Predictors of employment and lost hours from work in cancer caregivers

Abstract

This cross-sectional, descriptive study identified variables associated with caregivers who (1) were employed and (2) reported lost hours from work due to care demands. Family caregivers (N = 80) of persons with a primary malignant brain tumor participated in a 45–60 min telephone interview, answering questions regarding the impact of providing care on their emotional health and employment status. Younger caregivers were more likely to be employed. Caregivers were more likely to report lost hours from work when care recipients required assistance with Instrumental Activities of Daily Living (IADLs) and were closer to the time of diagnosis. Data suggest that interventions to assist caregivers in maintaining employment should target caregivers of persons with limitations in physical function and should include strategies to coordinate care to assist with IADLs.

Introduction

The way in which variables in the care situation affect family caregivers’ emotional and physical health has been well described [1,2]. The degree of caregiver involvement, care recipient’s disease and symptom severity, and the type of caregiver–care recipient relationship are examples of factors that have been shown to predict a wide range of negative outcomes for caregivers, including depressive symptoms, burden, altered immune function, and increased mortality rates [3–6]. Despite the breadth of data describing the relationship between care recipient variables and caregivers’ emotional and physical health, little is known, particularly in oncology, about how the physical demands and psychosocial issues that stem from providing care affect caregivers’ ability to sustain employment during the care situation.

In studies of elderly persons and persons with dementia, caregivers have reported reducing the number of hours worked, changing jobs, and experiencing reduced productivity as a result of providing care [7,8]. However, few studies have evaluated how providing care affects employment for caregivers of persons with cancer [9,10]. Studies in this area have focused on a specific treatment episode (e.g. neutropenia) or on a single phase of the care recipient’s disease. Yet, family members report continuing to provide care through active treatment and into survivorship, and the impact of providing care on employment is likely to go beyond a single time point in the care trajectory.

The ability to predict which groups of caregivers are at a risk of losing hours from work could help health-care providers target caregivers in need of support, increasing the efficiency of delivering supportive interventions to assist care-givers in balancing employment and care demands. Interventions such as these are particularly vital to caregivers who risk being forced to change employment to provide care. Changes in caregiver employment may have serious implications, such as reduced household income and loss of insurance for not only the caregiver but also the care recipient with cancer. Maintaining caregiver employment can help sustain the dyad financially and ensure medical coverage for the care recipient. The aims of this analysis were to identify variables associated with employment and lost hours from work for caregivers of persons with cancer. Specifically, we sought to determine if care recipient factors (physical function, neuropsychological function, months since diagnosis, and employment) together with caregiver factors (age, gender, relationship to the care recipient, presence of children in the home, depressive symptoms, burden, and number of secondary caregivers) were associated with whether or not caregivers were employed and were associated with caregivers who reported lost hours from work.

Background

Recently, investigators have focused attention on attempting to quantify the cost of informal care, particularly as a result of lost wages [8,11]. Providing care to a family member with dementia has been estimated to lead to a $10 709 average annual loss in wages for the caregiver [11], results that are higher than Small et al.’s [8] findings of an average of $556.41 per six months in lost wages for caregivers of persons with Alzheimer’s disease. In oncology, research quantifying lost wages for caregivers has been focused on a single, critical time in the care recipient’s illness. Calhoun et al. [12] estimated costs of chemotherapy-induced toxicity for women being treated for ovarian cancer and found that caregiver work loss accounted for 6–67% of the indirect costs of the toxic episode. One episode of neurotoxicity was estimated to have an indirect cost of $4220, of which $2837 was due to caregiver work loss.

Although attempts to quantify the cost of lost wages are helpful in determining the value of informal care, they are often limited by methodological concerns. In Small et al.’s sample [8], lost earnings were calculated based on the number of hours taken off work multiplied by the caregiver’s mean hourly income. However, time off work could have been paid time off, unpaid time off, or the result of rearranging hours to provide care. In addition, salaried individuals may have more flexibility in taking time off work without a reduction in pay, whereas hourly workers may feel more financial strain from missing work. Other authors have based lost earning estimates on modified labor force, employment, and earnings data [11,12], which may be too general to yield valuable estimates.

