Investigation of the impact of social, behavioral, and pre-existing clinical risk factors on preeclampsia

Abstract

Background

Preeclampsia is one of the leading causes of maternal morbidity and mortality, disproportionately impacting women from diverse backgrounds. Early identification of risk factors is critical for timely preventive interventions. This study examined the impact of socioeconomic factors, prior substance use, and pre-existing health conditions on the development of preeclampsia using the data from the NIH All of Us Research Program.

Methods

Participants aged 18 to 55 with at least one pregnancy in the NIH All of Us research program between 2018 and 2023 were included. Relative factors considered include socioeconomic status (education, income, health insurance), substance use disorders (alcohol, cannabis, cocaine, nicotine, opioid, stimulant), and pre-existing health conditions (blood clotting, depression, diabetes, hypertension, liver failure, renal failure, thyroid-stimulating hormone). Electronic health records and surveys were used to identify the study cohort and risk factors. We conducted multivariate analysis using a generalized linear model, along with subgroup analysis.

Results

A total of 12,098 individuals were identified with 14,354 pregnancy episodes after applying the inclusion and exclusion criteria, and 8.7% of the participants developed preeclampsia. They were predominantly Hispanic (43.5%), Non-Hispanic White (33.9%), and Non-Hispanic Black (14.5%), with an average age of 30.4 years (SD = 6.2). Among these individuals, 35% had a college or advanced degree, 38% reported income less than $50k, and 7% lacked health insurance. The commonly used substance use disorders were nicotine (8.5%), cannabis (3.7%), and opioids (2.2%). The leading pre-existing clinical conditions included depression (17%), followed by hypertension (8%), and diabetes (5%). Adjusted odds ratio from multivariate generalized linear model indicated that participants who had income less than $50k (OR:1.35, 95% CI: 1.05–1.74) are more likely to develop preeclampsia compared with those who had income higher than $100k. Other significant factors included stimulant use disorder (OR:2.04, 95% CI: 1.32–3.14), diabetes (OR:1.36, 95% CI: 1.07–1.70), and hypertension (OR:2.54, 95% CI: 2.12–3.05).

Conclusions

Low income, stimulant use disorder, diabetes, and hypertension were strongly associated with preeclampsia.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12884-026-09612-1.

Background

Preeclampsia is a complex, multisystem hypertensive disorder marked by the sudden onset of hypertension and proteinuria after 20 weeks of gestation in women with previously normal blood pressure [1, 2]. Despite evolving clinical criteria, preeclampsia remains a major contributor to maternal morbidity and mortality globally. In the United States, preeclampsia affects 5–8% of pregnancies with an even higher prevalence in developing nations [3]. As one of the leading causes of maternal mortality, preeclampsia contributes to over 70,000 maternal deaths and 500,000 fetal deaths worldwide each year [4]. Significant disparities exist in the United States. Black women experience preeclampsia at a 60% higher rate than White women [4]. Women in developing countries are seven times more likely to develop preeclampsia than those in developed nations, with mortality rates ranging from 10% to 25% [5, 6].

A wide range of social, behavioral, and pre-existing clinical factors influence the development of preeclampsia. Previous studies have reported that pre-existing clinical risk factors for preeclampsia include chronic hypertension, diabetes, kidney disorders, multifetal gestation, and substance use disorders [7–11]. Risk also increases with advancing maternal age, reflecting the higher incidence of comorbidities and vascular dysfunction among older pregnant individuals [12, 13]. Substance use before pregnancy is a significant public health concern in the United States [11]. The most frequently used substances are tobacco, alcohol, and marijuana, followed by cocaine and opioids [14–17]. Substance use before or during pregnancy frequently co-occurs with adverse socioeconomic conditions, including low income, limited educational attainment, and reliance on government-funded healthcare insurance, which independently and jointly contribute to increased obstetric risk [18–23]. Among these risk factors, the influence of substance use on preeclampsia remains comparatively underexplored [14, 24]. One of the significant challenges in analyzing the influence of substance use on preeclampsia lies in the limited understanding of the linkage between behavioral, clinical, and pregnancy outcomes data [25].

