Revisiting Shyness and Sociability in Schizophrenia: A Psychometric Examination of Measurement Invariance and Mean Level Differences

Abstract

Although there is a long and rich empirical history of demonstrating differences on psychological self-report measures between people with schizophrenia and healthy controls, the question of whether both groups respond to psychological measures in the same way has gone largely unexplored. That is, is there measurement equivalence, or invariance, across the samples? To our knowledge, there have been no published studies on measurement equivalency in personality measures across groups diagnosed with and without schizophrenia. Here we examined the question of measurement invariance on two widely used questionnaires assessing temperament, the Cheek and Buss Shyness and Sociability Scales (CBSHY and CBSOC, respectively) between 147 stable adult outpatients with schizophrenia and 147 healthy age- and sex-matched controls. Results supported measurement invariance of the CBSHY and CBSOC across our clinical and non-clinical groups. These findings suggested that stable adult outpatients with schizophrenia and age- and sex-matched controls respond to the shyness and sociability items in the same way. We found that adults with schizophrenia reported higher levels of shyness and lower levels of sociability than healthy controls, consistent with prior studies. Findings are discussed concerning their relevance more broadly to self-report assessments of personality and psychological traits in clinical populations.

Acknowledgments

We would like to thank the members of the Schizophrenia Outpatient Clinic, Cleghorn Clinic, and Hamilton Program for Schizophrenia for all of their efforts in this project. We would also like to thank Anna Swain, Jennifer Mullen, Jonathan Jin, and members of the Child Emotion Laboratory at McMaster University for their assistance.

Data availability statement

The data that support the findings of this study are available on request from the corresponding author, ZK. The data are not publicly available due to their containing information that could compromise the privacy of research participants.

Notes

1 We selected the 5 shyness items a priori based on their factor loadings from the study by Bruch et al. (1989). This shorter shyness scale was used to minimize participant burden, increase compliance, and make the number of items identical to the original Cheek and Buss (1981) sociability measure. The 5-item shyness scale has been used on many occasions over the years with nonclinical samples (see Brook & Schmidt, 2019 for details). We previously demonstrated that this 5-item shyness measure was conceptually and empirically linked to subscales of harm avoidance and novelty seeking - from the much lengthier Cloninger Temperament Character Inventory (TCI), which is routinely used with patients with schizophrenia - in a sample of stable outpatients with schizophrenia (Jetha et al., 2013). The rationale for the Jetha et al. (2013) was to provide the clinical community with a brief measure of shyness rather than the lengthy TCI in order to reduce patient burden and increase compliance. However, a limitation of Jetha et al. (2013) was that measurement invariance was not tested.

2 We modified the control group CFA models for both shyness and sociability to obtain acceptable fit with respect to the RMSEA because our values were higher than conventional critical values would dictate. However, some evidence and discussion in the literature has suggested that models with very low degrees of freedom and relatively small sample sizes produce artificially biased and large RMSEA values, and consequently, the RMSEA should not be reported under these conditions (Kenny et al., 2015). In our analyses, degrees of freedom were small (df = 5), and the sample size was not large (N = 147). Nonetheless, because the RMSEA is an extremely popular model fit statistic in the literature, we introduced correlated errors into our control group model for both the shyness and sociability CFAs based on modification indices and similarity in item meaning (Bandalos, 2018) to obtain adequate/good fit as a prerequisite for running MGCFAs. In an effort to allay concern about post hoc modifications through the introduction of correlated errors, we ran our shyness and sociability analyses without the specified correlated error terms and found our measurement invariance results were the same.

Additional information

Funding

This article was funded by Social Sciences and Humanities Research Council of Canada.