Study aims, design, conceptual model, setting, and research team training

Conceptual model.

Fig 1 displays the Socioecological-Biological Model of exposure to structural root causes and the resultant maternal physical and mental health outcomes that guide the study. Structural root causes produce adversity by generating unequal exposure to chronic stress at multiple levels (i.e., neighborhood, family, and individual) across the life course, resulting in deleterious biophysical outcomes such as an increase in inflammation, chronic medical conditions, and harmful epigenetic changes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Socioecological-Biological Model of exposure to structural root causes and maternal biological, health, and mental health implications and outcomes.

https://doi.org/10.1371/journal.pone.0348413.g001

Participant recruitment, enrollment, data collection, and ethics approval

Data collection.

Following enrollment, a 133-question survey, developed specifically for this study, is administered to all participants (Table 1). Throughout the survey, the research team reminds participants that they may decline to answer any question. After completion of the survey, participants’ biological samples (i.e., blood and urine) are collected. Participants provide a urine sample, and a certified phlebotomist performs a venous blood draw under sterile conditions while participants remain seated or supine to prevent dizziness or fainting. After data collection, participants are provided with postpartum medical and mental health resources, offered free Narcan and fentanyl test strips, compensated for their time and participation in the study ($125), and receive parking validation if needed. The data collection appointment takes an average of 60–90 minutes to complete.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Psychosocial and Contextual Measures Collected via Interviewer-Administered Questionnaires.

https://doi.org/10.1371/journal.pone.0348413.t001

Outcomes

Biospecimen handling and data management and protection

Data will be collected in a private area and entered directly into a Health Insurance Portability and Accountability Act (HIPAA)-compliant REDCap database. Biospecimens will be immediately labeled with a unique, study-specific identification number and the collection date and will not contain personal identifiers. On the day of collection, samples are transported to the UIC Department of Pathology Laboratory for processing within 4 hours of collection to obtain plasma, serum, and PBMCs, which are isolated by centrifugation, and then stored in a locked −80 °C freezer within the same laboratory. Every two months, the processed biospecimen samples are transported to the University of Illinois Urbana-Champaign for analysis; dry ice is used to preserve cold chain integrity and sample viability during transport. All data are being stored on the Nightingale computing cluster at the National Center for Supercomputing Applications at the University of Illinois, which is a high-performance computer cluster for sensitive data that operates as a HIPAA-compliant and secure system. The research team will preserve all sequencing data, secondary datasets, and associated codes in public repositories with a Digital Object Identifier. All finalized data will be stored and archived in the Illinois Data Bank; a public-access, file-based repository specifically designed to centralize, preserve, and provide reliable access to de-identified research data. All finalized data will be uploaded to the National Center for Biotechnology Information Gene Expression Omnibus repositories, in accordance with the National Institutes of Health Data Management and Sharing Policy.

Loss of confidentiality is mitigated through multiple safeguards. Authorized study personnel have access to identifiable participant information during recruitment and data collection; all data are subsequently de-identified using unique study identifiers prior to analysis. All data will be entered directly into a HIPAA-compliant REDCap database, de-identified using unique study codes, and stored on secure institutional servers. Electronic health record abstractions will be performed within UIC’s encrypted environment and only authorized research personnel will have access. Identifiers will be removed prior to data analysis and any public data sharing. All data will be de-identified before analysis. Biospecimens and sequencing data will be deposited only after institutional review and archiving, ensuring privacy protection.

Statistical analyses

Throughout, descriptive and bivariable statistics will follow standard approaches. Continuous variables will be presented as means with standard deviations or medians with interquartile ranges, while categorical variables will be summarized as frequencies and percentages. Two broad categories of analysis will be conducted for Aims 1 and 2. We will begin by defining the geographic footprint for each participant within the city of Chicago based on their history of home and work addresses, which is required to calculate aggregated exposure metrics related to environmental exposures and social risks. Neighborhood indices quantify social conditions, environmental pollution and hazards, and safety exposures, which will be appended to participant-level data using the patient’s address. The aggregated exposure for each participant will be joined to the biological metrics. Next, we will examine associations between exposures and biological outcomes within an Exposome-Wide Association Study framework using generalized linear models. Plasma steroid and biomarker concentrations will be evaluated using linear regression with Bonferroni or Sidak corrections; non-parametric Kruskal-Wallis and Dunn post hoc tests will be applied when distributional assumptions are not met. To further understand the mechanistic pathways through which neighborhood context may influence maternal health, we will conduct a set of mediation analysis to assess the paths between stress exposure, changes in biomarkers and maternal outcomes. Epigenomic sequencing data, including DNA methylation profiles, chromatin accessibility, and glucocorticoid receptor binding patterns derived from PBMCs, will be modeled using spatial regression to predict variation in epigenetic markers across multiscale exposures (i.e., Assay for Transposase-Accessible Chromatin with sequencing and spatial Cleavage Under Targets and Release Using Nuclease). A total of 200 postpartum participants across at least 40 census blocks provides 90% power to detect a medium effect size at 0.5 at α = 0.05, assuming an intraclass correlation of 0.05.

For Aim 2, data analysis will focus on evaluating spatial patterns of maternal health outcomes and self-reported survey data in relation to exposure to neighborhood-level structural root causes rather than individual-level risk alone. We will begin by assessing the extent to which maternal risk factors related to neighborhood context explain variation in health outcomes beyond idiosyncratic risk at the individual and family levels by assessing spatial autocorrelation in maternal health outcome variables collected via psychosocial self-reported data, clinical electronic health records, and biological markers. We will calculate relative distances among study participants, and the distance matrix will be joined with the survey data, clinical electronic health records, and biomarkers to conduct the spatial autocorrelation analysis. Findings from this aim will identify the degree to which neighborhood-level exposures contribute to disparities in maternal biological and health outcomes and will inform the development of place-based strategies to address structural determinants of postpartum health.