Recent studies have increasingly focused on the influence of social and behavioral factors of health on pregnancy outcomes [18, 19]. Studies have largely examined these risk factors separately, limiting the ability to capture their combined and potentially synergistic effects on the risk of preeclampsia [26–28]. A large systematic reviews of preeclampsia indicate that biomarkers, maternal history, and clinical factors remain the primary basis of risk stratification, with limited representation of non-clinical social and behavioral determinants [20]. An early warning preeclampsia model was developed that included psychological and clinical factors [21]. While many social determinants are associated with preeclampsia, most existing research focuses on individual exposures rather than integrated analysis that combines social, behavioral, and clinical factors [22]. Furthermore, different populations affected by preeclampsia may exhibit varying prevalence rates, patterns, or distribution of risk factors [23].

To accommodate the abovementioned research needs, we have conceptualized an integrative analytic approach that considers social, behavioral, and clinical determinants of preeclampsia simultaneously within large, diverse populations across the United States in the NIH All of Us Research Program [29]. We used generalized linear models to identify and interpret risk factors of preeclampsia and conducted subgroup analysis to examine differences in preeclampsia risk among race/ethnicity and age groups. Therefore, the overarching objective of the study is to identify socioeconomic, behavioral (prior substance use disorders), and clinical (pre-existing conditions) risk factors associated with preeclampsia, and conduct subgroup analysis to evaluate variation in preeclampsia risk across race/ethnicity and age groups.

Methods

Study population

Participants with at least an electronic health record of pregnancy between January 1, 2018, and December 31, 2023, in the National Institutes of Health (NIH) All of Us (AoU) Research Program were included. The pregnancy events include normal pregnancy as well as pregnancies with clinical conditions, including high-risk pregnancy, fetal abnormalities, normal delivery, incidental pregnancy, gestational hypertension, and gestational diabetes. These conditions were identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes defined in the 2023 official guideline from the Centers for Medicare and Medicaid Services (CMS) [30]. A single pregnancy era encompasses multiple pregnancy-related encounters, such as prenatal visits, labor and delivery records, follow-up consultations, and other medical interactions documented throughout the pregnancy journey. Individuals with pregnancies before January 2018 were excluded as AoU initiated participant enrollment in 2018 [26]. We considered the analysis of pregnancies occurring among individuals aged 18–55 years to ensure ethical consistency, and inclusion of clinically plausible pregnancies at advanced maternal ages [27]. In the case of multiple pregnancies, only the first recorded pregnancy for each individual was included in the analysis. A detailed inclusion and exclusion flow diagram is presented in Fig. 1.

Fig. 1.

Flowchart of identification of pregnant participants based on inclusion and exclusion criteria

Data collection

Launched in 2018, the NIH All of Us (AoU) Research Program is a collaborative initiative involving over 400 academic institutions. Its goal is to enroll at least one million participants across the United States to build one of the most diverse health databases in history. The program collects comprehensive data, including electronic health records (EHR), surveys, physical measurements, biospecimens, and wearable device information, to better understand how genetic, environmental, and lifestyle factors influence health and disease. The participants are from a diverse cohort of adults (18+) across the United States, with nearly half representing groups traditionally underrepresented in research, including diverse racial and ethnic backgrounds, varied socioeconomic status, and individuals with disabilities [28, 29]. The dataset is stored in a secure cloud-based researcher workbench platform. An institutional data use agreement (DUA) is required. As an authorized user under the DUA between Meharry Medical College and NIH All of Us Research Program, we can access diverse data that enable us to investigate a broad spectrum of factors that influence health and disease [31, 32]. This study utilized available deidentified data from Controlled Tier - Version 7.

