A Systematic Review and Meta-Analysis of Depression in Postnatal Women in the Case of a Low-Income Country; Ethiopia, 2020

Maternal mental health in the postpartum period is essential for the optimal development of the newborn and appropriate breastfeeding practices. Despite this, a shortage of concrete evidence exists regarding the magnitude of postpartum depression and associated factors. This meta-analysis study was therefore done to fill such a gap. and were investigated with no A manual search for a reference list of articles was also done. Relevant data were extracted using Meta XL package and analysis was done Stata-11 meta-prop package. Heterogeneity was checked with Cochran's Q-statistics and the Higgs I 2 test. Furthermore, sub-group and sensitivity analyses were done. Egger's test and funnel plots tests were engaged to identify publication bias. Depression Scale, OR:Odds Ratio, PHQ-9:Patient Health Questionnaire-9, PPD:Postpartum depression, PRISMA-P:Preferred Reporting Items Reviews and Meta-analysis, SRQ-20:Self Reporting Questionnaire-20.

More than one in five women was with postpartum depression and factors such as poor marital relations, history of depression, poor social support, domestic violence, unplanned pregnancy, family history of mental illness were related to it. Therefore, maternal postnatal care services should be geared to incorporate this public essential health concern.
Plain English Summary To date evidences showed that 10-20% of postpartum women develop depressive disorders worldwide and this is much higher in low income countries like Ethiopia. Therefore it is essential to have a pooled estimated data regarding depressive symptoms and related factors in postnatal women so that initiation of early intervention would be implemented. In this meta-analysis study we incorporated 16 studies that assessed 11400 postnatal women which were obtained from a detailed search of pub-med, Scopus and EMBASE databases.
From the sixteen studies data was extracted using a meta-XL package and main meta-analysis was done with STATA-11 meta-prop-package. Since Higgs I 2 statistics was suggestive of heterogeneity between included studies, further sub-group and sensitivity analysis was done.
The pooled estimated prevalence of depression among postnatal women was high; slightly higher than 1 in 5 women in Ethiopia were with postpartum depression. This pooled estimated prevalence was higher in studies assessed with SRQ-20 than those assessed with PHQ-9. Besides, postpartum depression was higher in the southern region of the country than Addis Ababa.
We also established that poor marital relation, unplanned pregnancy, previous history of depression, poor social support and domestic violence were among the top contributing factors for depression in postnatal women in Ethiopia. Therefore, postnatal maternal care services should integrate depression in postpartum period and also it's associated factors.

Background
Women with depression in the postpartum period show manifestations like sleep disturbance, change in appetite patterns, feeling of sadness, recurrent guilty feeling, crying, low energy, and unease feelings of anxiety and suicidal ideas (1). Nowadays postpartum depression is increasingly taking a lion share in the burden of disease among women of 15 to 49 years of age (2). Globally,10-20% of postnatal women develops depression at a point in time (3). A survey by the World Health Organization(WHO) identified that 20-40% of childbearing age women in the developing world sustained depression in the antepartum and postpartum period at a given time (4). Besides, 19% of postnatal women in low and middle-income countries (5,6) and one among twenty postnatal women in Ethiopia (7) develops postpartum depression.
Different studies across the world reported different prevalence rates for postpartum depression.
Postpartum depression in women has a great impact on the health of both the mother and her baby.
Early discontinuation of breastfeeding (60), negative emotion and poor health care utilization (61), cognitive impairment, and behavioral defects (62), poor mother-infant bonding resulting in rejection of the infant (63), and suicidal behaviors (64) are common complications of postpartum depression.
Furthermore, a community-based cross-sectional study in Ethiopia(23) revealed that maternal postnatal depression strongly affects a child's nutritional status leading to stunting and underweight as well as inappropriate complementary feeding practices of infants.

Search strategy
Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines(65), we performed both an electronic and manual search for eligible articles. Our search for electronic libraries in Scopus, PubMed, and EMBASE and manual exploration of the reference list of articles was the backbone of the current meta-analysis. In searching articles for postpartum depression among women in the postnatal period using the PubMed database, we used the coming searching terms: (Prevalence OR Epidemiology OR magnitude AND depression OR PPD OR "depressive disorder" AND mothers OR females OR women AND postpartum period AND factor OR "risk factor" OR determinant AND Ethiopia). Besides, the search for literature in EMBASE and Scopus was done per database-specific searching guidelines. Furthermore, the reference list of included studies was done and there was no specification for studies based on the study period.
Eligibility criteria's Before the start of data refining, we stated a defining inclusion and exclusion criteria. Articles were eligible for inclusion if (1) assessed the outcome of interest among women in the postpartum period, (2) the primary outcome of interest were prevalence of postpartum depression and its associated factors, (3) design of the study was community and institution-based cross-sectional, case-control and cohort study design, (3) and (4) the study should be piloted in Ethiopia. Previously studied reviews, studies on women with an already known psychiatric problem, studies on animals, editorials, and studies reporting depression solely during pregnancy were excluded. MN and MA screened the topics and abstracts of searched articles stored in an endnote reference manager depending on the primary and prespecified eligibility criteria independently. After that, the next stage was a detailed exploration of the screened articles in the first step by the two review authors stated above independently. Any disagreement between the two authors regarding the eligibility of articles was settled through discussion in line with available logical evidence.

