Epidemiology of postnatal depression and its associated factors in Africa: A systematic review and meta-analysis

Introduction Postnatal depression (PND) is a major cause of negative health-related behaviors and outcomes during infancy, childhood and adolescent period. In Africa, the burden of postnatal depression is high. However, it is under-investigated hence under-treated. To fill this information gap and to advise further interventions, we aimed at analyzing its epidemiology in Africa. Methods We searched observational studies conducted in Africa and published in between 01/01/2007 and 30/06/2018 in CINHAL, MEDLINE, PsycINFO, Psychiatry online, PubMed, SCOPES, and Emcare databases. We assessed the quality of the studies using the Newcastle Ottawa Scale (NOS) and included studies with good quality. We evaluated the heterogeneity using the Higgins I2 statistics. We used a random-effects model to pool estimates. We assessed publication bias using the funnel plot and Egger's test statistics and adjusted using Tweedie’s and Duval Trim and Fill analysis. The protocol has been registered in the PROSPERO (Protocol No. CRD42018100461). Results Nineteen studies involving 40,953 postnatal mothers were part of this systematic review and meta-analysis. The overall pooled prevalence of PND was 16.84% (95% CI: 14.49% –19.19%). The odds of having PND was higher among women with a poor obstetric condition (POR = 2.11; 95% CI: 1.11–4.01) and history of adverse birth and infant health outcomes (POR = 2.85; 95% CI: 1.29–6.25). Having a history of common mental health disorders (POR = 2.47; 95% CI: 1.51–4.04), poor social support (POR = 2.06; 95% CI: 1.05–4.05), lower economic status (POR = 2.38; 95% CI: 1.75–3.23), and those who had exposure to a different form of intimate partner violence (POR = 2.87; 95% CI: 1.60–5.16) had higher odds of PND. Conclusion While robust prevalence studies are scarce, our review indicated a high prevalence rate of postnatal depression. The analysis also identified postpartum women at increased risk of PND. Therefore, there is a need to design and escalate comprehensive strategies to decrease its burden, focusing on those women at risk of PND.

women", "postnatal mothers", "major depression", "psychological morbidity", "depressive symptoms", "major depressive disorder", "depressed mood", "clinical depression", and "depression". Example of the search strategy in MEDLINE: (exp POSTPARTUM DEPRESSION/) or (Depress � .tw,id.) AND (postnat � or postnatal wom?n or postpartum wom?n).tw,id.) [mp = title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]) AND ((exp Psychosocial Factors/ or exp Risk Factors/) or (risk � .tw,id.)) AND ((prospective cohort � or retrospective cohort � follow up � or longitudinal � or cross-sectional � or case-control � or nested-case control).mp.): all Sort by: PublicationDateFilters: Publication date from 2007/ 01/01 to 2017/12/31; Humans; English; Female; Field: Title/Abstract Included studies. We included all observational studies conducted in Africa, written in the English language, conducted from January 1, 2007, to May 30, 2018, and investigated postnatal depression and/or its associated factors. Studies were included in the current review if they used a standardized and validated tool to diagnose depression and diagnosed depression in the first two years of childbirth.
Excluded studies. This review excluded studies with poor quality and those conducted in a high-risk population (HIV).

Study selection
The title of all retrieved articles was screened, and those fit were imported to endnote software. The primary author did literature searching, title review, and duplicate removal. After excluding duplicated articles, abstracts and their full-text were independently reviewed by two groups of authors: group one (TYA, AFD) and group two (AGB, HFW). Differences were resolved through discussion. Articles deemed relevant during the abstract and full-text review were assessed for quality assessment.

Data extraction and quality assessment
The quality of included studies was assessed using the Newcastle Ottawa Scale (NOS) by two independent reviewers (TYA and AFD), and the third reviewer (HFW) was a tie break in case of discordance between the two reviewers. Articles that scored �7 points were considered as "good" quality and were included in the final review and meta-analysis [39]. Name of author, year of publication, country, study setting, study design, sample size, time of screening, tools used for screening depression, and estimates (prevalence, associated factors with their adjusted odds ratio and 95% confidence interval) information were abstracted and presented in a table.

Data synthesis and analysis
Data abstraction was done using a Microsoft Excel spreadsheet and was exported to Stata 14 for analysis. The prevalence of PND and odds ratios of risk factors obtained from each study were pooled after transforming the original estimates. Sub-group analysis was conducted based on the type of tool used for screening depression [40], geographical location, the income of countries, time of depression measurement, study setting, sample size, and year of publication. The magnitude of heterogeneity among included studies was quantitatively measured by I 2 and Higgins test (p-value <0.05). Sensitivity analysis was performed for checking the presence of an influential study. Publication bias was checked using a visual inspection of the funnel plot and Egger's regression test (P-value < 0.05). In the case of publication bias, Duval and Tweedie's Trim and Fill analysis in the random effect model was performed as an adjustment [41]. Finally, results were presented using tables and forest plots.

