor	49 (24.75%)	18 (22.22%)	31 (26.50%)	0.49	24 (27.59%)	25 (22.52%)	0.41	21 (21.43%)	28 (28.00%)	0.28
Co-residing	129 (65.15%)	47 (58.02%)	82 (70.09%)	0.080	51 (58.62%)	78 (70.27%)	0.088	55 (56.12%)	74 (74.00%)	0.008
Caregiver Activities and Preparedness
Work Status				0.32			0.93			0.052
Not working	97 (49.49%)	35 (43.21%)	62 (53.91%)		42 (48.28%)	55 (50.46%)		40 (40.82%)	57 (58.16%)
Full Time Work	65 (33.16%)	31 (38.27%)	34 (29.57%)		29 (33.33%)	36 (33.03%)		38 (38.78%)	27 (27.55%)
Part Time Work	34 (17.35%)	15 (18.52%)	19 (16.52%)		16 (18.39%)	18 (16.51%)		20 (20.41%)	14 (14.29%)
Pre-caregiving Work Status				0.20			0.22			0.96
Not working	54 (27.55%)	24 (30.00%)	30 (25.86%)		29 (33.72%)	25 (22.73%)		26 (26.80%)	28 (28.28%)
Full Time Work	103 (52.55%)	45 (56.25%)	58 (50.00%)		42 (48.84%)	61 (55.45%)		51 (52.58%)	52 (52.53%)
Part Time Work	39 (19.90%)	11 (13.75%)	28 (24.14%)		15 (17.44%)	24 (21.82%)		20 (20.62%)	19 (19.19%)
How Prepared to be a caregiver				0.80			0.51			0.20
Not at all/Not too well prepared	6 (3.03%)	3 (3.70%)	3 (2.56%)		4 (4.60%)	2 (1.80%)		5 (5.10%)	1 (1.00%)
Somewhat well prepared	22 (11.11%)	10 (12.35%)	12 (10.26%)		10 (11.49%)	12 (10.81%)		12 (12.24%)	10 (10.00%)
Pretty well/Very well prepared	170 (85.86%)	68 (83.95%)	102 (87.18%)		73 (83.91%)	97 (87.39%)		81 (82.65%)	89 (89.00%)
Caregiver Burden
Number of all tasks	7.59 (3.77)	4.38 (1.93)	9.81 (3.06)	<0.001	4.23 (1.57)	10.23 (2.75)	<0.001	5.44 (2.70)	9.70 (3.47)	<0.001
Number of ADL/IADL tasks assist with, mean (SD)	4.15 (2.30)	1.95 (1.08)	5.68 (1.57)	<0.001	2.31 (1.39)	5.59 (1.79)	<0.001	2.95 (1.82)	5.33 (2.11)	<0.001
Time Spent Caregiving, mean (SD)	10.21 (7.11)	6.05 (4.70)	13.09 (7.09)	<0.001	6.08 (4.64)	13.45 (7.05)	<0.001	4.71 (2.29)	15.60 (6.02)	<0.001
Zarit Subjective Burden, mean (SD)	11.81 (7.62)	10.12 (6.54)	12.98 (8.10)	0.009	10.91 (7.52)	12.51 (7.65)	0.14	11.99 (7.77)	11.62 (7.50)	0.73
Number of biweekly study visits, median	9	10	9					9	9
Patient Demographics
Age, mean (SD)	61.27 (11.60)	60.63 (11.84)	61.71 (11.47)	0.52	60.67 (11.67)	61.74 (11.58)	0.52	61.84 (11.18)	60.71 (12.03)	0.50
Gender				0.40			0.76			0.67
Male	100 (50.51%)	38 (46.91%)	62 (52.99%)		45 (51.72%)	55 (49.55%)		51 (52.04%)	49 (49.00%)
Female	98 (49.49%)	43 (53.09%)	55 (47.01%)		42 (48.28%)	56 (50.45%)		47 (47.96%)	51 (51.00%)
Race				0.43			0.15			0.050
White	101 (51.27%)	41 (50.62%)	60 (51.72%)		47 (54.02%)	54 (49.09%)		51 (52.04%)	50 (50.51%)
Black	81 (41.12%)	36 (44.44%)	45 (38.79%)		37 (42.53%)	44 (40.00%)		44 (44.90%)	37 (37.37%)
Other	15 (7.61%)	4 (4.94%)	11 (9.48%)		3 (3.45%)	12 (10.91%)		3 (3.06%)	12 (12.12%)
Health Insurance Status
Employer Insurance	60 (30.30%)	30 (37.04%)	30 (25.64%)	0.086	34 (39.08%)	26 (23.42%)	0.017	33 (33.67%)	27 (27.00%)	0.31
Marketplace Insurance	10 (5.05%)	2 (2.47%)	8 (6.84%)	0.17	1 (1.15%)	9 (8.11%)	0.026	5 (5.10%)	5 (5.00%)	0.97
Medicare Insurance	83 (41.92%)	31 (38.27%)	52 (44.44%)	0.39	32 (36.78%)	51 (45.95%)	0.19	39 (39.80%)	44 (44.00%)	0.55
Medicaid Insurance	64 (32.32%)	26 (32.10%)	38 (32.48%)	0.96	27 (31.03%)	37 (33.33%)	0.73	32 (32.65%)	32 (32.00%)	0.92
Other Public Insurance	7 (3.54%)	1 (1.23%)	6 (5.13%)	0.14	1 (1.15%)	6 (5.41%)	0.11	2 (2.04%)	5 (5.00%)	0.26
Any Other Insurance	17 (8.59%)	4 (4.94%)	13 (11.11%)	0.13	5 (5.75%)	12 (10.81%)	0.21	8 (8.16%)	9 (9.00%)	0.83
No Insurance	12 (6.06%)	4 (4.94%)	8 (6.84%)	0.58	4 (4.60%)	8 (7.21%)	0.45	4 (4.08%)	8 (8.00%)	0.25

