Antenatal screening timeline and cutoff scores of the Edinburgh Postnatal Depression Scale for predicting postpartum depressive symptoms in healthy women: a prospective cohort study

The present study was conducted as a part of a prospective cohort study which aimed to examine biological stress markers during pregnancy and their association with PPD. The target population of the study was healthy pregnant women who were receiving antenatal care at the Jikei University Hospital, which is a tertiary hospital located in central Tokyo and has approximately 800–900 deliveries per year. Between July 2014 and June 2015, pregnant women at 7–9 weeks of gestation were invited to participate in this study. The inclusion criteria were as follows: 1) healthy women (defined as women without any disease [internal diseases, mental disorders, or gynecological diseases] at diagnosis of the current pregnancy), 2) with fetal heartbeat confirmed by ultrasound, 3) receiving antenatal care at our institution from the first trimester, 4) planning to give birth at our institution, and 5) being able to answer the questionnaire written in Japanese. The exclusion criteria were women with coexisting complications (e.g., diabetes mellitus, thyroid diseases, hypertension), women with multiple pregnancies, and women who planned to deliver at other hospitals. It was pre-planned that women who miscarried or aborted at the first trimester and women who moved to other hospitals would be excluded from the analysis.

As regular antenatal care, the following information was collected from all the participants at the first visit (maternal age, parity, mode of conception, complications in their previous pregnancies, education, past medical history, smoking, and alcohol consumption) and during the second and third trimesters (lifestyle, jobs, physical and mental condition, and expected support from their family). In addition to the regular antenatal care, participants were requested to respond to the questionnaires at 8–10 weeks of gestation (first trimester), at 24–26 weeks of gestation (second trimester), at 35–36 weeks of gestation (third trimester), at 3–4 days postpartum, and at one month postpartum. Each questionnaire included the Japanese version of EPDS and questions about sleeping hours, working hours, exercise habits, and support from their husbands/partners. There have been several reports which suggested that the EPDS during early postnatal days could be a useful screening instrument for early-onset PPD [5, 20]. Therefore, we included EPDS 3–4 days postpartum in the present study.

The EPDS is a 10-item self-reporting screening tool for PPD with each item scored on a 4-point scale ranging from 0 to 3, and total scores ranging from 0 to 30. The original English version of EPDS has acceptable identical consistency and reliability [10], and the Japanese version was confirmed to be equivalent to the original English version by Okano and others [21] with acceptable internal consistency and test–retest reliability [11].

The primary outcome of the present study was PDS. In this study, we defined PDS as having a high EPDS (≥ 9) score at one month postpartum, which is considered to indicate a higher risk for PPD. The cutoff score of 8/9 at one month postpartum has been confirmed to be the most appropriate value for Japanese women (sensitivity 75%; specificity 93%) [21] and has been widely used to assess the risk of PPD in Japan [20, 22‐26].

First, we investigated the trends in EPDS scores during pregnancy and postpartum. The EPDS screening during pregnancy was performed with the aim of predicting PDS, whereas the postpartum EPDS was performed with the aim of assessing PDS at that time. In order to determine the optimal screening time and cutoff score, we developed receiver operating characteristic (ROC) curves for each trimester by plotting the sensitivity against the “1 – specificity” of each cutoff value and calculated the area under the ROC curve (AUC). A value of 0.7–0.8 indicates a reasonable predictive accuracy, 0.8–0.9, a satisfactory accuracy, and a value of 0.9 or above is interpreted as excellent [27]. The difference among the AUCs obtained from EPDS scores at the first, second, and third trimester was tested using the Stata’s roccomp command [28]. The optimal cutoff score was determined using the Youden index, which is one of the statistical methods to obtain the best cutoff value for continuous variables [29].

The demographic characteristics of the participants are presented in Table 1A. The average maternal age was 32.9 years (range 26–41). All participants were Japanese and married. The most common family type was a nuclear family (95.2%). The majority of the participants graduated from a university or graduate school and were working full-time or part-time until they entered maternity leave (72.8% and 77.3%, respectively). Twenty-one participants (16.7%) had a family history of mental illness.