As a result of these limitations, investigators in the area of dementia and general caregiving have turned their attention toward identifying factors associated with caregivers who report lost hours from work as a meaningful way of identifying caregivers at a risk of negative effects from providing care. In a large sample (N = 4592) of caregivers of frail elders, Covinsky et al. [7] reported that 22% of the sample either reduced their number of hours at work or quit working. Caregivers’ odds of reducing work hours were higher when the care recipient required assistance with Activities of Daily Living (ADLs), was diagnosed with a neurologic disorder (dementia or stroke), and when the caregiver was non-Caucasian, a daughter or daughter-in-law of the care recipient, or lived with the care recipient. Both Moore et al. and Small et al. reported that the severity of dysfunction in the care recipient was a consistent predictor of lost wages. Care recipients with later stage disease and increased dependence on ADLs required more hours of care, causing a greater impact on caregiver work hours. In fact, in Moore et al.’s sample, progression of dementia was associated with an additional annual loss in wages of $2000. This research has resulted in the identification of several variables to account for lost hours from work—care recipient ability, relationship to the care recipient, and disease progression.

In oncology, variables that are associated with lost hours from work have not been widely established. Grunfeld et al. [9] found that for care recipients with breast cancer, who were in the terminal phase of their disease, caregivers reported missed work and the inability to work regular hours due to providing care. The investigators’ work identified disease progression as an important predictor of lost hours from work, but the sample was limited to care recipients in the terminal phase of disease. Given et al. [10] reported an association between depressive symptoms and employment status in spouses of persons with cancer at the end of life. Interestingly, the relationship between depressive symptoms and employment status varied depending upon the relationship of the caregiver. For spouses, those with lower levels of depressive symptoms were more likely to be employed. Yet, for adult children, those with higher levels of depressive symptoms were more likely to be employed. The exact nature of this relationship remains unclear, as does the impact of other variables related to the emotional health of the caregiver on employment status and lost hours from work.

Another emotional response to providing care, caregiver burden, is closely linked to depressive symptoms [13] and is conceptualized as the impact of providing care on multiple aspects of the caregiver’s life, such as the caregiver’s health, self-esteem, schedule, finances, and feelings of abandonment [14]. As such, it is possible that measures of burden may also be associated with caregiver employment and lost hours from work. Caregiver mastery, caregiver’s sense of control over the care situation [15], has also been linked to caregivers depressive symptoms [4]. Caregivers with a stronger sense of mastery report lower levels of depressive symptoms when care recipients display neuropsychiatric symptoms (e.g. hallucinations or delusions). Links between caregiver mastery and employment status and lost hours from work have not been reported to date.

As a result of the previously described work, researchers have begun to identify care recipient and caregiver variables that may be associated with changes in employment for oncology care-givers. Because most prior studies in this area have focused on a single phase in the care recipient’s disease trajectory, the purpose of this study was to identify variables that predicted changes in employment for care recipients across the disease trajectory.

Methods

Recruitment and data collection

For this cross-sectional, descriptive study, family caregivers were defined as those who provided assistance to the care recipient on a regular basis. This assistance could include things such as helping with meals, driving to doctor’s appointments, doing laundry, or assisting with ADLs. To be eligible, caregivers had to be 21 years of age or older, able to read and speak English, providing care to someone with a primary malignant brain tumor (PMBT), and having regular and reliable access to a telephone. A total of 95 caregivers were recruited through two national brain tumor support groups, a metropolitan brain tumor treatment center, a statewide cancer registry, and a pre-existing research study (see Sherwood et al., 2006 for a complete description of recruitment procedures, including response rates [16]). For 15 caregivers, missing data precluded use of their responses; the sample used for the analysis in the first research question was 80 caregivers. Each caregiver participated in a 45–60 min structured telephone interview. Data for this analysis included responses to questions regarding the impact of providing care on caregivers’ emotional health and employment status. Human subject approval was obtained from the author’s institution as well as from the governing bodies of participating recruitment sites.

Measures

Dependent variables

The primary dependent variables were current employment status and lost hours from work. Employment status was categorized as either working (full or part time) or not working. Those working were then queried as to whether or not they had taken time off work to provide care in the month prior to the interview (yes or no). Lost hours from work that were not due to care demands were not categorized as lost hours from work for this analyses. Lost hours from work were treated as a dichotomous variable (yes or no), as the sample size precluded using lost hours as a continuous variable.

