Long-term QALY-weights among spouses of dependent and independent midlife stroke survivors

The participants in this study were from the Sahlgrenska Academy Study on Ischemic Stroke (SAHLSIS), the design of which has been reported previously [12, 13]. In brief, 600 patients with ischemic stroke before the age of 70 were consecutively recruited from four stroke units within western Sweden between 1998 and 2003. The hospitals have identical principles for acute stroke care, basal diagnostic set-up, and stroke unit care. The stroke survivors were age, sex, and geographically matched with 600 controls who were recruited from a population-based health survey or the population registry. For the 7-year follow-up, during 2006–2012, patients and controls were invited to respond to a questionnaire regarding background variables and self-rating instruments concerning health issues. In addition, stroke survivors included at the Sahlgrenska University hospital were invited to visit the research physician and research nurse. Cohabitant spouses of both stroke survivors and controls at the 7-year follow-up were invited to respond to a questionnaire regarding sociodemographic measures and a self-rating instrument for HRQoL. The recruited participants and drop-out rates have been presented in detail elsewhere [12]. In brief, at SAHLSIS baseline, 422 stroke survivors and 437 controls were cohabitant. In the 7-year follow-up, 299 stroke survivors and 344 controls were cohabitant, whereof 248 spouses of stroke survivors and 245 spouses of controls were recruited to the study. In this study, one patient had missing data for mRS, and thus, 247 dyads of stroke survivors were included. All respondents gave informed consent and approved merging of their data for the current study. The studies were approved by the Regional Ethics Committee in Gothenburg (Reference Number 413-04, 622-06).

Sociodemographic data for the study population were collected from the SAHLSIS database [12, 13]. The HRQoL was assessed using the Short Form-36 (SF-36) questionnaire (version 1) in an approved Swedish version [14]. The SF-36 consists of eight scales: physical functioning, physical role, bodily pain, general health, vitality, social functioning, emotional role, and mental health. The first four are physical scales and latter four are mental scales. To derive a preference-based measure of health from the SF-36, an algorithm developed by Brazier et al. [15] was used to revise the SF-36 questionnaire into the six-dimensional health state classification, SF-6D. The SF-6D consists of the following dimensions: physical functioning, role participation (combined role-physical and role-emotional), social functioning, bodily pain, mental health, and vitality. Each dimension has between four and six response levels, resulting in 18,000 different health states, which have been weighted directly or indirectly using the standard gamble method on a random sample of the general population in the United Kingdom.

For the stroke survivors, the modified Rankin Scale (mRS) [16] was used as a measure of global disability that also affects dependency. The scale is defined categorically with seven different grades: 0 indicates no symptoms at all, 1 indicates no significant disability despite symptoms (able to carry out all usual duties and activities), 2 indicates slight disability (unable to carry out all previous activities, but able to look after own affairs without assistance), 3 indicates moderate disability (requiring some help, but able to walk without assistance), 4 indicates moderately severe disability (unable to walk without assistance and unable to attend to own bodily needs without assistance), 5 indicates severe disability (bedridden, incontinent, and requiring constant nursing care and attention) and 6 indicates death. A score of 0–2 indicates independence, while a score of 3–5 indicates dependence [17]. Outcome was also assessed with the National Institutes of Health Stroke Score (NIHSS) [18], which provides a quantitative measure of stroke-related neurological impairment. A higher score indicates a more severe outcome. The ability to perform basic activities of daily living was assessed by the barthel index (BI) [19], with a score between 0 and 100, where a higher score indicates better abilities in self-care. Cognitive impairment was assessed by the Barrow Neurological Institute Screen for Higher Cerebral Functions (BNIS) [20, 21]. A lower score reflects lower cognitive function. Anxiety and depression were self-assessed by the Hospital Anxiety and Depression Scale (HADS-D for depression and HADS-A for anxiety) [22], and a high score reflects presence of anxiety/depressive symptoms. The BNIS was scored by the research nurse for patients who came to the Sahlgrenska University Hospital for the 7-year follow-up.

The distribution of the variables was presented as mean and SD or median and first (Q1) and third (Q3) quartile for continuous variables and as number and percentage for categorical variables. All significance tests were two-sided and conducted at the 5% significance level. For comparison between spouses of dependent and independent stroke survivors as well as dependent and independent stroke survivors, Mann–Whitney U test was used due to skewed data. However, SF-6D data were normally distributed and hence a t test was used for comparison between spouses of dependent and independent stroke survivors. To investigate the relationship between the spouses’ QALY-weights and the stroke survivors’ global disability, we used ordinary least squares regression with the health state value as the outcome variable and the spouses’ demographic features and stroke-related variables as explanatory variables. Due to high co-linearity between the mRS, NIHSS, and Barthel Index, only the primary explanatory variable mRS was included in the model to interpret the effect. The other explanatory variables in the models were the spouses’ age, sex, and educational level and the stroke survivors’ BNIS, HADS-D, and HADS-A. To investigate whether the dependency of the stroke survivors had any impact on their spouses’ QALY-weights, in model 1, mRS was included as dichotomized scores of dependency (mRS score 3–5) and independency (mRS score 0–2, set as a reference). To investigate which of the mRS scores had the highest impact on the spouses’ QALY-weights, in model 2, mRS was included as a categorical variable, where mRS score 0 was set as a reference. Due to few responders having mRS score 5, mRS scores 4 and 5 were merged into one category. Spouses’ age was categorized as <50, 50–65, and >65, where <50 was set as the reference. The model was constructed using a purposeful stepwise regression approach. A corresponding method but for logistic regressions has been described by Bursac et al. [23]. In the first step, the ANOVA was tested with one variable at a time to examine whether the F test had a P ≤ 0.25. If so, the variable was included in the large model (Age reference (<50) vs 50–65 P = 0.520 and vs >65 P = 0.016, Sex P = 0.851, Education P = 0.939, mRS reference (0) vs 1 P = 1, vs 2 P = 0.938, vs 3 P = 0.003, vs 4 P = 0.175, mRS reference (independence) vs dependence P < 0.001, BNIS P < 0.001, HADS-D P = 0.009, HADS-A P < 0.283). In the second step, the variables with P > 0.10 were removed from the large model, and only variables with P < 0.10 were included in the final model (Age reference (<50) vs 50–65 P = 0.322 and vs >65 P = 0.024, mRS reference (0) vs 1 P = 0.993, vs 2 P = 0.462, vs 3 P = 0.002, vs 4 P = 0.087, mRS reference (independence) vs dependence P < 0.001, BNIS P = 0.517, HADS-D P = 0.811). In the final step, the variables not included in the final model were reintroduced to the final model with one variable at a time to examine whether the F test had a P ≤ 0.25, although none of the variables met this criteria (BNIS P = 0.377, HADS-D P = 0.261, HADS-A P = 0.684, Spouses sex P = 0.730 and Spouses education P = 0.762). All the analyses were carried out in SPSS software (version 23, SPSS, Inc., Chicago, IL, USA).