Journal of Rehabilitation Medicine 51-8 | Page 23

Psychosocial well-being after stroke education, work status, caring responsibilities, social support, previous illnesses and rehabilitation services. Data collectors with healthcare backgrounds (RN or OT) administered the in- struments using a personal interview format. The assessors read the questions to the respondent and recorded the respondent’s answers in a web-based secure questionnaire using a tablet. At 6 months post-stroke (T2), a data collector who was blinded to the group allocation repeated the outcome measurements. Any changes in the patient’s living situation or health status since the first assessment were recorded. The term psychosocial refers to the interrelation between so- cial factors and individual thoughts and behaviours. Well-being generally refers to emotional reactions and subjective evalua- tions in response to events and includes a greater prevalence of positive than negative emotions and moods, satisfaction with life, sense of fulfilment, and positive relationships (21, 22). The primary outcome to evaluate psychosocial well-being was the General Health Questionnaire-28 (GHQ-28) (23). The GHQ- 28 has been translated into Norwegian (24) and evaluated as an appropriate tool for research purposes in studies to capture emotional stress (25). The GHQ-28 consists of the following 4 subscales addressing aspects of psychosocial well-being: somatic symptoms, anxiety and insomnia, social dysfunction and severe depression (23). Higher score on the GHQ-28 in- dicates higher distress levels. The continuous GHQ-28 score was calculated based on a Likert scoring of 0, 1, 2, 3, resulting in a scale of 0–84. The dichotomized GHQ-28 score was based on a case scoring of 0, 0, 1, 1, resulting in a scale of 0–28. Ba- sed on a comparable study (11) the cut-off was set at 5 in the dichotomized GHQ-28 variable. Scores <5 indicated a normal mood, and scores ≥5 indicated a low mood. Three secondary outcomes were explored. The Yale-Brown single-item questionnaire (Yale) measures the individual’s pre- sence or absence of depression (26). The Stroke and Aphasia Quality of Life Scale-39 (SAQOL-39g) addresses dimensions of stroke-specific health-related quality of life and is adjusted for people with aphasia (27). SAQOL-39g items are scored on a scale from 1 to 5, and mean scores are calculated for each dimension and for the total sum score. Higher scores indicate a higher quality of life. The SOC-13 measures the main concepts of SOC theory (16). In this study, the SOC-13 items were sco- red on a scale from 1 to 5. The total sum score was calculated after reversing the scores of the reversely formulated items, resulting in a scale from 13–65. Higher scores indicate higher levels of comprehensibility, manageability, and experiences of meaningfulness in life. The fatigue questionnaire (FQ-2), Lee’s fatigue scale (LFS-5) (28) and the Ullevaal Aphasia Screening test (UAS) (29) were included to describe the sample. Sample size calculations Sample size was determined based on the primary outcome measurement. With the power set at 80% and the significance level (α) at 0.05, the estimated sample size was 300 patients (150 per group). The calculations were based on a repeated measures logistic regression model of the output variable “normal mood” (GHQ-28). In this study, the power of finding a statistically significant difference between groups was 80% for an OR of 1.9 or higher. Randomization and blinding A computer-generated block randomization procedure was used to allocate the patients into either the intervention or control arm. 559 Participants were randomly assigned in blocks of 10, stratified by the recruiting centre with an allocation ratio of 5:5. An assistant independent of the research group prepared opaque randomization envelopes with 5-digit patient identification numbers printed on the outside and a note specifying intervention or control on the inside. Two regional study coordinators performed the randomiza- tion. Group allocations were communicated solely to the patient and the IP delivering the intervention. To maintain blinding during follow-up, a text message was sent from the study coordinator to the participants before the data collector contacted them with a reminder not to reveal their group allocation. Statistical methods The data were analysed using the Statistical Package for the Social Sciences (SPSS), version 25.0, for Windows (IBM Corp., Armonk, NY, USA) . All statistical tests were intention-to-treat analyses performed as 2-sided tests with a significance level of a=0.05. Missing data were imputed using multiple imputation by chained equations (MICE) in SPSS (30), and all reported results from the statistical analyses were combined results across 5 imputations based on Rubin’s rule (31). A logistic regression with time points (T1 vs T2) as the single independent variable was used to assess the odds of a normal mood (GHQ-28<5) and not having depression (Yale) from T1 to T2 separately per treatment group. The changes in mean scores on the SAQOL-39g and SOC-13 from T1 to T2 were examined for each treatment group using paired-sample t-tests. Logistic regression analyses were used to analyse the effects of the intervention on mood (GHQ-28) and depression (Yale) at 6 months. Multiple linear regression analysis was applied to determine whether participating in the intervention was statisti- cally significantly associated with the participants’ scores on the SAQOL-39g. Based on conceptual and theoretical assumptions, the following baseline characteristics were controlled for in the regression models in each analysis: group allocation, sex, age at admission, rehabilitation services at baseline, care responsibility, living arrangements, comorbidity, stroke severity (NIHSS), stroke aetiology, stroke symptom localization, depression, fatigue and SOC (SOC-13). The baseline value of the outcome measurement analysed was also added as a covariate because the results of the by-group analyses showed significant changes between time points. A variable of recruitment centre was in addition included in the logistic regression analysis of GHQ-28 to control for a potential effect of recruitment centre. The SAQOL-39g was analysed for both the original and log- transformed (2**) variables because of the non-normal distribu- tion of the data. The results were compared, and no significant differences were found. To simplify interpreting the results, only outcomes from the original data analysis are presented herein. Independent-samples t-tests were used to test for statistically significant differences in the mean scores on the SOC-13 bet- ween the intervention and control groups. Ethics The ethical approval of the study was provided by the Regional Committee for Ethics in Medical Research (REC) (2013/2047) and the Data Protection Authorities (2014/1026). The study fol- lowed the guidelines of the Declaration of Helsinki. When invited to participate, eligible participants received oral and written information about the study from trained clinical staff who also obtained informed consent. If patients showed signs of severe emotional distress during the interviews or intervention, the interviewer/intervention J Rehabil Med 51, 2019