A new perspective on the causal pathway between maternal mental health and neonatal adversity

Abstract

Purpose

Substantial evidence suggests a downstream impact of maternal mental health on birth outcomes. The roles of comorbid maternal physical health and familial confounding underlying this association remain unclear.

Methods

This cohort study included a random sample of children born 1997–2008 within a health maintenance organization (HMO) in Israel, their parents, and siblings. Outcomes were ICD-9 diagnoses of neonatal adversities (birth complications and congenital anomalies) and exposures were maternal diagnoses of mental health disorders. Odds ratios (ORs) and their 95% confidence intervals for the associations between maternal mental health diagnoses and measures of neonatal adversity were calculated using logistic regression, adjusting for maternal age, child’s year of birth, socioeconomic status, and maternal physical morbidity burden. We examined potential familial confounding using a negative control approach based on paternal exposure.

Results

In our sample of 74,533 children, 6,674 (9.1%) were born after birth complications and 14,569 (19.9%) with a congenital anomaly. Maternal mental health diagnosis around pregnancy was significantly associated with these measures of neonatal adversity after adjustment for potential confounders (birth complications: OR = 1.3 (1.2–1.4), p < 0.001; congenital anomalies: OR = 1.2 (1.1–1.3), p < 0.001). These associations became attenuated and non-significant after further adjustment for maternal physical morbidity burden. In a joint model, maternal and paternal diagnosis of a mental health disorder were independently associated with neonatal adversity (birth complications: OR_mat=1.3 (1.1–1.4), p < 0.001; OR_pat=1.2 (1.1–1.3), p = 0.004; congenital anomalies: OR_mat=1.2 (1.1–1.3), p < 0.001; OR_pat=1.1 (1.0–1.2), p = 0.01).

Conclusion

Physical health and familial factors play a role in the associations between maternal mental health and neonatal adversity.

Introduction

Maternal mental health (MH) disorders have been associated with several adverse birth outcomes in offspring, including prematurity (Männistö et al. 2016), stillbirth (Adane et al. 2021), and congenital anomalies (Schneid-Kofman et al. 2008; Adane et al. 2021). This risk is especially high among women who have been admitted to a psychiatric hospital or had MH care contact in the year before pregnancy (Langham et al. 2023).

The biological mechanisms proposed to account for the observational associations between maternal MH in pregnancy and birth outcomes predominantly focus on the role of factors occurring downstream to having a MH disorder, including stress hormones, inflammatory factors, changes in placental function (Cripe et al. 2011; De Paz et al. 2011; Sanchez et al. 2013), and medications (Janecka et al. 2018). While many studies investigating the birth sequelae of maternal MH conditions in pregnancy implemented careful adjustment to control for potential confounders, few considered the effects of maternal physical health. Meanwhile, there is substantial evidence for a pervasive comorbidity between mental and physical disorders (Momen et al. 2020; Zaks et al. 2023), including during pregnancy (Khachadourian et al. 2023). This means that pregnant women diagnosed with MH conditions are more likely to bear an increased burden of physical health problems that are themselves associated with neonatal adversity (Rahmati et al. 2020; Wang et al. 2021).

Additionally, family-level characteristics-including certain genetic underpinnings, environmental pollutants, and socioeconomic status - have been found to contribute to the risk of both maternal mental illness and neonatal adversity. Nevertheless, empirical evidence for the role of these factors remains limited, leaving the extent of their contributions and the ways they could confound the association unclear.

A better understanding of the pathways underlying the effects of maternal MH in pregnancy on offspring outcomes is critical to identify potential targets for preventing neonatal and child adversity (Kodesh et al. 2021). While maternal MH in pregnancy requires attention irrespective of its putative effects on offspring, elucidating the role of mediators and potential confounders in the associations may yield important modifiable risk factors. Therefore, we investigated the contributions of maternal physical health and background familial factors in the observational associations between maternal MH diagnoses and measures of neonatal adversity using a large, representative sample from Israel with detailed medical and demographic data.

Materials and methods

This study was approved by the institutional review board of the University of Haifa and the Helsinki Ethics Committee. Those bodies waived the need for informed consent because the study data were fully deidentified.