Index pregnancy

The index pregnancy was defined as the first recorded pregnancy episode for everyone. We first identified the initial hospital visit related to pregnancy using pregnancy-specific ICD-10 diagnosis codes recorded in the EHR. For each participant, the earliest pregnancy-related encounter was used to calculate the pregnancy start date, and all subsequent encounters occurring within a 10-month window were grouped into a single pregnancy episode. The 10-month interval was chosen to reflect the expected gestational period while accommodating heterogeneity in EHR documentation, including delayed prenatal care and post-delivery follow-up. To distinguish multiple pregnancy episodes for the same participant during the study period (2018–2023), we calculated the interval between the start date of one pregnancy and the next recorded pregnant visit. If this visiting interval exceeded 10 months, it was considered the start of a new pregnancy episode. This approach minimized misclassification due to overlapping or closely spaced pregnancies. The most frequently observed ICD-10 codes for pregnancy identification are normal pregnancy (Z34) and high-risk pregnancies (O09). Other pregnancy conditions are incidental pregnancy state (Z33.1), fetal abnormalities (O35, O36), normal delivery (O80), gestational diabetes (O24), and gestational hypertension (O13, O16). Less common conditions included viral diseases (O98, primarily HIV), abnormal glucose levels (O99.81), cardiomyopathy (O90.3), and maltreatment in pregnancy (O9A.3, O9A.4, O9A.5). Alcohol and drug use during pregnancy (O99.32) was also included for cohort identification, while substance use disorders before pregnancy were included as a potential predictor in the study outcome. Collectively, these diagnoses reflect the heterogeneity of pregnancy experience captured in the dataset with nearly 68% of pregnancy episodes involving either high-risk or non-routine conditions. The complete list of pregnancy-related ICD-10 codes and their corresponding frequencies across the cohort are presented in Supplementary Table 1.

Predictors

Multiple predictors were included in this analysis, including race, ethnicity, age, socioeconomic status, substance use disorders (SUDs), and pre-existing health conditions. Race and ethnicity were obtained from self-reported electronic health record data and analyzed using a composite variable. Individuals identifying as Hispanic or Latino were classified as Hispanic regardless of race, while Non-Hispanic individuals were categorized as Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Asian, or Non-Hispanic Other [33]. Socioeconomic variables, including education, household income, and insurance coverage, were derived from participant-reported basic survey questionnaire, whereas substance use disorders and medical conditions were identified using ICD-10 diagnosis codes from EHR in AoU. The SUDs condition before pregnancy included alcohol (F10), cannabis (F12), cocaine (F14), nicotine (F17), opioid (F11), and stimulant use (F15). The pre-existing medical conditions contain blood clotting disorders (D68.4-68.6, D68.8-68.9), depression (F31.3-31.5, F32, F33, F34.1, F43.2), diabetes (E13, E10, E11), hypertension (010), liver failure (B18, I85, I86.4, K70, K71.1, K71.3, K71.5, K71.7, K72, K74, K76.0, K76.2-K76.9, K77, Z94.4), renal failure (I12.0, I13.1, I13.2, N18, N19, N25.0, Z49.0, Z94.0, Z99.2), and thyroid-stimulating hormone (TSH) abnormalities (D69.4-D69.6). The selection of these risk factors was based on prior epidemiologic and clinical evidence, and their established relevance to preeclampsia [8, 34, 35].

After integrating multiple predictors, we observed substantial unresponsive or skipped results in socioeconomic variables, with more than 31% of participants not reporting or skipping income-related survey questions, followed by 1.4% for education and 2.2% for health insurance. As socioeconomic questions were self-reported, we grouped these unreported answers into an unknown category rather than applying multiple imputation in our analysis to minimize potential bias. Pre-existing behavioral and clinical variables were defined using EHR hospital visit records before the pregnancy start date, with an additional 24-hour observation window to capture relevant health information during observation encounters or documented with a short delay.

Outcome variable

The primary outcome was preeclampsia (O11, O14) occurring during pregnancy. A time window relative to the pregnancy visit date was defined to capture pregnancy-related preeclampsia. Specifically, preeclampsia outcome was identified if recorded within the period from 4 months before the first pregnancy start date to 10 months after the pregnancy start date (Fig. 2). The prior 4 months of this window accounted for delayed entry into prenatal care or diagnostic carry-forward from early prenatal encounters documented in the EHR. Postpartum preeclampsia was not included.

Fig. 2.

