Excluding items on salt and fluid restriction from the European Heart Failure Self-care Behaviour Scale - a confirmatory factor analysis

Abstract

Abstract Funding Acknowledgements Type of funding sources: None. Background Self-care is an important component in heart failure (HF) care, and worse, self-care is an independent predictor of mortality, HF hospitalization, and quality of life. The European Heart Failure Self-care Behaviour scale (EHFScB-9) is a validated scale with a nine-item scale and two subscales (autonomy-based adherence and consulting behavior). The items in the EHFScB-9 are based on evidence and consensus. However, in recent years two items regarding salt and fluid restriction have been questioned due to the lack of evidence on their effects. Aim The aim of the study was to evaluate the factor structure of the EHFScB-9 when items on fluid- (item 5) and salt restriction (item 7) were excluded. Method Confirmatory factor analysis (CFA) was used to evaluate the factor structure of the different models of EHFScB-9. Since the item responses of the EHFScB-9 are ordinal, the recommended estimation Weighted Least Squares with Means and Variances (WLSMV) was used. The WLSMV is based on polychoric correlations. The fit indices and recommended cut-off used to support model fit were: the root mean square error of approximation (RMSEA) (≤0.06), Comparative Fit Indices (≥0.95), and Tucker Lewis Index (>0.95). Reliability was evaluated by composite reliability. Results A total of 605 patients were included (mean age 66.9 years (SD=11.7), 71% (n=430) males). All response options were used, but a floor effect was evident for all items (i.e., extreme scores > 15%). The missing values for the items ranged from 2.0 to 2.6%. A total of 10 patients had missing values on all items of the EHFScB-9 and were not included in the CFA. The original two-factor 9-item version (Model 1 in table 1) presented good model fit in all fit indices, including RMSEA: 0.058 (90% confidence interval 0.043-0.073) (Table 1) and reliability of 0.678 in factor 1 "Adherence to regimen" and 0.878 in factor 2 "Consulting behavior". The RMSEA only slightly increased to 0.62 when the items on fluid- (item 5) and salt restriction (item 7) were both excluded (Model 1.2). The model presenting the best model fit was the model where item 5 (fluid restriction) was excluded from the model (RMSEA 0.047, 90% confidence interval 0.029-0.066) (Model 1.3 in the table). As expected, the reliability decreased in factor 1 ("Adherence to regimen") as the number of items was reduced. Conclusion Considering the clinical evidence and the model fit indices, both item 5 and item 7 can be excluded from the EHFScB. As expected, when reducing the number of items, the reliability decreases in factor 1 ("Adherence to regimen"). Nevertheless, the results open for an opportunity to exclude self-care items with less evidence from the EHFScB. The suggested version of the EHFScB needs to be further tested in other contexts, and adding items to factor 1 may be considered.