Background
It is estimated that 31–55% of caregivers of chronically ill elderly relatives experience depression
Factors that may be related to stress and depression in caregivers include hours of care, stressful life events, social support, age, and gender. Studies have found that hours of care were significantly related to caregivers' anxiety/depression and somatic symptoms
Social support may enhance the ability of the individual to cope with events or change the individual's cognitive appraisal of events
Age has been shown to have indirect effects on depression through its influence on perceived stress, the coping process, and perceived efficacy
Tsai et al. have suggested that stress and depression are emotional aspects of coping mechanisms and depression is the outcome of perceived caregiver stress
Although the associations between depression and self-esteem and marital satisfaction have been examined extensively, only a few studies have been conducted in the context of caregiving. Caregivers have been shown to have lower self-esteem
A recently developed Theory of Caregiver Stress
Methods
Sample
Data for the study were obtained from the Americans' Changing Lives (ACL) Survey: Wave 1, 1986 (N = 3,617), and Wave 2, 1989 (N = 2,867)
Only individuals with experience in caregiving to a chronically ill aged relative were included in the analyses reported here. This reduced the number of cases available for study to 335 from Wave 1 and 271 from Wave 2. The two samples were not completely independent because 99 cases were included in both waves; therefore, to ensure the independence of samples, these 99 cases were dropped from Wave 1. The final samples from Wave 1 and Wave 2 were thus 236 and 271, respectively.
Measures
Since the study was a secondary data analysis, indicators of the study variables (hours of care, age, gender, social support, stressful life events, perceived stress, depression, physical function, self-esteem and marital satisfaction) were selected from the Americans' Changing Lives Survey questionnaires, based on face validity. That is, the questions selected gave the appearance of measuring the content of interest. Exploratory factor analysis, confirmatory factor analysis, and internal consistency tests were then conducted to confirm the underlying structures of established scales and develop outcome measures for the current study.
Hours of care were the total hours estimated by the caregiver in the past year, categorized as less than 20 hours, 20 to 39 hours, 40–79 hours, 80–159 hours, and 160 hours or more. Providing more hours of care was expected to indicate more burden of caregiving.
Stressful life events were measured by a 12-item checklist of negative or undesirable events, such as being robbed or burglarized, losing a job, being physically attacked, or experiencing the death of spouse, death of a parent, death of a close friend/relative, serious illness, life-threatening illness/accident, divorce/separation, serious financial problem, death of children, and other such events. Respondents were asked to report whether they had experienced any of these events within the past 2 years. A simple score, the stressful life events index, was created by summing the number of events reported by each respondent. A high score reflected more stressful life events.
Social support was measured by two items: friends/relatives' love and care, and their willingness to listen. Alphas reliabilities were .73 and .74 for Wave 1 and Wave 2, respectively. Higher scores indicated greater support from friends/relatives. Demographic data included age, defined as the chronological age of the caregiver, and gender, coded as biological sex identity.
Perceived caregiver stress was measured by one item asking how much stress the caregiver felt about caring for or arranging care for the elderly relative. Responses were on a 5-point scale ranging from not stressful to very stressful; a higher score reflected more perceived stress. Other studies have shown that stress was associated psychosocial well-being, such as depression
Depression was measured by the 11-item Center for Epidemiological Studies Depression (CES-D) scale
Physical function was defined as consisting of functional health, number of chronic illnesses, and self-rated health. Functional health was measured by asking the caregiver whether the caregiver was bedbound, and whether the caregiver had difficulty bathing, climbing stairs, walking, or doing heavy housework, and the degree of difficulty of these tasks. Higher scores reflected a higher level of physical function. The number of chronic illnesses was the sum of the following: arthritis or rheumatism, lung disease, hypertension, heart disease, diabetes, cancer, circulation problems, stroke, fracture, and urinary incontinence. A low score on this measure indicated high physical function. Self-rated health was measured by a single item that asked caregivers to rate their own health on a 4-point scale ranging from poor to excellent. A high score reflected high physical function.
The caregiver's self-esteem was measured by five items: "I take a positive attitude toward self," "I am no good at all," "I see myself as a failure," "I have the feeling of being pushed around in life," and "I perceive myself able to solve problems." These items were measured on a 4-point scale ranging from strongly agree to strongly disagree. A higher score indicated higher self-esteem.
Marital satisfaction was also measured by five items: "Overall satisfaction with relationship," "love and affection expressed from spouse or significant other," "spouse treats me well," "thinking about divorce or separation," and "things happened that I can never forget." Higher scores indicated more marital satisfaction.
