Prevalence and Incidence of Postpartum Depression among Chinese Women: A Longitudinal Study

Purpose Postpartum Depression (PPD) is signicant public health and clinical concern regarding to women of reproductive age. This study aims to examine the prevalence, incidence and persistence of at 1 and 6 months after delivery among Chinese women. Method This is a prospective cohort study Participants were recruited Wuhan Women September 2018. Edinburgh Postnatal Depression Scale (EPDS) was used to assess possible (scores of EPDS >9) and high-level (scores of EPDS >12) of depression at 1 and 6 months after delivery. Prevalence and incidence with 95% condence intervals were calculated.

month, employment during pregnancy and multiparas were risk factors for elevated EPDS scores at 6 months(P<0.05)..

Conclusion
Data of incidences suggests that later-onset PPD after delivery. These results suggest that continuous postpartum monitoring maternal depression is helpful to identify high-risk groups as early as possible and reduce the incidence of PPD.

Background
Maternal postpartum depression (PPD) refers to a constant low mood, with the symptoms of feeling sad, worthless and hopeless, etc., discerned in mothers who have recently gone through their childbirth [1,2]. It is a common and serious mental disorder with long-lasting adverse consequences for the mother and child, as well as family harmony [3]. Two recent reviews found that the pooled prevalence of PPD based on self-report scales were 13.1% [4] and 19.7% [5], respectively. PPD negatively affects a mother's ability to engage with her family [6], mother-infant bonding and reduces breastfeeding [7]. It also constitutes a serious threat to the offspring' the physical, cognitive and psychological development during childhood and adolescence [8]. Given its high prevalence and impact, PPD has ranked as one of priorities of public health and clinical concerns.
There were numbers of studies have investigated the prevalence of PPD, however, the prevalence ranged widely because of variations in de nition, rating scale, and assessment time [9], which is not conducive to the formulation of policies and interventions. Evidence suggested that depressive symptoms are transient among some women who experience depressive symptoms during pregnancy or a one-off, short-lived episode after childbirth and recovers to previous good level of mental health, while among other women, depression is a persistent and/or recurrent problem [4]. This means that the time trajectories of PPD are heterogeneous and the prevalence of PPD may change over time. However, most of previous studies did not assess the prevalence of PPD over times, which thus were unable to provide information on the prevalence of PPD at several points in time [10]. In addition, previous studies did not distinguish between point prevalence and period prevalence [11], so it is not possible to explore the extent to which women recover from depression or experience persistent symptoms after childbirth. Moreover, because of di culties in de ne the onset of PPD by cross-sectional approach, it is unclear if the point estimates of PPD reported in many studies are new or the persistent ongoing episode [11], which mirror that data on incidence of PDD is scarce. In light of these issues, some researchers underlined the need of research on the prevalence and incidence of PPD using a sound methodology [11,14], such as longitudinal studies [13].
Although the causes of PPD have not been clari ed, it is clear that the presence of certain in uencing factors increase the risk of PPD [15]. These main risk factors include depression/anxiety during pregnancy or a previous depressive illness, life stress, lack of social support, poor marital status, poor family economic status, younger maternal age, lower educational level and smoking during pregnancy [1,3,16] Recent studies shown there were differences in risk factors for new onset and recurrent PPD, [17].
However, previous studies investigated risk factors of PPD did not distinguish women with a history of depression and/or postpartum depression from those with a rst-ever episode during the postpartum period [11], thus, it is remain unclear whether the risk factors of recurrent PPD and rst-ever episode PPD are different. Phillips et al. also found that women with recurrent PPD had more personality vulnerabilities and maternal-speci c negative attitudes than women with new onset PPD [18]. When PPD is the rst episode of depression in the mother's life, it may be unexpected and di cult for her or others close to her to identify, which may obstacle them from seeking for adequate treatment [19], and in turn leads to adverse consequences for the mother, child and family. Therefore, it is important to investigate the incidence of different subtypes of PPD, since it may provide new insights into the determinants of PPD.
Based on literature review, this study aims to examine 1) the prevalence and incidence of PPD at 1and 6 months' postpartum; 2) the persistence of these symptoms from 1 month to 6 months; and 3) the predictors of elevated EPDS scores from 1 month to 6 months.

