The diagnostic accuracy of the Edinburgh Postnatal Depression Scale without the self-harm item: Does culture matter?, Journal of Psychiatric Research

The diagnostic accuracy of the Edinburgh Postnatal Depression Scale without the self-harm item: Does culture matter?
To the Editor: We read with keen interest the recent article by Chen et al. (2023), in which the authors evaluated the performance of the Edinburgh Postnatal Depression Scale (EPDS) without the self-harm item, called EPDS-9, compared to the complete EPDS, called EPDS-10. They focused on identifying depression among people who are pregnant or postpartum. The authors concluded that the shortened EPDS-9 performs as well as the EPDS-10, suggesting it as a potential replacement for the full-length EPDS.
Our research partially supports the findings of Chen et al. (2023).
Our study sample comprises 1153 pregnant women and 309 postpartum women. These participants were enrolled from 11 healthcare centers located throughout Italy (masked citation). The characteristics of the participants are detailed in a separate publication (masked citation). Trained psychologists used unstructured clinical interviews and patient-rated Patient Health Questionnaire-9 (PHQ-9) and EPDS questionnaires to evaluate participants’ depression.
Our findings indicate a correlation of .998 between EPDS-9 and EPDS-10, observed in both the antepartum and postpartum groups. Only 1% of the participants were negative at EPDS-9 cutoff points of <10 but had a non-zero EPDS item 10 score, and 2% at EPDS-9 cutoff points of <13. Furthermore, EPDS-9 demonstrated excellent accuracy in distinguishing EPDS-10-based depression screening in both perinatal groups, in each of the four commonly used cutoff scores (Levies et al., 2020; Quip et al., 2023). We used the PHQ-9 as a criterion to compare the performance of the EPDS-9 versus EPDS-10, using a cut-off value of 13 (which is indicated as the most appropriate for the detection of major depression in perinatal people [Levi’s et al., 2019]). EPDS-9 and EPDS-10 demonstrated comparable sensitivity, specificity, and area under the curve (AUC) performances. In the antepartum group, both the EPDS-9 and EPDS-10 (a) show declining sensitivity with increasing cutoff values, (b) have high specificity across all cutoff values, and (c) have AUC values that suggest they perform reasonably well, though their performance declines with increasing cutoff values.
Comparison of AUC values between EPDS-9 and EPDS-10 suggests that there are no significant differences in performance between the two versions of EPDS at cutoff values of 10, 11, and 13. However, there appears to be a significant difference in performance at a cutoff value of 12, with the EPDS-10 performing better. Regarding the postpartum group, although the AUC remains relatively high for both EPDS-9 and EPDS-10 across all cutoff values, the equivalence tests showed a statistically significant difference at all cutoff values (see Table 1), indicating that there is a significant difference in overall test performance. Specifically, the EPDS-10 outperforms the EPDS-9 at all cutoff values. We also examined the predictive potential of EPDS-9 for responses to the EPDS self-harm item (item 10).
The AUC of EPDS-9 against self-harm responses varied depending on the frequency level, which could be an area for further study. Specifically, EPDS-9’s AUC against self-harm above the frequency of “hardly” ranged from 0.716 to 0.826, except for cutoff 13 in the antepartum group, where it dropped significantly to 0.288. This decrease in AUC at the cut-off point of 13 suggests that EPDS-9’s ability to predict self-harm responses decreases when this more conservative threshold is used. The AUC against self-harm above the frequency of “sometimes” and “often” ranged, respectively, from 0.712 to 0.826 and from 0.445 to 0.675. These variations emphasize the importance of