Methods for data extraction and quality assessment
The two declared authors (MN and MA) dig out the relevant data from the articles included in the final analysis autonomously using a standardized data extracted template. The included studies were extracted and summarized in the form of a table. Information's extracted and summarized in the table includes the author's name, publication year, study setting, study population, sample size, study design, and the assessment instrument for postpartum depression. Data from incorporated studies was extracted based on a template structured as suggested by PRISMA guidelines (65).
The modified Newcastle-Ottawa Scale (NOS) (66) was employed for the evaluation of the quality of studies. The domains of NOS scale in assessing the quality of studies include comparability between participants, representativeness of sample and sample size, statistical quality as well as ascertainment of cases.

2.4: Data synthesis and analysis
In this study, we engaged a random-effect model to work out the aggregate prevalence of postpartum depression and the associated factors for postpartum depression with their 95% CIs (67). Meta-XL version 5.3 (68) was employed to extract relevant data from included studies and the STATA11 Meta-prop package (69) was implemented to work out the pooled prevalence of postpartum depression and pooled odds ratio of the associated factors for Postpartum depression. Cochran Q-statistics and Higgs I 2 statistics (70) were also used to diagnose heterogeneity. The I 2 statistical value of and I 2 values of zero, 25, 50, and 75% connotes absence, little, moderate, and great heterogeneity respectively (70).
Furthermore, subgroup analysis and sensitivity analysis tests were also done to explore the source of heterogeneity between the included studies. A funnel plot test (71) and eggers publication bias test were used to detect publication bias.

Identification of studies
Our search with the pre-specified search strategies gives rise to an overall of 845 articles. Besides, 5 articles were obtained from the reference list of included articles making the total number of retrieved articles to be 850. Of this, we removed 24 duplicated studies before the further screening.
In the next stage, we excluded 789 of the articles simply by observing their titles. Therefore the leftover 37 articles had been completely inspected for eligibility of inclusion into the current systematic review and meta-analysis study nevertheless only 16 articles were tailored in the final meta-analysis since the rest 21 articles were also excluded because of varieties of methodological and technical flaws (Fig. 1).

Quality of Included Studies
In assessing the quality of included studies, we used the modified Newcastle Ottawa quality assessment as a gold standard. The quality of studies varies from 7 to 10 implying that none of the studies have poor quality. One of the studies has moderate quality(31) and the remaining fifteen were having good quality ( Additional file 1).
Subgroup analysis of the prevalence of postpartum depression by the tools used to measure postpartum depression.
Since the average effect size for the prevalence of postpartum depression was significantly influenced by the difference between the included studies, it was mandatory to conduct a subgroup analysis.
Therefore, we used the measurement tools for postpartum depression to perform subgroup analysis. Sub-group analysis of the prevalence of postpartum depression by the setting of the study.
Although a subgroup analysis was done based on the setting where the study was done, there was no significant difference in the pooled prevalence of postpartum depression between community-based studies (22%) and institution-based studies (21.7%).
Sub-group analysis of the prevalence of postpartum depression by the regional location of the study.
The pooled prevalence of post-partum depression was slightly higher in studies from the southern part of Ethiopia (22.6%) than the central region (Addis Ababa) (21.2%) and this difference was also significant statistically (p = 0.001).

