Screening for depression in pregnant women from Côte d'Ivoire and Ghana: Psychometric properties of the Patient Health Questionnaire-9

Background: Major depression in antepartum women is a considerable health problem. This article aims at exploring the psychometric properties of the 9-item Patient Health Questionnaire (PHQ-9) in West African pregnant women. Methods: In a cross-sectional survey, the PHQ-9 was administered to n=639 Ivorian and n=389 Ghanaian women in their last trimester of pregnancy (gestational age range: 28-40 weeks) in 2010-11. Statistical analysis applied methods from both classical test theory (CTT) and item response theory (IRT). Results: Internal consistency was Cronbach's α=.65 in Côte d'Ivoire and α=.68 in Ghana. Investigation of factorial validity by confirmatory factor analyses showed that unidimensionality of the PHQ-9 was sufficient. Rasch analyses resulted in excellent item infit and outfit measures. Yet, unidimensionality was questionable in residual principal component analyses. IRT analyses suggested that the response categories were not utilized as intended. Analysis of differential item functioning revealed interviewer-related item bias for several items in both samples. Item-person-fit was not ideal because the PHQ-9 items showed a low discriminability in the region of the latent trait where the majority of the women from the general population were located. Convergent validity was demonstrated by correlations between the PHQ-9 and two measures assessing anxiety and perceived disability. Limitations: Both samples were quite homogenous regarding residence in urban areas and gestational age. Conclusions: In our samples of African pregnant women, depression measured with the PHQ-9 does not appear as an entirely homogenous construct. However, the use of the sum score of the PHQ-9 is appropriate for depression screening purposes.