Cronbach's alphas for all measures were above the acceptable criterion of .70 in both waves except for self-esteem in Wave 1. However, that measure was on the margin of acceptance, at .68. Since Cronbach's alpha is a conservative estimate of internal consistency
Analytic procedure
Univariate and bivariate analyses were used to examine the descriptive findings. To test the appropriateness of the indicators for each latent variable in both waves, the following procedures were used. First, a single indicator was extracted when applicable (e.g., for social support, self-esteem, and marital satisfaction), and summary scale scores were used as single indicators. This strategy was used to reduce the number of parameter estimations in a complex model; it is considered appropriate when individual factor item loadings in a specific scale are high
Standardized factor loadings and measurement error variances for the measurement model predicting caregiver stress
Latent variable
Indicators
Factor loading (Measurement error)
Hours of care
Hours of care
1.00a (.00)b
Stressful life events
Number of stressful life events
1.00a (.00)b
Social support
Friend/relatives positive support
1.00a (.00)b
Age
Age
1.00a (.00)b
Gender
Gender
1.00a (.00)b
Perceived stress
Perceived caregiver stress
1.00a (.00)b
Depression
CES-D Depressed & positive mood
.75a (.44)
CES-D Somatic symptoms
.74 (.45)
CES-D Interpersonal
.39 (.84)
Physical function
Functional health
.55a (.70)
Numbers of chronic illness
.65 (.57)
Self-rated health
.77 (.41)
Self-esteem
Self esteem/mastery index
1.00a (.00)b
Marital satisfaction
Marital satisfaction index
1.00a (.00)b
Factors and measurement errors were from the completely standardized solution. All factor loadings and measurement errors were significant at .01 level in the preliminary measurement model.
a Parameter was fixed to 1.0 in the unstandardized solution.
b Parameter was fixed to 0 in the unstandardized solution.
Sample characteristics and comparisons by waves
Wave 1
Wave 2
Measure
- n
%
- M
- n
%
- M
- p
Hours of care
<20 hours
38
16.1
30
11.1
.45
20–39 hours
28
11.9
39
14.4
40–79 hours
35
14.8
48
17.7
80–159 hours
35
14.8
42
15.5
≥160 hours
100
42.4
112
41.3
Stressful life events
.57 (.69)
.50 (.64)
.13
Social support
7.65 (1.95)
7.91 (1.77)
.12
Age, in years
53.56 (16.36)
53.41 (14.43)
.91
Gender
Male
85
36.0
86
31.7
.31
Female
151
64.0
185
68.3
Perceived caregiver stress
Not at all stressful
46
19.5
57
21.0
.75
Not too stressful
62
26.3
78
28.8
Somewhat stressful
77
32.6
76
28.0
Quite stressful
26
11.0
35
12.9
Very stressful
25
10.6
25
9.2
Physical function
1) Functional health
Most severe impairment
7
3.0
9
3.3
.87
Moderately severe impairment
19
8.1
17
6.3
Least severe impairment
18
7.6
23
8.5
No impairment
192
81.4
222
81.9
2) Number of chronic illnesses
1.26 (1.28)
1.37 (1.34)
.36
3) Self-rated health
Excellent
34
14.4
42
15.5
.95
Very good
88
37.3
95
35.1
Good
65
27.5
79
29.2
Fair
38
16.1
45
16.6
Poor
11
4.7
10
3.7
Self-esteem
16.02 (3.11)
16.84 (2.86)
.00
Marital satisfactionb
-0.12 (3.66)
0.05 (3.61)
.67
Depression
16.00 (4.14)
15.15 (3.87)
.02
a Statistical significance is determined either by t-test or chi-square. b Score of marital satisfaction is standardized.
A covariance matrix derived from data in the Wave 2 sample was analyzed as input data in the process of model testing. Hypothesized models were tested using the maximal likelihood procedure in the LISREL statistics program. The model tests used absolute goodness-of-fit indices (Chi-square [χ2], the goodness-of-fit index [GFI], and the adjusted goodness-of-fit [AGFI]) and comparative fit indices (change in Chi-square [Δ χ2], the relative noncentral index [RNI] and the relative normed fit index [RNFI]). Values of GFI, AGFI, RNI, and RNFI between 0.90 and 1.00 were considered to indicate a good fit between the model and the data
An exploratory structural modeling method, specification search
Results
Descriptive findings
The characteristics of the two wave samples are summarized in Table
Respondents in both waves rated perceived caregiver stress similarly: 21.6% and 22.1% in Waves 1 and 2, respectively, reported "quite and very" stressful experiences, while over 45% in both waves reported that their experiences were not at all stressful or not too stressful.
Caregivers' physical function and marital satisfaction were also similar in both waves. Approximately 81% of the caregivers in each wave reported no functional impairment. Only approximately 20% of caregivers rated their health fair or poor, and the averages numbers of chronic illnesses were 1.26 and 1.37 for Waves 1 and 2, respectively. Scores on caregivers' marital satisfaction were -. 12 and .05 for Waves 1 and 2, respectively; the difference was not significant. However, caregivers in Wave 1 had significantly lower self-esteem scores than those in Wave 2 (16.02 vs. 16.84, p < 0.01). Depression also differed significantly (p < 0.05). Respondents in Wave 1 reported more depression than those in Wave 2, with mean scores on the 11-item CES-D of 16.00 (SD = 4.14) and 15.15 (SD = 3.87) for Waves 1 and 2, respectively.
