Validation of the short-form Generalized Self-Efficacy scale (GSE-6) in rehabilitation

Abstract

Introduction Self-efficacy (Bandura, 1997) plays a key role in psychological models explaining and predicting health behaviors (Schwarzer et al., 2011). This includes physical activity, nutritional behavior and substance consumption, which are addressed, for example, in interdisciplinary multimodal rehabilitation. The General Self-Efficacy Scale (GSE) is the most widely used instrument to assess this construct and is available in over 30 language versions (Schwarzer and Jerusalem, 1995). However, for economic reasons, it is not always possible to use the GSE in studies or clinical routine. Therefore, the short form GSE-6 was proposed, which consists of six out of the original ten items (Romppel et al., 2013). However, there is no validation except for the cited study, which was applied to a sample of the general population and patients with heart failure, only. Hence, the aim of this work is to determine psychometric properties and the validity of GSE-6 among rehabilitation patients. Methods The basis of the reported analysis is a nationwide study in Germany with 2530 insured persons of the Federal German Pension Insurance (Brunger et al., 2016; Brunger and Spyra, 2017). Patients were asked to fill in a questionnaire prior to their rehabilitation. In addition, mental impairment (PHQ-4), depression and anxiety (HADS), impairment due to pain (PDI), health behavior (IRES), social support (IRES), and subjective work ability (WAS) were assessed. The sampling was stratified according to the nine disease groups most common. Psychometric properties were calculated by comparing the short and the original version (GSE-6/GSE). By analogy with the original version, the value range of GSE-6 was transformed to 10 to 40 points. The concurrent validity is reported with Spearman correlations to the other impairments and resources mentioned. Multigroup confirmatory factor analyses (CFA) including gender, age and diagnosis groups demonstrate the factorial validity of both scales. Results Mean sum scores of the short and the original version were quite similar: GSE-6 28.1 (SD: 6.0) points; GSE 27.8 (SD: 6.0) points. In 92.9% of the rehabilitees, the GSE-6 sum score diverged by the GSE sum score with a maximum of ±2 points. Soil effects existed for 2.6% of the rehabilitant patients in GSE-6 (GSE: 2.2%), ceiling effects for 0.5% (GSE: 0.4%). The mean discriminatory power of the items ranged from rix = 0.70 in GSE-6 to rix = 0.74 in GSE. The internal consistency according to Cronbach's alpha was 0.89 for GSE-6 and 0.93 for GSE. GSE-6 and GSE were correlated with rs = 0.98. The correlations of GSE-6 and GSE to other impairments and resources were in a comparable range: depression (rs = 0.61, rs = 0.62), anxiety (rs = 0.56, rs = 0.57), mental impairment (rs = 0.56, rs = 0.58); social support (rs = 0.41, rs = 0.42); impairment due to pain (rs = −0.40, rs = 0.40); subjective ability to work (rs = 0.37, rs = 0.37) and health behavior (rs = 0.30, rs = 0.31). Stratified analyzes by gender and diagnosis groups provided comparable results. Multigroup CFA confirmed the one-dimensional structure of both scales independently of gender, age and diagnosis group. Conclusions There were only minor differences when using the GSE-6 compared to the original version. The psychometric properties are slightly less favorable on the short scale - as expected due to the lower number of items - but are also at a high level. Thus, an application of the short version in rehabilitation across all major diagnosis groups seems possible without relevant losses compared to the original version. In the future, it allows for a 40% reduction in time compared to the original GSE version and may facilitate the assessment of self-efficacy in studies and for screening purposes in clinical routine.