Sample

We used a population-based cohort from Meuhedet, a large health maintenance organization (HMO) in Israel. We obtained data on a random sample of 19.5% (95,978) of all children born in the HMO between January 1, 1997, and December 31, 2008, as well as their siblings and parents. In order to ensure reliable coverage of parental health records before the child’s birth, we excluded all individuals born before January 1, 1999. Other exclusion criteria included lack of at least one medical diagnosis in parental records (indicating administrative lapse in recording), maternal age at child’s birth lower than 13 or higher than 55, or paternal age lower than 13, non-singleton pregnancies, and lack of covariate data. All children were followed up through January 2015.

Exposures

Exposure was presence of any ICD-9 MH diagnosis (ICD-9: 290–319) in the mother recorded around pregnancy, defined here as within 635 days before child’s birth (spanning the entire pregnancy period and the preceding 12 months, assuming a term pregnancy). Extending the exposure period to 12 months before pregnancy permitted ascertainment of chronic disorders for which the mother may not have sought help during pregnancy itself, but which may still have affected the fetus. In additional analyses we assessed the differential effects across the most prevalent MH diagnoses in our sample (Khachadourian et al. 2023): anxiety, dissociative, and somatoform disorders (ICD-9: 300); personality disorders (ICD-9: 301); other disorders with specific symptoms (ICD-9: 307); and depressive disorders (ICD-9: 311). All exposures were coded as binary variables, indicating presence/absence of the relevant ICD-9 codes in maternal health records. The same criteria were used to ascertain paternal diagnoses of MH disorders.

Outcomes

Outcomes were classified into two categories of neonatal adversity: birth complications (ICD-9: 660–669, 678–679, 760–779) and congenital anomalies of the child (ICD-9: 740–759). Both outcomes were coded as binary variables, indicating presence/absence of the relevant code in the child’s health records obtained from the HMO. We considered diagnoses made within three years of childbirth, allowing for the delay in the diagnosis of certain congenital anomalies. Since the broad ICD categories of birth complications and congenital malformations encompass also the milder forms of the diagnostic categories, we conducted sensitivity analyses considering only the ICD-9 codes with prevalence of < 1.5% in our sample, representing the rarer/more severe conditions (birth complications– ICD-9: 661, 663–665, 669, 764–773, 775–779); congenital anomalies– ICD-9: 740–753, 756–759).

Covariates

The covariates included child’s year of birth, maternal age at child’s birth, family socioeconomic status (SES), and the number of distinct maternal medical comorbidities during the exposure period. Adjusting for the calendar year of childbirth allowed us to mitigate confounding due to temporal changes in the diagnostic rates of psychiatric disorders and neonatal adversities. Controlling for maternal age further enabled us to account for confounding due to the variation in rates of mental illness in women across periods of life, and the higher rates of neonatal adversities among offspring of very young and older women (Cavazos-Rehg et al. 2015). While healthcare in Israel is provided irrespective of demographic and economic position, accounting for family SES allowed us to account for potential socioeconomic differences, e.g. healthcare seeking behaviors in women across SES strata. Family SES was defined as number of electronic devices per person and per capita income in the living area. Lastly, our earlier work showed that women with a MH diagnosis in pregnancy are more likely to receive many other physical health diagnoses (Khachadourian et al. 2023), and that some of these diagnoses could be independently associated with neonatal adversity (Razaz et al. 2017; Rahmati et al. 2020). Therefore, we adjusted for the burden of physical health diagnoses, defined as the total number of distinct ICD-9 level-3 diagnoses recorded during the exposure period. Information about all covariates was obtained from the Meuhedet HMO records.

Statistical analysis

To estimate the effects of maternal MH disorders on each neonatal outcome, we constructed a series of analogous logistic regression models iteratively controlling for covariates. In all models, we adjusted for child’s year of birth and maternal age (model 1); next, we additionally adjusted for family SES (model 2); finally, we also adjusted for the number of distinct medical comorbidities during the exposure period as a proxy measure for maternal burden of physical comorbidities (model 3).