Preeclampsia observation window. Note: (− 4 months) pre-pregnancy period included to account for delayed EHR recording

Statistical analyses

We conducted a multivariate regression analysis using a Generalized Linear Model (GLM) with a binomial family and a logit link function for the binary outcome of preeclampsia. The model included main effects for all 34 covariates, including race/ethnicity, age, socioeconomic status, behavior, and clinical factors to estimate adjusted associations with the risk of preeclampsia. The descriptive statistics report each continuous variable with median and interquartile range (IQR), and categorical variables were presented as frequencies with percentages. Multiple categorical variables were encoded by using one-hot encoding during the analyses. The GLM model results are presented as odds ratios (ORs) with corresponding 95% confidence intervals (CIs), and p < 0.05 was considered statistically significant. To account for multiple comparisons across demographic, social, behavioral, and clinical variables, p-values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure. Continuous variables were compared using Welch’s t-test, and categorical variables were compared using the chi-square test. Furthermore, subgroup analyses were conducted based on race/ethnicity and age groups to identify potential differences in preeclampsia risk among individual groups. All statistical data analyses were performed using Jupyter Notebook on the NIH All of Us researcher workbench.

Results

Study population

A total of 12,098 pregnant individuals with 14,354 pregnancy episodes were identified. Analyses were restricted to the single pregnancy episode for each individual, resulting in 12,098 observations for descriptive, statistical, and subgroup analyses after applying the inclusion and exclusion criteria. Participants were predominantly Hispanic (43.5%) and Non-Hispanic White (33.9%), with a mean maternal age of 30.4 years (SD = 6.2). Approximately one-third of the participants held an advanced or college degree, 37.8% were reported their income was less than $50k, and uninsured (6.6%). The most common SUDs before pregnancy were nicotine (8.5%), cannabis (3.7%), and medical conditions were depression (17.1%) and hypertension (8.1%). A total of 1,054 individuals (8.7%) experienced preeclampsia across 1,248 multiple pregnancy episodes (10.3%). After adjusting for multiple comparisons, race/ethnicity, social (education status, annual income), stimulant use disorder, pre-exiting conditions (diabetes, hypertension, renal failure) were significantly associated with preeclampsia. A detailed descriptive statistics and correction for multiple testing of the study population is provided in Table 1.

Table 1.

Social, behavioral, and clinical characteristics of the study population, with correction for multiple testing

Variables	All pregnant Individuals (N = 12,098)	P Value
Age (years), mean (SD)	30.4 (6.2)	0.431
Race/Ethnicity, n(%)		< 0.001
Hispanic	5259 (43.5%)
NH Asian	411 (3.4%)
NH Black	1758 (14.5%)
NH White	4101 (33.9%)
NH Other	102 (0.8%))
Unknown	467 (3.9%)
Socioeconomic Factors
Education Status, n (%)		< 0.001
Advanced Degree	1878 (15.5)
Below Eleven Grade	1298 (10.7)
College Graduate	2346 (19.4)
High School Graduate	3245 (26.8)
Some colleges, 1–3 years	3161 (26.1)
Unknown	170 (1.4)
Income Level, n (%)		< 0.001
Higher than 100k	2043 (16.9)
Less Than 50k	4594 (37.97)
50k to less than 75k	953 (7.9)
75k to 100k	751 (6.2)
Unknown	3757 (31.1)
Health Insurance Coverage, n (%)		0.174
No	531 (4.4)
Yes	11,298 (93.4)
Unknown	269 (2.2)
Prior Substance Use Disorders, n(%)
Alcohol Use	213 (1.8)	0.988
Cannabis Use	449 (3.7)	0.331
Cocaine Use	80 (0.7)	0.998
Nicotine Use	1031 (8.5)	0.141
Opioid Use	261 (2.2)	0.957
Stimulant Use	177 (1.5)	0.01
Pre-existing Clinical Conditions, n(%)
Blood Clotting	415 (3.4)	0.841
Depression	2074 (17.1)	0.887
Diabetes	610 (5.0)	< 0.001
Hypertension	982 (8.1)	< 0.001
Liver Failure	409 (3.4)	0.945
Renal Failure	101 (0.8)	< 0.001
Thyroid-stimulating Hormone	208 (1.7)	0.468
Outcome
Preeclampsia, n (%)	1054 (8.7)