Protocol registration
The protocol for this systematic review and meta-analysis has been registered in the PROS-PERO (Protocol No. CRD42018100461).

Data reporting
The PRISMA [42] statement for reporting a systematic review and meta-analysis was used to present the study inclusion, exclusion, and reasons for exclusion in a diagram (Fig 1). Finally, the reporting of this systematic review and meta-analysis result followed the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) statement [43].

Description of studies
The electronic database search retrieved 965 records and 104 records were duplicates. The titles and abstracts of 861 articles were assessed, and 823 articles were removed. Full text of 38 articles was evaluated, and 18 articles were excluded. The reason for exclusion includes: conducted on restricted population [44,45], different objective from the current review [46, 47], lack of access to full articles[48-53], conducted in a different population with this review [54], the article was written in a non-English language [55], and were reviews and reports [56-60]. Twenty articles were assessed for quality, and one study was excluded because it had poor quality on NOS [61]. Finally, 19 articles (40,953 study participants) with good quality were included in this systematic review and meta-analysis (Fig 1).
Of 19 studies, five were from Ghana [62-66], three were from Ethiopia [7,67], two were from Egypt [68,69], two were from Tanzania [70,71], two were from South Africa [72,73], and one was from each of these countries (Malawi [74], Sudan [75], Zimbabwe [76], Zambia [77], and Cotedevior). Included studies were published in the year from 2007 to 2018 with a sample size ranging from 159 in South Africa to 16,560 in Ghana. Twelve (63.16%) and seven (36.84%) studies were cross-sectional and cohort studies, respectively. Nine studies were conducted in the community, while ten studies were conducted in health institutions. The time of screening for PND varied from 4 weeks to 56 weeks. Postpartum depression was measured using PDQ in 1 study, DSM-IV in 2 studies, EPDS in 6 studies, PHQ -9 in 7 studies, CES-D in 2 studies, and SRQ-20 in 1 study ( Table 1).

The pooled prevalence of postnatal depression
The pooled prevalence of postnatal depression before adjusting for publication bias was 16.84% (95% CI: 14.49-19.19, I 2 = 98.7%, Eggers test = 0.001) (Fig 2). However, after adjustment, the final pooled prevalence was found to be 17.8% (95% CI: 13.9%, 21.7%). There is no significant difference in estimates between the original and the trimmed prevalence (Figs 3 & 4). The extent of heterogeneity among the included studies was high. A meta-regression showed that the effect of the difference in sample size explained 12% (P-value = 0.09) of the total observed variation. We also further did and report estimates from a sub-analysis considering other possible sources of variations.

Factors affecting postnatal depression
In the random effect model, the following factors were identified as a risk factor for PND. The odds of PND was higher among women with a history of the poor obstetric condition (such as Hyperemesis Gravidurum and cesarean section delivery) (Pooled odds Ratio (POR) = 1.72;95% CI: 1.36, 2.17; I 2 = 70.7%). A history of adverse birth and infant health outcomes (such as low birth weight, preterm, stillbirth or infant loss after delivery) (POR = 2.38; 95% CI: 1.56, 3.64; I 2 = 74.5%) were associated with an increased odds of PND. A history of maternal health problems during pregnancy (such as hypertension, gestational diabetes, HIV/AIDS, TB, and other health problems during pregnancy) (POR = 2.75; 95% CI; 1.89, 3.98; I 2 = 81.7%) was associated with the risk of PND. Women with inadequate support from husband, family or any other source during pregnancy (POR = 2.06; 95% CI; 1.05, 4.05; I 2 = 83.2%) were more likely to had PND compared to those who had good support.
The result also showed that low economic status (explained by financial hardship or low perceived wealth, experienced hunger in the past months) was significantly increased the odds of PND (POR = 2.57; 95% CI: 1.97, 3.34; I 2 = 12.9%). Exposure to different forms of violence, such as physical, sexual, and psychological, was also significantly affected the occurrence of PND. Accordingly, the odds of depression among mothers who had exposure to a different form of IPV was 2.68 (POR = 2.68; 95% CI: 1.84, 3.89; I 2 = 85.5%) times higher (Figs 5 & 6).