To account for familial correlations due to the presence of siblings in the dataset, all models were stratified by maternal ID to obtain robust standard errors. Since children from larger families were more likely to be included in the dataset (if one child is randomly selected, their siblings are also included), in all models we also adjusted for this sampling strategy by using inverse probability weights determined by the number of children born within the cohort years in the family (see (Janecka et al. 2018) for details on weights calculation).

To investigate familial confounding, we estimated the effects of paternal MH on each neonatal outcome by constructing a logistic regression adjusted for child’s year of birth, parental ages at child’s birth (Janecka et al. 2019) and family SES. We included both parental diagnoses in the model to mitigate potential inflation of paternal results due to assortative mating (Madley-Dowd et al. 2020). The negative control design involves comparing of effects of paternal and maternal diagnoses, leveraging the underlying assumption that paternal exposures during pregnancy cannot directly influence the fetus. As such, an equal magnitude of association between maternal and paternal MH disorders and neonatal adversity would suggest that family-level factors (e.g., genetics or shared environment) drive the association between maternal MH and neonatal adversity. An effect for maternal but not paternal MH disorder, would suggest direct maternal influence of on the fetus.

In the secondary analyses, (a) we tested the effects of specific MH diagnoses (focusing on the most prevalent diagnoses in our sample, affording sufficient power) on birth complications and congenital anomalies. We also (b) interrogated the associations between maternal MH in pregnancy and the rarer/more severe neonatal adversities. The statistical procedures were analogous to those performed in the main analysis, i.e., we used logistic regression with weights, calculated standard errors, and iteratively added the covariates.

Results

Our primary sample consisted of 95,978 children, including the random sample of the birth cohort and their siblings. After removing individuals based on our exclusion criteria, our analytical sample consisted of 74,533 individuals, including 36,534 females (49% of the cohort), and 37,999 males (51% of the sample). The sample demographics are presented in Table 1. Among these individuals, 6,674 (9.1%) experienced complications at birth, and 14,569 (19.9%) were born with any type of congenital anomaly. The majority of the birth complications related to perinatal jaundice (ICD-9: 774, 5.3%), umbilical cord problems (ICD-9: 663; 1.4%), and congenital infection (ICD-9: 771; 1.4%), while the most common congenital anomalies were musculoskeletal deformities (ICD-9:754; 8.2%) and other congenital anomalies of limbs (ICD-9: 755; 5.3%). After excluding the codes with a prevalence ≥ 1.5% in the sample, birth complications and congenital anomalies were diagnosed in 4.1% and 6.8% of the sample, respectively. The prevalence of each included ICD-9 diagnosis is presented in Supplementary Table 1.

Table 1.

Sample demographics

Maternal mental health diagnosis around pregnancy period
	Yes	No
	(N = 3,284)	(N = 72,182)
	N (%)
Offspring female	49.95%	48.97%
Offspring birth complications	375 (11.6)	6,299 (9.0)
Offspring congenital anomalies	776 (24.1)	13,793 (19.7)
	Mean (SD)
Maternal age at child’s birth	31.4 (5.3)	29.7 (5.4)
Paternal age at child’s birth	34.1 (6.0)	32.3 (6.1)
Number of maternal physical health diagnoses in pregnancy	20.3 (17.5)	10.7 (10.1)
Median SES (median absolute deviation)	7.0 (4.4)	7.0 (4.4)

Maternal diagnosis of a mental health disorder is associated with birth complications and congenital anomalies

Main analysis

Maternal diagnosis of any MH disorder around the pregnancy period was significantly associated with birth complications and congenital anomalies after iterative adjustments for child’s year of birth, maternal age, and SES (birth complications: OR = 1.3 (1.2–1.4), p < 0.001; congenital anomalies: OR = 1.2 (1.1–1.3), p < 0.001). See Table 2: models 1 and 2.

Table 2.

