Older adults with cancer are vulnerable to functional decline, which places greater onus on caregivers. Few studies have prospectively examined burden in caregivers of older cancer patients. The objective of this study was to determine the factors associated with high caregiver burden.
In total, 100 caregivers of patients aged ≥65 years with cancer, who were recruited at a single institution, completed questionnaires gauging their perception of the patient's physical, emotional, and social health. The association between these items, cancer-related factors, sociodemographic factors, and caregiver burden (measured using the Caregiver Strain Index [CSI]) was determined through multivariate analysis.
The median patient age was 70 years (range, 65-91 years), 70% of patients had advanced disease, and 98% were receiving treatment. Caregivers were mostly women (73%), spouses (68%), and lived with the patient (79%). The median amount of care provided was 10 hours per week. The mean CSI score (± standard deviation) was 3.1 ± 3.2. Most caregivers (75%) reported some burden, with 15% reporting high caregiver burden (CSI score, ≥7). In multivariate analysis, employed caregivers (odds ratio, 4.5; 95% confidence interval, 1.1-18.4; P = .04) and those caring for patients who required more help with instrumental activities of daily living (Older Americans Resources and Services-Instrumental Activities of Daily Living score, <12 of a possible 14; odds ratio, 12.4; 95% confidence interval, 2.4-62.5; P < .001) were more likely to experience high caregiver burden (CSI score, ≥7).
Cancer treatment is increasingly administered in an outpatient setting, which results in reliance on informal caregivers—predominantly family—to assist with day-to-day activities.[1, 2] Caregiver responsibilities can include bathing, transportation, medication management, and monitoring for and managing treatment side effects.[1, 3, 4] This can place burden on caregivers, many of whom feel inadequately prepared for this role.[5-7]
Caring for a loved one can be rewarding, but it also can be associated with significant caregiver burden. Caregiver burden is defined as a subjective experience that the individual perceives as stressful and occurs when an imbalance exists between caregiving demands and caregiver resources to cope with those demands. Caregivers are at higher risk of depressive symptoms,[9-11] are more likely to experience difficulties with sleep and fatigue,[11-13] and are less likely to practice preventive health care than noncaregivers. Older caregivers experiencing burden may also be at higher risk of death than noncaregivers.
Caring for older adults with cancer may involve additional challenges, which may put these caregivers at particularly high risk of burden. Increasing age is associated with a decline in physiological reserve, which can result in increased need for functional assistance at baseline and in periods of stress, such as during treatment. Furthermore, comorbidities, frailty, and geriatric syndromes are more common in older adults with cancer.[16-18] Given the aging population and the increasing prevalence of cancer, a better understanding of the unique challenges that face this population of patients and caregivers is needed. Unfortunately, few studies have examined burden specifically in caregivers of older adults with cancer. We sought to determine the factors associated with high burden in caregivers of older adults with cancer in this hypothesis-generating pilot study.
MATERIALS AND METHODS
Patients aged ≥65 years with a diagnosis of cancer and their primary caregiver, as identified by the patient, were approached in clinic. Patients without a caregiver were excluded. One hundred patient-caregiver dyads were enrolled (the patients included 50 with solid tumor malignancies and 50 with hematologic malignances). Non-English–speaking patients and caregivers were excluded because some measures have been validated only in English. Approval from the institution's ethics and research board and the patient's attending physician was obtained. Informed consent was obtained from patients and caregivers.
Caregiver sociodemographics, relationship to the patient, cohabitation with the patient, and time spent caring for the patient were assessed through a structured interview with the caregiver. Patient sociodemographics and cognitive function were obtained by patient interview. Details about cancer diagnosis and treatments were gathered by chart review.
Geriatric Assessment Measures
A geriatric assessment serves to provide a comprehensive picture of a patient's health, encompassing several domains, including cognition, functional status, comorbidity, nutrition, psychological state, and social activity and support. The validated instruments used (Table 1) were chosen based on the feasibility and acceptability of a self-administered geriatric assessment that was developed in a prior study and fully described in a previous report.
