Retest of the reliability and construct validity of the Caregiver Strain Index positive (CSI+) instrument:

Background and objective: This study reports a retest of the reliability and validity of the Caregiver Strain Index including the same additional positive items as tested by Al-Janabi et al. (2010) using a new, large and heterogeneous sample of informal care givers in the Netherlands. This study included more characteristics of the caregiver, care recipient, care situation and other burden measures and gathers therefore more information on the impact of caring on different aspects of the caring experience. This study analyses which evidence of construct validity can be obtained of the CSI instrument after addition of five positive items (CSI+) and if its reliability is retained in a new heterogeneous sample of informal caregivers of the Dutch society. Methods: An online questionnaire was used to gather data among informal caregivers of 18 years or older selected from a representative sample of the Dutch population in terms of age and gender (n=1244). Construct validity was used to test if the new CSI+ instrument is measuring what is intended to do. The construct validity of the CSI+ instrument is tested by clinical validity and convergent validity. Tests of the clinical validity of the CSI+ instruments include chi-square tests, univariable and multivariable tests. Chi-square tests were used to assess whether the characteristics of informal caregivers, care recipients and care situation had a significant impact on the answer of the positive items. Univariable and multivariable analyses were performed to test if the sum scores of the positive items, CSI scores and CSI+ scores were associated with the characteristics of interest. The clinical validity was also assessed by performing subgroup analysis including only the youngest and oldest caregivers. Also a subgroup analysis was performed of caregivers with the lowest and the highest CarerQol-VAS score. Convergent validity was assessed on the basis of correlation patterns between the score of each of the positive items of the CSI+ instrument and other burden scales which were also included in the digital questionnaire (i.e. CarerQol-7D, CarerQol-VAS, PU, SRB and ASIS). The correlation patterns were tested by using Spearman’s correlation coefficients. Reliability was assessed by measuring the internal consistency of the CSI and CSI+ instrument. Factor analysis of the CSI+ scale was performed in order to explore any underlying clustering of items. The numbers of factors were determined by analyzing Kaiser’s criterion and scree plot analysis. Factor loads of the CSI+ items were found with principal axis factoring (PAF) method, whereupon the extracted factors were rotated using direct Oblimin with Kaiser Normalization. Results: The findings of the tests of the clinical validity in this study were mostly in accordance with the hypotheses based on other studies of the CSI and CSI+ instrument. Multivariable analysis showed that being a young caregiver, having a full time job, caring for a partner or parent, being in poor health, caring for someone in poor health, caring for someone with a mental or chronic disease, caring for someone in need of continuous surveillance, caring for a large part of the week, spend a large amount of hours a week on providing care and living together with the care recipient were all significantly associated with higher and CSI scores. The same characteristics of the multivariable analysis of the CSI were also significantly associated with higher CSI+ scores in the multivariable analysis, except, caring for a male was also significantly associated with higher CSI+ scores, while caring for a parent and caring for someone with a chronic disease were not found to be significantly associated with higher CSI+ scores. The findings of the convergent validity were also in accordance of the results of previous study and in line with the hypotheses of our study. The convergent validity of the CSI+ was stronger than the convergent validity of the CSI. The reliability of this study was good (Cronbach’s alpha=0.81). Factor analysis resulted in a three factor solution and showed that only one positive item ‘enough time to self’ load onto the same factor as nine negative items. The other four positive items load onto the same factor, without any negative dimension of the CSI. Conclusion: The original CSI is still a valid and a reliable instrument to measure the caregiver burden. However, tests of the clinical and convergent validity shows that adding five positive items to the original CSI instrument gives better results. The factor analysis that suggested that the positive and negative items of the CSI+ represents three different dimensions of care; practical burden of care, satisfaction and emotional strain. Four positive items might represent a self contained aspect of care. The findings of our study support the inclusion of positive items in the CSI+ instrument to measure the caregiver burden and meet the need of a subjective burden instrument which combines both negative and positive aspects of informal care.