NH Non-Hispanic, Other: Middle Eastern or North African, Native Hawaiian or Other Pacific Islander; Unknown: I prefer not to answer, PMI: Skip, None of these, None Indicated

P values represent the statistical significance of differences in participant characteristics after correction for multiple testing using the Benjamini–Hochberg FDR method

Predictors

Table 2 summarizes the statistically significant odds ratios (ORs) for variables associated with preeclampsia. The ratio of preeclampsia events to included predictor variables was approximately 31, which is higher than the recommended events-per-variable threshold and supports reliable multivariable estimates. Hispanic (OR: 1.36, 95% CI: 1.14–1.61) and NH Black (OR: 1.52, 95% CI: 1.24–1.86) individuals had higher odds of developing preeclampsia compared with NH White participants. Compared with household income higher than $100k, income less than $50k (OR: 1.35, 95% CI: 1.05–1.74) was significantly associated. Furthermore, prior stimulant use disorder (OR:2.04, 95% CI: 1.32–3.14), and pre-existing health conditions including diabetes (OR: 136, 95% CI: 1.07–1.70) and hypertension (OR:2.54, 95% CI: 2.12–3.05) were strongly associated with increased risk of preeclampsia.

Table 2.

Social, Behavioral, and Clinical factors odds ratio (OR) association with Preeclampsia

Variables	OR (95% CI)	P value
Age	0.994 (0.983 to 1.004)	0.237
Race/Ethnicity (Reference: NH White)
Hispanic	1.356 (1.139 to 1.612)	< 0.001
NH Asian	1.089 (0.742 to 1.6)	0.66
NH Black	1.516 (1.239 to 1.855)	< 0.001
NH Other	0.322 (0.101 to 1.024)	0.055
Unknown	1.187 (0.8571 to 1.643)	0.303
Socioeconomic Factors
Education Status (reference: Advanced Degree)
Below Eleven Grade	0.984 (0.733 to 1.32)	0.915
College Graduate	0.96 (0.757 to 1.218)	0.738
High School Graduate	0.869 (0.667 to 1.133)	0.3
Some colleges, 1–3 years	1.026 (0.799 to 1.318)	0.839
Unknown	1.437 (0.898 to 1.298)	0.13
Income Level (reference: Higher than 100k)
Less Than 50k	1.352 (1.049 to 1.744)	< 0.001
50k to less than 75k	1.122 (0.906 to 1.651)	0.189
75k to 100k	0.992 (0.708 to 1.389)	0.961
Unknown	1.627 (1.256 to 2.109)	< 0.001
Health Insurance Coverage (reference: Yes)
No	1.09 (0.829 to 1.433)	0.535
Unknown	0.988 (0.673 to 1.449)	0.95
Prior Substance Use Disorders
Alcohol Use	0.716 (0.443 to 1.163)	0.394
Cannabis Use	0.973 (0.718 to 1.319)	0.862
Cocaine Use	0.942 (0.459 to 1.933)	0.871
Nicotine Use	0.977 (0.781 to 1.222)	0.838
Opioid Use	0.825 (0.529 to 1.285)	0.394
Stimulant Use	2.038 (1.322 to 3.142)	< 0.001
Pre-existing Clinical Conditions
Blood Clotting	1.164 (0.769 to 1.759)	0.473
Depression	0.913 (0.775 to 1.075)	0.273
Diabetes	1.356 (1.072 to 1.702)	< 0.001
Hypertension	2.539 (2.117 to 3.047)	< 0.001
Liver Failure	1.922 (0.677 to 1.256)	0.605
Renal Failure	1.42 (0.892 to 2.259)	0.139
Thyroid-stimulating Hormone	0.588 (0.317 to 1.091)	0.092