Note: The groups are not balanced by a median split because there were multiple people exactly “at” the median. Baseline is defined as the first time a caregiver and patient filled out a biweekly survey, or visit 1 of the study. For significance tests, all continuous variables were compared using ANOVA while binary or categorical variables were compared using Person’s chi-squared test. The median values for the number of ADL/IADL tasks, number of all caregiving tasks, and time spent helping are 4, 7, and 8.5 hours, respectively.

Regarding objective burden, the average number of ADL/IADL tasks caregivers helped with at baseline was 4.2 (out of 10), the average number of total tasks caregivers helped with out of 24 tasks possible at baseline was 7.6, and the amount of time spent providing care was approximately 10.2 h in a 24-h period on caregiving-related tasks. Average hours changed by no more than 1–2 h a week on average over time.

When examining caregivers by the intensity of objective caregiver burden across the three measures, we observe similar patterns. A greater proportion of caregivers at or above median objective burden are male, White or report other race, have a high school diploma/GED as highest level of education, have health status of fair/poor, and feel pretty well/very well prepared. Notably, a higher proportion of caregivers who live with the patient report providing care for the median or greater value of two of the measures of objective burden used: completing ADLs or IADLs and time spent caregiving. A higher proportion of Black caregivers were in the lower intensity group (e.g., below median) compared to the higher intensity group whereas White caregivers were evenly split across the intensity groups.

Table 3 shows that average economic cost outcomes varied across study visits and illustrates that study attrition was steep due to patient death (from 198 to 98 dyads at visit 12). Median out-of-pocket costs per visit were $54 and $111 conditional on having any out-of-pocket costs. Costs differed by duration in the study. Average 2-week out-of-pocket expenditures were $185.05 for caregivers who remained in the study for 8 + visits and $132.44 for those with only 2–3 visits. Among all caregivers, 32% reported ever taking on some debt to manage their caregiving duty. Those with debt reported an average of $7980 in debt, and 54% of caregivers reported ever being in an adverse economic situation. Finally, whereas just under 50% of caregivers reported working for pay at the start of the study, at the final interview, less than one-third reported working for pay (Table 3).

