Article Text
PDF
Statistics from Altmetric.com
QUESTION
Question:
Does the Edinburgh Postpartum Depression Scale (EPDS-3) anxiety subscale perform as well as the full 10 item EPDS scale in screening for postnatal depression, and how do other brief screening alternatives compare?
Patients:
Mothers (n = 199; 14–26 years) who made ⩾1 visit to the Colorado Adolescent Maternity Program with an infant ⩽6 months old.
Setting:
Primary care paediatric clinic for children of adolescent mothers, Colorado, USA; time period not stated.
Test:
EPDS-3 brief screening tool for postnatal depression. The EPDS-3 is a subscale of the full 10 item EPDS which includes the three anxiety related items. Other brief alternatives that were compared were the depressive EPDS subscale (EPDS-7) and the two items of the EPDS that resemble the Patient Health Questionnaire (EPDS-2). The score on each scale was multiplied by 10 and divided by the number of items used, so that a cut-off of ⩾10 could be used for all scales. In addition, a cut-off of ⩾3 was used for the EPDS-2 raw score. Mothers were assessed during well-child visits, and could be assessed more than once during the study period.
Diagnostic standard:
Full 10 item EPDS; a score ⩾10 indicated a need for referral for evaluation of depression.
Outcomes:
Sensitivity, specificity, negative predictive value, positive predictive value, internal reliability (Cronbach’s alpha), stability (kappa; based on 97 women who were screened twice, on average 2.1 months apart).
METHODS
Design:
Diagnostic cohort study.
MAIN RESULTS
20.6% (41/199) of the mothers screened positive for elevated depression symptoms and were referred for depression evaluation (EPDS score ⩾10). Compared with the full EPDS, the EPDS-3 (anxiety) subscale had a high sensitivity (95%) and negative predictive value (98%). The EPDS-3 displayed better sensitivity and negative predictive value than alternative brief forms of the EPDS (EPDS-2 and EPDS-7; see table 1). The EPDS-3 showed lower specificity and positive predictive value than the other brief alternatives (table 1). The reliability and stability (based on 97 repeated assessments) of the EPDS-3 was comparable with the full EPDS (EPDS-3: Cronbach alpha = 0.8; kappa = 0.6; full EPDS: Cronbach alpha = 0.9; kappa = 0.5).
CONCLUSIONS
The EPDS-3 may be a useful, time efficient screening tool for detecting postnatal depression in primary care.
ABSTRACTED FROM
Kabir K, Sheeder J, Kelly LS. Identifying postpartum depression: Are 3 questions as good as 10? Pediatrics 2008;122:E696–E702.
Commentary
There has been considerable interest in finding the briefest possible methods of identifying depression and anxiety in both primary and secondary care.1 The 2007 NICE guidelines recommended use of an adaptation of the Patient Health Questionnaire-2 (PHQ2).2 Remarkably there were no publications on these ultrashort methods in perinatal settings at the time of this recommendation but two since that time. Kabir and colleagues have taken a reductionist approach by comparing three abbreviated versions of the EPDS against the EPDS 10 item version; one version contained just the seven depression questions (EPDS-7), one contained three anxiety questions (EPDS-3) and one version had two questions which were very similar to the PHQ2. There is no attempt to calculate optimal cut-offs using an receiver operating curve but thresholds were chosen to minimise false negatives.
The original EPDS identified 41 probable depressions and 158 probable non-depressions. The authors do not present summary accuracy statistics but a post hoc calculation using the fraction correct statistic (overall accuracy) is helpful. This shows that the 7 item depression version identified 91% of these probable cases and non-cases, compared with 87% for the EPDS-2 (cut-off ⩾3), 83% for the 3 item anxiety version and 80% for the EPDS-2 (cut-off ⩾2). The authors actually recommend the EPDS-3 because of its high negative predictive value (NPV) of 98%. The NPV can be misleading if judged alone because a test is most useful as a rule-out method when nearly all those without the index condition test negative. This is the basis for the “clinical utility index” which in this case reveals the EPDS-7 to be the optimal rule-in and rule-out method although the EPDS-2 (at a cut-off ⩾3) was not far behind.
A recent similarly powered study compared the PHQ2 with the EPDS and found the positive predictive value (PPV) of the PHQ2 was only 24% but the NPV was 99%.3 Together these studies offer preliminary evidence that ultrashort methods can be used as a first step rule-out screen in perinatal settings but both studies are limited by the absence of a robust diagnostic interview as a gold standard.
Footnotes
Source of funding: None stated.
Footnotes
Competing interests: None.