Perinatal Mental Illness in the Middle East and North Africa Region—A Systematic Overview

Aims: Perinatal mental illness (PMI) is associated with a high risk of maternal and infant morbidity. Recently, several systematic reviews and primary studies have explored the prevalence and risk factors of PMI in the Middle East and North Africa (MENA) region. To our knowledge, there has been no critical analysis of the existing systematic reviews (SRs) on this topic in the MENA region. Our systematic overview primarily aimed to synthesize evidence from the published SRs on PMI in the MENA countries focusing on a) the prevalence of PMI and b) the risk factors associated with PMI. Methods: We conducted a systematic overview of the epidemiology of PMI in the Middle East and North Africa region by searching the PubMed, Embase, and PsycInfo databases for relevant publications between January 2008 and July 2019. In addition to searching the reference lists of the identified SRs for other relevant SRs and additional primary studies of relevance (those which primarily discussed the prevalence of PMI and/or risk and protective factors), between August and October 2019, we also searched Google Scholar for relevant studies. Results: After applying our inclusion and exclusion criteria, 15 systematic reviews (SRs) and 79 primary studies were included in our overview. Studies utilizing validated diagnostic tools report a PMI prevalence range from 5.6% in Morocco to 28% in Pakistan. On the other hand, studies utilizing screening tools to detect PMI report a prevalence range of 9.2% in Sudan to 85.6% in the United Arab Emirates. Wide variations were observed in studies reporting PMI risk factors. We regrouped the risk factors applying an evidence-based categorization scheme. Our study indicates that risk factors in the relational, psychological, and sociodemographic categories are the most studied in the region. Conversely, lifestyle-related risk factors were less studied. Conclusions: Our systematic overview identifies perinatal mental illness as an important public health issue in the region. Standardizing approaches for estimating, preventing, screening, and treating perinatal mental illness would be a step in the right direction for the region.


Introduction
Mental illness in women during the perinatal period (start of pregnancy until one year postpartum) can have a significant impact on maternal and infant morbidity [1,2] Perinatal mental illness (PMI) include depression, anxiety, and postpartum psychoses, the latter of which usually manifests as bipolar disorder [1]. Studies also report suicide as one of the major causes for maternal mortality in several countries [3,4]. The mental health of women during the perinatal period also has a profound impact

Population of Interest
Our population of interest included pregnant and postpartum women up to one year after delivery living in any of the 20 MENA countries (namely, Algeria, Bahrain, Djibouti, Egypt, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Pakistan, Palestine, Qatar, Saudi Arabia, Sudan, Syria, Tunisia, the United Arab Emirates, and Yemen). These countries account for 8% of the global population [22]. According to the World Bank data of 2018, the crude birth rate in the MENA region was 23/1000 people [23].

Primary Outcomes
Our primary outcomes of interest included all epidemiological data from the SRs concerning perinatal anxiety; mania; bipolar disorder; postpartum blues; depression; psychoses, including schizophrenia; and psychiatric emergencies namely, self-inflicted injuries, suicides, and all other forms of mental illness. We also gathered conflicts of interest reported by the SR authors and that of the authors of the primary studies included in the SRs.
The search criteria were limited to reviews, systematic reviews, and meta-analyses concerning PMI. The search criteria used for each database is included as Supplementary Materials, Table S4. In addition to searching the reference lists of the identified SRs for other relevant SRs and additional primary studies of relevance (those which primarily discussed prevalence of PMI and/or risk and protective factors), between August and October 2019, we also searched Google Scholar for relevant studies during this period.
After removing duplicate publications with Endnote [24], independent title/abstract screening followed by independent full text screening were conducted by A.A., and S.C. (Sonia Chaabane), with Rayyan software [25,26]. Independent data extraction was carried out by A.A., and S.C. (Sonia Chaabane). Discrepant inclusions and extraction of SRs were discussed by A.A., S.C. (Sonia Chaabane), A.J., K.C., and S.D. under the supervision of the senior author.

Quality Assessment
The quality of the included SRs was assessed using the 11 criteria listed in the checklist of the measurement tool to assess the methodological quality of systematic reviews (AMSTAR) [27]. The quality of the primary studies reported in the SRs was assessed using the Population, Intervention, Comparison, Outcomes, Timing, Setting (PICOTS) framework [28] (interventions and comparators not being relevant in our overview).