Table 3.

Caregiver Unadjusted Outcomes over the Biweekly and Final Interview (Visit 11) and Ever (at least one instance over visit 1–11)

	Biweekly Interviews											Ever
	1	2	3	4	5	6	7	8	9	10	11
	N=198	N=180	N=159	N=143	N=131	N=120	N=111	N=96	N=91	N=87	N=98	N=198
Out-of-pocket Expenditures, mean (SD)	189.33 (457.08)	221.56 (484.41)	167.16 (311.31)	195.65 (492.35)	224.77 (672.79)	164.57 (492.34)	157.84 (364.13)	149.44 (312.70)	225.68 (356.52)	204.49 (429.74)	172.85 (447.49)
Out-of-pocket Expenditures, conditional on any	241.85 (504.45)	297.62 (541.30)	219.66 (340.55)	266.46 (558.51)	303.55 (767.33)	240.84 (580.97)	206.12 (404.43)	202.06 (349.17)	293.38 (381.63)	265.53 (473.49)	364.49 (596.48)
Took on Any Debt, %	26 (13.13%)	22 (12.22%)	27 (16.98%)	23 (16.08%)	14 (10.69%)	16 (13.33%)	18 (16.22%)	14 (14.58%)	12 (13.19%)	16 (18.39%)	20 (20.41%)	64 (32.32%)
Amount of Debt, conditional on any	1575.12 (4061.62)	7229.78 (21399.65)	17018.23 (76677.02)	1286.76 (3557.39)	25443.62 (82607.87)	9314.07 (26523.50)	3214.84 (9916.88)	5264.62 (16465.36)	356.09 (422.87)	3085.12 (7845.67)	6724.16 (22433.00)
Economic Situation, %
I am in good shape/okay	110 55.84%	95 53.37%	90 56.60%	78 54.93%	77 58.78%	69 57.98%	65 59.09%	57 60.64%	55 60.44%	55 63.22%	59 60.20%
I am barely getting by	59 29.95%	58 32.58%	52 32.70%	45 31.69%	35 26.72%	35 29.41%	31 28.18%	26 27.66%	26 28.57%	22 25.29%	22 22.45%
I am falling behind/in  serious financial trouble	28 14.21%	25 14.04%	17 10.69%	19 13.38%	19 14.50%	15 12.61%	14 12.73%	11 11.70%	10 10.99%	10 11.49%	17 17.35%
Work for Pay, %	98 49.49%	84 46.67%	76 47.80%	68 47.55%	63 48.09%	57 47.50%	52 46.85%	44 45.83%	43 47.25%	41 47.13%	28 28.57%	57.07%

Note: Biweekly interview #1 reported here is the second study visit (as OOP costs were not collected at the first visit). Final Biweekly interview #11 corresponds to the 12^th total visit, that is the final interview for caregivers of care recipients who completed the study period (e.g., 12 visits were completed). A final interview also would occur due to patient death, patient moved into professional caregiving situation, or a caregiver change after initial interview, as well as 12 visits See Figure 2.

Table 4 shows the estimated marginal effects of objective and subjective burden on all caregiver outcomes. Objective burden was measured at baseline and the economic outcomes were defined at the per-visit level, from visit 1 to visit 12 (visit 2 to 12 for OOP costs). The top panel of the table shows a model of objective burden measured as a count of ADL and IADL tasks provided over a 24-h period. The middle panel shows a model that measures objective burden as all reported caregiving tasks provided over a 24-h period (see Table 1 for details). The bottom panel shows a model that measures objective burden as hours of caregiving provided in a 24-h period. Full model results appear in the Appendix.

