Postpartum severe maternal morbidity: A longitudinal analysis of the association with economic segregation and neighborhood deprivation in Georgia, USA 2016–2022

Abstract

Severe maternal morbidity (SMM) is a composite measure comprising a spectrum of pregnancy-related complications that have substantial implications for short- and long-term maternal health. Existing research focuses primarily on the occurrence of SMM at delivery, but growing evidence points to risks throughout the post-partum period. While racial, ethnic, economic, and geographic disparities in SMM are well documented, the underlying drivers of inequities, particularly the roles of socio-spatial stratification and access to health services, are not well understood. Using a longitudinally-linked cohort of 637,932 hospital deliveries between 2016 and 2022 we examined two questions: (1) Does the race-specific association between neighborhood deprivation (NDI) or the index of concentration at extremes for income (ICE) and SMM vary by time period of SMM (delivery, 3 months after delivery, or up to 12 months postpartum)? (2) Are these associations modified by policy-relevant health service indicators, including payor, proximity to Federally Qualified Health Centers, and availability of OB-GYN’s? Half of all SMM events occurred in the postpartum year. NDI was not significantly associated with SMM at delivery for Hispanic, Non-Hispanic White or Black individuals, but was positively associated with SMM between 3 and 12 months for all race/ethnic groups. Living in areas of concentrated affluence was protective against postpartum SMM for non-Hispanic Black individuals while concentrated poverty was associated with higher postpartum SMM risk for Hispanic and non-Hispanic Black women. Place-based structural disadvantage is a key determinant of postpartum SMM and disparities in SMM rates.

1. Introduction

Severe maternal morbidity (SMM) is a constellation of life-threatening conditions or complications arising during pregnancy, childbirth, or the postpartum period. These conditions, which range from organ failure and sepsis to severe hemorrhage and eclampsia, often result in extended hospital stays, intensive care, long-term disability, and substantial psychological and emotional tolls on affected individuals and families (Callaghan et al., 2012; Korst et al., 2021). Moreover, SMM events are often considered “near-miss” indicators for maternal mortality, making them critical for identifying vulnerable populations and informing targeted prevention interventions (Firoz et al., 2013). Public health surveillance of SMM is therefore essential not only for clinical quality improvement but also for the development of population-level strategies aimed at reducing maternal morbidity and mortality.

Recent SMM research in the United States and other high-income countries has yielded several important observations. First, racial and ethnic disparities persist across the continuum of maternal health, with Black and Indigenous women disproportionately experiencing both SMM and maternal mortality in the U.S. context (Adams et al., 2024; Creanga et al., 2014; Darville et al., 2025; Hirai et al., 2022a; Leonard et al., 2019; Liese et al., 2019). These disparities are not fully explained by individual-level health characteristics, pointing to the powerful influence of structural racism and systemic inequities as drivers of between-race differences (Gao et al., 2023; Jeffers et al., 2023; Liu et al., 2019). Second, geographic disparities in SMM risk have been documented across rural and urban settings and among U.S. states or sub-national regions, highlighting the influence of structural, environmental and health service factors in shaping maternal outcomes (Admon et al., 2023; Guglielminotti et al., 2021; Hirai et al., 2022a; Kozhimannil et al., 2019; Roese et al., 2024; Tsamantioti et al., 2024). Finally, recent work has underscored the role of contextual influences—including hospital quality, regional health care infrastructure, and social determinants of health (SDOH)—as important contributors to SMM risk (Adams et al., 2024; Geddes-Barton et al., 2025; Guglielminotti et al., 2021; Mari et al., 2023; Neerland et al., 2024).

While a growing body of literature has explored the incidence and drivers of SMM during the delivery hospitalization, with few exceptions (Boghossian et al., 2024a; Boghossian et al., 2024b; Carmichael et al., 2022; Druyan et al., 2023; Kaufman et al., 2023), little attention has been given to severe maternal health complications that arise in the months following childbirth. This gap in knowledge is particularly concerning given increasing evidence that the postpartum period carries risks not only for readmission, but for severe morbidity and mortality that may equal or even exceed those encountered during delivery. For example, more than half of all pregnancy-related deaths in the U.S. occur more than 1 week after delivery and nearly a third occur more than 42 days postpartum (Centers for Disease Control and Prevention, 2024; Gunja et al., 2024). Further, racial and socioeconomic disparities in severe morbidity may actually be wider in the postpartum period as compared to the time of delivery (Boghossian et al., 2024a). Unlike the delivery hospitalization, where individuals are closely monitored and supported by health professionals, access to health care, surveillance, and professional support during the postpartum year in the United States is often lacking or fragmented, particularly relative to other high-income countries with universal coverage. Data limitations and challenges in longitudinal tracking of maternal outcomes have further constrained research on postpartum SMM.

Several factors underscore the importance of extending SMM surveillance into the postpartum year. The postpartum period involves physiologic transitions —such as uterine involution, hormonal and endocrine fluctuations, hemodynamic shifts, sustained hypercoagulability, and immune reconstitution (Chauhan and Tadi, 2025)— that heighten vulnerability to complications such as infection, hemorrhage, cardiomyopathy, and thromboembolism. These biologic risks are oftentimes compounded by barriers to postpartum care and the fact that those suffering postpartum complications in areas with poor access to maternity care often present to emergency or non-obstetric settings where recent pregnancy status may be unknown or not recognized as clinically relevant. Additionally, the postpartum period is characterized by a complex interplay of social and emotional transitions, including the demands of newborn care, sleep deprivation, changes in support systems, and heightened risk of mental and behavioral health conditions all of which may influence an individual’s wellbeing and ability to seek timely care (Gelabert et al., 2024; Modak et al., 2023; Wickramaratne et al., 2022).