Synthesis
All available data on the prevalence and risk factors associated with PMI of the population of interest was synthesized. We created evidence tables from the extracted data. Country level prevalence as estimated by different tools was separated for antepartum and postpartum women where possible. From the available studies, we also mapped the risk factors which have an impact on perinatal mental health. For the purpose of grouping the identified risk factors, we used a broad categorization scheme put forward by Furber et al. [29]. This categorization scheme lists all "ever" identified potential risk factors for mental illness under primary and secondary risk categories. We matched the risk factors identified by the primary studies in our overview with those in the categorization scheme and included them in the appropriate primary and secondary categories.

Results
Our initial search identified 11 SRs [8,[30][31][32][33][34][35][36][37][38][39]. Our manual search identified an additional four SRs [40][41][42][43], making a total of 15 SRs on the epidemiology of PMI in at least one of the 20 MENA countries included in our overview. From the 15 SRs, we originally identified 134 primary studies. Twenty-eight primary studies  featured more than one SR. Six additional primary studies [52,[72][73][74][75][76] relevant to the overview were identified by a manual search. All in all, 79 primary studies  from the 15 SRs, their reference lists, and those identified by additional manual search were included in the qualitative synthesis.
A PRISMA flowchart summarizing the search and inclusion of the systematic reviews and the primary studies is provided in Figure 1.

Characteristics of the Included SRs
Relevant data were available for 13 countries, namely Bahrain, Egypt, Jordan, Kuwait, Lebanon, Morocco, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, Tunisia, and United Arab Emirates. Prevalence data were available for all countries except Kuwait. Twelve SRs [8,30,31,[34][35][36][38][39][40][41][42][43] had prevalence data for one or more mental disorders such as anxiety, depression, postnatal blues, suicidal ideation, and suicides in perinatal women. Two SRs [33,37] focused exclusively on risk factors of PMI, and one SR focused on suicide [32] and its relative contribution to pregnancy-related mortality. One SR included risk factors from Kuwait [37]. All data extracted from the SRs and their primary studies are tabulated in detail in Supplementary Materials, Tables S5 and S6.

Quality Assessment of the Included SRs
The quality of the SRs was assessed using the original version of the AMSTAR recommendations (Table 1) as it was found to be more appropriate for observational studies. While all SRs provided the list of included studies including their characteristics, none of them provided the list of excluded studies. No SR reported conflicts of interest in their included studies. Thirteen SRs (86.7%) [8,[30][31][32][33][34][35][36][37][38]40,42,43] had a comprehensive literature search (defined as searching at least two databases). Five of them (33.3%) [30,32,33,36,38] actively looked for grey literature. Only three SRs (20%) [32,36,38] deployed two persons in data extraction (either independent extraction or one screen and another check procedure) and had a consensus procedure in place for disagreements. Four SRs (26.7%) [8,32,36,38] had included a publication bias assessment.  [42] −  [40] − Categories for risk of bias are as follows: +, yes; −, no; N/A, not applicable. * No single review provided a list of excluded studies; ** conflict of interest of all included studies was never stated.

Suicidal Ideation and Suicides
The SR on suicide contribution to pregnancy-related mortality [32], through its meta-analysis, estimates 0.4% (95% CI: 0.1-0.9) contribution to pregnancy-related deaths for the two countries in the region (Tunisia and Egypt) where specific data is available. This ranged from 0.3% in Egypt [107] (95% CI: 0.04 to 1.1) to 0.6% (95% CI: 0.1-1.9) in Tunisia [90]. When the numerator is changed to include suicides, falls, drowning, poisoning, and burns, the average proportional contribution to pregnancy-related deaths becomes 3.5% (95% CI: 0.4-9.4) for Jordan, Egypt, and Tunisia. The country specific estimates range from a low of 0.6% (95% CI: 0.1-1.9) in Tunisia [90] to a high of 6.2% in Jordan [122] (95% CI: 2.5 to 12.4) with 5.8% (95% CI: 4.1-7.9) estimated in Egypt [107]. The authors acknowledge that these proportions are likely to be underestimates because of underreporting and nonrecognition of suicides as causes of pregnancy-related deaths in eligible studies.
There was only one other SR [39] comprising one primary study [82] from Pakistan which discussed suicide, suicidal ideation, and suicide attempts. The study reported that 11% of pregnant women screened between 20-26 weeks of gestation had suicidal ideation; 45% of those or 5% of pregnant women overall had attempted suicide in that study. None of the other SRs discussed suicidal ideation and suicides as part of their review of PMI.