Independent variables

Care recipient variables

Care recipient’s physical function was evaluated by the Involvement with ADL and instrumental Activities of Daily Living (IADL) scale (α = 0.93) [17], an 11-item instrument that summarizes caregiver reports of care recipient dependencies and has shown construct and content validity in varied samples of adults [18,19]. Caregivers identified, during the past 2 weeks, the level of assistance the care recipient required in six ADLs and five IADLs using a 4-point Likert-type scale ranging from 0 (needs total assistance) to 3 (able to perform independently). Continuous total scores for ADL and IADL were generated by summing the level of assistance required for each activity over the total number of activities; higher scores indicated higher levels of care recipient function. [Note: Assistance with ADLs and IADLs was defined as assistance that was provided only as a result of the brain tumor or its treatment.]

Care recipient’s neuropsychological function was measured using the Neuropsychiatric Inventory-Questionnaire (NPI-Q) (α = 0.78) [20]. The NPI-Q is a 12-item measure in which the caregiver is asked to indicate the presence of 12 abnormal care recipient behaviors (e.g. delusions and hallucinations). Each behavior was rated as either present or absent and scoring for the NPI-Q consisted of summing individual items to generate a total continuous score, higher scores indicating better neuropsychological function (possible range 0–12). Validity for the NPI-Q has been established in persons with neurodegenerative disorders and in older hospitalized patients on acute care floors [21].

Months since the care recipient’s diagnosis was treated as a continuous variable. Care recipient’s employment status was dichotomized as either working (full or part time) or not working.

Caregiver factors

Age and gender were treated as continuous and dichotomized (male as referent) variables, respectively. Relationship to the care recipient was dichotomized as ‘spouse’ or ‘other’. Presence of children in the home was dichotomized as yes or no.

Caregivers’ depressive symptoms were assessed using the Center for Epidemiologic Studies-Depression scale (CES-D) (α = 0.85) [22]. The CES-D has proven to be a valid measure of depressive symptoms in adults [23] and is a 20-item scale that assessed the respondent’s current level of depressive symptoms on a 4-point Likert-type scale. Scoring on the CES-D consisted of summing individual items to produce a total score, higher numbers indicating the presence of more depressive symptoms (possible range 0–60).

Caregiver burden was assessed via the Caregiver Reaction Assessment (CRA) scale, which has been validated in various caregiver populations [14,24]. The CRA is a 24-item instrument that asks caregivers to indicate their level of agreement with statements using a 5-point Likert-type scale. The CRA assesses caregivers’ perceptions of the impact of providing care on their self-esteem, schedule, finances, feelings of abandonment, and health. The schedule, finances, and health subscales were chosen for this analysis since these domains were thought to most likely affect employment. The five items on the schedule subscale (α = 0.75) assessed the impact of providing care on the caregiver’s usual activities, including whether providing care had forced them to eliminate activities and interfered with relaxation. The finance subscale of the CRA (α = 0.87) was used to assess caregiver burden related to the financial implications of providing care. This subscale contains three items that measured caregivers’ perception of the adequacy, difficulty, and strain of their financial situation. The 4-item health subscale (α = 0.70) measured the caregiver’s energy and physical capability to provide care. Scores were generated by summing individual items, higher numbers indicating higher levels of burden.

Caregiver mastery was assessed by the 7-item Caregiver Mastery scale (α = 0.73) [15]. Caregivers used a 5-point Likert-type scale to indicate their perception of how certain they were about what to do in providing care, how they perceived themselves as able to handle most of the problems they faced in the care situation, and how well they believed that they were mastering the challenges in caregiving. Item scores were summed to generate a total mastery score, with higher scores indicating higher levels of mastery (possible range 7–35).

A number of secondary carers were assessed as a count of the total number of individuals defined as ‘anyone who assists in providing care’

Analysis

In the descriptive analyses, the entire sample was utilized (N = 95). For the first research question, to identify predictors of whether or not caregivers were employed, 15 cases with missing data were removed to yield a sample size of 80. For the second research question, to identify factors associated with caregivers who missed hours from work to provide care, caregivers who were not working and who did not change employment due to care demands were removed from the analysis to yield a sample size of 61. Descriptive analyses were used to portray the sample and box plots supported normality of variables’ distributions. Linearity of logit for the fitted models was verified.