Sensitivity analysis
We performed a leave out a sensitivity analysis to detect the source of heterogeneity. However, our result showed that the average prevalence of postpartum depression when each study was left out from the analysis ranges from 21.08% (18.47, 23.68) to 22.52% (20.34, 24.69). Therefore, the result was not outweighed by the influence of a single study (Table 2). Publication bias A scatter plot of the logit event rate of postpartum depression on the X-axis and its standard error on the Y-axis was done and showed that it seems as there was a publication bias since the graph was slightly asymmetrical (Fig. 4). However, the eggers publication bias test revealed that there was no significant publication bias (B = 71.2, SE = 44 and P-value = 0.13).
Furthermore, the pooled odds ratio of family history of mental illness (24,28,31), use of substance (21,27,34) and low income (27,28,59), stressful life event (22,59)    Possible grounds for the difference could be due to a long time gap between the previous study and the present study. The difference in the time point for the assessment of postpartum depression could also bring the difference. The previous study includes studies that screened PPD only in the three months of the postpartum period but in the present study assessment period was far beyond this.
Moreover, a socio-economic and cultural difference between high-income study subjects of the earlier study and low-income study subjects of the present study could happen and sources the variation.
Contrary to the subgroup analysis result of a worldwide meta-analysis study (8)  However, no significant variation in the magnitude of PPD was noticed between the community and institutional settings of the studies. This was consistent with a review and meta-analysis study in India(10) in which little but insignificant variation was observed across the community and institutional setting studies.
The pooled odds ratio of the poor marital relationship among the above-mentioned studies was 3.56 that implies women with poor marital relationships were 3.56 times at higher risk of developing PPD than women who good marital relationships. A meta-analysis study in India has a similar conclusion supporting this (10). The possible reason for this could be poor marital relationship increasing the burden of psycho-social responsibility of child care and other household responsibility in the postpartum period as compared to the shared psychosocial responsibility in women which good marital relation.
Besides, the pooled odds ratio for unplanned pregnancy for the nine studies reported above was found to be 3. 48. This showed that women who gave birth from an unplanned pregnancy were 3.5 times more likely to be depressed than women with a planned pregnancy. A consistent finding was reported in studies in turkey (17), Qatar (14), northwestern Brazil (73), and Iran (74). The possible reason for this could be the absence of psychological readiness in mothers with unplanned pregnancy predisposing them to be vulnerable to physiological, psycho-social challenges of pregnancy and the postpartum period.
The previous history of depression (21,27,28,34) and stressful life events (22,59) was also an associated factor for the development of PPD with a pooled estimate odds ratio of 4.33 and 3.5 respectively. A consistent finding was observed in an Indian meta-analysis study (10). Furthermore, other several studies (6,75,76) reported that the presence of depression during pregnancy, facing stressful life events during pregnancy, and prior history of depression as related to postpartum depression which was also in line with the current study.
Besides domestic violence was also having a significant association with the development of PPD with the estimated pooled odds ratio of 3.77 (21,22,27,34). A report by world health organization in 2013 (77) suggested a similar conclusion in that intimate partner violence in women predisposes to various mental health problems in the postpartum period.
Poor social support with a pooled odds ratio of 4.5 (21,(32)(33)(34)(35) also increases the vulnerability of postpartum depression. This was also supporting the findings of the Indian review study (10). Multiple earlier studies also reported that good interpersonal relationships across the social support networks increase flexibility to stress and subsidize enormously women from developing postpartum depression Furthermore, the pooled odds ratio of family history of mental illness (24,28,31), use of a substance (21,27,34) and low income (27,28,59) and perinatal complications (28, 59) was 4, 4.67, 2.87, 3.5 and 3.8 respectively. A systematic review and meta-analysis study in which 17 articles were reviewed and postpartum depression between 3 and 52 weeks postpartum periods were assessed (83) reported that substance use as a contributing factor for postpartum depression in line with the current study.

Strengths And Limitations
To start with the strength, this study utilized a prespecified search strategy through the mentioned libraries that are intended to reduce the assessor's bias. The subsequent strength was that the independent extraction of data and quality assessment of the included studies by two autonomous reviewers also lessen the reviewer's bias. Furthermore, the employment of subgroup analysis and sensitivity analysis to distinguish the foundation of heterogeneity was also strength. On the opposing side, the limits of this meta-analysis study emanate from the presence of a high difference between the included studies that might upset the deduction of the study results. Besides, the use of two or three studies in the sub-group analysis might diminish the soundness of estimate.

Conclusion
This review and meta-analysis study obtained a high pooled estimated prevalence of postpartum depression (21.9%) suggesting that more than one in five women were having postpartum depression. The pooled estimate of postpartum depression was higher in studies that used SRQ-20 (24.6%) than studies that used PHQ-9 (18.9%. Moreover, the pooled prevalence of postpartum depression was slightly higher in the southern part of Ethiopia (22.6%) than Addis Ababa (21.2%).
However, no significant difference in pooled prevalence was noticed across study settings

Availability of Data and Materials
All relevant data concerning this systematic review and meta-analysis study is incorporated in the manuscript.

Competing Interests
We authors have no competing interests for this meta-analysis study.

Funding
The cost afforded for this research work was covered by the authors themselves.      A forest plot for the pooled adjusted odds ratio of unplanned pregnancy.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download. Additionalfile1.docx