Associations between maternal diagnosis of a MH disorder and birth complications and congenital anomalies

	Maternal Mental Health Diagnosis (Exposure)
Outcome	Model 1			Model 2			Model 3
	OR	95% Cl	p	OR	95% Cl	p	OR	95% Cl	p
Birth Complications	1.3	1.1–1.4	< 0.001	1.3	1.2–1.4	< 0.001	1.1	0.9–1.2	0.40
Congenital Anomalies	1.2	1.1–1.3	< 0.001	1.2	1.1–1.3	< 0.001	1.1	1.0–1.2	0.14
Adjustments	Maternal age +			Model 1 +			Model 2 +
	Child’S year of birth			Family SES			# Maternal physical health diagnoses

Restriction to rare outcomes

When the neonatal outcomes were re-defined to only include the rarer/more severe ICD-9 codes (those with prevalence of < 1.5%), the odds remained similar (birth complications: OR = 1.4 (1.2–1.6), p < 0.001; congenital anomalies: OR = 1.3 (1.2–1.5), p < 0.001). See Table S4: models 1 and 2.

Stratification by maternal mental health diagnosis

Considering the effects of specific maternal MH disorders on neonatal adversity, we observed that anxiety, dissociative, and somatoform disorders were associated with birth complications (OR = 1.6 (1.4–1.7), p < 0.001) and congenital anomalies (OR = 1.3 (1.1–1.4), p < 0.001). These results remained consistent across the models with and without adjustment for SES. Personality disorders were associated only with birth complications (OR = 2.0 (1.4–2.7), p = 0.04). There were no statistically significant effects of depressive disorders or other MH disorders with specific symptoms on either measure of neonatal adversity. See Table S5: models 1 and 2.

Adjustment for maternal burden of physical health comorbidities rendered the effects of maternal mental health diagnosis on neonatal adversity non-significant

Main analysis

The associations between maternal MH and neonatal adversity– which remained significant through adjustment for child’s year of birth, maternal age at child’s birth and family SES– were no longer statistically significant after adjusting for maternal burden of physical health comorbidities. See Table 2: model 3. A higher number of maternal physical health diagnoses was associated with risk of neonatal adversity (birth complications: OR = 1.2 (1.1–1.3), p < 0.001; congenital anomalies: OR = 1.1 (1.0–1.2), p = 0.02).

The association between the number of physical health diagnoses and neonatal adversity was present both in women with and without a MH diagnosis, suggesting that the effect was not driven by increased health surveillance in women with MH disorders (birth complications: OR_{MentalHealth−YES}=1.01 (1.00–1.02), p = 0.04; OR_{MentalHealth−NO}=1.02 (1.02–1.02), p < 0.001; congenital anomalies: OR_{MentalHealth−YES}=1.01 (1.00–1.01), p = 0.01; OR_{MentalHealth−NO}=1.02 (1.01–1.02), p < 0.001; NB. Those ORs are additive on a multiplicative scale, i.e. an increase in risk associated with n disorders would be 1.01ⁿ for women with, and 1.02ⁿ for women without a diagnosis of MH disorder).

Restriction to rare outcomes

Consistently, when considering only the rarer/more severe types of birth complications and congenital anomalies, adjusting for maternal physical health comorbidities resulted in attenuation of the effect sizes and significance levels of both types of neonatal adversity (birth complications: OR = 1.2 (1.0–1.4), p = 0.06; congenital anomalies: OR = 1.2 (1.0–1.3); p = 0.04). See Table S4: model 3.

Stratification by maternal mental health diagnosis

A similar pattern was seen for nearly all specific maternal MH diagnoses, with only maternal anxiety, dissociative, and somatoform disorders remaining significantly associated with birth complications after adjustment for maternal burden of physical health diagnoses (albeit with substantially reduced effect sizes). See Table S5: model 3.

There is evidence for familial confounding in the association between maternal mental health diagnosis and neonatal adversity

Main analysis

In a joint model testing the effects of paternal and maternal MH diagnoses concurrently, both were significantly associated with the study outcomes (birth complications: OR_mat=1.3 (1.1–1.4), p < 0.001; OR_pat=1.2 (1.1–1.3), p = 0.004; congenital anomalies: OR_mat=1.2 (1.1–1.3), p < 0.001; OR_pat=1.1 (1.0–1.2), p = 0.01). While the magnitudes of association did not significantly differ between maternal and paternal MH diagnoses and neonatal adversity, the effect (point estimates) of maternal MH was consistently higher. See Table 3.