Although many high-income countries have universal health coverage across the lifespan, in the U.S. context health care access varies by employment, geography, and pregnancy status, compounding these challenges. Historically, Georgia’s Medicaid coverage for pregnancy-related services ended 60 days after childbirth leaving many individuals without insurance during a time of elevated health risk (Johnson, 2023). Georgia first implemented a 6 month postpartum extension and then, along with virtually all other states, extended coverage to the full 12 months (Kaiser Family Foundation, 2025).

Importantly, the role of social determinants of health in shaping maternal outcomes is increasingly recognized. These determinants—such as neighborhood-level poverty, racial segregation, housing stability, and healthcare access—shape health not only during pregnancy and delivery, but also in the weeks and months after childbirth (Gao et al., 2023; Geddes-Barton et al., 2025; Mari et al., 2023; Neerland et al., 2024; Roese et al., 2024). As proposed in the adaptation of Glass and McAtee’s conceptual model to maternal health (M. R. Kramer et al., 2019), the community context contains and represents risk regulating experiences that may produce distinct health patterns by place, race/ethnicity, and life course stage. During the prenatal period, frequent health care encounters may buffer the effects of some social risks, facilitating early detection and intervention for emerging complications, a pattern typical of high-income health systems during pregnancy. However, in the postpartum period, when routine medical touchpoints decrease, community-level factors and resources (such as social services, peer support networks, home visiting programs, community health clinics) may play a more critical role in addressing unmet needs and preventing maternal morbidity, particularly in the US context where access to postpartum care can be fragmented or absent. Measurement of the neighborhood characteristics related to structural racism and affecting upstream markers of health (diabetes, hypertension) is being refined by researchers (Dyer et al., 2023; Wien et al., 2023) while recent work concluded that ‘embodied stress’ stemming from one’s social environment is a major path driving maternal hypertensive disparities (Keith and Martin, 2024). Consequently, understanding how area-based measures of social stratification are associated with maternal outcomes beyond delivery is essential for identifying leverage points for intervention.

Two indices commonly used to quantify contextual social stratification and socioeconomic conditions are the Index of Concentration at Extremes (ICE) and the Neighborhood Deprivation Index (NDI). ICE is designed to capture spatial social polarization by simultaneously measuring concentrations of affluence and poverty within geographic areas (Jeffers et al., 2023; Krieger et al., 2018; Larrabee Sonderlund et al., 2022; Mari et al., 2023). NDI, in contrast, is a composite measure reflecting multiple dimensions of disadvantage—including income, education, employment, and housing—aggregated at the community level (Adams et al., 2024; M. R. Kramer et al., 2014; Messer et al., 2006; Vos et al., 2014). Both ICE and NDI provide complementary proxy measures of how place-based social and community environments may contribute to adverse maternal health outcomes, offering potential targets for geographically focused health interventions (M. R. Kramer et al., 2019).

In this context, our research aims to contribute to the growing but still limited literature on postpartum severe maternal morbidity by analyzing a large, longitudinal dataset of individuals who delivered in Georgia, USA. Specifically, we seek to (1) describe the prevalence of SMM during delivery and throughout the postpartum year; (2) examine associations between two area-based social determinants—ICE and NDI—and the risk of SMM by timing (i.e., during delivery hospitalization vs. postpartum); and (3) explore whether these associations are modified by policy-relevant health system factors. These include access to Federally Qualified Health Centers (FQHCs), insurance payer type at delivery, and area-level concentration of OB-GYNs, which may influence both exposure to risk and the capacity for timely response to emergent maternal health conditions. Given the shift in health service surveillance from the prenatal to postpartum periods, a feature that varies across health systems, we hypothesize that community based social drivers of health proxied by NDI and ICE may have added influence on severe maternal morbidity in the postpartum year that may be buffered by the availability of supportive primary care services.

2. Methods

2.1. Population and data source

We created a retrospective, longitudinal cohort of Georgia, USA-resident mothers of all live birth and fetal death deliveries occurring between 2016 and 2022 as identified in deterministically-linked hospital discharge and vital statistics datasets. Specifically, ICD-10 codes consistent with delivery were identified in the Georgia Hospital Discharge Abstracts and linked via a unique person-level ID to associated vital statistics including birth and fetal death records. Ectopic pregnancies, molar pregnancies, and induced pregnancy terminations were excluded. Follow-up time was observed by identifying and linking all subsequent admissions to any Georgia healthcare facility from the date of delivery through 1-year post-partum for each person. To insure equivalent 12-month follow-up time for all individuals, only deliveries occurring from 2016 to 2021 were retained, but post-partum events through 2022 were included. The longitudinal dataset was arranged to be person-centric, including all longitudinal information about individuals as the unit of analysis, excluding records with missing address information.