Table 4.

Marginal Effects of Objective and Subjective Burden on Economic Cost of Caregiving Outcomes

	(1)	(2)	(3)	(4)
	Out-of-Pocket Expenses	Incurring any Debt	Caregiver Worked for Pay	Adverse Economic Situation^
Baseline Number of ADL/IADL Tasks Provided	−27.224 ***	0.006	−0.023 *	0.019
	(6.436)	(0.008)	(0.010)	(0.014)
Baseline Zarit Subjective Burden	7.791 ***	0.002	−0.002	0.020 ***
	(1.654)	(0.003)	(0.003)	(0.005)
Baseline Number of Total Tasks Provided##	−16.920 ***	0.013 **	−0.015 *	0.017 *
	(3.818)	(0.005)	(0.006)	(0.008)
Baseline Zarit Subjective Burden	9.317 ***	0.002	−0.002	0.020 ***
	(1.883)	(0.003)	(0.003)	(0.005)
Baseline Time Spent Caregiving	−1.119	0.006 **	−0.008 ***	0.012 **
	(2.270)	(0.002)	(0.003)	(0.004)
Baseline Zarit Subjective Burden	3.130	0.003	−0.003	0.022 ***
	(2.099)	(0.003)	(0.003)	(0.005)
N	1367	1368	1368	1360

^†p< 0.10

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

^{^}Adverse economic situation is defined as reporting “barely getting by” or “falling behind/in serious financial trouble” compared to being in “good or okay shape”

^##Total caregiver tasks are shown in Table 1.

Notes: Covariates include baseline caregiver demographics (age, sex, race, education), self-reported health, caregiver lives with the patient, and caregiver worked prior to becoming a caregiver (with the exception of the stopped working outcome); baseline patient demographics (age, sex, race), patient’s insurance status and time indicators.

First, considering the number of ADL/IADL tasks caregivers provided, each additional ADL/IADL task is associated with a $27.22 decrease in out-of-pocket expenses (p < 0.001) and a 2.3 percentage point decrease in probability of a caregiver working for pay (p < 0.05). We also find evidence that every one unit increase in caregiver baseline Zarit Subjective Burden score is associated with $7.79 increase in out-of-pocket expenses (p < 0.05) and a 2.0 percentage point increase in probability of reporting an adverse economic situation (p < 0.001), controlling for all other variables in the model.

Second, considering the number of all caregiving tasks provided, we find that each additional task is associated with a $16.92 decrease in out-of-pocket expenditures (p<0.01); a 1.3 percentage point increase in probability of incurring any debt (p<0.01); 1.5 percentage point decrease in probability of a caregiver working for pay (p<0.05); and, 1.7 percentage point increase in probability of reporting an adverse economic situation (p < 0.05). We also find evidence that every one unit increase in caregiver baseline Zarit Subjective Burden score is associated with a $9.32 increase in out-of-pocket expenditures (p < 0.01) and a 2.0 percentage point increase in probability of reporting an adverse economic situation (p < 0.05), controlling for all other variables in the model.

Third, considering the time spent providing care, each hour of care is associated with a 0.6 percentage point increase in probability of incurring debt (p < 0.01); 0.8 percentage point decrease in probability of working for pay (p < 0.0001); and, a 1.2 percentage point increase in probability of reporting an adverse economic situation (p < 0.01). We also find evidence that every one unit increase in caregiver baseline Zarit Subjective Burden score is associated with a 2.2 percentage point increase in probability of reporting an adverse economic situation (p < 0.001), controlling for all other variables in the model.

Notably, after controlling for objective burden capturing tasks requiring caregiving assistance, caregiver baseline Zarit Subjective Burden score was associated with increased out-of-pocket expenses. Finally, we observe some differences in economic outcomes by race. Caregivers of “other” race were less likely to be working for pay compared to White caregivers across all three models (see Appendix Tables A1, A2, A3).