Study Design and Participants
The present study is part of a prospective prenatal cohort study at Wuhan Women and Children Medical Care Center (Wuhan, Hubei, China). Details of the study design and methodology have been reported in our previous work [20,21]. The participated women completed face to face interviews at 1 and 6 months  2009) and has been demonstrated a satisfactory psychometric properties [23]. According to the previous studies in mainland China and Hong Kong [23,25], we used the cut-off value proposed by Cox et al., where score > 9 indicates possible depressive symptomatology, and has been recommended for community-based screening to identify individuals who require further follow-up, while a score > 12 indicates high level of depressive symptomatology [22].

Other variables
Based on the literature review, we collected various of co-variables of PPD in present study, which included the following four aspects: socio-demographic factors, pregnancy-related factors, deliveryrelated factors and newborn-related factors. Socio-demographic factors included maternal gestational age (continuous data), educational level (≤ junior school, senior school or ≥ college), annual family income (< 30,000, 30,000 ~ 49,000, 50,000 ~ 99,000, 100,000 ~ 190,000, or ≥ 200,000), and pre-pregnancy employment status (yes or no). The pregnancy related factors were passive smoking during pregnancy (yes or no), accidental pregnancy (yes or no), parity (primparous or multiparous), gestational and employment status during pregnancy (yes or no). The delivery-related factors included mode of delivery (vaginal delivery or cesarean delivery), and preterm birth (preterm birth described as a living birth before 37 weeks gestation (yes or no) [26]. The newborn-related factors included birth weight (< 2500 g, 2500 ~ 3999 g, or ≥ 4000 g), infant gender (male or female).

Statistical Analysis
At each observation time (1 and 6 months after delivery), the continuity-corrected method was used to calculate the prevalence and incidence with 95% CIs. Point prevalence was calculated by dividing the number of cases at each time by the total number of participants with no missing outcome data at that time. Incidence was determined by dividing the number of new cases since the preceding observation time by the number "at risk" for the condition. Persistence over time was calculated as the proportion of participants who had the condition at the rst 2 observation times. The denominators for these calculations were all women with the outcome at the preceding time. Analyses were conducted for the overall sample and disaggregated by EPDS scores difference at 1 month to 6 months. Speci cally, all women divided into the following three groups: new-onset group (EPDS scores<10 at 1month and ≥ 10 at 6months), persistent group (both EPDS scores ≥ 10 at 1 month and 6 month), improved group (EPDS scores ≥ 10 at 1month and < 10 at 6months and non-PPD group (both EPDS score < 10). Demographic variables were examined sequentially among the subtypes.
Then, generalized linear mixed effects model were used to identify any factors that might predict a high EPDS score at 6 months' postpartum. Factors which we included were based on bivariate analyses of the potential predictors, and only those found to be signi cantly predictive on bivariate analysis (P < 0.1) were retained in the nal model. Baseline EPDS was examined in the nal model as a predictor for later elevated score. The goal of these analyses was to identify any factors that might be useful in predicting the EPDS scores at 6 months based on 1 month scores.

