Predictive accuracy of Edinburgh postnatal depression scale assessment during pregnancy for the risk of developing postpartum depressive symptoms: A prospective cohort study

J. L. Meijer, C. Beijers, M. G. Van Pampus, T. Verbeek, R. P. Stolk, J. Milgrom, C. L.H. Bockting, H. Burger

Research output: Contribution to journalArticleAcademicpeer-review

43 Citations (Scopus)

Abstract

Objective: To investigate whether the 10-item Edinburgh Postnatal Depression Scale (EPDS) administered antenatally is accurate in predicting postpartum depressive symptoms, and whether a two-item EPDS has similar predictive accuracy. Design: Prospective cohort study. Setting: Obstetric care in the Netherlands. Population: One thousand six hundred and twenty women from the general population. Methods: Mean values, area under the receiver operating characteristics curve (AUC), sensitivity, specificity and predictive values of antenatal EPDS for the likelihood of developing postpartum depressive symptoms were calculated. Analyses were repeated for each trimester, several cut-off values and a two-item EPDS (low mood and anhedonia). Main outcome measures: Postpartum depressive symptoms, defined as EPDS score ≥10. Results: Mean EPDS scores were significantly higher during each trimester in women with postpartum depressive symptoms than in those without the symptoms (P < 0.001). Using the prevailing cut-off (≥13), the AUC was reasonable (0.74), sensitivity was 16.8% (95% CI 11.0-24.1), positive predictive value was 41.8% (95% CI 28.7-55.9), specificity was 97.8% (95% CI 97.0-98.5) and negative predictive value was 92.7% (95% CI 91.3-94.0). Using a lower cut-off value (≥5), sensitivity was 70.8% (95% CI 62.4-78.3) and specificity was 65.4% 4 (95% CI 62.9-67.8), but positive predictive value was low (15.9%, 95% CI 13.1-19.0). Negative predictive value was exceedingly high at 96.0% (95% CI 94.6-97.2). Results were similar during the second and third trimester. The predictive accuracy of the two-item EPDS appeared inferior. Conclusions: The EPDS was not sufficiently accurate in predicting risk of postpartum depressive symptoms. Nevertheless, when using the ≥5 cut-off value, it may be adequate for initial screening, followed by further assessments and possibly antenatal intervention when positive. Furthermore, when negative, women may be reassured that postpartum depressive symptoms are unlikely. A two-item version showed poor predictive accuracy.

Original languageEnglish
Pages (from-to)1604-1610
Number of pages7
JournalBJOG: An International Journal of Obstetrics and Gynaecology
Volume121
Issue number13
DOIs
Publication statusPublished - 1 Dec 2014

Keywords

  • Antenatal depression
  • postpartum depression
  • predictive accuracy

Cite this