Table 1. Geriatric Assessment Domains and Measures
Indicator of nutritional status over previous 6 mo
Veit & Ware 198324
Evaluation of anxiety, depression, loss of behavioral/emotional control, and positive affect using a 6-point Likert scale
Social activity and support
Stewart & Ware 199220
Measures social activity limitations because of health using a 5-point Likert scale
MOS-Social Support Survey
Stewart & Ware 199220
Measures perceived social supports using a 5-point Likert scale
All domains were obtained using a self-administered questionnaire that was completed by the caregiver using a touch-screen interface, with the exception of cognition, which was administered to the patient by the research team. Caregivers responded based on their assessment of the patient. Responses were obtained independently from the patient. We used the caregiver's assessment of the patient's health rather than the patient's because we hypothesized that the caregiver's perception of the patient's health may differ from the patient's self-assessment but may be more important to caregiver burden than the patient's perception of these domains.
Caregiver burden was assessed using the Caregiver Strain Index (CSI), a validated, 13-item, self-administered measure that evaluates 5 domains of caregiving: employment, financial, physical, social, and time. It has a reliability coefficient α of .86 and has good construct validity. Answers are dichotomous (yes/no) on the CSI, and higher values indicate more burden (score range, 0-13). A CSI score ≥7 is considered high caregiver burden, and this threshold for caregiver burden has been used in other studies of cancer caregivers.[12, 27]
Descriptive statistics were used to summarize patient and caregiver demographics, clinical characteristics, caregiver strain scores, and geriatric assessment measures. We evaluated the association between caregiver burden and patient, caregiver, and caregiving factors, as well as geriatric assessment measures. A model for caregiver burden was developed using unconditional logistic regression. The caregiver strain score was dichotomized (scores ≥7 vs scores <7) to identify those caregivers who were experiencing high levels of caregiver burden. The mean values were used to dichotomize all quantitative variables that were normally distributed. Median values were used to dichotomize variables that were not normally distributed (caregiver age, hours of caregiving per week, caregiver-reported Karnofsky performance status [KPS], and Blessed Orientation-Memory-Concentration [BOMC] test). Variables that reached a P value < .1 in bivariate analysis (Table 2) were considered in the multivariate analysis using the stepwise method. Two variables were included in the final model (caregiver employment status and caring for patients who needed more help with instrumental activities of daily living [IADLs]). The likelihood-ratio test was used to check for possible interaction terms between factors. Calibration of the final model was checked using the Hosmer-Lemeshow test, and the discrimination of the model was assessed according to the area under the receiver operating characteristics (ROC) curve. Statistical analyses were performed using the software packages SAS version 9.2 (SAS Institute, Cary, NC) and STATA SE version 12.0 (StataCorp, College Station, Tex).
Patient and Caregiver Characteristics
One hundred patient-caregiver pairs were included (for nonparticipants, see Fig. 1). The median patient age was 70 years (range, 65-91 years) (Table 3). Patients had a variety of tumor types, including lymphoma (26%), breast cancer (19%), gastrointestinal malignancies (15%), and myeloma (10%). Most patients (70%) had advanced disease and were receiving treatment (chemotherapy, 62%; targeted therapy alone, 34%).
The median age of caregivers was 66 years (range, 28-85 years) (Table 3). Most caregivers were women (73%), and almost a quarter (23%) were employed. Spouses and children comprised the majority of caregivers (68% and 18%, respectively), and other relatives and friends made up the remainder. Almost 80% of caregivers were living with the patient.
Caregiver Assessment of Patient
Patients had an average (± standard deviation) of 2.4 ± 1.8 comorbid conditions, and 26% had had at least 1 fall in the past 6 months, as reported by caregivers (Table 4). Caregivers reported that patients had a mean ± standard deviation Medical Outcomes Study (MOS)-Physical Function score of 53.7 ± 27.9 and a mean ± standard deviation KPS of 80.5 ± 16 (indicating that the average patient was able to do “normal activity with effort, some symptoms of disease”). Sixty-four percent of patients required some assistance with IADLs (Older Americans Resources and Services [OARS]-IADL score, <14]) and the mean ± standard deviation OARS-IADL score was 11.9 ± 2.3, indicating that, on average, patients were either completely dependent for 1 IADL or required some help with 2 IADLs.