The hypothesized model
When the hypothesized model was tested to determine whether depression mediated the relationship between perceived caregiver stress and caregiving outcomes (physical function, self-esteem, and marital satisfaction), the statistics showed a moderate fit between the model and the data (χ2 = 237.22; d.f. = 73; GFI = .89; AGFI = .84; RNFI = .85).
More hours of care and female gender predicted greater caregiver stress, accounting for 7% of the variance in stress, as shown in Figure
The hypothesized model
The hypothesized model
The data-derived model
Since the hypothesized model fit the data only moderately well, a data-derived model was built to compare with the hypothesized model. After all the insignificant paths in the hypothesized model had been dropped, paths were added at each step to improve the goodness-of-fit statistics based on the modification index and pre-set assumptions. The resulting data-derived model had a good fit with the data (χ2 = 147.73; d.f. = 74; p = .00; GFI = .93; AGFI = .90; RNFI = 1.00), close to that of the measurement model, with an insignificant difference (Δ χ2 = 33.58; Δ d.f. = 34; p = ns).
The strength and direction of the relationships among the latent variables are shown by the standardized coefficients in Figure
The data-derived model
The data-derived model
Cross-validation
In order to test the robustness of the paths across the samples, the data-derived model was cross-validated by Wave 1 sample. The results are summarized in Table
Summary of cross-validation for the data-derived model
Comparison to previous model
Steps and Purpose
χ2
df
p
χ2diff
dfdiff
p
Step 0 Factor loadings, path coefficients, factor variance, and covariance were all set to be inequality across group
369.14
148
.00
Step 1 Constrain factor loadings
376.21
152
.00
7.07
4
n.s.
Step 2 Constrain factor loadings, and the paths between endogenous variables (beta linkages)
390.92
159
.00
14.71
7
<.01
Step 3 Constrain factor loadings, the paths between endogenous variables (beta linkages), and the paths between exogenous variables and endogenous variables (gamma linkages)
396.04
163
.00
5.12
4
n.s.
Step 4 Constrain factor loadings, path coefficients, and error variance
401.39
169
.00
5.35
6
n.s.
Step 5 Constrain factor loadings, path coefficients, error variance, and factor variance
413.63
179
.00
12.24
10
n.s.
Discussion
The hypothesized model postulated that hours of care would be the most important factor in perceived caregiver stress, and other factors would include stressful life events, social support, age, and gender. Higher perceived caregiver stress was expected to result in more depression, which in turn would lead to poorer health function, lower self-esteem, and lower marital satisfaction. These expectations were only partially supported by the data.
The data-derived model suggested that age had an indirect effect on perceived caregiver stress, through hours of care. Stone et al. found that older caregivers tended to assume the role of primary caregiver in attending to their chronically ill relatives
Age was also found to predict physical function. George and Gwyther noted that spouse caregivers were more susceptible to diminished physical function than other caregivers and this was probably caused by their older age
Stressful life events failed to predict perceived caregiver stress; instead, stressful life events predicted caregivers' outcomes through depression. One explanation for this might be that stressful life events and caregiver stress work independently in predicting caregivers' outcomes. One other study found that stressful life events did not influence health outcomes through perceived stress; rather, they were a confounding factor in predicting health
Hours of care and perceived caregiver stress were expected to play important roles in caregiver outcomes. However, while hours of care predicted caregiver stress, hours of care was not the only nor the most important determining factor. Further, perceived caregiver stress explained only a small amount of the variance in depression. This is consistent with Pruchno et al's finding that caregiving had little impact on depression or the physical health of the caregiver
Unexpectedly, in this study social support had no impact on perceived caregiver stress; but this is not unprecedented: similar results were reported by Lawton et al.
Research has shown that low social support makes people more vulnerable to depression, and that has been clearly demonstrated for the elderly
The data-derived model showed that, as expected, depression mediated the relationship between perceived caregiver stress and self-esteem and marital satisfaction. However, the relationship between depression and physical function was not confirmed by the cross-validation. This is inconsistent with the findings of Pruchno et al.
The study was limited to the variables in the original Americans' Changing Lives survey, constricting our choices in operationalizing constructs. Further, the study was cross-sectional and consequently was limited in testing the causal relationships depicted in the model. Although the findings provide preliminary evidence of causal relations among the variables, better examination of causality will require longitudinal data.
Conclusion
In spite of its limitations, the study shows the importance of psychological mediators in the care of a chronically ill relative. The question of how caregivers manage to avoid adverse outcomes or why some caregivers are at risk for adverse outcomes can be answered in part by understanding the role of depression. Clearly, to avoid adverse outcomes, clinical interventions should target caregivers who are experiencing depression.
List of abbreviations used
Americans' Changing Lives: ACL
Center for Epidemiological Studies Depression: CES-D
The goodness-of-fit index: GFI
The adjusted goodness-of-fit: AGF
The relative noncentral index: RNI
The relative normed fit index: RNFI
Declaration of competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Author 1, PT, developed the research proposal, carried out the data analysis, interpreted the data and participated in the sequence alignment of the manuscript. Author 2, MMJ, participated in developing the research proposal and interpreting the data.