Association of Antepartum and Postpartum Air Pollution Exposure With Postpartum Depression in Southern California

Key Points Question Is maternal ambient air pollution exposure associated with increased risks of postpartum depression (PPD)? Findings In this cohort study of 340 679 pregnant women, antepartum and postpartum exposures to ozone, particulate matter less than or equal to 10 μm, and particulate matter less than or equal to 2.5 μm and its constituents (organic matter and black carbon) were associated with increased risks of PPD. Meaning These findings suggest that long-term antepartum and postpartum air pollution exposure is a potentially modifiable environmental risk factor for PPD and an important public health issue to address for improved maternal mental health.


Introduction
Postpartum depression (PPD) is the main perinatal form of major depressive disorder and is one of the most frequent childbirth complications. 1 Postpartum depressive symptoms affect approximately 10% to 20% of women worldwide, 1,2 and approximately 9% of mothers experience PPD based on validated diagnostic interviews. 3Mothers with PPD experience symptoms such as depressed mood, anxiety, and anhedonia 4 ; have a higher risk of morbidity and mortality from suicide; and may be more likely to commit infanticide. 5,6Infants born to mothers with PPD may be at a higher risk of developing cognitive, emotional, and psychological impairments and behavioral abnormalities. 5,7Identifying modifiable environmental risk factors is important due to increased susceptibility during the antepartum and postpartum periods.
Previous research has shown that air pollution exposure may contribute to several health concerns, including neuropsychological disorders, [8][9][10] and could result in subsequent poor mental health.[19] Inconclusive results may be partially due to differences in study populations, sample sizes, methods of outcome assessments, exposure windows, and study regions with different levels and compositions of particulate matter less than or equal to 2.5 μm (PM 2.5 ).For example, 2 studies performed in the US had very small samples (n = 598 and 180) and, thus, may lack statistical power. 17,18Maternal depression or depressive symptoms in previous studies were defined using questionnaires, such as the Edinburgh Postnatal Depression Scale (EPDS) and Center for Epidemiologic Studies-Depression Scale, without further assessment and clinical diagnosis of PPD.
No previous study has examined PM 2.5 chemical compositions, which have large spatiotemporal variations and affect health differently. 20Furthermore, air pollution exposures in prior studies focused on the antepartum period or were estimated based only on one-time surveys (eg, 6 weeks, 6 months, or 12 months post partum) without considering different durations of time-varying exposures, potentially leading to bias in the exposure estimates.Pregnancy is a sensitive period, as women experience major metabolic changes, and pregnancy-induced hyperventilation may lead to inhalation of more ambient toxic substances. 21Considering susceptible windows of exposure could be as important as considering the intensity of exposure.Investigating the critical exposure time windows and different air pollution constituents could help to reveal etiologic mechanisms and aid in the design of targeted interventions or guidance on behavioral changes for high-risk women during and after pregnancy. 22In this study, we aimed to (1) investigate the associations between PPD and maternal residential exposure to air pollution (ie, PM 2.5 , particulate matter less than or equal to 10 μm

Study Population
This retrospective cohort study was approved by the Kaiser Permanente Southern California (KPSC) institutional review board with an exemption of informed consent, as the research was considered minimal risk for participants.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.Data were analyzed between January 1 and May 10, 2023.

Outcome
Postpartum depression was first assessed using the EPDS for participants during their postpartum visits. 25Participants with positive screening results based on the EPDS (score Ն10) during postpartum visits were referred to a clinical interview for further assessment and follow-up care, including diagnosis and treatment. 26Compared with the PPD diagnosis solely based on questionnaires or diagnostic codes in EHRs, the completeness and accuracy of PPD identification can be improved by supplementing clinical codes with pharmacy use records. 27Therefore, PPD diagnosis in this study was defined by using a combination of International Classification of Diseases, Ninth and 10th Revision diagnostic codes for depression and prescription medication records (eTable 1 in Supplement 1) from the date of delivery to the first 6 months after delivery in KPSC EHRs.

Exposure Assessment
Daily ambient air pollution measurements for NO 2 , O 3 (an 8-hour window of 10 AM-6 PM), PM 2.5 , and PM 10 from 2007 to 2017 were obtained from the US Environmental Protection Agency's monitoring stations.Monthly averages were then calculated and spatially interpolated between stations using empirical Bayesian kriging (EBK). 28,29Publicly available monthly data (1 × 1-km resolution) on PM 2.5 total mass and constituents, including sulfate, nitrate, ammonium, organic matter, and black carbon, were derived from the fine-resolution geoscience-derived model, which combined satellite remote sensing data, ground-based measurements, and chemical transport modeling to provide validated PM 2.5 outputs over North America. 30,31More details are described in the eMethods in Supplement 1 and elsewhere. 32formation on residential changes during pregnancy (address, start date, and end date) was abstracted from KPSC EHRs.All maternal residential addresses were geocoded.Monthly air pollution estimates during pregnancy were spatiotemporally linked to each woman based on the geocoded residential history; air pollution exposures during postpartum periods were estimated based on the residential address at delivery.We then temporally interpolated the monthly air pollution metrics to generate daily values and calculated trimester-specific and postpartum exposures by averaging the air pollution measurements in each specific period (more details are provided in the eMethods in Supplement 1).Long-term antepartum and postpartum air pollution exposure was defined as the period from conception to the date of PPD diagnosis within 6 months after delivery.

