Does the Edinburgh Postnatal Depression Scale measure the same constructs across time?

Abstract

The Edinburgh Postnatal Depression Scale is the most widely used measure for screening for depression in perinatal populations. A weakness is that the factor structure of the scale is inconsistent across studies. It is unclear the degree to which this inconsistency results from variability arising from the Edinburgh Postnatal Depression Scale (EPDS). The present study aimed to determine whether the EPDS factor structure remained stable in the same individuals reporting on their levels of distress across two testing occasions. Data were analysed for 636 postpartum inpatient females who were administered the EPDS at admission and discharge from a psychiatric mother and baby unit. Exploratory factor analyses (EFAs) and confirmatory factor analyses (CFAs) were conducted separately on the admission and discharge data to determine the optimal factor structure at each time point. The EFAs and CFAs supported a two-factor model at admission and a three-factor model at discharge. Given that the EPDS did not demonstrate an invariant number of factors, no further tests of measurement invariance were conducted. The EPDS does not appear to be invariant from admission to discharge. These findings suggest that individuals may respond differently to items depending on their level of distress. Potential implications for the EPDS in terms of comparability of scores across groups/time and its screening abilities are discussed.

Factor structure of the Edinburgh Postnatal Depression Scale in a population-based sample.

“…The factor structure in the present study was consistent with some other studies (EFA in Cunningham, Brown, & Page, 2015;Pop, Komproe, & van Son, 1992;Zhong et al, 2014) or similar (Kubota et al, 2014 ;Lee King, 2012;Reichenheim, Moraes, Oliveira, & Lobato, 2011;Tuohy & McVey, 2008). All but two of these papers had sample sizes over 400 which may suggest that as the pattern of correlations becomes more stable, items one (laugh) and two (enjoyment) separate out into a separate factor of anhedonia.…”

“…For example, in contrast to our results, Cunningham et al (2015) did not find longitudinal measurement invariance in their study of the EPDS at two time points in the postpartum with their factor structures differing by severity of EPDS scores. Interestingly when EPDS scores were lower, and more comparable with this study, the factor structure was exactly the same as in this study.…”

“…Although Bina and Harrington (2015) found that anhedonia and depression factors correlated at .84 limiting the discriminant validity of these factors individually (a cut-off of .85 is commonly used as problematic for discriminant reliability (Brown, 2015), in this study the correlations between factors were appropriate for retaining them individually. Correlations were higher between anxiety and depression than between anhedonia and depression, providing tentative evidence the concept of dysphoria.…”

“…The little existing research has shown that prevalence of, and risk factors for, postnatal depressed mood and anhedonia differ across race and ethnicity (Liu & Tronick, 2014). In the field of psycho-cardiology anhedonia has been related to poor cardiovascular outcomes whilst depressed mood has not (Davidson et al, 2010).Therefore careful enquiry into these two specific symptom profiles may be an important direction in perinatal research in predicting future mental and other health problems (Truijens et al, 2014).Although Bina and Harrington (2015) found that anhedonia and depression factors correlated at .84limiting the discriminant validity of these factors individually (a cut-off of .85 is commonly used as problematic for discriminant reliability (Brown, 2015), in this study the correlations between factors were appropriate for retaining them individually. Correlations were higher between anxiety and FACTOR STRUCTURE OF THE EPDS Considering the depression items, item ten [self-harm] showed the lowest loadings (0.29 -0.37) on any factor as may be expected for an item about self-harm but it clearly fit with depression rather than as an item on its own.…”

“…limiting the discriminant validity of these factors individually (a cut-off of .85 is commonly used as problematic for discriminant reliability (Brown, 2015), in this study the correlations between factors were appropriate for retaining them individually. Correlations were higher between anxiety and FACTOR STRUCTURE OF THE EPDS Considering the depression items, item ten [self-harm] showed the lowest loadings (0.29 -0.37) on any factor as may be expected for an item about self-harm but it clearly fit with depression rather than as an item on its own.…”