2.2. Measures

The outcome of interest – severe maternal morbidity (SMM) – was identified in delivery and postpartum admissions based on ICD-10 and procedure codes using 20 diagnoses defined by the Alliance for Innovation of Maternal Health (AIM) (v05-14-2024) and recommended by the CDC (Alliance for Innovation on Maternal Health, 2024); we are not including blood transfusion as an SMM-defining event because of its high prevalence and very poor specificity (Friedman et al., 2022; Hirai et al., 2022b; Snowden et al., 2021). We categorized postpartum SMM into events following the delivery hospitalization through 3 months and 3–12 months after delivery to distinguish acute postpartum complications from later morbidities more likely influenced by care transitions, insurance loss, and social and neighborhood context. The 3-month cut-point approximates the end of routine obstetric surveillance. Women with SMM-defining events at 2 or more time periods were identified.

The Index of Concentration at Extremes (ICE) and the Neighborhood Deprivation Index (NDI) are frequently used area-based measures of social and economic context that tap into complementary aspects of the historical and experienced environment. ICE focuses on neighborhood spatial social polarization by capturing the relative inequality of those with very high versus very low income and providing insight into social class exclusion or isolation, while saying little about middle-income populations (Krieger et al., 2018). NDI, on the other hand, is a theoretically grounded, validated contextual indicator of area-based social and economic capital in a designated area originally developed for understanding community drivers of maternal and child health outcomes (Messer et al., 2006). It focuses on cumulative material deprivation and resource scarcity across the socioeconomic spectrum and thus offers a broader view of levels of material deprivation, but little insight into affluence as distinct from absence of deprivation. We consider each to provide complementary insight for understanding social environments during and after pregnancy.

ICE measures small area segregation or polarization based on household income levels. Using data from the American Community Survey we calculated for each Georgia census tract, in each year, the ICE value using the standard formula: $\frac{A_i-P_i}{T_i}$ where A_i is the count of households in a given census tract with an income greater than $125,000/year (approximately 80th percentile at start of study follow-up); P_i is the count of households with an income less than $20,000 (approximately the 20th percentile); and T_i is the total household count in census tract (i). The resulting scale ranges from −1 to +1, with negative values suggesting concentration of poverty, and positive values indicating concentration of affluence or wealth. In keeping with the notion that ICE identifies areas of “concentration at extremes” we categorized the continuous value into three groups: concentrated poverty (ICE < −0.2); concentrated affluence (ICE > +0.2); and a referent category of economically diverse or heterogeneous areas. These cut points were selected to balance the indices focus on polarized extremes (departures from zero) while maintaining a meaningful sample size in each direction. A value of ICE was assigned to each pregnant person based on their geocoded census tract of residence at delivery.

NDI was calculated based on principal component analysis loadings of 8 variables (% families below federal poverty level; % female headed households; % families with income <$30,000; % families receiving public assistance; % households with overcrowding; % unemployment; % adults without a high school degree; % adults employed in management or professional positions) from ACS measures for each census tract and year. The raw NDI was standardized so that a value of zero is the average level of neighborhood deprivation in Georgia, and each unit is a one-standard deviation change from this average; higher values represent greater “deprivation”.

Large and persistent racial and ethnic disparities in maternal health outcomes are well-described, and are not a result of biologically essential traits, but instead, of socially produced stratification and inequity producing differences in exposure prevalence, lived experience, and resource allocation (M. R. Kramer and Hogue, 2009). To recognize and describe both within- and between-race/ethnic patterns in SMM, we therefore stratify all results on maternal race/ethnicity, focusing here on the three largest groups in Georgia: non-Hispanic Black, non-Hispanic White, and Hispanic women. In addition, we abstracted other covariates from vital statistics and hospital discharge records at the time of delivery including maternal age, maternal education, and medical risk factors including hypertension and diabetes. In addition, we categorized each individual’s neighborhood using the National Center for Health Statistics rural/urban classification based on county of residence (Ingram and Franco, 2012).

Candidate effect measure modifiers included individual-level payor type at delivery (private, Medicaid, self-pay), county-level rural/urban status, number of Federally Qualified Health Clinics (FQHC’s), and number of obstetrician-gynecologists (OB-GYN’s) per 10,000 women of reproductive age obtained from the Health Resources and Service Administration Area Health Resource Files (Health Resources and Services Administration, 2025).

2.3. Analysis

To answer our first research question, covariates were aggregated by timing of SMM event (e.g. during delivery, after delivery hospitalization but before 3 months, or 3 months–12 months postpartum) including SMM event counts and rates per 10,000 deliveries. To describe geographic variation in delivery and postpartum SMM rates, we aggregated events and births by race/ethnicity and by county of residence at the time of delivery. To stabilize rate estimates in the context of sparse counts for mapping geographic variation in SMM, we applied non-spatial empirical Bayesian smoothing to rates for each county, suppressing counties with fewer than 50 live births over the study period.