Table 3.

Associations between maternal + paternal diagnosis of a mental health disorder and birth complications + congenital anomalies

Outcome	Exposure	OR	95% Cl	p
Birth Complications	Maternal MH Diagnosis	1.3	1.1–1.4	< 0.001
	Paternal MH Diagnosis	1.2	1.1–1.3	0.004
Congenital Anomalies	Maternal MH Diagnosis	1.2	1.1–1.3	< 0.001
	Paternal MH Diagnosis	1.1	1.0–1.2	0.01
Adjustments	Maternal age + Paternal age + Child’s year of birth + Family SES

Restriction to rare outcomes

Among the rarer/more severe neonatal outcomes, we observed significant effects only for maternal but not paternal MH diagnosis (birth complications: OR_mat=1.4 (1.2–1.6), p < 0.001; OR_pat=1.1 (1.0–1.3), p = 0.13; congenital anomalies: OR_mat=1.3 (1.2–1.5), p < 0.001; OR_pat=1.1 (0.9–1.2), p = 0.35). See Table S6.

Stratification by maternal mental health diagnosis

We observed a similar pattern of results for the specific MH disorders, whereby the effects of most maternal MH conditions were stronger than those of paternal MH conditions on neonatal adversity. See Table S7.

Discussion and conclusions

Our findings replicated results from earlier studies demonstrating that maternal diagnosis of a MH disorder in pregnancy is associated with birth complications and congenital anomalies in offspring. We further demonstrated that physical health around pregnancy accompanying MH conditions may be an important factor in this relationship, and that the confounding role of background familial factors cannot be ruled out. Taken together, our results underscore the complexity of the effects of maternal MH in pregnancy on the child, and question the notion that the associations with off-spring outcomes arise predominantly due to the downstream effects of maternal MH itself.

We observed that the risk of neonatal adversity associated with maternal diagnosis of a MH disorder is attenuated (but not completely eliminated) after adjustment for her burden of physical health diagnoses. These results are consistent with our (Khachadourian et al. 2023) and others’ (Momen et al. 2020) earlier results demonstrating high comorbidity between mental and physical health, including during pregnancy– with likely knock-on effects on offspring outcomes. While those earlier studies have demonstrated that women with a known MH diagnosis are more likely to have a series of different physical conditions, our current study is the first one to directly demonstrate how this comorbidity impacts the offspring. Previous studies have suggested the role of increased inflammation (Cripe et al. 2011; De Paz et al. 2011; Liu et al. 2014; Humberg et al. 2020; Wu et al. 2021) or lifestyle factors, e.g. smoking (Aliaga et al. 2019) as potential mediators of the observational association between maternal MH and neonatal outcomes. Our findings demonstrating the importance of physical health are not in contrast to these suggestions, as factors like maternal inflammation, smoking, and the burden of physical health conditions could exist on the same causal pathway between maternal MH and neonatal outcomes. Future studies should consider whether these putative mediators converge on similar physiological changes.

Furthermore, as paternal exposures cannot directly influence the fetus– creating a so-called negative control– performing analogous analyses in relation to paternal diagnosis of a MH disorder allowed us to highlight the potential familial confounding underlying the observed associations. Unlike the analyses considering maternal physical health, these analyses were specifically tailored to elucidate potential confounding effects, i.e. the extent to which the association between maternal MH and neonatal adversity could be non-causal in nature, and instead arise due to familial factors. Such familial factors can include both genetic and environmental correlates of mental illness, including e.g., parental socioeconomic factors (Meltzer-Brody et al. 2017), urbanicity (Vassos et al. 2012) or pollution (Peterson et al. 2015) in the area– each of which has also been linked with offspring adversity (Aliaga et al. 2019). While we observed that the maternal point estimates were consistently higher than the paternal ones, the differences were not statistically significant. As with maternal physical health, future studies incorporating a wider range of measures of the home environment and genetic factors should further examine the biological processes through which familial factors could act to affect the child. In the secondary analyses we observed that the role of familial confounding was lower for the relatively rare birth complications and neonatal anomalies, suggesting that these likely more severe outcomes are to a lesser extent influenced by familial factors.