Discussion
This systematic review and meta-analysis assessed the prevalence and factors affecting postnatal depression among postnatal women in Africa. In this review, a small number of studies with a high level of heterogeneity found that one in five postpartum mothers were likely to have PND in the continent. A history of the poor obstetric condition, adverse birth, and infant health outcomes, exposure to a different form of IPV, poor maternal health condition, and social support was found to increase the odds of PND. This estimated prevalence was in line with a review conducted on perinatal depression in low and middle-income countries that showed nearly one-fifth (19%) of postnatal women had depression [26]. However, this review differs from the current in terms of geographical coverage, population, and year of publication. The highest pooled prevalence was reported from two studies in Egypt that might be due to a small sample size and their cross-sectional nature [78]. The Sub-Saharan countries had a high prevalence of PND likely due to economic struggle that leads to stress [79]. Postnatal depression prevalence appeared to increases the time of screening after delivery increased. It has been implicated that depression symptoms increased in the first two to 12 weeks because of hormonal fluctuation and the new environment of maternity [35,80]. However, the interpretation for these estimates should account for the window of measurement as a more comprehensive window predicts more significant prevalence. The pooled estimate of PND in Africa also differed by sample size included in the studies. Studies with larger sample size estimated lower prevalence as compared to studies with small sample sizes, and this is mostly true as precise estimation depends on the adequacy of sample size. The other variation in pooled estimate was accounted for the study type, cross-sectional and community-based studies relatively estimated bigger prevalence. Overestimation is a major limitation of cross-sectional studies. Similarly, as depression affects health-seeking behavior, women with depression tend to stay at home, and this might slightly increase prevalence estimates in community-based studies. The current estimate was higher compared with a review of longitudinal studies on perinatal depression that showed 13.1% of postnatal women had signs of depression [81]. In the previous review, longitudinal studies that followed women starting from pregnancy to postpartum period were included. During this course of follow-up, those who were depressed during pregnancy might have treated, and this might reduce the risk of recurrent depression in the postnatal period.
In this review, a history of poor obstetric and adverse birth and infant health conditions has significantly increased the odds of PND. This finding is supported by a systematic review and meta-analysis conducted in low-and middle-income countries [3]. The possible reason might be women with poor obstetric conditions are likely to feel guilty or ashamed [82], which leads to depression. Similarly, a history of adverse birth and infant conditions conceptualized as a stressful experience is mostly linked with depressive symptoms [1,83,84].History of maternal mental health problems significantly increased the odds of current PND. This finding was in line with a systematic review and meta-analysis of common perinatal mental disorders conducted in low and middle-income countries [6]. Women with a history of mental disorder are more likely to lose their positive affect [85], practice rumination and develop a negative cognitive style in their life that could also persist throughout the continuum of pregnancy [86,87].
In the current review, poor social support was positively associated with the odds of PND. This finding was supported by a systematic review and meta-analysis carried out on the prevalence of postpartum depression and its effect in low and middle-income countries [57]. It has also been found that women who had weak support from close families at delivery or in the care of the newborn are less satisfied, stressful, and at a higher risk of depression [31].Being in lower economic status increased the risk of postnatal depression in this review. This finding was replicated in a systematic review and meta-analysis conducted in low and middle-income countries [3]. Women in low socio-economic status could become underprivileged due to scarcity of financial resources and insufficient health insurance, which leads to stress [27].
Exposure to different forms of IPV has increased the odds of postnatal depression in this review. This finding was supported by a systematic review and meta-analysis conducted in low and middle-income countries [88]. Exposure to IPV (physical, sexual, economical) could result in physical and social isolation; emotionally affect the abused women leading to low selfesteem, and disgusting lives [89,90]that end up with depression. Depression during the postnatal period affects infant growth [14], breastfeeding practice [91] and under nutrition [22] through negatively affecting women's interpersonal and parenting behavior [92]. The intimacy and interaction of the mother with her child would be affected, and the mother would fail to cope with her caring responsibility [93][94][95]. As a result, the child might suffer from diarrheal and other common childhood diseases [3]that might also lead to death. Depression also affects the mother's quality of life [49].

Epidemiology of postnatal depression
So, depression screening during the postpartum period has significant importance. Findings from the current study, therefore, are helpful for clinicians, programmers, and policymakers to think of ways to integrate maternal mental health with routine maternal health services. The screening and treatment of depression should start by identifying an appropriate screening tool and setting the ideal time of testing. Early detection and treatment of PND would help to minimize its further re-occurrence and potential adverse health outcomes among the mothers and their children.
This review has the following limitations. Only English language reviews were included, and this might introduce publication bias. Moreover, the extent of heterogeneity among the included studies was high, which can be attributed to differences in methodology, study period, type of screening tool, and other unexplained variations. The other limitation of this review is that as only studies from 10 African countries were included, it might be difficult to generalize to the whole countries in Africa.

Conclusion
This systematic review and meta-analysis found that PND could be arguably prevalent in the African continent based on a small number of published studies with inherently heterogeneous estimates. We also found that the prevalence of PND varied across different characteristics of the studies included in the review. Postnatal women with a history of the poor obstetric condition, adverse birth and infant health outcomes, mental health conditions, poor social support, and exposure to a different form of IPV were more likely to have PND. Based on this review, an early screening of postnatal depression and taking prompt intervention would save the mother and her baby from different forms of morbidity. This review also implicated that there is a shortage of robust studies in Africa to produce generalisablity evidence.