Introduction

Maternal mental health disorders are common problems worldwide (Gold and Marcus 2008) and include anxiety and depression, psychotic disorders, and a range of other mental disorders. These disorders constitute a prominent public health issue given their scale; for instance, depression alone affects up to 25% of pregnancies (Gelaye et al. 2016) and their association with myriad adverse outcomes for mother and offspring (Pierce et al. 2019; Zhong et al. 2018). A large body of evidence highlights that women with a range of different mental disorders are at distinctly higher risk of delivering a preterm or low birthweight baby (Ding et al. 2014; Grigoriadis et al. 2018; Grote et al. 2010) and of having children with poorer developmental outcomes (Reupert et al. 2013). In addition, the results of a limited set of studies suggest that poor maternal mental health poses an increased risk for early life mortality (Webb et al. 2005; Zhong et al. 2018).

The mechanisms whereby maternal mental health impact on adverse fetal and infant outcomes are complex, not fully understood, and likely include alterations to the intrauterine environment as well as behavioural pathways. Depression and anxiety, for example, may alter the cortisol response to stress, which can affect fetal growth (Field et al. 2006; Zorn et al. 2017). For some women, mental health disorders are also associated with inadequate antenatal care, smoking, and misuse of illicit substances and drugs—all common risk factors of perinatal mortality (Judd et al. 2014; Weaver et al. 2003). However, only a limited number of studies have evaluated the association of maternal mental health disorders and stillbirth and infant mortality outcomes, with inconsistent findings (Di Prinzio et al. 2020; Heun-Johnson et al. 2018; Pavlov et al. 2014; Surkan et al. 2016; Vigod et al. 2014; Zhong et al. 2018).

To our knowledge, only two systematic reviews (Jacques et al. 2019; Webb et al. 2005) have focused on the risks for mortality, and none has comprehensively investigated the link between maternal pre-conception and prenatal (occurring before or during pregnancy) mental health disorders and perinatal and infant mortality. A 2005 review focused exclusively on the association between schizophrenia and stillbirth but did not include any other mental health disorders, and identified only six studies, all with important methodological limitations (Webb et al. 2005). A 2019 review examining maternal depression and infant hospitalization and mortality identified only three in-scope studies that examined infant mortality outcomes (Jacques et al. 2019). Notably, most of the included studies in these reviews (three in the first and two in the second) included maternal mental health disorders that occurred in the postnatal period or after the fetal/infant outcome, exposing the possibility of reverse causality bias.

The current systematic review and meta-analysis aimed to provide a comprehensive evaluation of published research that has examined the association between maternal prenatal mental health disorders and stillbirth and infant mortality. The resulting evidence can support prevention and early interventions that target the high proportion of women with mental health disorders.

Material and methods

Protocol and registration

We followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) (Moher et al. 2009) and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) (Stroup et al. 2000) guidelines. The review protocol has been published (Adane et al. 2020) and registered with the PROSPERO (# CRD42020159834).

Search strategy

The details of the search terms and strategies are available in the published protocol (Adane et al. 2020) and as a supplementary file (Supplementary Box 1) but, in summary, we searched MEDLINE, Embase, PsycINFO, and Scopus electronic databases using keywords tailored to each database, from their inception to December 2019. These searches were limited to studies published in English language and on humans only, with no restriction on the year of publication. The reference lists and citations of relevant articles were also checked manually for additional studies. Three authors were contacted for additional information, but only one provided further details (Di Prinzio et al. 2020).

Inclusion and exclusion criteria

Original studies examining the association of any prenatal mental health disorder (occurring at any time before or during pregnancy) and stillbirth (death of a fetus at 20 or more weeks of gestation), neonatal (the first 28 days of life), post-neonatal (the period after 28 days of life and before 1 year), or infant (under 1 year) mortality were eligible. The use of either a diagnostic measure of mental health or symptom screening measure formed part of the eligibility criteria. However, correspondence, theses, reviews, editorials, case-only studies, and conference abstracts were excluded. We also excluded studies where the onset of maternal mental health disorders in relation to the date of childbirth was unclear and/or when postnatal mental health disorders were combined with prenatal disorders.

Study selection

First, all citations identified through database searches were imported into an Endnote library, and duplicates were removed. The titles of the remaining records were screened, with eligible studies uploaded to Rayyan (Ouzzani et al. 2016) and abstracts and full-texts independently evaluated by two of the authors (AAA and HDB). Disagreements between assessors were resolved by face-to-face discussion.