To address our second research question, we estimated race/ethnicity stratified random intercept logistic regression models with SMM for a given time period as the dependent variable, either ICE or NDI as the exposure, and observations clustered at the county level. Our interest is in absolute risk differences, which were calculated from logistic models via marginal standardization, and 95 % confidence intervals were estimated via bootstrapping. While many individual covariates are plausible predictors of the outcome, SMM, it is unlikely that any individual covariates are causes of neighborhood NDI or ICE. However, it is plausible that unmeasured upstream variables (e.g. statewide historical patterns of development, resource allocation and social stratification) induce correlations between aggregate individual covariates such as age or education, making them act as confounders. There are also area-based covariates that may confound the independent effect of ICE or NDI on SMM including degree of rurality, availability of FQHC’s and OB-GYN’s. We estimated models adjusting for key individual and area level covariates to evaluate these patterns. Our inference target is on the fixed effect estimates for ICE and NDI conditional on covariates and random effects. Detailed model specifications are summarized in the Supplement.

Finally, to produce exploratory results for our third question of whether health-service factors modify the association between socioeconomic context and SMM, we introduced two-way interactions into adjusted race-stratified models between each exposure (NDI or ICE) and each of the effect modifiers including payor at delivery, rural/urban status, FQHC’s, and OB-GYN’s per capita. Meaningful departure from additivity was assumed to suggest different magnitude of association of NDI or ICE as a function of modifiable health resource context. Tests of significance were for alpha of 0.05.

The project was reviewed and approved by Emory University Institutional Review Board. All analysis was conducted in R 4.4 (R Core Team, 2024).

3. Results

After exclusion of 17,223 (2.6 % of total) deliveries with missing address information, there were 637,932 deliveries identified between 2016 and 2021, with one year of postpartum follow-up. In this cohort, the rate of SMM at delivery was 80.1/10,000 (N = 5161 events) (Table 1). SMM was more common among deliveries to non-Hispanic Black women and those of lower education level, with Medicaid, and residing in rural areas. Although the delivery hospitalization was the time period with the highest SMM rate, the majority of SMM events happened across the two postpartum time periods with 3031 SMM between the delivery hospitalization discharge and 3 months postpartum and 2227 between 3 and 12 months postpartum (overall postpartum SMM rate of 83.8/10,000) (Table 1; Fig. 1).

Table 1.

Distribution of deliveries and severe maternal morbidity (SMM) events by time period and covariates, Georgia, 2016–2021.

Characteristic	All Deliveries N = 637,932a	No SMM N = 627,084a	Delivery Only SMM N = 5,161a	3 mo PP SMM N = 3,031a	3mo to 1 yr PP SMM N = 2,227a	Delivery & PP SMM N = 429a
Maternal age	28.0 (24.0, 33.0)	28.0 (24.0, 33.0)	30.0 (25.0, 35.0)	29.0 (24.0, 34.0)	28.0 (24.0, 33.0)	30.0 (24.0, 34.0)
Maternal race/ethnicity
NH White	286,469 (48 %)	282,735 (48 %)	1833 (38 %)	986 (34 %)	826 (39 %)	89 (21 %)
NH Black	230,545 (39 %)	225,003 (38 %)	2477 (51 %)	1655 (58 %)	1106 (52 %)	304 (73 %)
Hispanic	78,287 (13 %)	77,291 (13 %)	528 (11 %)	237 (8.2 %)	210 (9.8 %)	21 (5.1 %)
Missing (N)	42,631	42,055	323	153	85	15
Maternal education
Less than 9th grade	17,284 (2.7 %)	17,028 (2.7 %)	133 (2.7 %)	60 (2.0 %)	61 (2.8 %)	2 (0.5 %)
9th – 11th grade	59,001 (9.3 %)	57,788 (9.3 %)	487 (9.9 %)	349 (12 %)	333 (15 %)	44 (11 %)
HS diploma/GED	200,123 (32 %)	196,244 (32 %)	1734 (35 %)	1105 (37 %)	856 (39 %)	184 (45 %)
Some college or higher	356,641 (56 %)	351,435 (56 %)	2585 (52 %)	1493 (50 %)	949 (43 %)	179 (44 %)
Missing (N)	4883	4589	222	24	28	20
Gestational diabetes	34,409 (5.4 %)	33,586 (5.4 %)	371 (7.2 %)	244 (8.1 %)	174 (7.8 %)	34 (7.9 %)
Hypertensive disorders of pregnancy	63,396 (9.9 %)	60,697 (9.7 %)	1496 (29 %)	594 (20 %)	437 (20 %)	172 (40 %)
Rural/Urban status
Large central metro	61,249 (9.6 %)	60,068 (9.6 %)	616 (12 %)	308 (10 %)	213 (9.6 %)	44 (10 %)
Large fringe metro	304,564 (48 %)	299,509 (48 %)	2523 (49 %)	1361 (45 %)	966 (43 %)	205 (48 %)
Medium metro	68,106 (11 %)	66,963 (11 %)	516 (10.0 %)	346 (11 %)	229 (10 %)	52 (12 %)
Small metro	97,486 (15 %)	95,854 (15 %)	671 (13 %)	492 (16 %)	404 (18 %)	65 (15 %)
Micropolitan	63,478 (10.0 %)	62,407 (10.0 %)	474 (9.2 %)	328 (11 %)	236 (11 %)	33 (7.7 %)
Noncore	43,049 (6.7 %)	42,283 (6.7 %)	361 (7.0 %)	196 (6.5 %)	179 (8.0 %)	30 (7.0 %)
Insurance payor
Private	280,302 (46 %)	276,552 (46 %)	2030 (41 %)	1007 (35 %)	595 (28 %)	118 (29 %)
Medicaid	303,154 (49 %)	297,018 (49 %)	2588 (53 %)	1796 (62 %)	1469 (68 %)	283 (69 %)
Self pay	31,650 (5.1 %)	31,143 (5.2 %)	297 (6.0 %)	109 (3.7 %)	92 (4.3 %)	9 (2.2 %)
Missing (N)	22,826	22,371	246	119	71	19
FQHCs per 100,000	1.13 (0.44, 2.27)	1.12 (0.44, 2.27)	1.20 (0.44, 2.44)	1.18 (0.34, 2.30)	1.17 (0.34, 2.44)	1.17 (0.45, 2.36)
Neighborhood Deprivation Index (NDI)	−0.19 (−0.78, 0.51)	−0.19 (−0.78, 0.50)	−0.09 (−0.72, 0.65)	0.03 (−0.61, 0.77)	0.07 (−0.56, 0.79)	0.20 (−0.56, 0.91)
Missing (N)	242	235	5	0	1	1
Index of Concentration at Extremes for Income
1-ref	400,008 (63 %)	393,391 (63 %)	3134 (61 %)	1850 (61 %)	1384 (62 %)	249 (58 %)
2-poverty-concentration	125,549 (20 %)	122,854 (20 %)	1146 (22 %)	811 (27 %)	610 (27 %)	128 (30 %)
3-wealth-concentratin	112,299 (18 %)	110,768 (18 %)	878 (17 %)	370 (12 %)	232 (10 %)	51 (12 %)
Missing (N)	76	71	3	0	1	1