Because the sample size precluded using all potential predictors in the model, univariate analyses were performed to determine potential predictors of change in employment and employment status (see Table 1). Variables that were considered clinically significant based on previous literature in the area and those that were at or near statistical significance were considered potential predictors of the outcome variables of interest. Multivariate logistic regression in SAS version 8.2 (Cary, North Carolina) was used with stepwise selection, entry criteria was p<0.25, variables were kept in the model if p<0.10. All potential interactions were explored and none were found to be significant.

Table 1.

Univariate analysis to identify potential predictor of outcome variables

Variables	Lost hours from work (logistic)				Current caregiver employment (logistic)
	Estimate	p-Value	OR	95% CI	Estimate	p-Value	OR	95% CI
Change in care recipient employment	−0.71	0.17	0.49	0.18–1.36	1.80	0.01	6.05	1.66–22.10
Relationship to care recipient	0.07	0.89	1.07	0.39–2.91	−0.71	0.18	0.49	0.17–1.39
Care recipient functional status
ADL	−0.35	0.010	0.71	0.55–0.92	0.16	0.14	1.18	0.95–1.46
IADL	−0.43	0.001	0.65	0.50–0.84	0.19	0.08	1.21	0.98–1.50
Care recipient neuropsychological status	−0.09	0.31	0.92	0.78–1.08	0.04	0.68	1.04	0.88–1.22
Type of treatment (chemo and radiation versus radiation only)	−0.04	0.93	1.04	0.38–2.85	−0.13	0.77	0.86	0.35–2.17
Presence of children in the home (yes/no)	−0.27	0.58	0.76	0.29–1.99	0.19	0.69	1.21	0.49–3.00
Number of secondary carers	−0.23	0.46	0.79	0.42–1.48	0.20	0.57	1.22	0.61–2.44
Caregiver age	0.04	0.11	1.04	0.99–1.09	−0.10	0.0003	0.91	0.86–0.96
Caregiver gender (female)	0.88	0.10	2.40	0.84–6.84	0.71	0.18	2.04	0.72–5.74
Time since diagnosis	0.01	0.18	1.00	1.00–1.02	0.01	0.40	1.01	0.99–1.02
Caregiver education
Some college versus completed high school	−0.15	0.85	0.86	0.17–4.27	0.46	0.52	1.59	0.38–6.63
Completed college versus completed high school	−0.92	0.23	0.40	0.08–1.81	0.36	0.59	1.43	0.39–5.27
Completed graduate degree versus completed high school	−0.65	0.40	0.52	0.12–2.33	0.81	0.24	2.25	0.58–8.77
Caregiver race	−0.15	0.86	0.87	0.16–4.59	0.22	0.81	1.25	0.20–7.87
Care recipient tumor type	0.48	0.31	1.62	0.64–4.11	−0.37	0.49	9.69	0.29–1.65
Caregiver depressive symptoms	0.03	0.19	1.04	0.98–1.10	−0.01	0.62	0.99	0.94–1.04
Caregiver mastery	−0.02	0.72	0.98	0.88–1.09	0.03	0.56	1.03	0.93–1.14
Caregiver burden—finances	0.10	0.21	1.10	0.95–1.28	0.01	0.91	1.01	0.88–1.16
Caregiver burden—schedule	0.13	0.02	1.14	1.02–1.29	−0.11	0.06	0.90	0.80–1.01
Caregiver burden—health	0.05	0.51	1.05	0.91–1.22	−0.02	0.76	0.98	0.85–1.12

Results

Sample

Descriptive statistics for the sample as a whole are provided. The majority of the caregivers were female (74%, N = 70) (see Table 2), Caucasian (94%, N = 89), middle aged (M = 51.4 yrs, SD = 11.7 yrs), and spouses of the care recipient (74%, N = 70). The most common tumor type for care recipients was an astrocytoma (67%, N = 64) and the median time since diagnosis was 19 months (range 2–216). Most caregivers were employed either full time or part time (58%, N = 61) and those who were employed were likely to be in a company with more than 50 employees (61%, N = 37). Approximately $\frac{1}{3}(N=21)$ of the employed caregivers reported lost hours from work as a result of providing care, and about $\frac{1}{3}$ of those who lost hours at work reported that decreasing hours at work had affected their insurance and retirement benefits. Approximately $\frac{1}{3}$ (32%, N = 30) of the caregivers were the primary insurance carrier for the care recipient. Whereas the majority of caregivers were employed, the majority (73%, N = 72) of care recipients were not employed. Over $\frac{1}{2}$ of the care recipients were not employed as a direct result of the tumor (59%, N = 56). For those care recipients who changed employment due to the tumor, 25% (N = 14) retired, 32% (N = 18) quit, 29% (N = 16) took paid leave, and 14% (N = 8) took unpaid leave.

