Validation of the Caregivers’ Satisfaction with Stroke Care Questionnaire: C-SASC hospital scale

Interest in assessing caregivers’ (spouse, partner, child, sibling, etc.) satisfaction with the hospital care given to stroke victims—an important indicator of quality of care [1‐6]—is increasing [1]. Caregivers are often dissatisfied [6], many citing a lack of communication about services the patient has received [7‐9]. Caregiver well-being could be improved through in-hospital delivery of stroke care that meets their needs and demands which, in turn, could relieve caregiver stress.

Such state of dissatisfaction despite the importance of the caregivers’ role could be explained in part by the lack of a validated stroke-specific caregiver satisfaction measure. Since the Satisfaction with Stroke Care (SASC) Questionnaire (Appendix 1) is validated for stroke patients [3, 10‐13], our study investigates its reliability and validity for caregivers after adjusting the original SASC hospital scale accordingly. We tested the resulting 11-item caregivers’ satisfaction with hospital care scale (C-SASC hospital scale; Appendix 2) on stroke patient caregivers in the Netherlands.

Stroke patients were identified through the stroke service facilities, and caregivers were identified through the participating stroke patients. Stroke patients who agreed to participate were asked to nominate their principal informal caregiver. This person was defined as “the person who helps you the most but who is not paid to do so.” Patients who were subsequently diagnosed with transient ischemic attack (TIA) instead of stroke were excluded, which led to an initial sample of 915 patients. Because 91 of these 915 stroke patients died during their hospital stay, the final sample consisted of caregivers of 824 patients.

We used demographic data that was collected by staff during hospital admission of the patient, which included gender, education, and relationship to the stroke patient (partner, child, sibling, other).

The original SASC for patients was validated by Boter and colleagues [1] in the Netherlands using forward and backward translation. We divided the first item of the original SASC, “I have been treated with kindness and respect by the staff at the hospital,” into two items to inquire about both the caregiver and the stroke patient. To obtain more information about satisfaction with the hospitalization process and hospital information received, we adjusted the SASC hospital scale and added two items. The resulting C-SASC hospital scale (Appendix 2) consists of 11 items measuring caregivers’ satisfaction with inpatient stroke care. Caregivers indicated their agreement with each item on a four-point scale ranging from 0 (strongly disagree) to 3 (strongly agree); higher total scores indicate greater satisfaction [3].

Reliability of the instrument was assessed by determining the statistical coherence of the scaled items, which reflects the degree to which they measure the intended aspect of satisfaction. Validity is the degree to which a scale measures what it is intended to measure; here we focused on the construct validity of the questionnaire. Construct validity is supported if (1) instruments purported to assess the same concept (e.g., satisfaction) correlate substantially with one another (convergent validity) and (2) instruments purported to assess a different concept exhibit lower correlations (divergent validity). We evaluated construct validity by comparing the C-SASC hospital scale scores with caregivers’ reports (score of 1–10) on the staff’s attitude towards them and the extent to which their wishes had been taken into account (convergent validity). We evaluated divergent validity by correlating the overall C-SASC scores with those of stroke caregivers’ health and quality of life (QoL). We used the EuroQol questionnaire (EQ-5D) to measure the QoL of stroke caregivers, which uses a simple generic measure to aggregate QoL into a single index [15]. The EQ-5D comprises five questions about the current health state in five dimensions: mobility, self care, usual activities, pain/discomfort, and anxiety/depression [15, 16]. Responses collapse into one of three levels: no problems (score of 1), moderate problems (2), and extreme problems (3). To assess health of stroke caregivers the EQ-5D questionnaire was followed by a the visual analogue scale (VAS), a graph representation similar to a thermometer that ranges from 0 (worst imaginable health state) to 100 (best imaginable health state) [16].

We used descriptive statistics to analyze the caregivers’ demographic characteristics. Analysis of variance (ANOVA) was used to examine differences between demographic characteristics of stroke caregivers in relationship to SASC. Cronbach’s α served as a measure of homogeneity reflecting the (weighted) average correlation of items within a scale [17]. In general, reliability is considered to be good if α > 0.80. The construct correlation patterns (trough convergent and divergent validity) were calculated with Pearson’s correlation coefficients.

To verify the one factor structure of the 11 C-SASC hospital scale items and to seek relationships between observed variables and their underlying latent constructs, confirmatory factor analysis was performed twice using LISREL software [18] by (1) listwise deletion of items with missing responses and (2) replacing missing responses with mean values. The models were tested in LISREL with four indices of fit using the cutoff criteria proposed by Hu and Bentler [19]. First, the overall goodness-of-fit test assessed the discrepancy between the model and the sample covariance matrix by means of a normal-theory weighted least squares test. Plausible models had low, preferably non-significant, χ ² values. Second, the root mean square error of approximation (RMSEA) reflects the estimation error divided by the degrees of freedom as a penalty function. RMSEA values below 0.06 were taken to indicate small differences between the estimated and observed models. Third, we used the standardized root mean square residual (SRMR), which is a scale-invariant index for global fit that ranges from 0 to 1. SRMR values lower than 0.08 were considered to indicate a good fit. Fourth, we calculated the incremental fit index (IFI), which compares the independent model (i.e., observed variables are unrelated) to the estimated model. We considered IFI values larger than 0.95 indicators of a good fit. If confirmatory analysis with the full model (11 items) did not meet the four indices proposed by Hu and Bentler [19], we were prepared to exclude items one by one, following the modification indices provided by LISREL and the strength of the loadings, until the cutoff criteria for the four indices were met.