SMM: Severe maternal morbidity.

PP: Post-partum.

FQHC: Federally Qualified Health Center.

^aMedian (Q1, Q3); n (%).

Fig. 1.

Distribution of severe maternal morbidity by timing, Georgia 2016–2022.

The SMM rate was relatively evenly split between delivery and postpartum for non-Hispanic Black and White women, but for Hispanic women, SMM was more common at delivery than during the postpartum period. SMM rates increased between 2016 and 2022 for non-Hispanic Black women, with more modest recent increases for non-Hispanic White and Hispanic women (Fig. 2). Geographically, for non-Hispanic White and Black mothers, delivery SMM was more concentrated in metropolitan counties whereas postpartum SMM was more concentrated in non-metro counties (Fig. 3). In contrast, the higher rate of delivery as compared to postpartum SMM among Hispanic women was evident across counties, but the metro-nonmetro pattern was less clear because of data sparsity in many rural counties. The leading diagnoses that comprise SMM varied by timing and race/ethnicity. At delivery, disseminated intravascular coagulation (DIC) was the leading diagnosis for all race/ethnic groups, but in the postpartum, sepsis, acute respiratory distress syndrome and acute renal failure were leading diagnoses (Table 2).

Fig. 2.

Race/ethnic specific SMM rates by year of delivery, Georgia, 2016–2022. Note: Data summarized by year of delivery. For this reason, “Delivery SMM” goes through 2022, but “Post-partum SMM” is only reported through 2021 to allow for full year of post-partum follow-up.

Fig. 3.

Within-race geographic distribution of delivery and postpartum severe maternal morbidity, Georgia, 2016–2022.

Table 2.

Top 5 Diagnoses that Comprise Severe Maternal Morbidity (SMM) by race/ethnicity and timing of event.

Pregnancies at risk:	Total Births	Hispanic	NH Black	NH White
	637,932	78,287	230,545	286,469
	SMM rate per 10,000 pregnancies
At delivery
Disseminated intravascular coagulation	23.0	17.9	29.1	18.4
Acute renal failure	15.0	9.5	24.4	9.1
Acute respiratory distress syndrome	14.8	14.7	18.8	11.5
Sepsis	13.1	15.2	16.8	9.3
Hysterectomy	12.4	11.1	13.4	11.7
up to 3 months post-partum
Sepsis	20.4	17.5	25.7	17.3
Pulmonary edema/Heart failure	11.4	3.4	21.1	6.5
Acute respiratory distress syndrome	9.3	4.6	16.1	5.7
Acute renal failure	7.4	4.2	11.9	5.0
Air & thrombotic embolism	5.6	3.4	8.3	4.4
3 months to 1 year post-partum
Sepsis	18.0	16.4	23.3	15.3
Acute respiratory distress syndrome	9.3	7.7	13.0	7.2
Acute renal failure	8.7	4.2	14.0	6.1
Pulmonary edema/Heart failure	4.3	1.0	9.0	1.9
Ventilation	4.0	1.5	5.9	3.1

There were no meaningful differences in estimates of association between either exposure and SMM for unadjusted versus adjusted models (adjusted results presented). Increasing Neighborhood Deprivation Index (NDI) had varying association with SMM depending on maternal race/ethnicity and timing in relation to delivery (Table 3a; Fig. 4a). There were no significant associations between NDI and SMM at delivery for any race/ethnic group. For SMM between delivery and 3 months, each 1-SD increase in NDI was associated with an additional 3.0 (95 % CI −0.3–6.7) and 5.0 (95 % CI 1.3–9.4) SMM events per 10,000 deliveries for non-Hispanic White and Black women respectively; there was no association for Hispanic women. NDI was associated with larger increases in SMM between 3 and 12 months postpartum for all three race/ethnic groups, with risk differences per 1-SD change in NDI ranging of 5.2 per 10,000 (95 % CI 0.8–9.8) for non-Hispanic White women, 11.2 per 10,000 (95 % CI 6.4–15.9) for non-Hispanic Black women, and 10.9 (95 % CI 2.6–19.3) for Hispanic women. To contextualize these risk differences, the overall average 3-month to 1-year postpartum SMM rate for non-Hispanic White and Black women was 28.8 and 50 per 10,000 (Table 1). Thus, there is approximately a 20 % increase in SMM occurrence for each 1-SD increase in NDI.