Covariates
Pregnancy-related covariates and potential confounders were selected a priori based on the existing literature 2,4,[17][18][19] were obtained from KPSC EHRs, and included maternal age, self-reported race and ethnicity, and educational level; smoking status during pregnancy; exposure to passive smoking; season of conception; and year of infant birth (2008-2016).Block group-level median household income was estimated using the 2013 census tract based on patients' residential address at delivery (categorized as quartiles). 33

Statistical Analysis
The distribution of selected population characteristics and air pollution metrics were assessed.We was applied to examine the correlation between air pollution metrics.The association between PPD and air pollution was examined using time-to-event models, and women without PPD were censored at 6 months after delivery.Specifically, we used a discrete-time approach with pooled logistic regressions to estimate time-varying associations between air pollution exposure and PPD during each period, including the entire pregnancy, trimesters, and postpartum periods.A discrete-time approach is more flexible than traditional approaches that require the proportional hazards assumption (eg, Cox proportional hazards regression) and may be useful for handling large data sets with time-dependent variables.In discrete-time logistic regression models, 34 we included time (the gestational month) as a covariate in a flexible manner (polynomials), as suggested by prior research. 16In trimester-specific analysis, we included all trimester average exposures in a single model to reduce bias as suggested in a previous study. 35Odds ratios (ORs) and 95% CIs of PPD associated with per-IQR increment of each air pollutant were estimated.
The primary analysis was adjusted for maternal age, race and ethnicity, education, median neighborhood household income, smoking during pregnancy, season of conception, and year of infant birth.County was fitted as a random effect to account for potential PPD spatial clustering.In sensitivity analyses, we examined the influence of further controlling for pregnancy-related comorbidities and preterm birth, preconception air pollution levels, green space exposure, multiple deliveries, fixed cohort bias, and urban and rural status (more details are provided in the eMethods in Supplement 1).
Due to potential differences in susceptibility of air pollution effects on health across population subgroups, 36 we performed stratified analyses to examine whether the hypothesized associations differed by maternal characteristics, including maternal age, race and ethnicity, educational level, neighborhood household income, and prepregnancy body mass index categories.Cochran Q tests were used to measure the heterogeneity among subgroups.Adjustment for multiple comparisons was not made for subgroup or sensitivity analyses; thus, those results should be interpreted as exploratory.Participants with missing or invalid data for the exposure (eg, living in regions where air pollution could not be calculated) or confounding variables were excluded from the analysis.All analyses were conducted using SAS, version 9.4 statistical software (SAS Institute Inc).A 2-sided P < .05 was considered statistically significant.

Results
A total of 340 679 singleton births (eFigure in Supplement 1), including 25 674 mothers with PPD (7.54%), at 6 months after delivery were included in our study.Among women with PPD, 9238

Correlations Among Air Pollutants
Summary statistics and Pearson correlation coefficients between air pollution metrics are shown in eTable 2 and the eResults in Supplement 1. Kriged PM 2.5 was positively and moderately correlated with PM 10 (r = 0.66) and NO 2 (r = 0.64), while O 3 was negatively correlated with PM 2.5 (r = −0.11)and NO 2 (r = −0.33)and weakly correlated with PM 10 (r = 0.23).The PM 2.5 mass concentrations from the EBK model and the chemical constituent model were highly correlated (r = 0.82), and both were moderately or highly correlated with PM 2.5 constituents (r = 0.58-0.93).