Data extraction

Using a standardized excel sheet, one of the researchers (AAA) extracted the data from articles that passed the full-text evaluation, and the second author (HDB) verified the results. Extracted data included the first author’s last name, country and year of publication, study population and design, exposure and outcome assessment, and association/s as well as confounders adjusted for. When available, adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were extracted. For studies which did not report ORs, crude measures were manually calculated when enough data were available. Some studies provided relative risk (RR) estimates, which were converted to ORs using a previously developed package (Barendregt 2016).

Assessment of methodological quality and risk of bias

We used the Newcastle-Ottawa quality assessment scale (Wells 2001), which includes the following: the selection of the study groups, the comparability of the groups, and the ascertainment of either the exposure (for case-control studies) or outcome of interest (for cohort studies). We also employed the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach (Guyatt et al. 2008) to assess the degree of risk of bias for each included study (Higgins et al. 2011). Two of the authors (AAA and HDB) independently performed the quality and risk of bias assessments. When the involved authors differ in the results of the quality assessments, an agreement was reached during a face-to-face discussion.

Data analysis

We have included studies with several differences between them, such as in the type of maternal mental disorders investigated, diagnostic classification system, mortality outcome measure, geographical location, and time period. As such, we do not expect them to share a common effect (fixed effect). Therefore, we employed a random-effects model (Harris et al. 2008), which accounts for between-study heterogeneity and provides an estimate of the average effect, to calculate pooled ORs with 95% CIs for the overall and subgroup meta-analyses of the associations between maternal mental health disorders and stillbirth/infant mortality outcomes. The maternal mental health disorders were grouped into depression and anxiety (including major depressive disorder) and severe mental illness (including schizophrenia, bipolar, schizoaffective/affective psychosis, and psychotic disorders not otherwise specified). We undertook further subgroup analyses for each outcome (stillbirth, perinatal, and infant mortality) separately and collectively as a ‘stillbirth/infant mortality’ variable within each maternal mental health disorder group. Eight studies only reported perinatal mortality (includes stillbirth and neonatal death) as an aggregate outcome, so the perinatal mortality subgroup includes these studies and those studies that separately reported stillbirth and neonatal mortality. Additional sensitivity and subgroup analyses were conducted by excluding each study and also based on whether an adjustment was made for smoking (yes/no), the number of offspring mortality in women with mental health disorders (< 20 vs. ≥ 20 cases), year of publication (≤ 2010 vs. > 2010), and risk of bias based on the GRADE approach (low vs. moderate). We also ran a sensitivity analysis by excluding studies using symptom only screening measures for depression and anxiety. A few of the included studies contributed two or more estimates for stillbirth, neonatal, post-neonatal, and infant death as they assessed different mental disorders. So to check the effect of non-independent effect sizes, we repeated the primary analyses by including an average measure (obtained via a fixed effect model) for studies with two or more estimates. Analyses were conducted using STATA 15 (StataCorp, College Station).

Assessment of heterogeneity and publication bias

Between-study heterogeneity was examined using the I2 statistic (Higgins et al. 2003) and was considered low (I2 < 50%), moderate (I2, 50–74.9%), or high (I2 ≥ 75%). The risk of publication bias was graphically examined using funnel plot asymmetry and further quantified using the Egger’s test (p value ≤ 0.05 suggests significant publication bias) (Harbord et al. 2009).

Results

Search results

A total of 4487 records were identified through the database and manual searches. After duplicates were removed, 3756 were screened by title and abstract. Full-text assessments were made on 78 articles with 50 of these excluded for reasons such as no assessment (n = 22) or separate reporting (n = 6) of early mortality, and no assessment of prenatal mental health (n = 6) or inclusion of postnatal mental health cases (n = 7). Finally, 28 studies (Ban et al. 2012; Bansil et al. 2010; Bitew et al. 2017; De Lange et al. 2008; Di Prinzio et al. 2020; Gardosi et al. 2013; Hanlon et al. 2009; Heun-Johnson et al. 2018; Hizkiyahu et al. 2010; Howard et al. 2003; Howard et al. 2007; Huang et al. 2017; MacCabe et al. 2007; Mei-Dan et al. 2015; Navaratne et al. 2016; Nilsson et al. 2002; Pare-Miron et al. 2016; Pavlov et al. 2014; Raisanen et al. 2014; Simoila et al. 2018; Vigod et al. 2014; Webb et al. 2006; Webb et al. 2010; Weobong et al. 2014; Wrede et al. 1980; Yedid Sion et al. 2016; Zax et al. 1977; Zhong et al. 2018) were included in this systematic review (n = 78,492,940), with all but one (Hanlon et al. 2009) contributing at least one estimate for the meta-analyses (Fig. 1).

