Abstract
Recent research has suggested that the shared characteristics and co-occurrence among depression, anxiety, and female sexual dysfunction may represent a shared underlying liability (i.e., the internalizing spectrum, which traditionally accounts for the overlap between depression and anxiety in psychopathology research). To date, however, whether common covariates shared by these symptom domains might instead account for the interrelationships has not been examined. Three such potential confounders include intimate relationship quality, social support, and selective serotonin reuptake inhibitor (SSRI) use. We therefore aimed to examine whether and to what extent controlling for these covariates affects the structure of an internalizing spectrum model that includes sexual problems. Participants (n = 525, mean [SD] age = 32 [11.1]) were women who participated in an online self-report survey and were in a current intimate relationship. Hierarchical exploratory structural equation models of the internalizing spectrum were compared before and after controlling for relationship quality, social support, and SSRI use and were markedly similar, indicating that the model was robust. This study provides further evidence that the internalizing spectrum can account for the relationships among depression, anxiety, and low sexual function in women, which has potential implications for diagnosis and treatment.
Notes
1 Morin et al. (Citation2016) also recommended fitting additional models for comparison to the hierarchical ESEM. We fit a CFA model as a baseline for comparison to determine whether the additional parameters estimated in the ESEM framework were necessary. A three-factor CFA model without cross-loadings between factors provided poorer fit to the data than the ESEM model (CFI = .94, RMSEA = .07, SRMR = .06, and a sample size adjusted Bayesian information criterion [BIC] value 74 points larger than the ESEM model). Given BIC in particular favours parsimony, these results indicated that the cross-loadings in the ESEM model are important to include. We subsequently fit a bifactor ESEM model to compare to the hierarchical ESEM model that was the primary focus here: A bifactor ESEM had similar model fit to the hierarchical ESEM (CFI = .99, RMSEA = .05, SRMR = .02); although the smaller BIC value indicated that the bifactor model had better fit (∆BIC = 14 points), these fit indices have all been found to be biased towards the bifactor model, which has a tendency to overfit data due to its flexibility (Greene et al., Citationin press). Given the hierarchical model aligns with extant research on these relationships in a SEM framework, we retained the hierarchical ESEM model as the focus in subsequent analyses.
2 We also tested an alternative approach to controlling for the covariates by regressing all of the observed variables in Model 1 on relationship quality, social support, and SSRI use simultaneously. This does not completely remove the shared variance between the indicators in the model and the covariates—as the model is fit to a new covariance matrix that treats the covariates as three additional observed variables—but it does represent a conceptually similar comparison to Model 2 conducted in an SEM framework. As for Model 2, the differences in standardized factor loadings for the indicators compared to Model 1 were small (Mdiff = .02, SDdiff = .05, rangediff = − .09–.13). The differences in the factor loadings for fear and distress on internalizing were in the same range (λdiff = .12 and .06, respectively), whereas the sexual problems loading on internalizing was reduced by λdiff = .32. This SEM model was also highly similar to the factor structure in Model 2 (Mdiff = .01, SDdiff = .05, rangediff = − .08–.13). Imposing the factor loadings, means, and variances from Model 1 onto the SEM model resulted in a non-significant Satorra-Bentler scaled chi-square difference test (χ²(33) = 19.81, p = .966), and a tendency towards improvement in model fit from CFI = .97, RMSEA = .06, and SRMR = .02 to CFI = .99, RMSEA = .03, and SRMR = .06.