Finally, regarding sensitivity analyses, OOP model results are consistent when the extreme outlier is included (Appendix Table A4), and including the final study visit for those caregivers whose patient died does not change the results from the main results presented in Table 4 (please see Appendix Table A5).

6. DISCUSSION

Both Yabroff and Kim [19] and Girgis and Lambert [33] have stated that without examination of the costs to caregivers of patients with cancer, we risk developing interventions that add to caregivers’ overall burden. This study is the first to prospectively and longitudinally examine a comprehensive set of economic costs to US caregivers in the context of advanced cancer.

The majority of caregiver participants incurred economic costs due to caregiving. Moreover, these costs remained steady and persistent in the (up to) 6-month study period. In adjusted models, the relationship of objective burden and economic outcomes varied by the specific measure of objective burden used. Time as the measure of objective burden had a statistically significant and negative association with three different economic cost outcomes at the p < 0.05 level (incurring any debt, being in an adverse economic situation, and a lower chance of working for pay). Number of total caregiving tasks was statistically significant and negatively associated with incurring any debt and out-of-pocket expenses at the p < 0.05 level; and finally, the number of ADL/IADL tasks was negatively associated with out-of-pocket expenses at the p < 0.05 level and working for pay the 10% level. Thus, for researchers interested in the relationship between objective burden and caregiver economic costs, time intensity may be the most closely related objective burden measure. In addition, we found that subjective burden had its own independent relationship with caregiver economic costs. A higher Zarit Burden score (indicating higher strain) was consistently associated with higher out-of-pocket expenses and having an adverse economic situation. Black caregivers did not have differential work changes, as found in past literature [16], although caregivers of “other” race were less likely to work.

This study has limitations. Being a cohort study in two different US states where participants were identified through health systems, patients are typically insured and in active cancer treatment. Thus, for those caregivers who share budgets and insurance with the cancer patient, they likely have some protection against economic costs. Second, in this study, patients were selected and recruited based on an expected limited life expectancy and costs were captured up through 6 months of study start. As such, we are not capturing the longer-term economic impacts of cancer nor following up on long-surviving advanced cancer patients and their caregivers. Third, we are not able to discern intangible economic costs or time costs in this study, and there is more and more discussion of “time toxicity” in cancer care that caregivers may also face, that is, the overwhelming time needed to aide a patient with treatment [34]. We do include perceived economic costs, such as feeling one is in an adverse economic situation, as well as direct costs, such as out-of-pocket costs, and indirect costs, such as spillovers to work, which is a strength. Fourth, our data does not let us ascertain the extent to which unpaid caregiving use was a financial coping strategy for the patients and caregivers. We know that a sizeable portion (43.5%) used home health care but we are not able to identify whether patients obtained their preferred help in the home or how much financial considerations drove the type of care (unpaid, family versus paid) obtained in the home. We also do not examine presence of unmet needs, although we do control for subjective burden which could be higher when there are unmet needs [5]. Fifth, and finally, we do not consider endogeneity and instead estimate associations of caregiver intensity at an initial study visit (e.g., baseline) and economic costs in the 24 weeks thereafter. Addressing endogeneity using a quasi-experimental method such as instrumental variable estimation could lead to different results, and we cannot quantify the direction of the bias in the current estimates. Nevertheless, despite the potential limitations, we think this analysis adds value because we are able to present novel, descriptive longitudinal data that fills a gap in our understanding of how caregivers of persons with advanced cancer fare economically in the short term using direct and direct measures of economic costs. Specifically, similar to the literature on patients with cancer, it is common for caregivers to experience direct and indirect economic costs: they accrued debt, reported out-of-pocket expenses, and quit work due to the caregiving role. Furthermore, perceptions of economic strain were present: almost half of caregivers report at least once that they were in an adverse economic situation.