The mean ± standard deviation patient Mental Health Inventory score, as assessed by the caregiver, was 76.9 ± 16.3. Five percent of patients scored above the threshold for possible cognitive impairment (BOMC score, ≥11).
Caregivers reported that patients were well supported (mean ± standard deviation MOS-Social Support Survey [MOS-SSS] score, 90.6 ± 12.7), but that patients' social activities were limited by their health (mean ± standard deviation MOS-Social Activity score, 54.7 ± 19.4). Thirty-five percent of patients had an unintentional weight loss ≥5% at assessment.
Caregivers reported providing a median of 10 hours of care per week (Table 5). The highest quartile provided ≥35 hours per week. Most caregivers (61%) had provided care for at least 1 year. The mean ± standard deviation caregiver burden score, as measured by the CSI, was 3.1 ± 3.2 (range, 0-13). Seventy-five percent of caregivers endorsed some level of caregiver burden, and 15% reported high levels of burden (CSI score, ≥7).
On bivariate analysis, high levels of burden were more likely among caregivers who were non-Caucasian (odds ratio [OR], 4.76; P = .033), employed (OR, 3.77; P = .024), and aged <66 years (OR, 5.56; P = .011) and those who cared for patients who had an unintentional weight loss ≥5% (OR, 3.88; P = .025) and solid tumors (OR, 8.67; P = .01). Caregivers who provided care for at least 10 hours per week (median) and 35 hours per week (highest quartile) were more likely to experience high caregiver burden (OR, 3.11 [P = .07] and 4.0 [P = .017], respectively). The caregiver's perception of the patient's physical and social function also correlated with caregiver burden. High levels of burden were observed among those who cared for patients who had a KPS <90 (OR, 5.56; P = .011), required more assistance with IADLs (OARS-IADL score, <12; OR, 20.0; P < .001), had lower perceived physical function (MOS-Physical Function score, <53.7; OR, 6.25; P = .007), had more limitations in social activities because of their health (MOS-Social Activity score, <54.7; OR, 6.25; P = .007), and had less social support (MOS-SSS score, <90.6; OR, 2.94; P = .073). Of the caregivers who reported that the patient was completely independent (KPS, 100%; or OARS-IADL score, 14), none experienced high levels of caregiver burden. Full results of the bivariate analysis are provided in Table 2.
Table 2. Bivariate Analysis of Factors Related to High Caregiver Burden
OR (95% CI)
Abbreviations: BOMC, Blessed Orientation-Memory-Concentration test; MOS, Medical Outcomes Study; OARS-IADL, Older Americans Resources and Services-Instrumental Activities of Daily Living; [Ref], reference category.
P < .1 on univariate analysis (this variable was included in the multivariate analysis).
For geriatric assessment factors, scores indicating better health were used as the reference category.
Age <70 vs ≥70 [Ref]
Sex: Men vs women [Ref]
Ethnicity: Hispanic vs non-Hispanic [Ref]
Race: Nonwhite vs white [Ref]
Not married vs married [Ref]
College vs ≤high school [Ref]
Greater than college vs ≤high school [Ref]
Employed vs nonemployed [Ref]
Annual household income: <$100,000 vs ≥$100,000 [Ref]
Abbreviations: CI, confidence interval; KPS, Karnofsky performance status; MOS, Medical Outcomes Study; OARS-IADL, Older Americans Resources and Services-Instrumental Activities of Daily Living; SD, standard deviation.