Table 1B presents delivery outcomes of the participants. The average gestational age at delivery was 38.5 weeks (range 31–41). The modes of delivery were natural vaginal delivery (90, 69.8%), vacuum or forceps delivery (14, 10.8%), planned cesarean section (13, 10.1%), and emergency cesarean section (12, 9.3%). Forty-six (35.7%) participants requested and received labor analgesia. Forty-nine (38.0%) had delivery complications.

The trends of EPDS scores are shown in Fig. 2. The mean EPDS scores in the first, second, and third trimesters, at 3–4 days postpartum, and at one month postpartum were 3.51 (standard deviation [SD]: 2.67), 3.25 (SD: 3.10), 3.02 (SD: 3.06), 4.20 (SD: 4.46), and 3.89 (SD: 4.12), respectively. The mean EPDS score was highest at 3–4 days postpartum. On the other hand, the mean EPDS score was lowest in the third trimester.

The ROC curve for each trimester was developed based on the data from participants who answered the questionnaire both at that trimester and one month postpartum (Fig. 3). The statistical test for the AUCs at the first, second, and third trimester revealed a statistically significant difference for predicting PDS among them (P = 0.01). The AUC at the second trimester was higher among the AUCs during the antenatal period (0.89, 95% CI: 0.82–0.96). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each of the possible cutoff scores of EPDS at the second trimester are shown in Table 2. A cutoff score of 3/4 had a quite high sensitivity (92.9%); however, the specificity was low (63.8%). On the other hand, cutoff scores of 5/6 and 6/7 had higher specificity (85.7%, 92.4%), whereas the sensitivity was low (71.4%, 50.0%). The Youden index indicated that the cutoff score of 4/5 was reasonable for predicting PDS (sensitivity: 85.7%; specificity: 77.1%; PPV: 33.3%; NPV: 97.6%).

The number of women with an EPDS score ≥ 9 at one month postpartum was 15/123 (12.2%). Antenatal and perinatal risk factors for predicting PDS are summarized in Table 3. Family history of mental illness was the only statistically significant risk factor for PDS found in the univariate analysis.

Previous studies on the EPDS during the antepartum period generally aimed to validate the diagnostic accuracy of the EPDS for antepartum depression [14, 15, 30]. Other studies aimed to predict PDS using antenatal EPDS score; however, they obtained EPDS scores only once or twice during pregnancy at various screening times, or the EPDS was validated using only certain cutoff scores [4, 17, 19]. In our study, among the EPDS scores in the first, second, and third trimester, the EPDS score at the second trimester had the highest predictive ability for PDS. Generally, pregnant women have a more stable physical condition during the second trimester [31]; morning sickness has a non-negligible impact on women’s metal condition during the first trimester [32]; prenatal stress and physical discomfort (such as leg edema, labored breathing, and added weight) have a negative impact on health-related quality of life during the third trimester [33]. Thus, EPDS scores during the second trimester might reflect women’s mental condition more clearly without being affected by their physical condition.

There have been several studies which reported that the cutoff score of 4/5 could be used for the initial antenatal screening to detect women at risk of developing PDS, defined as postnatal EPDS score ≥ 10 [18, 19]. Our study obtained results similar to these studies. In their studies, as well as ours, this low cutoff score enabled reaching a high NPV over 96% [18, 19], which means that women scoring < 5 on the antenatal EPDS can be reassured that it is very unlikely that they will develop PDS.

The antenatal EPDS is reported to perform better for prediction of PDS when combined with other predictors/risk factors, such as a prior history of major depression before pregnancy and low partner support [4, 19]. In our study, EPDS ≥ 5 at the second trimester and family history of mental illness were identified as risk factors for PDS. There was no significant relationship between PDS and other risk factors such as socioeconomic problems, obstetric factors, and newborn conditions in our analysis unlike other studies. This may be because of the difference in background characteristics of the study populations as discussed in the Strengths and Limitations section below.

The strength of our study is that EPDS scores were obtained at the first, second, and third trimesters of pregnancy and one month postpartum. This enabled us to investigate the optimal time for screening to identify women at risk of developing PDS.

Nonetheless, our study has some limitations. The participants of the present study were pregnant women who were seen at a university hospital located in central Tokyo, who generally had good educational attainment, high socioeconomic status, and familial support. In addition, we only enrolled healthy women without co-existing diseases. This resulted in relatively low frequency of PDS and may jeopardize the generalizability of our study findings.