Screening for anxiety disorders in third trimester pregnancy: a comparison of four brief measures

Abstract

Evidence regarding the accuracy of existing anxiety screeners used in pregnancy is limited. This study compares the psychometric characteristics of the Generalized Anxiety Disorder 2- and 7-item Scales (GAD-2 and GAD-7), the anxiety subscale of the Edinburgh Postnatal Depression Scale (EPDS-3A) and the two anxiety items of the Antenatal Risk Questionnaire (ANRQ-2A). Nine hundred fifty-four women completed the screening measures and anxiety modules of a diagnostic reference standard (SAGE-SR) in the third trimester. Test performance characteristics of each measure was assessed using Receiver Operator Characteristic (ROC) analysis. We applied four previously recommended criteria to ascertain the value of each measure for widespread clinical use: area under the curve (AUC ≥ 0.8, Youden's index ≥ 0.5, negative predictive value (NPV) ≥ 0.8 and positive likelihood ratio (LR +) ≥ 4.0). Prevalence for any SAGE-SR anxiety disorder was 3%. All measures yielded an acceptable AUC of ≥ 0.8, Youden's index of ≥ 0.5 and NPV of ≥ 0.8. Only the EPDS-3A, at a cut-point ≥ 5, also achieved a LR + of ≥ 4.0 (4.35) but at this cut-point sensitivity was less than 0.75. The ANRQ-2A, at its optimal cut-point of ≥ 6, was the only measure to additionally attain both a sensitivity and specificity of ≥ .75. This study expands the evidence base for brief anxiety screening measures in the maternity setting and provides empirical support for the use of the EPDS-3A and ANRQ-2A in routine screening programmes. Studies assessing the performance of these measures in samples with higher disease prevalence and broader socio-economic status are warranted.

Keywords: Perinatal anxiety; Screening; Test accuracy; Validation.