No. of comorbid conditions
Mean ± SD
2.4 ± 1.8
MOS-Physical Function score
Mean ± SD
53.7 ± 27.9
Mean ± SD
11.9 ± 2.3
Percentage completely independent: Score 14
Mean ± SD
80.5 ± 16.0
No. of falls
Mean ± SD
0.5 ± 0.9
Percentage with ≥1 falls in past 6 mo
MOS-Social Activity score
Mean ± SD
54.7 ± 19.4
MOS-Social Support Survey score
Mean ± SD
90.6 ± 12.7
Mental Health Inventory score
Mean ± SD
76.9 ± 16.3
Blessed Orientation-Memory-Concentration score
Mean ± SD
3.0 ± 3.4
Percentage scoring ≥11
Unintentional weight loss, %
Table 5. Caregiver Burden
Abbreviations: SD, standard deviation.
Caregiver strain index score, n = 100
Mean ± SD
3.1 ± 3.2
Score, % of patients
Caregiving time, h
Mean ± SD
21.7 ± 30.5
Duration of caregiving, n = 100
On multivariate analysis, employed caregivers (OR, 4.5; 95% confidence interval, 1.1-18.4; P = .04) and those caring for patients who required more help with IADLs (OARS-IADL score, <12; OR, 12.5; 95% confidence interval, 2.4-62.5; P < .001) remained significantly associated with high caregiver burden (Table 6). There was no interaction between employment status and IADL dependency. There was good calibration (Hosmer-Lemeshow test; P = .85) and discrimination (ROC = 0.85).
Table 6. Multivariate Model of Factors Associated With High Caregiver Burden
OR (95% CI)
Abbreviations: CI, confidence interval; OARS-IADL, Older Americans Resources and Services-Instrumental Activities of Daily Living; OR, odds ratio.
Caregiver employed vs not employed
Caregiver perspective of patient IADL dependence
OARS-IADL score <12 vs ≥12
Our current study demonstrates that most caregivers in this cohort experienced some level of caregiver burden, with 15% reporting high levels of burden. Screening for caregiver burden is not routine, and caregiver burden is rarely addressed in the clinical setting.[28, 29] Given the exponential growth of older adults with cancer, the prevalence of caregiving burden, and the potential for higher mortality among caregivers who express caregiver burden, identifying the factors associated with high caregiver burden is important for targeting interventions at those caregivers most in need. Although many factors were associated with caregiver burden, employed caregivers and those caring for patients who required help with IADLs, as assessed by the caregiver, had a higher risk of high caregiver burden on multivariate analysis.
Compared with studies of caregivers of cancer patients of all ages, caregiver burden in our study was lower (mean CSI score, 3.1 vs 4.2-6.36,12,27,30-32), suggesting that caregivers of older adults with cancer potentially may experience less burden than other caregivers of cancer patients. Caregiver burden may have been lower in our study because the caregivers generally were older than those in other studies (mean age, 63.1 years vs 43.8-63.1 years). This is consistent with studies demonstrating that older caregivers have better mental health and experience less psychological distress because of caregiving.[33, 34] This may be because older caregivers perceive caregiving as less overwhelming and their activities as less restricted than younger caregivers.[34, 35] We recognize, however, that caregiving may have a greater physical impact on older caregivers, especially on those who are already vulnerable (ie, those with serious health conditions, poor health, and disability).[36, 37]
Caregivers who perceived care recipients as more dependent with their IADLs were more likely to experience high caregiver burden. This is consistent with prior studies, which also indicated that caregivers of patients who required more daily assistance had higher caregiver burden.[38-40] One possible explanation for this is that caregiving restricts the caregiver's ability to socialize and participate in valued activities.
