Data for this study were obtained from a cross-sectional baseline survey of caregivers to patients enrolled in the Guided Care randomized trial (see Wolff et al. for more details23
). This study enrolled 308 caregivers identified by patients enrolled in the Guided Care trial (N
904 patients). Patients were eligible for the study if they had been seen by a participating physician within the previous year and were predicted to be in the highest 25% of Medicare utilization services for the next year using the hierarchical condition categories (HCC) model.24
Each patient who reported receiving assistance with ADLs, IADLs or health care tasks (HCTs) was asked to identify the person (family or unpaid friend) who assisted them the most. These caregivers were then contacted and, upon completing informed consent, were administered an in-person caregiver interview by a professional survey research interviewer. Each caregiver enrolled in the study provided care for only one patient enrolled in the study. Patients identified 353 eligible caregivers, and response rate among eligible caregivers was 86% (n
Scope of Caregiver Assistance Assistance was measured using caregiver-reported assistance with HCTs, IADLs and ADLs (e.g., “Do you help [person] with [task]??”). ADLs included dressing, eating, using the toilet, bathing, and transferring between a bed and chair. IADLs for this study included using the telephone, heavy housework, light housework, managing money and shopping. HCTs included eight tasks related to health care which all patients needed to perform and may have asked a caregiver to assist with (such as scheduling medical appointments), and 4 tasks which were considered “optional” because not all patients performed the task and therefore were never in the position of asking the caregiver for assistance (such as obtaining community services; see Table for a complete list).
Caregivers in the study were categorized into three assistance groups: (1) helps with only HCT (N
43), (2) helps with IADL and HCT only (N
128), and (3) helps with ADL, IADL, and HCT (N
127). The ten caregivers who did not fit one of these three groups were added to the group which was most similar to their assistance pattern in a hierarchical fashion (see Table for more detail). Sensitivity analyses showed that the inclusion of these ten caregivers did not influence results.
Health Care Task Difficulty
Our HCTD scale was based on the Oberst Task Difficulty scale14,17
and included eight HCTs applicable to all patients in the study (see Table ). This scale has previously been validated in multi-morbid older adults and is strongly associated with health-related quality of life and quality of care.25
For each HCT, caregivers were asked if they assisted the patient with the task and, if so, how much difficulty they had performing the task. Each item was then coded as 0
no difficulty or does not help, 1
some difficulty, and 2
a lot of difficulty. In the measurement of caregiver HCTD, “does not do task” was considered to be conceptually equal to “does task with no difficulty.26
” A summary score was created by adding values across the eight questions (range, 0–16). Due to the skew in the distribution of responses (skewness
2.12), caregivers’ HCTD summary scores were categorized as follows: no difficulty (HCTD=0), low difficulty (HCTD=1), medium difficulty (HCTD=2), and high difficulty (HCTD=3+).
We used an eight-item scale, adapted from the chronic disease self-management efficacy scale,27
in which caregivers reported how certain they were that they could handle different stressful caregiving situations using a scale of 0 to 10 (10 being very certain). The caregiving situations presented included managing general caregiving problems, maintaining a positive mood, improving mood, dealing with frustration, dealing with demands of caregiving, managing worries, getting information, and finding resources. A summary self-efficacy score was created by averaging all eight items.
The modified Caregiver Strain Index (CSI) was used to measure strain25,28
and the Center for Epidemiological Studies Depression scale (CESD) was used to measure depression.29
The CSI is a 13-item index (range, 0–24) used to screen for caregiver strain. The CESD is a 20-item scale (range, 0–60) used to self-report the presence of depressive symptoms. A score of 16 or above on the CESD is suggestive of clinically significant depression.
Independent Variables Caregiver variables assessed were age, gender, relationship to the patient, education, employment status, co-residence with the patient, co-residence with a child under the age of 18, average hours of care provided in a typical week, number of years spent caring for the patient, and self-rated health. We additionally assessed the caregiver-reported quality of their relationship with the care recipient and the caregiver’s level of satisfaction with caregiving on a scale of 1 to 10 (10 being the best possible).
Patient variables assessed in this study were age, gender, race, financial situation, and education. Patient health variables included the SF-36,30
number of chronic conditions out of the 14 most common chronic conditions in the Medicare population,31
number of days confined to bed and in a nursing home in the previous month, and the HCC risk for future utilization.24
Statistical Analysis Patients and caregivers across the three assistance groups and across the four difficulty groups were compared on the variables described above, using Cuzick’s non-parametric test for trends.
To test the hypothesis that caregiver self-efficacy was associated with health care task difficulty, ordinal multivariate logistic regression was used to determine the odds of caregivers reporting higher versus lower levels of difficulty for every unit increase in self-efficacy. This model can be considered as an extension of the logistic regression model for binary response variables to variables with ordered categories. Inherent in this model is the assumption of proportional odds (which was upheld in this analysis based on approximate likelihood ratio tests (p
0.05). To control for potential confounding, we adjusted for the caregiver and patient characteristics related to caregiver self-efficacy which were also hypothesized to be related to HCTD. These variables included both caregiver characteristics (age, gender, education, health, and reported quality of relationship with patient), patient health-related quality of life (measured using the SF3632
), and the number of assisted HCTs. The number of assisted HCTs was included in the model to ensure the estimated coefficients reflected the relationship of each independent variable with HCTD, regardless of the number of HCTs being performed.
To test the hypothesis that HCTD is associated with depression and strain symptoms in the caregiver, multivariate linear regression models were constructed. Models were adjusted for the same covariates used in the previous model which have been shown to affect depression or strain and were theorized to affect HCTD. To clarify the magnitude of these effects, effect sizes (ES) were calculated using Hedges’ d
, taking into account the covariates adjusted for in the model. The ES statistic describes the average change in standard deviation units of the outcome variable associated with one unite change in the independent variable (0.2, small ES; 0.5, medium ES; 0.8, large ES33