Latent structure and factor invariance of somatic symptoms in the patient health questionnaire (PHQ-15)

Somatic symptoms are highly prevalent in primary care although insufficiently understood. The Patient Health Questionnaire (PHQ-15) is a valuable screening test but it has not yet been possible to unequivocally demonstrate its latent structure and measurement invariance.

A total of 1,255 patients from 28 primary care centres suffering symptoms of anxiety, depression or somatisation participated in a clinical trial. They completed the PHQ-15 at baseline and 374 retook it at three months. Exploratory structural equation modelling (ESEM) was used to compare three models: 1) a single global factor for somatisation, 2) four specific correlated factors, and 3) a bifactor model integrating the first two models.

A multi-group invariance analysis of the best-fit model was performed: the bifactor model (χ²=25.17, df=23, p = 0.34, RMSEA=0.009, CFI=1.00, TLI=0.999). Strict invariance was good for both gender (RMSEA = 0.046, CFI = 0.973, TLI = 0.963) and age (RMSEA = 0.048, CFI = 0.964, TLI = 0.962). Configural and metric invariance were confirmed for moment of assessment, but scalar invariance was not.

The two main limitations were the sample (primary care patients with emotional disorders), which was not representative of the general population, and the utilisation of ESEM (vs. confirmatory factor analysis), which did not allow a second-order factor model to be tested.

PHQ-15 showed a bifactor structure, providing both a single global measure of somatisation and specific measures of pain, gastrointestinal, cardiopulmonary and fatigue factors. Its factor invariance with regard to both gender and age was confirmed.