Crucially, neither maternal physical health nor familial factors linked with MH diagnoses fully explained the effects of maternal MH on offspring adversity, suggesting their potential joint effects on the risk of neonatal adversity, likely along with other factors not included in the current study. The effects of anxiety, dissociative, and somatoform disorders on neonatal adversity remained statistically significant after adjustment for the maternal burden of physical health. Considering the strong link between anxiety and depression (Meulen et al. 2021), the absence of such an effect of depression is unexpected and should be replicated prior to interpretation. Future research should further investigate the specific parental factors that in our analyses were captured by maternal burden of comorbidities– including specific diagnoses that contribute to these associations, and the finer underlying mechanisms. In the meantime, studies exploring the potential effects of maternal MH problems in pregnancy should account for the presence of comorbid physical health diagnoses– which, as we have previously shown (Khachadourian et al. 2023), are considerably more common in pregnant women with a MH diagnosis.

The strengths of this study include the use of a large, population-based sample, and reliance on data from one of the largest Israeli HMOs, limiting bias due to ascertainment or differential access to healthcare across individuals in our sample. Using registry-based medical records enabled us to ascertain both exposure and outcome with a high degree of confidence– mitigating the problems associated with e.g., retrospective recall of MH during pregnancy by the parents or missing the neonatal outcomes that might not have been clinically visible immediately at birth. Availability of rich paternal health data allowed us to implement our strategy of estimating the risk of both maternal and paternal exposures jointly, attenuating potential confounding due to assortative mating (Madley-Dowd et al. 2020), whereby propensity for a MH disorder in a parent increases the probability of choosing a partner who is also more likely to suffer from it. Finally, inclusion of a wide range of covariates available through the HMO records and accounting for the sampling strategy represent additional strengths of our work.

However, we also acknowledge the limitations, including lack of a replication sample; use of a relatively crude measure of physical health burden (number of diagnoses)– which does not account for the severity of the recorded diagnoses; availability of only a compound, area-level measure of family SES; and medical data being limited to ICD-9 codes, precluding testing the potential physiological underpinnings of the observed effects. Similarly, the measures of pregnancy complications and congenital anomalies explored in the analyses consisted of a series of ICD-9 codes; therefore, while our conclusions hold more broadly, it is possible that the effects of familial factors and physical health will differ for the individual conditions included under these outcome definitions. In the analyses considering the burden of physical health diagnoses, we also could not distinguish their putative mediating effects (whereby MH disorder leads to increased burden of physical health disorders, which then in turn lead to offspring adversity(Wu et al. 2021) from confounding effects (whereby MH is the consequence and not a cause of physical health disorders). In the future, determining the up- and downstream nature of the associations between mental and physical health will be critical for the identification of potentially modifiable factors influencing risk of neonatal adversity. Additionally, the MH diagnoses used in this study likely do not capture the less severe cases, which could lead to under-ascertainment of the exposure. Furthermore, the severity of congenital anomaly diagnoses ascertained through ICD-based classifications was highly variable, and in the future, other classifications (e.g. by EUROCAT) should be explored. Finally, our study included children with a follow-up period ending in 2008, which restricts our ability to incorporate trends in diagnostic rates of maternal MH disorders and neonatal adversities that may have emerged in the past decade, potentially impacting the generalizability of our findings.

In conclusion, our results indicate that maternal diagnosis of a MH disorder around pregnancy is associated with an increased risk of neonatal adversity, including birth complications and neonatal anomalies. These associations reflect at least in part the contributions of maternal physical health and familial factors associated with MH disorders.

Supplementary Material

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00737-024-01541-2.

Article highlights.

• Maternal mental health conditions are associated with adverse neonatal outcomes (birth complications and congenital anomalies).

• This relationship was largely attenuated after accounting for the higher physical morbidity in women with mental health problems.

• We also demonstrated that factors beyond the direct mother-to-fetus causal route likely contribute to this relationship.

• We offer a new perspective on the mechanisms linking maternal mental health with birth complications and congenital anomalies in offspring.

Data availability

Data access rules do not permit public sharing of the data. Interested researchers should discuss access options with Arad Kodesh and Stephen Levine.