A comparison of univariate and meta-analytic structural equation modeling approaches to reliability generalization applied to the Maslach Burnout Inventory.

ABSTRACT

Introduction: Reliability is a property of tests scores that varies from sample to sample. One way of generalizing reliability of a test is to perform a meta-analysis on some reliability estimator. In 2011, a reliability generalization meta-analysis on the Maslach Burnout Inventory (MBI) was conducted, concluding that average alpha values for the MBI dimensions ranged from 0.71 to 0.88. In the present study, we aimed to update the average reliability values of the MBI by conducting a literature search from 2010 until now and comparing to statistical procedures of meta-analysis: the Univariate approach, that were used in the previous study, and a novel meta-analytic approach based on structural equation modeling. Method: An estimation of average reliability was done based on 69 independent primary reliability coefficients for the Univariate approach. The average reliability was based on 9 independent studies in the case of the Meta-analytic Structural Equation Modeling (MASEM) approach. Given that MASEM has the additional capability of testing the internal structure of a test, we also fitted several models. Results: The data was well-suited to the bifactor model, revealing the dominance of the general factor over the domain-specific ones. Acceptable overall alpha and omega coefficients were achieved for the two of the MBI dimensions, having depersonalization reliability estimates below recommendations. Discussion: In general, the MBI can be viewed as a highly interconnected three-factor scale, being its appropriate for research purposes.