In our study, caregivers who were employed were more likely to have high levels of caregiver burden, likely because of competing demands from caregiving and work. This association is consistent with some studies,[3, 42, 43] but not others.[44, 45] Research suggests that employment can be either protective or detrimental to caregivers. Although work can be a source of additional stress and demands, it can also be an important source of income, social support, and respite from caregiving. Differential effects of employment on caregiver burden may depend on age, because younger caregivers may be less well established in their careers and less financially secure.[35, 47]
Many studies have explored the factors associated with burden in caregivers of cancer patients in general. Fewer studies have investigated the factors associated with caregiver burden solely among those who care for older adults with cancer. Other studies of those caring for older adults with cancer also identified higher caregiver burden among employed caregivers and among those caring for patients who were more dependent.[48, 49] One study, however, did not identify an association between IADL dependency and caregiver burden. Other factors associated with burden and/or mental health among caregivers of older cancer patients include patient symptoms or distress,[40, 48, 50] cancer type, proximity to surgical treatment, patient physical function, trajectory of illness (eventual death or survival), caregiver age, caregiver sex (husbands > wives), caregiver health status, and relationship of the caregiver to the patient (spouse > nonspousal).[48, 49] Many of these factors also were identified in our study at the bivariate level, although no significant association between caregiver burden and caregiver sex, relationship to the patient, or the patient's emotional state were observed in our study. Differences may be caused by variations in methodology, instruments used, cultural norms surrounding caregiving, and the population of caregivers and patients studied. In addition, several of those studies included only patients who had early stage disease or included prespecified groups of caregivers (eg all spouses aged ≥65 years).[40, 49]
We acknowledge the limitations of our study. First, we did not differentiate between physical and psychological caregiver burden, which may have demonstrated a differential pattern in caregiver burden in this cohort of caregivers. We also did not capture potentially important information about caregivers, such as their health status and the presence of informal and formal social supports. Our study included all caregivers and patients of all stages to allow wider applicability of these results to the clinical setting. However, as in previous studies, most of the caregivers in our cohort were women and were either spouses or children, and we acknowledge that specific subgroups of caregivers may have unique needs. In addition, although we used the caregiver assessment of the patient as an indicator of high caregiver burden, we recognize that high caregiver burden itself may impact negatively on caregiver perception of the caregiving tasks and the caregiver recipient's health. Although this is one of the largest studies to our knowledge investigating the factors associated with high burden in caregivers of older adults with cancer, given the modest sample size and small number of caregivers with high caregiver burden, the confidence intervals are wide and the model may be “overfit”. Furthermore, only 2 factors were identified in the final model, and given our modest sample size, some factors may not have been identified. Thus, because this was a single-institution, hypothesis-generating study, we acknowledge the need to validate these findings.
Hopefully, identifying caregivers of older adults at high risk of caregiver burden will allow researchers and clinicians to target interventions at those most in need. Interventions have included psychological counseling, teaching caregiver skills, support groups, and respite care. Two large meta-analyses, one on caregivers of patients with cancer and another on caregivers of older adults (predominantly those with dementia), indicated that most interventions had a small-to-moderate positive effect on a variety of outcomes, including caregiver perception of caregiving,[2, 52] knowledge, coping and self-efficacy,[2, 52] quality of life, and, ultimately, caregiver burden.[2, 52] Few studies have tested interventions specifically among caregivers of older adults with cancer. More studies are needed to determine whether these interventions produce similar results in caregivers of older adults with cancer or whether tailored interventions are needed to target the unique needs of these patients and caregivers.
In conclusion, although most caregivers of older adults with cancer do experience some caregiver burden, only a small proportion report high levels of burden. Caregivers who care for patients who require help with IADLs and who are employed are most likely to experience high levels of caregiver burden. These risk factors can help identify those caregivers at highest risk of burden and serve as a platform to test interventions to help alleviate this burden.
This work was supported by a National Institutes of Health grant R21 AG041489 and by a City of Hope Excellence Award to Dr. Hurria.
CONFLICT OF INTEREST DISCLOSURES
Mr. Loscalzo reports personal fees from Novartis and Lilly and royalties from City of Hope for SupportScreen Technology, and he has a patent pending for SupportScreen Technology. Ms. Clark reports personal fees from Novartis and Lilly. Mr. Rinehart reports personal fees from SupportScreen Technology and has a patent pending for SupportScreen Technology. Dr. Hurria reports support from GlaxoSmithKline and Celgene for investigator-initiated research and has acted as a consultant to GTX and Seattle Genetics.