La Center for Epidemiological Studies-Depression Scale (CES-D) est un outil d’évaluation de la symptomatologie dépressive. Différentes études ont été réalisées sur la version courte en dix items, y compris chez les adolescents. Elles ont mis en évidence une structure en deux facteurs Affect négatif (huit items) et Affect positif (deux items). L’objet de la présente étude est de valider la version française de la CES-D10 sur un échantillon de 269 adolescents et de tester une structure unifactorielle en tenant compte d’un artefact méthodologique supposé sur le facteur Affect positif. Deux analyses ont été réalisées : une analyse factorielle confirmatoire afin de choisir le modèle présentant le meilleur ajustement suivie d’une analyse confirmatoire multigroupes sur la structure retenue permettant ainsi de tester son invariance selon le genre. Les résultats montrent que la CES-D10 présente une consistance interne acceptable. Le modèle le mieux ajusté aux données est celui à un seul facteur avec corrélation des erreurs pour les deux items associés au facteur Affect positif dans de précédentes recherches. L’invariance structurelle selon le genre est confortée, ce qui permet d’éviter un biais d’interprétation des différences constatées entre les filles et les garçons.

Screening depressivity among adolescents is a key public health priority. In order to measure the severity of depressive symptomatology, a four-dimensional 20 items scale called “Center for Epidemiological Studies-Depression Scale” (CES-D) was developed. A shorter 10-item version was developed and validated (Andresen et al.). For this brief version, several authors supported a two-factor structure – Negative and Positive affect – but the relationship between the two reversed-worded items of the Positive affect factor could be better accounted for by correlated errors.

The aim of this study is triple: firstly to test a French version of the CES-D10 among adolescents; secondly to test the relevance of a one-dimensional structure by considering error correlation for Positive affect items; finally to examine the extent to which this structural model is invariant across gender.

The sample was composed of 269 French middle school adolescents (139 girls and 130 boys, mean age: 13.8, SD=0.65). Confirmatory Factorial Analyses (CFA) using the LISREL 8.52 were conducted in order to assess the adjustment to the data of three factor models: a one-factor model, a two-factor model (Positive and Negative affect) and a one-factor model with specification of correlated errors between the two reverse-worded items. Then, multigroup analysis was conducted to test the scale invariance for girls and boys.

Internal consistency of the CES-D10 was satisfying for the adolescent sample (⍺=0.75). The best fitting model is the one-factor model with correlated errors between the two items of the previous Positive affect factor (χ²/dl=2.50; GFI=0.939; CFI=0.894; RMSEA=0.076). This model presented a better statistical fit to the data than the one-factor model without error correlation: χ²_diff (1)=22.14, p<0.001. Then, the one-factor model with correlated errors was analyzed across separate samples of girls and boys. The model explains the data somewhat better for boys than for girls. The model’s overall χ²(68) without equality constraints from the multigroup analysis was 107.98. The χ²(89) statistic for the model with equality-constrained factor loadings was 121.31. The change in the overall Chi² is not statistically significant. This result implies that the model is, therefore, invariant across gender. The mean scores were higher for girls than boys: 9.69 versus 7.19; t(267)=4.13, p<0.001.

To conclude, and waiting for further research using the French version of the CES-D10 for adolescents, it appears that this short scale is generally acceptable and can be a useful tool for both research and practice. The scale invariance across gender has been demonstrated but the invariance across age must be tested too.