Table 2.

Descriptive analysis of sociodemographic characteristics

	Current caregiver employment					Lost hours from work
	Full time	Part time	Not working	χ2	p-Value	Yes	No	χ2	p-Value
Gender				6.09	0.05			2.19	0.16
Male	16	3	6			4	15
Female	24	18	27			17	25
Marital status				8.87	0.08			1.80	0.51
Never married	2	1	1			2	1
Married	36	16	31			17	36
Divorced	2	4	0			2	3
Widowed	0	0	1
Education				2.17	0.92			2.68	0.48
Completed high school	4	3	6			4	4
Some college	8	5	7			6	7
Completed college	14	6	12			5	15
Completed graduate degree	14	7	8			6	14
Race				2.89	0.62			1.21	0.79
Caucasian	38	20	31			20	37
Other	2	1	2			1	3

Research question 1: Which care recipient and caregiver factors are associated with whether or not caregivers are employed?

The first research question sought to identify variables in the care situation which were associated with whether the caregiver would be employed (see Table 3). Overall fit indices for the final model were good (χ² = 1.58, p = 0.99). The only variable to significantly affect whether or not caregivers were employed was caregiver age (OR, 0.91; CI, 0.86–0.96). For every one year increase in age, caregivers were 10% less likely to be employed. There was also a trend for caregiver employment to vary as a function of the level of assistance the care recipient required with IADLs (OR, 1.23; CI, 0.97–1.56). Higher IADL scores indicated better care recipient function; therefore,. as the level of assistance required decreased, caregivers were 23% more likely to be employed. Caregiver employment was neither significantly related to the care recipient’s ability to perform ADLs, neuropsychological function, or employment status, nor was it related to caregiver gender, relationship to the care recipient, presence of children in the home, caregivers’ depressive symptoms, burden, mastery, or number of secondary caregivers.

Table 3.

Odds ratio estimates of the relationship between care situation variables and whether or not caregivers are employed, N = 80

Variable	Point estimate	CI	p-Value
Caregiver age	0.91	0.86–0.96	0.0002
Care recipient IADL	1.23	0.97–1.56	0.08

IADL, Instrumental Activities of Daily Living.

Research question 2: Which care recipient and caregiver factors are associated with whether or not caregivers will report lost hours from work?

The final parsimonious model is shown in Table 4, which yielded adequate goodness-of-fit indices (χ² = 10.92, p = 0.14). The variable with the strongest association with lost hours from work for the caregiver was the care recipient’s ability to perform IADLs. As care recipients were more limited in their ability to perform IADLs (a lower IADL score), caregivers were 44% times more likely to report lost hours from work (OR, 0.56; CI, 0.42–0.76). The number of months since diagnosis was also associated with caregiver employment. For every one month from diagnosis, caregivers were 2% more likely to report lost hours from work (OR, 1.02; CI, 1.01–1.04). Care recipients’ neuropsychological function was not associated with caregivers who reported lost hours from work. In addition, none of the caregiver variables affected whether caregivers reported lost hours from work.

Table 4.

Odds ratio estimates of the relationship between care situation variables and whether or not caregivers reported lost hours from work, N = 61

Variable	Point estimate	CI	p-Value
Months since diagnosis	1.02	1.02–1.04	0.043
Care recipient IADL	0.56	0.42–0.76	<0.001

IADL, Instrumental Activities of Daily Living.