Fig. 1
figure 1

Study identification and selection flow chart for the systematic review and meta-analysis of maternal mental health disorders, and stillbirth/infant mortality

Full size image

Characteristics of included studies

The included studies are described in Table 1 and summarized in Box 1. The majority of studies were from European countries (n = 12) and the USA (n = 5). The most common maternal mental health disorders examined were depression (n = 8) and schizophrenia and related disorders (n = 5). The most common outcome was stillbirth (n = 17). The vast majority (n = 24) of studies were cohort studies, while two each were case-control and analytical cross-sectional studies. The sample size (births) of included studies ranged from 56 to 32,156,438 (Table 1).

Table 1 Characteristics of studies included in the systematic review and meta-analysis of maternal mental health disorders, and stillbirth/infant mortality
Full size table
Box 1 Descriptive summary of studies included in the systematic review and meta-analysis
Full size table

Included studies varied considerably in terms of confounder or covariate adjustments. While most studies adjusted for maternal age, very few studies fully adjusted for other maternal sociodemographic and lifestyle factors (Supplementary Table 1).

Methodological quality and risk of bias

Based on the Newcastle-Ottawa quality assessment scale, 17 studies scored eight or nine, and the remaining 11 studies received scores of six or seven out of nine (Supplementary Tables 2 and 3). Using the GRADE approach, the risk of bias in most studies was low (n = 15) or moderate (n = 12) with only one study having a high risk of bias (Supplementary Table 4).

Results of the meta-analyses

Depression and anxiety

Using 16 estimates (i.e., ORs) from 13 unique studies (a study can assess more than one child outcome; so it can contribute more than one estimate) including 39,526,270 births, the pooled OR for the association between maternal depression and anxiety and stillbirth/infant mortality was 1.42 (95% CI, 1.16–1.73; I2, 76.7%). There were similar associations for stillbirth and perinatal mortality, but the association for infant mortality was marginally lower and spanned the null value (Fig. 2). The pooled ORs (95% CI) were 1.46 (1.17–1.82: I2, 75.1%) for the association between maternal depression and stillbirth/infant mortality, 1.51 (1.17–1.96: I2, 81.3%) for the association between maternal depression and stillbirth, 1.46 (1.16–1.84: I2, 76.4%) for the association between maternal depression and perinatal mortality, and 1.36 (0.90–2.05: I2, 54.6%) for the association between maternal depression and infant mortality.

Fig. 2
figure 2

Summary of pooled odds ratios with 95% confidence intervals for studies included in the meta-analysis

Full size image

Severe mental illness

Using 27 estimates from 13 unique studies which included 36,038,106 births, the pooled OR for the association between severe maternal mental illness and stillbirth or infant mortality was 1.47 (95% CI, 1.28–1.68; I2, 62.3%). We found similar associations with stillbirth (OR, 1.42; 95% CI, 1.20–1.69; 12 estimates from 10 unique studies and 33,365,387 births; I2, 68.5%), perinatal mortality (OR, 1.51; 95% CI, 1.28–1.79; 19 estimates from 12 unique studies and 33,557,786 births; I2, 72.6%), and infant mortality (OR, 1.49; 95% CI, 1.20–1.85; 13 estimates from seven unique studies and 6,812,821 births; I2, 46.2%) (Fig. 2).

Any maternal prenatal mental health disorders

Overall, using 52 estimates from 27 unique studies including 78,316,822 births, the pooled OR for the association between any maternal prenatal mental health disorders (listed in Table 1) and stillbirth/infant mortality was 1.59 (95% CI, 1.43–1.77; I2, 75.5%). We found similar associations for stillbirth and perinatal mortality and a stronger association for infant mortality, particularly for post-neonatal mortality and sudden infant death syndrome (Fig. 2 and Fig. 3).

Fig. 3
figure 3

Pooled odds ratios for the association between maternal prenatal mental health disorders, and stillbirth/infant mortality

Full size image

Hanlon et al. (Hanlon et al. 2009), which is not included in the meta-analyses, assessed common mental health disorders using the Self-Reporting Questionnaire (SRQ-20). The study did not find a statistically significant association between maternal antenatal SRQ score (≥ 6) and stillbirth (RR, 1.7; 95% CI, 0.6–5.5) or neonatal mortality (RR, 0.8; 95% CI, 0.2–3.0), perhaps because of low statistical power.