Risk factor comparison by subgroup analysis

We further stratified participants into different racial/ethnic and age groups to examine the potential association between each risk factor and preeclampsia within subgroups (Supplementary Tables 2 and 3). The race/ethnicity groups include Hispanic, NH Black, NH White, and the age groups are 18–25, 26–35, 36–55 years. Pre-existing hypertension and stimulant use disorder were consistently and strongly associated with preeclampsia across all racial/ethnic and age groups. Among NH Black individuals, participants who had an income less than $75k (OR: 5.54, 95% CI: 1.15–26.75) were more likely to develop preeclampsia compared with those who had an income higher than $100k, participants with an education of high school showed lower odds of association (OR: 0.46, 95% CI: 0.23–0.93) comparing with advanced degree participants. NH White participants with renal failure (OR: 3.0, 95% CI: 1.09–8.21), and Hispanic individuals with diabetes (OR: 1.89, 95% CI: 1.33–2.70) had higher odds to develop preeclampsia. Age was a significant factor across all racial and ethnic groups. Among younger participants (aged 18–25 years), diabetes (OR: 1.97, 95% CI: 1.04–3.68) was significantly associated with increased risk of preeclampsia. In the age group of 26–35, Hispanic (OR: 1.40, 95% CI: 1.10–1.79), and NH Black participants (OR: 1.87, 95% CI: 1.41–2.49) had significantly higher odds compared with NH White. Among those aged 36–55 years, participants having an annual income below $50k (OR: 1.68, 95% CI: 1.04–2.72) were more likely to develop preeclampsia compared with income above $100k. Diabetes (OR: 1.59, 95% CI: 1.03–2.43) was a statistically significant factor associated with preeclampsia.

Discussion

This study examined the association of preeclampsia risk during pregnancy through the combined lens of social, behavioral (substance use disorders), and clinical (pre-existing health) factors. The preeclampsia prevalence observed in this study (8.7%) is at the upper bound of reported U.S. estimates (5–8%) [3]. It reflects the diverse and socially vulnerable populations represented in the NIH All of Us Research Program. Our analyses indicated that social disadvantage, prior SUDs, and pre-existing clinical conditions jointly contributed to the risk of preeclampsia. The integration of these variables will improve risk stratification and help identify individuals who could benefit from enhanced prenatal monitoring or targeted interventions. Pre-existing hypertension emerged as the dominant and most consistent predictor, along with metabolic comorbidities and SUDs. Collectively, these findings underscore that preeclampsia risk arises from the intersection of social, behavioral, and clinical vulnerabilities rather than a single etiologic pathway.

Our study indicated persistent racial/ethnic and social disparities in preeclampsia risk, with Hispanic and NH Black participants showing higher odds of preeclampsia compared with NH White individuals. Lower household income (<$50k) was associated with increased preeclampsia risk relative to income levels exceeding $100k, reflecting the influence of socioeconomic disadvantage. These findings are consistent with previous studies [36, 37]. Subgroup analyses further revealed heightened vulnerability among NH Black participants, particularly those with household incomes below $75k.

Behavioral risk factors were associated with heterogeneity across racial/ethnic and age groups. Stimulant use emerged as a prominent risk factor among Hispanic and NH White individuals, with effect sizes comparable to or exceeding those of several clinical comorbidities, consistent with previous findings [38]. Pre-existing health conditions played an important role in developing preeclampsia. Hypertension emerged as the most dominant and consistent predictor across all racial/ethnic and age groups, consistent with prior literature [8, 39, 40]. Diabetes was also significantly associated with elevated risk, supporting evidence linking vascular and metabolic dysfunction to hypertensive disorders of pregnancy [14]. Together, these findings underscore the contribution of baseline cardiometabolic health to the development of preeclampsia.

Survey responses were collected at a single time point, which were not always completed during pregnancy. However, survey completion generally occurred close to the index pregnancy date (median difference = 0.0 years; IQR: 6 months before to 4 months after the index pregnancy period). Although some participants completed the survey up to 4.8 years before pregnancy, while others completed it up to 4.4 years after delivery. Because socioeconomic status is dynamic, a single time-point measurement may not adequately capture the socioeconomic conditions most relevant to preeclampsia. Additionally, socioeconomic variables were self-reported, introducing random measurement error that may attenuate observed associations. The time gap between the socioeconomic survey and the index pregnancy period is likely to impact the association between education and preeclampsia, which is not statistically significant.