Quality Assessment of the Primary Studies
The primary studies included in the 15 SRs for this overview were assessed using the PICOTS framework. A summary of the quality of the primary studies is presented in Table 3, and a detailed listing is provided as Supplementary Materials, Table S7. We identified that all SRs consistently reported the population covered by the respective primary studies. However, the exact point in the perinatal period was clearly defined in 44 (55.5%) studies only [10,[44][45][46][47][48][49][50][51][52][53][54][55][56][57][58][59][60][61][62][63][65][66][67][68]70,71,77,81,83,86,89,90,95,97,98,101,107,111,112,115,117,118,121]. Mentioning the exact point of measurement is important, for example, to distinguish serious mental illness from postpartum blues, which are often mild and appear only in the first few weeks after delivery. Among the 73 primary studies included in the SRs, the timing, defined as at least the period (year and months) in which the study was carried out, was explicitly mentioned in four studies (5.5%) only [90,97,107,121]. We believe that this is a serious omission because of the possibility of seasonal affective disorders in perinatal women [122].
Non-reporting of the setting introduces bias within the sample as it fails to distinguish between clients with advanced conditions who report to a health facility when compared to the milder forms and/or those with poor health-seeking behaviors who may not report to a health facility.
All six additional primary studies (not mentioned in the SRs but identified by manual search) [72][73][74][75][76]120] clearly defined the population of interest. However, only three of them [74][75][76] defined the specific time period of the study and three [73,75,76] clarified the study setting.
Though not a usual part of the quality assessment, it is worthwhile to reflect on the validity of the tools used for estimating prevalence in the primary studies. With the exception of the Aga Khan University Anxiety and Depression Scale (AKUADS), all other tools (screening and diagnostic) were developed using samples from western settings. Tsai et al. have discussed this issue in the context of African settings [124]. They discuss an "etic approach" in which the construct of mental illness is promoted irrespective of cultures and an "emic approach" which emphasizes on the evaluation of mental illness constructs within a specific cultural context. The fields of mental illness and perinatal illness have long advocated for the need for integrating the etic and emic validation criteria to obtain more reliable prevalence estimates and to study risk factor associations [125]. We did not find a discussion on the use of such a hybrid approach in data collection in any of the primary studies included in the SRs. Four of the primary studies in our overview used the Arabic and Urdu version of the EPDS, but their validation process is unknown [54,75,101,103].  [83] provided a combined prevalence of anxiety and depression, and one study [98] chose to label the illness studied as "perinatal mental disorders".
The tool most commonly used in the primary studies (N = 7) [55,58,65,68,75,95,112] to estimate the prevalence of mental illness during the antepartum period was the Edinburgh Postnatal Depression Scale (EPDS). The EPDS cutoffs varied widely and ranged from 10-13 as opposed to the standard cutoff of 13 [126]. Four studies [60,83,99,115] used the Aga Khan University Anxiety Depression Scale (AKUADS), four studies [82,90,117,118] used the Hospital Anxiety and Depression Scale (HADS), and two studies [70,76]  With the various screening tools used in the region, we found the antepartum mental illness prevalence to range from 11.5% in a community setting (measuring all mental disorders using AKUADS) [98] to 75% (measuring depression using EPDS) [95] in a hospital setting in Pakistan. The only study from Jordan [65] utilizing the EPDS in a hospital setting estimated a prevalence of 19% antepartum depression. We noted wide variations in the cutoffs used in various studies and hence the "prevalence", leading to difficulties in interpreting the reported data. Restricting the data reported from the use of diagnostic tools in Pakistan, a community-based study by Rahman et al. [67] using SCAN estimated the prevalence of all forms of antepartum mental illness to be 25%, and the study by Sadaf et al. [111] estimated a prevalence of 10% antepartum depression in a hospital-based sample using HAM-D. In Morocco, Alami et al. [51] estimated a prevalence of 19.2% antepartum depression also in a hospital setting, using MINI.
Prevalence data on diagnostic interviews using MINI and SCAN were available for four countries only, namely Morocco, Pakistan, Saudi Arabia, and Sudan. However, the prevalence of postpartum depression in Sudan using MINI [101] was obtained from a two-stage sampling design with pre-diagnostic screening by EPDS and hence cannot be seen as true prevalence. Similarly, in Pakistan, the study of Rahman and Creed [66] has reported a postpartum depression prevalence range of 62% to 95% during different time points in the postpartum period. However, these proportions are for those women who were already diagnosed with depression during the antenatal period. This prevalence reported in the SRs hence cannot be taken as true prevalence. The postpartum depression prevalence measured using MINI in a hospital settings varied between 5.6% in Morocco [47] and 10.2% in Saudi Arabia [84] during the time period from 3 months to 1 year postpartum. In Pakistan, the prevalence of combined postpartum mental disorders varied between 25% from 6 weeks to 3 months [108] to 28% [67] from 3 months to 1 year postpartum, using the WHO SCAN. With the EPDS, the range of postpartum depression varied widely between 9.2% in Sudan [101] to 85.6% in United Arab Emirates (UAE) [56].
We grouped the available risk factor data and provided a summary in Table 4. We have attempted to summarize the available odds ratio and relative risk reported by the various studies in the summary table. The majority of the studies which reported odds ratio and relative risk were cross-sectional studies. They had not been designed to assess risk factors. SRs reporting these studies have not provided an adequate description of the statistical analysis in the primary studies to determine whether the identified risk factors are independently associated with the outcome. They were then considered as potential risk factors as their independence has not been ascertained. Given the measurement issues (choice of instrument, cutoff points, and lack of theoretical framework), any attempt to study association without controlling variables by individual studies is a major barrier in deriving quality evidence [128,129]. Studies in the region continue to study relational (spousal and others), psychological, and sociodemographic factors in detail while overlooking lifestyle, environmental and occupational factors. Detailed information on the various factors as extracted from the primary studies are detailed in Supplementary Materials, Table S8.
While individual studies had identified associations between postpartum mental health and antepartum depression, stressful life events, mode of delivery, wanted/unwanted pregnancy, number of children, age at marriage, breast-feeding practice, and health of the infant, all SRs found very few studies of good quality to generate conclusive evidence.
This section may be divided by subheadings. It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn.  Total  128  3  7  11  1  10  10  6  44  9  8  5  14 * Independence of these potential risk categories/factors cannot be ascertained with available data; ** odds ratio and relative risk of potential risk factors as indicated in the SRs/primary studies; NA, not available; CI-Confidence interval.