Associations Between Air Pollution and PPD and Critical Exposure Windows
The associations between antepartum and postpartum exposure to air pollution and the risk of PPD are shown in the Figure .Positive associations were observed between PPD and kriged PM 2.5 , PM 10 , and O 3 .The adjusted OR per IQR increase was strongest for O 3 (1.09;95% CI, 1.06-1.12),followed by PM 10 (1.02; 95% CI, 1.00-1.04)and PM 2.5 (1.02; 95% CI, 1. 00-1.03).No statistically significant association was observed between NO 2 and PPD.For PM 2.5 chemical constituent models, exposure to PM 2.5 total mass, organic matter, and black carbon were associated with increased PPD risks.The strongest association was observed for PM 2.5 black carbon (OR, 1.04; 95% CI. 1.00-1.09).In sensitivity analyses (eTable 3 in Supplement 1), the results were similar to the primary analysis after controlling for space, preconception air pollution, or the random effect of multiple deliveries.
Associations of air pollution with PPD were slightly decreased in magnitude after further adjusting for pregnancy-related comorbidities and preterm birth, while slightly stronger after restricting to urban areas or controlling for fixed cohort bias compared with the primary analysis.
In terms of antepartum air pollution exposure (Table 2), we only found kriged O 3 during the entire pregnancy to be associated with PPD (adjusted OR, 1.07; 95% CI, 1.04-1.10).When considering PM 2.5 constituents, we found an increased PPD risk with PM 2.5 organic matter and black carbon during the entire pregnancy but no associations with other PM 2.5 constituents.The trimester-specific results showed that increased risk was detected in the first and third trimesters for kriged O 3 and in the third trimester for PM 10 and PM 2.5 .Associations of O 3 and PM exposure with PPD were slightly stronger after combining air pollution exposure in the third trimester and the postpartum period.
Moreover, associations between different PM 2.5 constituents and PPD differed by exposure windows (Table 2).Increased risks of PPD were observed for PM 2.5 black carbon during the first trimester and PM 2.5 nitrate and ammonium during the second trimester.

Effect Modification by Maternal Characteristics
In subgroup analyses (eTable 4 in Supplement 1), significant heterogeneity in associations with different air pollutants among population subgroups was observed.Overall, the associations between antepartum and postpartum air pollution exposure and PPD were stronger among mothers aged 25 to 34 years, African American or Hispanic mothers, mothers with higher education (college or more), and mothers with underweight.While Cochran Q tests did not reveal any significant heterogeneity for other characteristics, we observed slightly higher associations among mothers who lived in low-income neighborhoods (quartiles 1 and 2).

Discussion
To our knowledge, this study is the first to examine the association between antepartum and postpartum air pollution exposure, particularly PM 2.5 chemical components, and PPD using a timeto-event framework.In this large obstetric population of 340 679 women in Southern California, we found that long-term antepartum and postpartum exposures to O 3 , PM 10 , and PM 2.5 and its chemical constituents (organic matter and black carbon) were associated with an increased risk of PPD (from 2% for PM 2.5 and PM 10 to 9% for O 3 ).Overall, an increased risk of PPD was associated with PM 10 and PM 2.5 exposure during the late pregnancy and postpartum periods and with O 3 exposure during the entire pregnancy and postpartum period.
8][19] Only 1 previous study assessed postpartum NO 2 but reported no association with PPD. 18Most existing studies were conducted in regions with high air pollution concentrations, which were approximately 2 to 3 times that of our study. 4,17,19The only other study with air pollution levels similar to those in our study was conducted in urban Los Angeles. 18However, this study used the Center for Epidemiologic Studies-Depression Scale rather than PPD diagnosis and had a small sample size (n = 180).Our results are partially consistent with previous findings.We found that antepartum and postpartum exposures to PM 2.5 , PM 10 , and O 3 were associated with an increased PPD risk, although the sensitive windows were different: O 3 during the entire antepartum and the postpartum period and PM 10 and PM 2.5 during the late pregnancy and postpartum period.We did not observe an elevated risk of PPD to be associated with NO 2 exposure.Compared with the limited studies, our study adds to the knowledge of associations between PPD and low levels of air pollution.
No prior research has explored the difference in concentrations and components of PM 2.5 .For trimester-specific exposures, we found that first-trimester PM 2.5 black carbon and second-trimester Previous studies have proposed potential mechanisms that may support our findings, such as dysfunction of the HPA axis, neuroinflammation, and oxidative stress.For example, past research has shown that when exposed to PM 2.5 , the HPA axis becomes overactivated and cannot normally regulate the body's stress response through negative feedback 12,13,18,37 due to decreased glucocorticoid receptor levels and the subsequent release of extra cortisol. 37,38Particulate matter in the upper respiratory epithelium 39 causes an inflammatory response that releases proinflammatory cytokines, such as interleukin 1α (IL-1α), IL-1β, IL-6, and tumor necrosis factor-α. 7,37,38,40The subsequent release of these proinflammatory cytokines can lead to increased activation of the HPA axis.As a reactive oxygen species, long-term O 3 exposure can cause oxidative stress, leading to continued neuroinflammation, 10,41,42 thus increasing proinflammatory cytokine levels.
Associations of demographic factors with PPD are mixed and complex. 6In our stratified analysis, the risk of PPD associated with air pollution exposure was significantly higher among mothers aged 25 to 34 years, African American or Hispanic mothers, mothers with higher education, and mothers with underweight, suggesting that these population subgroups may be more vulnerable to air pollution and at an increased risk for PPD.Effective screening and prevention strategies (eg, purposeful use of air filters, avoidance of outdoor activity during smoggy days, and avoidance of major sources of air pollution) focusing on the most influential time windows (ie, late pregnancy and after delivery) could be recommended, especially for vulnerable subpopulations, to optimize the benefits of reducing air pollution exposure on maternal mental health.