Discussion

Data from 80 caregivers of persons with a PMBT were analyzed to examine the impact of providing care for a family member with cancer on caregiver employment. Over $\frac{1}{2}$ of the sample was employed, and providing care resulted in lost hours from work which ultimately affected insurance and retirement benefits for a portion of the sample. The first portion of the analysis focused on identifying variables that influence whether or not the caregiver was employed. Caregiver age was the only variable that was associated with whether or not the caregiver was employed; older caregivers were less likely to be employed than younger caregivers. As the mean age of the sample was 51, older caregivers may have been more likely to be retired or to have taken early retirement after the care recipient’s diagnosis. Contrary to other studies [7], care recipient variables such as physical and neuropsychological dysfunction did not appear to affect overall employment rates, although there was a trend for caregivers of persons with more limitations in performing IADLs to be less likely to be employed. Caregiver employment status was not affected by the presence of children in the home, number of secondary caregivers, caregiver gender, relationship to the care recipient, depressive symptoms, or burden, nor was it affected by months since the care recipient’s diagnosis, or the care recipient’s employment status. Lack of an association between these variables and employment status could have been due to the small sample size.

Further analysis sought to determine variables that were associated with caregivers who reported lost hours from work. The care recipient’s ability to perform IADLs appeared to have the greatest impact on whether or not the caregiver lost hours from work due to the demands of providing care. Similar to findings in caregivers of persons with dementia, caregivers were more likely to report lost hours from work when care recipients required assistance with IADLs [7,8]. Intuitively, impairment in performing IADLs increases time demands for the caregiver and caregivers are forced to lose work hours to provide more care. In addition, an increase in IADL assistance may signal disease progression, leading to the caregiver reducing work hours in order to spend more time with the care recipient, particularly if the caregiver is the spouse. Finally, the odds of caregivers losing hours from work increased over the months following the care recipient’s diagnosis.

It is interesting to note that lost hours from work was not affected by the presence of children in the home, number of secondary caregivers, caregiver age, relationship to the care recipient, caregivers’ depressive symptoms or burden, or by care recipients’ neuropsychological dysfunction or employment. In particular, the absence of a link between care recipients’ neuropsychological function and caregivers’ odds of lost hours from work was unexpected. Previous studies both in this sample and in the dementia caregiving literature cite a strong link between care recipient neuropsychological status and caregiver outcomes (particularly emotional health such as level of depressive symptoms) [2,16]. It may be that the risk of losing work hours is less amenable to change than emotional health in the presence of care recipients’ neuropsychological dysfunction or that sample size precluded finding a significant relationship.

In summary, caregivers were more likely to be employed when they were younger and more likely to report lost hours from work when care recipients required assistance with IADLs and were further from the time of diagnosis. Data suggest that interventions to assist caregivers in maintaining employment should target caregivers of persons with functional impairments and should include strategies to coordinate care to assist with IADLs. Future work should focus on tracking changes in employment over time and identifying variables that increase or decrease the likelihood of losing hours from work, particularly as they relate to the physical demands and psychosocial outcomes of care. Future work should also be done to detail how changes in employment affect insurance and out-of-pocket costs for persons with cancer.

Limitations

The study was limited by its cross-sectional design, which allowed us to begin identifying relationships between providing care and employment status, but precluded more sophisticated analyses to determine the temporal nature of these relationships. The sample was small, restricted to a single cancer site, primarily Caucasian, and well educated. Given the possible associations between some cancers and socioeconomic status, predictors of changes in employment may be quite different for other groups of caregivers, particularly given Covinsky et al.’s [7] findings that changes in employment differed based on ethnicity of caregivers of frail elders. Because the sample size was small, we were unable to treat lost hours from work as a continuous variable, and rather dichotomized the sample into those who did and did not report lost hours from work. Future investigations should build on this preliminary work by recruiting a large enough sample size to allow evaluation of lost hours as a continuous outcome.

Next, care recipient physical function and neuropsychological status were obtained by care-giver interview. Research has shown that caregivers who have high levels of depressive symptoms and burden may not always accurately reflect care recipient symptoms [25]. Future work should include more objective measures of care recipient disability. In addition, future studies should query caregivers as to whether or not changes in employment affected their insurance and retirement benefits.

Concerning the choice of variables included in the analysis, it is likely that multiple factors affect employment, not all of which may have been included in these analyses (such as care recipient clinical variables). Future research should be done with sample sizes large enough to accommodate including a large number of potential predictors. Finally, data regarding occupational level (e.g. professional, skilled trade, or hourly versus salaried workers) were not obtained. As higher-level jobs are likely to offer more flexibility to the caregiver in changing work hours to provide care, investigating the role of job level in lost hours from work may provide further insight into the effect of providing care on employment.