Discussion
Our systematic overview synthesizes available data on PMI prevalence and the associated risk factors. The overview points out that PMI is a major public health issue in the region. In the MENA region, there is limited data which are diagnostic in nature to be able to reliably assess country level prevalence. The published SRs and associated primary studies have measured various types of mental illness at different times during pregnancy and the postpartum period; have utilized different screening scales; and, even when using the same scales, have used different cutoff points, all of which makes the data highly heterogenous and difficult to compare. The SRs with meta-analyses in our overview have not questioned the use of screening tools for determining prevalence and have not taken into consideration the sensitivity and specificity of the respective screening tools in attempts to better estimate true prevalence. Our concern is shared by other researchers [128,129] who have questioned the use of self-reporting screening tools for prevalence estimation. A standardized framework to group or to clearly define risk factors in the studies is also lacking. The region's SRs on the topic have been of varying quality, with all SRs consistently omitting to report on the conflicts of interest in individual studies or to present a list of excluded studies and with most of them failing to discuss publication bias as recommended by international guidelines. Few published SRs follow the recommended procedures for data extraction and resolving disagreements in inclusion of studies. The majority of the SRs do not analyze/discuss the individual studies using the PICOTS framework, making it difficult to understand if PICOTS items have not been reported by the SRs or if the individual studies themselves did not carry out their research accordingly or report their approach explicitly.
However, the data available from the SRs and the included individual studies provide key insights on perinatal mental health in the region. The SR of Fuhr et al. [32] helps provide context to this discussion. The MENA region that our overview discusses includes most countries in the Eastern Mediterranean (EMRO) region as defined by the World Health Organization and as adopted by Fuhr et al. [32]. It has been found that the EMRO region has a lower than average proportion of global pregnancy-related deaths attributable to suicide but a higher than global average proportion of pregnancy-related deaths due to injuries in general or when deaths due to suicide, falls, drowning, poisoning, and burns are combined together. This discrepancy is unique to the EMRO region, implying an underreporting of suicides. Preexisting mental illness is a risk factor for suicides; underreporting of suicides may be due to prevailing stigma towards suicides and its associated underlying mental illness. The fact that other SRs did not attempt to systematically include and discuss suicides in their scope of study indicates either a low recognition of suicides as a manifestation of perinatal illness and/or a lack of data to study this in depth.
The SRs which provide data on the prevalence of mental illness focus heavily on depression in the perinatal period. Antepartum data based on diagnostic criteria are available from Morocco and Pakistan only, making it difficult to generalize the findings for the region. For the postpartum period though, meaningful diagnostic data was available for only three countries (Morocco, Pakistan, and Saudi Arabia) with a wide-ranging prevalence. We also note that the diagnostic tools used different definitions for depression/depressive disorders in particular due to the changes in International classification of diseases (ICD) and Diagnostic and Statistical Manual of Mental Disorders (DSM) classification over the last decade. Due to these shortcomings, making any interregional/national comparison presents a challenge. However, in general, the finding of high prevalence in some countries of the region is of concern and of public health importance. This reported prevalence in these countries meanwhile needs to be seen in the context of the risk factors identified in the overview.
Very few primary studies included in the SRs outline associated factors with antepartum mental illness. However, for postpartum mental illness, specifically depression, the SRs included in the review consistently found positive association with social support (from spouse and in-laws), education levels, and financial stability. Recently, more women in the MENA region are completing higher education and are entering the labor market but are still expected to maintain their central role in managing the family [11]. Under these circumstances, family support becomes extremely critical for women to be able to maintain optimal mental health. Education and wealth are likely to be factors which prevent the women from falling into a vicious cycle of stress during pregnancy and after giving birth. There is a need for more research in the region exploring risk factors for perinatal mental health related to lifestyle, occupation, and environmental exposures.
Our overview has important implications for practice and future research. Firstly, our overview highlights the lack of a standardized methodology to estimate the prevalence of PMI in the region. This may prevent appreciation of the true magnitude of the problem and may make comparison between and within countries difficult. Such comparisons could help countries understand the effectiveness of their identified approaches to addressing PMI. The research capacity on PMI should be strengthened by regional institutions such as the World Health Organization in partnership with leading academic institutions in the region. Secondly, our overview provides synthesis of available data on PMI for individual countries in the region. This data can enable countries to recognize PMI as a significant public health problem and can help them develop evidence-based locally informed national guidelines for preventing, screening, and treating PMI. Countries with no available data should identify this as a major gap which must be addressed at the national level. Researchers from these countries should be encouraged to conduct good quality research on PMI. Also, regarding the potential risk factors associated with PMI, certain categories appear to be over-studied. Future research in the region should focus on risk factor categories which have either been poorly studied or not studied at all.