From a public health perspective, these findings have important implications for improving preeclampsia risk identification and prevention. The results suggest that incorporating socioeconomic and behavioral factors alongside traditional clinical risk factors may enhance risk stratification and screening efforts, enabling earlier identification of women at elevated risk for adverse pregnancy outcomes. These findings highlight the need to move beyond a solely biomedical approach and adopt a more comprehensive framework that recognizes the influence of social and behavioral determinants on maternal health. Effective prevention of preeclampsia will require not only robust clinical surveillance and evidence-based medical management but also policies and interventions that improve access to high-quality prenatal care, behavioral health services, and social support resources. Such efforts may be particularly important for women from low-income, underserved, and socially marginalized communities, who continue to experience a disproportionate burden of preeclampsia and related maternal health complications.

Limitations

This study has several limitations. First, approximately 31% of the participants had missing income data. As the socioeconomic survey was self-reported, the unreported responses were categorized as unknown rather than being imputed to minimize potential bias, which potentially affects the observed associations [41]. Second, socioeconomic survey data were collected at a single time point and may not capture changes related to the pregnancy period. Third, substance use disorders identified through ICD-10 codes likely underestimate the true prevalence, as these conditions are often underreported during pregnancy due to social stigma. Finally, our analysis was limited to twelve pregnancy-related conditions defined according to CMS ICD-10 guidelines. Future studies will incorporate a broader spectrum of pregnancy complications and maternal health conditions to provide a more comprehensive characterization of preeclampsia and its associated risk factors.

Conclusion

In summary, this study identified significant risk factors for preeclampsia during pregnancy through the integration of social, behavioral, and clinical conditions. To our knowledge, it is the first study of preeclampsia to apply this multifactorial approach and leverage data from the NIH All of Us Research Program. Our findings demonstrate the potential of big data analytics to enhance risk stratification of preeclampsia using statistical and subgroup approaches, thereby supporting personalized prenatal care. As healthcare systems increasingly adopt advanced analytics, this approach can play a critical role in improving risk stratification, personalized prenatal care, and promoting more efficient allocation of maternal health resources.

Abbreviations

AoU

All of Us

NIH

National Institutes of Health

CMS

Centers for Medicare and Medicaid Services

ICD-10

International Classification of Diseases, Tenth Revision

SUD

Substance Use Disorder

HER

Electronic Health Record

NH

Non-Hispanic

GLM

Generalized Linear Model

IQ

Interquartile Range

OR

Odds Ratio

CI

Confidence Interval

SD

Standard Deviation

FDR

False Discovery Rate

HRSA

Health Resources and Services Administration

HHS

Health and Human Services

Authors’ contributions

All authors contributed to this study. AC supervised all stages of the work and provided critical scientific and methodological guidance on study design, data management, analysis, interpretation, and manuscript reviews. US conducted the data analysis and manuscript writing under the supervision of AC. TG, BA, LA, and BS provided critical review and revisions to improve the intellectual content of the manuscript. All authors read and approved the final manuscript.

Funding

The research is supported by the Health Resources and Services Administration (HRSA) of the U.S. Department of Health and Human Services (HHS) under cooperative agreement UR6MC50342, which began on September 30, 2023, Maternal Health Research Collaborative. The information, content, and/or conclusions are those of the author. They should not be construed as the official position or policy of, nor should any endorsements be inferred by HRSA, HHS or the U.S. Government.

Data availability

This research was conducted using data from the National Institutes of Health All of Us Research Program. Access to the data is available upon application and approval of an institutional data use agreement with the NIH All of Us Research Program.

Declarations

Ethics approval and consent to participate

This study used de-identified data from the National Institutes of Health All of Us Research Program (Controlled Tier – version 7). Access was granted through Meharry’s Data Use and Registration Agreement with All of Us, and analyses were conducted by authorized users in the All of Us Researcher Workbench. All of Us participants provided informed consent at enrollment.

Consent for publication

Not applicable. The approval for dissemination of this work was obtained through the All of Us Research Program’s publication review process.

Competing interests

The authors declare no competing interests.