Study update and discussion

As of March 11, 2026, 101 participants have been enrolled in the study and completed data collection. The study offers several advantages and valuable contributions to maternal health research. This study employs an explanatory sequential mixed methods design, consisting of an initial quantitative phase followed by a qualitative phase to further explain and contextualize the numerical findings. By integrating these methods, the study achieves greater interpretive depth, transforming abstract statistical outputs into contextually grounded conclusions. In the context of maternal health, this design is particularly valuable because it moves beyond clinical metrics to capture the living experiences of perinatal women. This focus is critical given that the study is enrolling a postpartum population that has high exposure to stressors and poor health outcomes [26,53]. The study predominantly enrolls Black and Hispanic women residing in Chicago’s West and South Side neighborhoods, where systemic poverty and violence manifest as toxic stress, resulting in a cascade of adverse maternal health outcomes [27,54,55]. While disadvantage is often linked to poor maternal health, the underlying and intersecting mechanisms and pathways by which social and environmental exposures affect maternal health are largely unknown.

Significant investments have been made to develop technologies, standardize reagents for quantitative measurements, and establish computational pipelines to analyze and interpret epigenetic data. Despite these advances, few studies have leveraged genetic and epigenetic data from at-risk populations, especially among perinatal women, to examine how social stressors and protective factors drive the molecular changes underlying health outcomes. Investigating these intricate pathways requires significant statistical power; accordingly, the study’s sample size of 200 participants is notably large for studies involving the collection and analysis of biological specimens [56–58]. Biomarker and molecular epidemiological studies are frequently constrained by the logistical complexity, cost, and participant burden associated with biological sample collection and processing, often resulting in small sample sizes that necessitate careful statistical optimization to maximize the information extracted from limited biological specimens [59]. More broadly, low statistical power has been identified as a pervasive problem across the biological and neurosciences, with median power estimated at only 8–31%, driven largely by small sample sizes [60].

The larger sample size provides substantially greater statistical power to detect meaningful effects and reduces the risk of inflated effect sizes and irreproducible findings. Another strength of our approach is the multi-dimensional nature of our biological data, encompassing the Cleavage Under Targets and Release Using Nuclease method for histone profiling, a steroid hormone panel, inflammatory cytokines, DNA methylation, and DNA accessibility. Multi-omics and multi-layered epigenomic approaches are increasingly recognized as essential for capturing the complexity of biological regulation that single-assay designs cannot achieve [61]. Importantly, by collecting these diverse molecular data from a cohort of this size, we are well-positioned to identify environmental factors that produce coordinated, dose-dependent biological signatures across multiple systems. For example, graded exposure to neighborhood violence may manifest as a dose-dependent increase in cortisol that, in turn, drives downstream changes in inflammatory cytokine levels, DNA methylation, and shifts in chromatin accessibility, which is consistent with the biological embedding model [62–65]. Our design allows us to trace these cascading/interconnected effects across biological layers within the same individuals, enhancing our capacity to identify mechanistic pathways linking environmental exposures to molecular outcomes.

This study has several limitations that should be acknowledged. A key limitation of this study is its cross-sectional design; because survey and biological data are collected at a single time point, the results provide only a snapshot of the population. This allows for the identification of correlations but precludes drawing conclusions about causality, trends, or potential timing biases. To address this, future research should employ longitudinal designs to track changes over time and establish clearer causal relationships between exposures, biological markers, and clinical outcomes.

Another limitation is that measuring individual exposures based on the history of home and work locations is subject to uncertainty compared to the use of travel diaries and wearable devices. Utilizing spatially weighted kernels centered on home and work addresses, combined with mobility information from self-reported survey data, can potentially attenuate this limitation and provide a more realistic measure of participants’ mobility, thereby improving the quantification of their exposure to different factors. We also plan to assess the sensitivity of the findings to changes in spatial weighting to quantify uncertainty ranges related to specific exposure factors.

Biomarker variability related to physiological stage and ancestry may influence study findings. Previous work identified a ten-cytokine blood signature for depression in postmenopausal women, with four cytokines differing among Black pregnant women. These biomarker patterns may differ in postpartum women due to hormonal changes. Coupling cytokine detection with epigenetic and other molecular data will help identify novel biomarkers of maternal health. Additionally, genetic ancestry and ethnicity will be considered as factors that may influence baseline levels of blood biomarkers.

Despite the limitations, this study will generate comprehensive multilevel data, spanning clinical, biological, and socio-environmental factors, from postpartum women living in high-risk, low-resource communities. By utilizing a multi-omics approach with a robust sample size of 200 participants, this study sets a new methodological benchmark for capturing how social stressors are embedded biologically. It moves the field beyond isolated biomarker analysis to a more integrated systems biology model, tracing how environmental toxins and systemic stressors create cascading molecular signatures across histone profiling, DNA methylation, and inflammatory pathways. The findings could provide the biological evidence needed to advocate for targeted investments in high-risk neighborhoods. By demonstrating a dose-dependent relationship between neighborhood violence and molecular health markers, the research supports policies that treat urban poverty and systemic violence as public health crises requiring structural intervention to improve long-term maternal outcomes. For health providers, this study has the potential to highlight the necessity of trauma-informed perinatal care that accounts for a patient’s living environment as much as their clinical metrics. Finally, in collaboration with our Community Advisory Board and community partners, we are committed to transparency as we share our findings with maternal health stakeholders through academic publications, professional conferences, and direct community engagement and outreach.