Table 3a.

Race-stratified, adjusted results for NDI and SMM.

	Delivery			3 months			1 year
	RD	95 % CI		RD	95 % CI		RD	95 % CI
NDI
NH White	−2.3	−6.6	2.0	3.0	−0.3	6.7	5.2	0.8	9.8
NH Black	0.9	−3.5	5.4	5.0	1.3	9.4	11.2	6.4	15.9
Hispanic	−7.3	−14.5	0.3	0.9	−4.1	6.7	10.9	2.6	19.3

Fig. 4a.

Adjusted association of NDI with SMM by race/ethnicity and timing of SMM. Note: all estimates are rate differences adjusted for maternal age, hypertensive disorder of pregnancy, gestational diabetes, payor, maternal education, birth year, county-level FQHC’s per capita, rural/urban, OB-GYN’s per capita. All post-partum estimates are also adjusted for presence of SMM event at delivery. Deliveries 2016–2021, with postpartum follow-up through 2022.

In contrast to the monotonic directionality implied by the NDI scale, the ICE illustrates two opposite poles of concentration: the spatial concentration of poverty and the spatial concentration of affluence, with economically mixed places representing a referent category. The patterns of association are varied by race/ethnicity and timing. For non-Hispanic White women, there is no significant association between concentrated poverty or affluence and SMM at any time period (Table 3b; Fig. 4b). For non-Hispanic Black women living in concentrated poverty, there is increased SMM as compared to those in referent neighborhoods, and there is lower SMM for those living in concentrated affluence (ICE >0.2); these differences are significant and stronger at 1-year postpartum than at 3-months postpartum. Hispanic women had significantly higher SMM between 3- and 12-months postpartum if they lived in concentrated poverty, but otherwise there were no significant associations.

Table 3b.

Race-stratified, adjusted results for ICE and SMM.

	Delivery			3 months			1 year
	RD	95 % CI		RD	95 % CI		RD	95 % CI
NH White
ICE: poverty concentration	−3.8	−12.8	5.4	5.2	−2.4	12.6	7.3	−3.6	16.8
ICE: affluence concentration	6.0	−2.8	15.0	1.6	−4.6	7.7	−1.7	−10.2	6.7
NH Black
ICE: poverty concentration	4.6	−6.2	15.0	5.6	−3.4	14.4	12.1	0.4	23.8
ICE: affluence concentration	9.2	−8.9	24.8	−13.5	−24.5	0.0	−28.1	−44.3	−11.3
Hispanic
ICE: poverty concentration	2.2	−15.9	21.3	4.1	−8.3	18.0	29.9	12.4	51.2
ICE: affluence concentration	17.6	−0.7	37.5	4.5	−7.0	18.1	−6.3	−21.1	9.0

All estimates are risk differences per 10,000 pregnancies adjusted for maternal age, hypertensive disorder of pregnancy, gestational diabetes, payor, maternal education, rural/urban, FQHC’s per capita, OB-GYN’s per capita and birth year. All post-partum estimates are also adjusted for presence of SMM event at delivery. Sample restricted to deliveries 2016–2021, with postpartum follow-up through 2022.

Fig. 4b.

Adjusted association of ICE with SMM by race/ethnicity and timing of SMM. Note: all estimates are rate differences adjusted for maternal age, hypertensive disorder of pregnancy, gestational diabetes, payor, maternal education, birth year, county-level FQHC’s per capita, rural/urban, OB-GYN’s per capita. All post-partum estimates are also adjusted for presence of SMM event at delivery. Deliveries 2016–2021, with postpartum follow-up through 2022.

In a final set of models, we tested race-stratified models for significant and meaningful additive interaction between each key exposure (NDI and ICE) with health service context measures including individual payor, county-level high vs low availability of FQHCs to approximate safety access to primary care services, metropolitan vs non-metropolitan residence, and OB-GYN’s per capita in each county. Given a rare outcome and sparse data, we excluded Hispanic strata, and combined 3-month and 1-year postpartum SMM into a single composite indicator. The result was 16 interaction models tested (2 exposures, 4 modifiers, 2 race/ethnic groups). Of those, only two had statistically significant interactions. For non-Hispanic White women, the association of both NDI and ICE were modified by the number of FQHC’s in the resident county such that more FQHC’s modestly attenuated the ill effects of NDI or ICE in the postpartum period (Supplemental Tables 1 and 2). Given the large number of interactions tested, and modest evidence for departure from additivity, these results should be interpreted as exploratory.