Heterogeneity and publication bias

The between-study heterogeneity was high for the associations of maternal depression and anxiety and stillbirth/infant mortality, and any maternal mental health disorders and stillbirth/infant mortality. A high degree of between-study heterogeneity remained in subgroup analyses according to the types of mortality outcomes, particularly for stillbirth and perinatal mortality. These could be partly due to differences in the types of maternal mental health disorders, confounders adjusted for, sample size, and year of publication, as shown in the stratified and sensitivity analyses (Supplementary Table 5). We found no evidence of publication bias for studies evaluating the associations of maternal depression and anxiety and stillbirth/infant mortality; severe maternal mental illness and stillbirth/infant mortality, and studies assessing any maternal prenatal mental health disorders and stillbirth (Supplementary Fig. 1).

Sensitivity and subgroup analysis

We did not find any significant changes for the associations between maternal depression and anxiety and severe mental illness with stillbirth/infant mortality outcomes when conducting sensitivity analyses that removed each study one by one (Supplementary Fig. 2). However, we found considerable differences in pooled ORs in the stratified analyses based on whether the adjustment was made for smoking, year of publication, and degree of risk of bias using the GRADE approach. While the pooled effect estimates for maternal depression and anxiety were stronger for studies adjusted for smoking and those published after 2010, those estimates for severe maternal mental illness were stronger if studies did not adjust for smoking, published in 2010 or earlier, and had a moderate risk of bias (Supplementary Table 5). Overall, we found slightly stronger associations and lower between-study heterogeneity in the sensitivity analyses excluding studies that used symptom only screening measures for depression and anxiety: stillbirth/infant mortality (OR, 1.47; 95% CI, 1.18–1.82; I2, 74.0%), stillbirth (OR, 1.67; 95% CI, 1.35–2.08; I2, 65.1%), perinatal mortality (OR, 1.48; 95% CI, 1.19–1.85; I2, 75.5%), and infant mortality (OR, 1.22; 95% CI, 0.59–2.53; I2, 55.3%). We found similar results to the primary analyses in the sensitivity analyses using the average measure for studies with two or more estimates (Supplementary Fig. 3 and Fig. 4). However, the between-study heterogeneity slightly increased (Supplementary Fig. 3), and the pooled estimate for neonatal death slightly attenuated and spanned the null value (Supplementary Fig. 4).

Discussion

Summary of main findings

In this systematic review and meta-analysis, which included 28 studies comprising over 78 million births, we found a moderate and statistically significant higher risk of stillbirth and infant mortality for women with prenatal mental health disorders (including high and low prevalence disorders), compared with women without prenatal mental health disorders. These findings persisted in sensitivity and subgroup analyses according to the types of maternal mental health disorders and fetal or infant mortality outcomes, with results typically indicating a 40–50% increase in the risk of stillbirth and infant mortality. However, considerable heterogeneity was observed in effect sizes across studies included in the meta-analysis, and very few studies evaluated the impact of common maternal mental health disorders, particularly anxiety disorders.

Research in context and potential biological mechanisms

The previous meta-analyses found that women with depression and anxiety disorders were 40 to 80% more likely to have preterm birth and low birthweight (Ding et al. 2014; Grigoriadis et al. 2018; Grote et al. 2010), the leading causes of perinatal mortality. The findings of this systematic review and meta-analysis extend the evidence base by including a broader range of mental health disorders and further corroborated the detrimental impact of maternal prenatal mental health disorders. The underlying reasons for the higher risk of adverse perinatal outcomes among women with prenatal mental health disorders are not yet fully understood, but they are likely to include a complex array of mechanisms that include direct disease effects, behavioural factors and comorbidities, and increased risk of pregnancy complications.