Results
Overall Prevalence, Incidence and Persistence Table 1 shows the prevalence and incidence of possible depressive symptomatology and high level of depressive symptom for participant at 1 and 6 months after delivery. The prevalence of XX was higher at Note: CI, con dence interval; NA, not applicable; a More than 9 on the Edinburgh Postnatal Depression Scale (EPDS); b More than 12 on the EPDS.
Comparisons of Basic characteristics among subtypes Table 2 displays comparing the basic information strati ed by women with an elevated EPDS at 6 month postpartum (new-onset group), women with EPDS less than 10 on both occasions (non-PPD group), women with EPDS at more than 10 on both occasions (persistent group), and women with EPDS declined at 6 month postpartum (improve group). Univariate analysis showed that there were signi cant differences among subgroups on the maternal age, parity, employed during pregnancy and birth weight of infant (all P < 0.05). Factors that may predict a high EPDS score at 6 months' postpartum included a baseline EPDS higher than 1, multiparas and employment during pregnancy. Of these factors, a baseline EPDS higher than 1 was risk factor (OR = 6.5, 95%CI: 6.3-6.7), multiparas and employment during pregnancy were protective factors (OR = 0.4, 95%CI: 0.2-0.6 and OR = 0.2, 95%CI: 0.1-0.5, respectively). In the mode, maternal age and birth weight of infant did not predict or prevent an increase in EPDS scores at 6 moths. (Table 3)

Discussion
In this longitudinal cohort study, we examined the prevalence, incidence, and persistence of postpartum depressive symptomatology over the 6 moths postpartum among Chinese Han women. The rate of possible depressive symptomatology (the score of >9 on the EPDS was judged for possible depressive symptoms) was high at 1 month, while the rate for high depressive symptomatology was relatively low. In addition, incidences of possible depressive symptomatology and high depressive symptomatology at 6 months were moderate. Besides, our generalized linear mixed effects model analysis found that the factors associated with scores uctuation from 1 month to 6month included employed during pregnancy, parity (≥2) and EPDS score >9 at 1 month.
The rate of possible depression symptomatology at 1 month was 37.7%, which is higher than the overall Chinese immigrant women prevalence of 24.4% reported by Dennis et al [27] and the prevalence of 13% reported by longitudinal study in Western countries [28,29]. This rates also signi cant higher than the average prevalence of 13.1% reported from a recent systematic review conducted by Underwood et al [4]. It is worth noting that these longitudinal studies mentioned above were measured at 6-8 weeks after childbirth and counted in the Western countries, that making their results di cult to interpret or compare with our data. Furthermore, of those with possible depressive symptomatology at 1 month, more than one-fth women continued to have symptoms at 6 months. This result proves our previous hypothesis that postpartum depressive symptoms for mother are dynamic after childbirth, becoming a mother has on emotional health considered as an event within a process, rather than simply a turning point. These rates were lower than the 19.4% of Italian women 8 weeks postpartum reported by Abdollahi et al. and other recent longitudinal studies, while they were higher than that of New Zealand women at 9 months postpartum.
In our study, the rates for high depressive symptomatology (the score of >12 on the EPDS was judged for high depressive symptoms) were 15.2% at 1 month and 13.1% at 6 months after child birth, respectively.
They were lower than the 19.4% of Italian women 8 weeks postpartum reported by Abdollahi et al. [30], as well as other recent longitudinal studies [31,32]. But they were higher than the 9 months postpartum rate of New Zealand women [33]. Differences in measurement time and cultural criteria may account for the higher rate of PPD in the present study, especially for possible depressive symptomatology. Firstly, the difference of the time periods used in the research of PPD may have great in uence on the prevalence of PPD. Generally speaking, the rates obtained in 4 to 6 weeks postpartum are higher than those conducted closer to delivery (eg. at 1week) [34] or later postpartum (eg.at 3 month to 12 month. Secondly, special cultural beliefs and postnatal practices in China may have caused the higher rate in our survey. "Doing the month" is the most signi cant cultural event associated with childbirth for Chinese mother. Chinese women remain in seclusion with activity, dietary restrictions and rest while her mother or mother-in-law takes care of both the baby and the household during this time [35,36]. Recent studies indicated that the traditional Chinese practice of postpartum care, such as "Doing the month" may be a potential in uencing factor for PPD. For instance, a lot of women argue that traditional period "Doing the month" causes con icts among family members involved and often restrict the smooth transition