4. Discussion

This study provides novel insights into the burden and determinants of severe maternal morbidity (SMM) throughout the perinatal continuum, with a specific emphasis on the postpartum period. In a longitudinal cohort of over 600,000 deliveries in Georgia, USA between 2016 and 2021, we found that nearly half of all SMM events occurred after discharge from the delivery hospitalization, underscoring the importance of extending maternal health surveillance and intervention strategies well beyond childbirth. Moreover, our findings suggest that neighborhood-level social disadvantage—as measured by the Neighborhood Deprivation Index (NDI) and the Index of Concentration at Extremes (ICE)—plays a more pronounced role in shaping postpartum SMM risk compared to delivery-related SMM, with varying impacts across racial and ethnic groups.

Three major findings emerge from our analysis. First, the postpartum period represents a critical window for maternal health risk, with SMM rates slightly exceeding those observed during delivery hospitalizations. While delivery SMM rates were highest among non-Hispanic Black women, all racial/ethnic groups experienced substantial postpartum morbidity, including serious complications such as sepsis, acute renal failure, and cardiopulmonary conditions.

Second, neighborhood deprivation (NDI) was not significantly associated with SMM at delivery for any racial/ethnic group but was associated with significantly elevated SMM risk in the postpartum period—especially between 3 and 12 months postpartum—for all groups studied. For Hispanic and non-Hispanic Black women, a one standard deviation increase in NDI was associated with more than 10 or more excess SMM cases per 10,000 deliveries during late postpartum, nearly double the effect observed for non-Hispanic White women and Hispanic women.

Third, ICE revealed that concentrated poverty was associated with increased postpartum SMM, while concentrated affluence appeared protective for non-Hispanic Black women. In exploratory analyses, these relationships were not meaningfully modified by health system factors such as payor type, rurality, or availability of OB-GYNs. However, greater availability of Federally Qualified Health Centers (FQHCs) modestly attenuates the harmful effects of deprivation and concentrated poverty in some models among non-Hispanic White women.

While prior studies have established racial and geographic disparities in SMM during delivery (Creanga et al., 2014; Hirai et al., 2022a), relatively few have extended these analyses into the postpartum period. Our findings align with emerging research suggesting that postpartum morbidity may be equal to or exceed delivery-related complications, particularly among marginalized populations (Boghossian et al., 2024b; Druyan et al., 2023). That the strongest effects of NDI and ICE on SMM emerged in the postpartum period speaks to the relatively greater sensitivity of environmental influences on maternal health as obstetric surveillance wanes, a finding potentially relevant beyond Georgia or the US. The strength of the neighborhood effects specifically among non-Hispanic Black and Hispanic mothers highlights the consequences of structural racism in organizing U.S. spatial and social environments. Racialized housing and credit markets, exclusionary zoning, and uneven distribution of public goods have concentrated deprivation and limited access to supportive postpartum infrastructures in communities of color (Desmond and Wilmers, 2019; Hailu et al., 2022). While prenatal and delivery care provide scheduled surveillance and clinical safeguards, postpartum individuals—particularly those losing Medicaid coverage—must navigate recovery within these structurally constrained environments (Crear-Perry et al., 2021). These findings point to the need for structural and policy interventions that address racialized neighborhood disadvantage, not solely extension of clinical care.

The observed associations between NDI and postpartum SMM mirror those found in studies of infant mortality and adverse perinatal outcomes (M. R. Kramer et al., 2014; O’Campo et al., 2008), extending the influence of neighborhood deprivation beyond pregnancy into the extended postpartum year. Similarly, our findings on ICE support previous literature demonstrating the health consequences of spatial social polarization (Jeffers et al., 2023; Krieger et al., 2018), particularly the exacerbation of racial health disparities in contexts of concentrated poverty.

Of note, we found limited evidence that commonly proposed health system mitigators (e.g., insurance type, FQHC access, OB-GYN density) meaningfully modified the effects of neighborhood deprivation or segregation on SMM risk (although in some cases they had main effects on SMM and the racial gap (Adams et al., 2024, 2025). Given the large number of interaction models tested, and relatively modest magnitude of effects, these findings are exploratory but may suggests that while health care access is essential, it may be insufficient to buffer against the structural disadvantages embedded in neighborhoods.

These findings have substantial implications for maternal health surveillance and intervention design. First, they affirm the critical importance of extending maternal care and monitoring well beyond the immediate postpartum window. More than 5200 SMM events in this cohort occurred after delivery through 1 year postpartum, time periods traditionally not covered by standard maternal care protocols. Policy-makers and healthcare systems must therefore align maternal health benefits and service delivery models to reflect this extended risk period (Smith et al., 2025; Steenland and Wherry, 2023).

Second, the role of community-level deprivation in shaping maternal outcomes highlights the need for population health strategies that extend beyond clinical settings. Investments in housing, transportation, food security, and community health infrastructure—particularly in areas of concentrated poverty—may offer meaningful returns in maternal morbidity reduction. Place-based interventions that target structural determinants of health are likely to yield greater gains than those focusing solely on individual-level behavior change or health service use.