There are several hypothesised pathways through which fetal health may be compromised by intrauterine exposure to psychological distress, including pathways that impact fetal neurodevelopment, placental function, and physical growth (Kinsella and Monk 2009; Lewis et al. 2016). Recent studies, for example, have linked maternal prenatal anxiety and depression with disruptions in the placental 11β-hydroxysteroid dehydrogenase type 2 (11B-HSD2) enzyme and cortisol transfer to the fetus (Cottrell et al. 2014; Duthie and Reynolds 2013; O’Donnell et al. 2012), inflammatory pathways, and altered physiological responses (Christian 2014; Seth et al. 2016). While these effects have generally been associated with greater morbidity, a link with mortality risks is less clear (Ding et al. 2014; Field and Diego 2008). Unhealthy lifestyles and behaviours are highly correlated with psychiatric disorders, and their interplay plausibly accounts for part of the association between maternal mental health disorders and perinatal morbidities and mortality. Women with mental health disorders tend to have poor health-seeking behaviour (late and less frequent antenatal care) (Ben-Sheetrit et al. 2018; Lin et al. 2009) and are more likely to misuse illicit substances, alcohol, and drugs, smoke tobacco, and have poor nutrition (Barker et al. 2013; Frayne et al. 2019; Judd et al. 2014; Teasdale et al. 2019; Weaver et al. 2003), all of which can have profound consequences for fetal development. Most prenatal mental disorders are likely to persist in the postnatal period and directly compromise the quality of essential parenting, which may contribute to the increased infant mortality risk (Chen et al. 2010). Recent evidence also indicates that some mental disorders and/or their medications increase the risk of pregnancy complications such as gestational diabetes and pre-eclampsia (Galbally et al. 2020; Qiu et al. 2009; Uguz 2017), which are key risk factors for perinatal morbidity and mortality (Hutcheon et al. 2011; Schmidt et al. 2001). However, the evidence about the association of mental disorders or antipsychotic treatments and pregnancy complications is inconsistent and likely to include complex pathways (Vigod et al. 2015; Vigod et al. 2014). Collectively, these underscore the need for comprehensive support, care, and careful treatment for women with a mental health disorder.

Strengths and limitations

The strengths of this systematic review and meta-analysis are that it is based on a pre-published protocol and employed independent reviews and quality assessments on records obtained via electronic searches in multiple databases (supplemented with a manual search). To the best of our knowledge, this is the first systematic review and meta-analysis that has comprehensively evaluated the association between maternal prenatal mental health disorders and stillbirth or infant mortality—providing the most robust evidence to date on this topic.

The review does, however, have some limitations. Although a restriction was not imposed on the year of publication, electronic search results were limited to English language and hence relevant non-English studies could have been excluded. Relatively fewer studies were available with different maternal mental health disorders and assessments. This precluded us from a comprehensive investigation of the source of heterogeneity and publication bias. Most studies included in the systematic review did not adequately control for potential socioeconomic and lifestyle factors and, among those that did, there was variation in the choice of modelled covariates (Supplementary Table 1). In addition, there is always a possibility of residual confounding. However, a statistically significant association remained (albeit attenuated for severe maternal mental illness) in the meta-analysis, which included studies that adjusted for smoking. Additionally, one of the largest studies (Zhong et al. 2018) included in this systematic review demonstrated a significant association between maternal psychosis and stillbirth (OR, 1.37; 95% CI, 1.23–1.53) after adjustment for several factors including sociodemographic (age, race, and income) and lifestyle (smoking and alcohol/substance abuse) factors. The other limitation is that the studies included in the review did not examine or provide separate data for mentally ill mothers with and without psychotropic medications, so we could not evaluate the effect of the use of psychotropic drugs.

Implications

The findings of this review have important implications for public health, particularly given the high prevalence of maternal mental health problems (especially anxiety and depression) in populations worldwide. Efforts to prevent chronic and acute prenatal mental health disorders and treat those suffering from mental health disorders are likely to be critical in reducing the scale of stillbirth and infant mortality and the consequent grief experienced by families and communities. Focusing interventions on the earliest stages of family planning and prenatal period is important as many mental health problems are chronic at the time of conception (O'Hara and Wisner 2014). In terms of policy, holistic healthcare models, with a multidisciplinary approach, are preferable (Frayne et al. 2019; Galbally et al. 2013; Tachibana et al. 2019). Evidence-based treatments for women with severe mental illness, particularly depression, may reduce the rate of adverse birth outcomes, including perinatal and infant mortality, but this requires further investigation as the evidence in this area is conflicting (Ban et al. 2012; Jarde et al. 2016; Ross and Grigoriadis 2014). The fact that most of the studies included in this review were not designed to assess the associations of maternal prenatal mental health disorders and perinatal or infant mortality highlights the need for high-quality prospective cohort studies to better understand the issue.

Conclusions

The findings of this comprehensive systematic review and meta-analysis have demonstrated that women with prenatal mental health disorders are at a higher risk of perinatal and infant mortality, compared with women without prenatal mental health disorders, with consistent effects across a range of mental health disorders. Prenatal screening and access to treatment for mental health disorders with better and co-ordinated antenatal, obstetric, and psychiatric care may reduce the risk of perinatal and infant mortality. Further well-designed studies are required to better understand the nature of the problem, underlying mechanisms, and to confirm whether medical treatments for women with prenatal mental illness reduce perinatal and infant mortality.