Whereas in this study we examine the economic costs of caregiving in advanced cancer for up to 24 weeks, the effects of financial strain can have long-term negative consequences and can lead to financial hardship for the caregiver [35]. Temporary or permanent absences from the labor force result in lost wages and employment benefits, and can potentially reduce future retirement income [36]. In certain industries and jobs, taking paid or unpaid leave is not possible, and caregivers risk losing their jobs or forgoing promotions [7]. Specifically, recent evidence shows that female caregivers in particular do not return to work upon leaving for caregiving [37]. Future work should quantify the long-term costs of caregiving for patients with advanced cancer, for example, long after caregiving has ceased. This work should include examining long-term economic costs and economic costs to caregivers when survival is long for the advanced cancer patient, which is particularly important because more people diagnosed with advanced and metastatic cancer are living longer with their cancer [38].

Clinical and Policy Implications

Historically, research documenting financial toxicity of cancer care has focused on patients [9, 12, 13, 39], and yet the current evidence suggests caregivers also experience pervasive direct and indirect economic costs. It is critical that clinicians avoid making assumptions about the role of economic strain in patients’ and caregivers’ lives based on insurance status or socioeconomic status, as insurance coverage for patients does not immunize caregivers from experiencing direct and indirect economic costs. In the context of busy clinical practices, one clinic intervention to address the pervasive caregiver economic costs observed could be to incorporate brief patient and caregiver screeners for economic strain as a part of the delivery of cancer care or palliative care. If these screeners are repeated over time, it would more accurately portray the dynamic changes in caregiving costs that occur over time (through survivorship or up to patient death) and allow provider teams to help match caregivers and patients with necessary programs and supports based on their current situation. However, it is likely not feasible to address caregiver financial strain fully in the clinic, especially when caregivers are not in the same health care system as the patient.

A more promising intervention to address caregiver economic costs would be to implement national policy reforms. Policy change, such as insurance redesign, could cap out-of-pocket costs for caregivers on an annual basis. Coverage of paid home care and respite care would also be beneficial to caregivers, allowing them to maintain work or child care duties and reduce the out-of-pocket costs of accessing these typically uninsured services. Work place benefit change could address the indirect economic costs (e.g., quitting work) of caregiving through paid family leave or unpaid leave with job guarantees [40]. Furthermore, especially for working caregivers, making home- and community-based care a standard employer health insurance benefit would likely reduce out-of-pocket expenses and help caregivers balance their work responsibilities, potentially preventing premature exits from the labor force. Expansion of home health care could occur from national policy change, employer-led benefit change, or both.

Finally, policy change needs to consider equity and address inequality. Whereas wealthier, working, and White caregivers may report spending more out-of-pocket to pay for caregiving, they may not experience adverse financial strain from such expenditures. Wealthier and White caregivers are also likely to be able to take unpaid family leave compared to caregiving families who are Black or who have limited or no access to comparable benefits [41]. Such distributional effects mean that we need to identify policy levers to target scarce resources to the caregivers who will be least resilient to weather the economic costs of caregiving in face of a cancer diagnosis.

7. CONCLUSION

This study extends previous research by describing a comprehensive set of economic costs over 24 weeks for caregivers of persons with advanced cancer, as well as examining the association between objective burden and economic costs. The results showed that higher objective burden was associated with a higher likelihood of incurring debt and being in an adverse economic situation and a lower likelihood of working for pay. Higher subjective burden was independently associated with higher OOP expenses and a higher likelihood of being in an adverse economic situation. Whereas this paper focuses on quantitative economic costs, other papers associated with this study will use triangulated data collected through interviews, direct observations, surveys, and receipts of economic costs to provide a deeper understanding of how caregiver’s experiences of economic costs affects their lives (see Thomson, under review). Overall, these findings indicate that policy change is needed to reduce the direct and indirect economic costs of caregiving for caregivers and patients with advanced cancer.

Data Availability.

The de-identified datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.