of women in their maternal role [35]. However, the relationships between the traditional Chinese practice of postpartum care and PPD were inconsistent [37,38]. These ndings suggest that research is needed to elucidate psychosocial, cultural, and systemic factors that may result in a longer period of vulnerability for this population, who may need ongoing monitoring and support for the entire postpartum year. Thirdly, the cut-off points differed across studies is another potential reason for different rate of PPD among studies. In our study, we selected two cut-off scores of the EPDS which focus on the cognitive and affective features of depression [39]. These means that for such Chinese, they may have a tendency towards somatic rather than cognitive presentation and may be missed using a higher cut-off values. In contrast, lower cut-off score is appropriate for identifying at-risk groups [23,25]. Previous studies shown the incidence of PPD ranged from 3.4% to 34% worldwide [40]. In our study, incidences of possible depressive symptomatology and high depressive symptomatology at 6 months were 11.8% and 7.2%, respectively. This result concurs with the most previous studies, which reported the incidence ranges between 12 and 20 % [41]. Our result further con rmed that a signi cant number of PPD onset during postpartum. Whereas for many women depressive symptoms remit within the rst few months postpartum, the high incidence of depressive symptoms beyond the early postpartum period indicates that depressive symptoms are not necessarily limited to the early postpartum period. This highlights the need for more incidence studies to be conducted, in order to identify the predictors for new onset of PPD so the treatment can be administered effectively. According to the results of generalized linear mixed effects model analysis in our study, the factors associated with scores uctuation from 1 month to 6month included employed during pregnancy, parity (≥2) and EPDS score >9 at 1 month. The most frequently described risk factor for developing PPD is a personal history of postpartum/nonpuerperal depressive episodes [42]. Beck et al. estimated 30% increased risk of developing PPD among with women with a personal history of major depressive episode. In our study, the association between personal history of a postpartum depressive symptom at 1 month and 6 month was even higher (OR=6.5, 95%CI: 6.3-6.7). These results indicate that the association between depression during pregnancy/past episodes of depression and the PPD has been recognized, the extent to which depression during pregnancy/past episodes of depression confers and may modify PPD risk has still unclear, which need further study [42]. In our study, working during pregnancy was found to be a protective factor (OR=0.2, 95%CI: 0.1-0.5). The reasons for this result may include the following two aspects: rstly, working mothers usually have higher income level. Adequate nancial means for raising an infant indicates a low level of stress, which can reduce the incidence of depressive symptoms; secondly, work can enable mothers to get more social support in addition to their families. However, the association between working and PPD has been inconclusive. Furthermore, we found multiparas to be at lower risk at 6 months postpartum (OR=0.4, 95% CI: 0.2-0.6). This result was also found in Satoh et al's study [12]. They attributed this to decreased obstetric events of multiparas. Compared to primiparas, multiparas have more con dence in playing the role of mother may be another potential reason. In contrast, however, Fiala et al. [42]found primiparas had a lower risk at 6 months postpartum in their study. In view of the lack of research in this area, the above contradictions need to be con rmed and speci ed by further research.

Limitations
A signi cant limitation of our study is that the depressive symptoms of all participants were measured at 1 month and 6 months after childbirth, and antenatal depressive symptoms were not assessed. Therefore, when identifying the in uencing factors of PPD, the effects of antenatal depression were not taken into account. Our data are based on EPDS, a self-report scale, which is not the same as a clinically con rmed PPD diagnosis. As a result, we cannot report the rate of newly identi ed depression, but only the rate of women who have an increased risk of depression. Furthermore, due to the use of hospitalbased study design, the selection bias is inevitably, which may limit the generalizability of our results.

Conclusion
Understanding the development trajectory of maternal PPD is helpful to identify the most bene cial moments of intervention. Our ndings showed that the prevalence of maternal depressive symptoms decreased from 1 moth to 6 months after childbirth. Furthermore, our study highlights the signi cant number of PPD onset during postpartum. Our results point to the need to extend that recommendation to continued evaluation of women beyond the early postpartum period and throughout the rst year after childbirth. Declarations