Evidence for a Somatic and Non-Somatic Factor Structure in the Patient Health Questionnaire-8 in a Military Sexual Assault Sample.

INTRODUCTION

The Patient Health Questionnaire-8 (PHQ-8) is a measure of depression symptom severity that is the 8-item version of the more widely used Patient Health Questionnaire-9 (PHQ-9). However, the PHQ-8 lacks the question about suicide ideation and is often used when questions about suicide ideation cannot be administered. A recent review of the literature on the PHQ-9 indicates mixed findings on factor structure, with evidence for both a unidimensional model and a 2-factor model of somatic and non-somatic symptoms. To date, few studies have explored the factor structure of the PHQ-8, and none to our knowledge have examined this in military samples. This secondary analysis examined this in a sample of military sexual assault survivors given their heightened risk for depression.

MATERIALS AND METHODS

Service members and veterans who experienced assault (N = 346; 49.1% female) completed the PHQ-8 in a previously published study. The parent study was approved by the Utah State University Institutional Review Board (IRB) and secondary analyses were exempted from IRB review by the Arizona State University IRB. Five structural models were tested using confirmatory factor analysis, including 1 unidimensional factor model and 4 2-dimensional factor models. The following goodness of fit statistics were compared between models: Chi-squared testing, Comparative Fit Index (CFI), Tucker Lewis Index (TLI), root mean square error of approximation (RMSEA), Bayesian Information Criterion (BIC) and standardized root mean square residual (SRMR). Strong model fit was determined by a CFI and TLI ≥ .95, RMSEA ≤ .06, and SRMR ≤ .08.

RESULTS

The 2-dimensional model with anhedonia, depressed mood, feelings of worthlessness, concentration difficulties, and psychomotor agitation/retardation specified on the non-somatic factor, and sleep difficulties, fatigue, and appetite changes specified on the somatic factor had the most optimal fit (X2 [df] = 46.19 [19], CFI = 0.98, TLI = 0.97, RMSEA = 0.06, SRMR = 0.03, BIC = 6,130.98). Other models had adequate fit, though the fit for the unidimensional model was statistically inferior.

CONCLUSION

The use of 2-factor models of depression might be superior compared to the unidimensional model in samples of military sexual assault survivors which may provide clinical utility in treating specific depression symptom clusters. Studies that wish to examine potential differences in outcomes as a function of somatic and non-somatic depressive symptoms may consider this model. Future studies should examine model fit in samples that may not have been exposed to military sexual assault.