Strengths and Limitations
The main strengths of this study include the large and diverse obstetric population; the high-quality and rich clinical data, especially PPD ascertainment; a wide range of air pollutants and exposure windows examined; detailed residential history in combination with well-validated air pollution models; and comprehensive individual-and neighborhood-level covariates that allowed us to control for a series of potential confounders, explore effect modification by maternal characteristics, and conduct several sensitivity analyses to check the robustness of our findings.Some limitations of our study also should be acknowledged.First, although residential mobility during pregnancy was considered, postpartum exposures were estimated solely based on maternal address at delivery, which may have led to exposure misclassification.Nonetheless, based on the residential history during pregnancy for this population, most women who relocated may have moved within the same subregion (median distance, 6 km), which would not significantly change their exposure levels. 32In addition, potential exposure misclassifications may also exist since indoor and personal exposure levels could not be estimated without data on personal monitors or activity patterns.Second, it is common for individuals with mental health disorders to have delays in diagnosis and initial treatment. 43Although we used follow-up data from multiple time points and detailed medical records, a proportion of the women with depression may not have been identified in a timely manner, which may have led to an underestimation of the associations.Future research may also consider PPD events over a longer postpartum period.Third, although several covariates were adjusted for, some residual or unmeasured covariates were inevitable due to data unavailability, such as psychiatric history, adverse life events, and marital status, which may affect mental health and potentially bias the estimates.Moreover, previous studies have reported associations between other environmental factors and PPD 44,45 ; thus, research is needed to explore the joint effects of air pollution and other related exposures, such as green space, noise, and meteorologic factors.
Furthermore, studies conducted in other regions and populations are warranted as air pollution levels and compositions could vary in different regions and be associated with different health outcomes.

[
PM 10 ], nitrogen dioxide [NO 2 ], ozone [O 3 ], and PM 2.5 constituents) in a large pregnancy cohort in Southern California and (2) identify windows of susceptibility to antepartum and postpartum air pollution exposure.

used χ 2
tests to determine the difference between PPD and non-PPD groups.Pearson correlation JAMA Network Open | Environmental Health Antepartum and Postpartum Air Pollution Exposure and Postpartum Depression JAMA Network Open.2023;6(10):e2338315. doi:10.1001/jamanetworkopen.2023.38315( 23,24artum and Postpartum Air Pollution Exposure and Postpartum DepressionThis cohort included women who had births between January 1, 2008, and December 31, 2016, at KPSC facilities.The KPSC serves approximately 19% of the population in Southern California and represents its sociodemographic diversity, which provides valid inference for epidemiologic research.23,24Womenwho were not a KPSC member, who gave birth at less than 20 weeks or more than 47 weeks gestation, who did not have a residential address, who'd had multiple births, and who had deliveries resulting in a stillbirth were excluded from the study.The KPSC electronic health JAMA Network Open.2023;6(10):e2338315. doi:10.1001/jamanetworkopen.2023.38315(Reprinted) October 18, 2023 2/12 Downloaded From: https://jamanetwork.com/ on 10/23/2023

Table 1 .
Selected Population Characteristics by Postpartum Depression (PPD) Status, 2008-2016 Abbreviation: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared).aIncludes Pacific Islander, American Indian or Alaska Native, and mothers with multiple races and ethnicities specified by the patient.

Table 2 .
Postpartum Depression Associated With Air Pollution Exposure in Specified Periods of the PregnancyAbbreviations: NO 2 , nitrogen dioxide; O 3 , ozone; OR, odds ratio; PM 2.5 , particulate matter Յ2.5 μm; PM 10 , particulate matter Յ10 μm.Units of measurement are units μg/m 3 for PM 10 , PM 2.5 total mass, and PM 2.5 constituents, and parts per billion for NO 2 and O 3 .ORs and 95% CIs were calculated per IQR increment for each air pollutant.The base model was adjusted for maternal age, race and ethnicity, education, median neighborhood household income, smoking during pregnancy, season of conception, and year of infant birth; county was fitted as a random effect.PM 2.5 nitrate and ammonium exposures were associated with an increased risk of PPD.For long-term exposures during pregnancy and postpartum periods, PM 2.5 organic matter and black carbon were the main components associated with PPD.