Limitations
This overview has limitations. The data from primary studies were extracted from the SRs, and quality was assessed based on their findings as reported in the SRs. Extreme variations in PMI definitions, methods of assessment, and study quality precluded us from performing a meta-analysis to arrive at a pooled prevalence. The primary studies were reviewed only if outliers or errors were suspected. The studies included in our analysis have come from only 13 out of the 20 MENA countries, and the majority of these have been reported from one country: Pakistan. We were not able to gather any data for the relatively low-income countries of the region. Though we scanned extensively for both published and unpublished literature, all retrieved studies were in English, which is also acknowledged as a limitation.

Conclusions
There are very few SRs or primary studies with sufficiently good quality data of diagnostic nature from the MENA region to arrive at an informed prevalence of PMI. There is scope for strengthening the quality of future primary studies and SRs. Our data suggests that PMI is a major public health issue in the region. Social influence, particularly spousal support, is seen as a key aspect in improving perinatal mental health in the region. Programs and policies in raising awareness of husbands and in-laws through mass media and community gatherings as well as during antepartum and postpartum consultations can be useful interventions. Due to the low priority given to mental health, particularly suicides during pregnancy and the postpartum period, there appears to be low awareness and a high level of stigma associated with this. Additional efforts are needed to better capture data on perinatal suicides and other self-inflicted injuries in the region to be able to develop evidence-guided population-based policies.