Third, although safety-net healthcare access through FQHCs may offer some protection, particularly among certain subgroups, our findings suggest that broader structural interventions will be needed to truly shift risk trajectories. NDI and ICE capture small area material and social processes and context that may be indicative of availability or accessibility of quality affordable childcare, flexible work opportunities or other supports in the postpartum period that facilitate health and wellness. Importantly, our findings also lend support to Medicaid expansions and postpartum coverage extensions, which could help bridge the care gap during this high-risk postpartum period. It will be important to assess the impact of this extended coverage and variation in effects by race/ethnicity or neighborhood. Analysis of Medicaid-covered births indicated that more than three-quarters of enrollees with full-year Medicaid coverage had at least one outpatient visit from 60 days to 12 months postpartum and utilization was higher among enrollees with diagnosed physical or mental/behavioral health conditions or pregnancy/delivery complications (Smith et al., 2025). However, utilization was more common for non-Hispanic white than non-Hispanic Black and Hispanic enrollees and for rural than urban enrollees. The emerging federal changes in Medicaid coverage could imperil potential gains from these extensions and potentially worsen existing disparities.

This study is strengthened by the use of a large, population-based, longitudinally linked cohort with one year of follow-up, allowing for the capture of delayed-onset SMM events and robust stratification by race/ethnicity. The use of validated neighborhood-level indices (NDI and ICE) enables an in-depth examination of contextual social determinants, while multilevel modeling methods account for clustering and allow for estimation of tract-level effects.

Nonetheless, several limitations must be acknowledged. First, SMM by definition is just the tip of the iceberg of maternal health, and therefore the full impact of structural factors on postpartum maternal health is understated. Second, our reliance on hospital discharge data may underestimate or misclassify certain SMM events. Although we excluded blood transfusion—a frequently debated indicator implicated in exaggerating truly severe morbidity prevalence—we may still be capturing events with variable clinical significance. Third, our longitudinal approach was achieved through deterministic linkage of administrative datasets, and any false negative or positive links would result in misclassification of measures. In unpublished validation of this linkage strategy in Georgia data, record linkage specificity was high but sensitivity was low suggesting most links are “real”, but some hospital discharges for a given individual may not be appropriately linked (M. Kramer and Carter, 2005). This poor sensitivity was more pronounced for Hispanic women. The rate of SMM at delivery is similar for Hispanic women who did and did not link, but the NDI was higher (more deprived) for Hispanic women with successful longitudinal linkage; this likely results in bias towards the null. Third, we rely on maternal residential address at delivery to assign exposure. While missing address information was rare (only 2.6 %), it is likely women move over the postpartum year. Most moves over short periods are not likely to result in dramatically different levels of NDI or ICE (M. Kramer, 2016; M. R. Kramer et al., 2014), but some misclassification is possible. Fourth, while NDI and ICE are well-validated proxies for neighborhood deprivation and segregation, they do not capture all relevant features of the social environment, such as exposure to violence, social capital, or local health systems. Fifth, we were limited in our ability to fully assess effect modification due to statistical power constraints and limited variation in some health system indicators across rural areas. Sixth, we included fixed effects for year in an effort to adjust for changes during the Covid-19 pandemic, but there may still be residual variation during this period. Finally, causality cannot be inferred from these observational data, and unmeasured confounding at both individual and area levels remains a possibility.

Future studies should explore mechanisms through which neighborhood deprivation and spatial social polarization contribute to elevated postpartum SMM risk. Potential pathways include reduced access to follow-up care, differential quality of care, psychosocial stress, or environmental exposures. There is evidence that embodied stress related to place-based social environments is associated with pregnancy-associated hypertensive disorders (Keith and Martin, 2024), and further examining this and other pathways may highlight place or person-based strategies to reduce risk. Mixed-methods research that integrates quantitative modeling with qualitative exploration of lived experience may illuminate context-specific barriers and facilitators to postpartum health.

Additionally, evaluation of recent policy changes—such as Georgia’s expansion of Medicaid postpartum coverage to 12 months—will be essential to assess whether expanded insurance access mitigates disparities in postpartum morbidity. Further research should also explore whether community-based interventions, such as home visiting programs or neighborhood-level investments, as well as community-clinical collaborations can buffer the impact of social deprivation on maternal health outcomes and improve care during the postpartum period. Community-clinical collaborations are those that leverage existing community trust, address structural barriers, and integrate clinical and social care. Those that have shown promise in promoting maternal health and reducing adverse maternal health outcomes include nurse-based home visiting, embedding community health workers and doulas within FQHCs, integrating behavioral health supports within primary care (including through telemedicine), and mobile health apps that can facilitate screening, remote monitoring, and connection to providers (Filene et al., 2013; Olds et al., 2014).

Finally, to promote health equity, it is crucial that future research designs incorporate an explicit structural lens, interrogating the role of racism, policy, and disinvestment in shaping both place and health. Disaggregated analyses that center the voices and experiences of marginalized populations, particularly Black and Hispanic communities, will be essential to designing responsive and effective interventions.

5. Conclusion

This study highlights the postpartum period as a critical yet under-recognized window for maternal health risk and one that is strongly shaped by neighborhood social conditions. Our findings underscore the necessity of integrating social determinants of health into both clinical practice and public health strategy, and of advancing structural interventions to address the root causes of maternal morbidity. Only through sustained, cross-sector efforts can we reduce preventable suffering and advance equity in maternal health outcomes.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.socscimed.2026.118967.

Data availability

The authors do not have permission to share data.