Geospatial analysis of preterm and small-for-gestational age births in Washington D.C.

Authors: Alina Peluso, Kate Michel, Kristin L. Atkins, Shannon Gopaul-Balser, Tara Gibbons, Rachel K. Scott

Abstract

Background

This study is based on the recognition that adverse pregnancy outcomes significantly affect maternal and infant health, leading to increased morbidity and mortality. These outcomes are shaped by a complex interplay of individual-level factors—like maternal age and education—and community-level influences, including socio-economic status and access to healthcare. Understanding these determinants is crucial for developing effective public health strategies, especially for marginalized populations, by identifying high-risk areas and informing targeted interventions that address both individual and structural barriers.

Methods

We utilized geospatial analysis to explore the association between individual- and community-level factors and adverse pregnancy outcomes, specifically preterm birth (PTB) and small-for-gestational-age (SGA) birthweight in Washington, D.C. We used Empirical Bayes smoothing methods to calculate rates of adverse birth outcomes from 2010 to 2018 at the U.S. Census tract–level. Spatial scan statistics were used to investigate if adverse birth outcomes clustered in specific areas. ANOVA tests were conducted for individual- and community-level factors within identified clusters.

Results

Spatial analysis identified significant high-risk clusters for PTB and SGA infants primarily in southeastern Washington, D.C., particularly in Wards 7 and 8. Individuals residing within these clusters experienced a 47% increased risk of PTB (RR = 1.467) and a 56% increased risk of SGA (RR = 1.560) compared to those outside clusters. Space–time analysis revealed temporal variation, with PTB clusters persisting from 2011 to 2014 and SGA clusters extending through 2017. Compared to low-risk clusters, high-risk clusters had younger birthing individuals (mean age ~26.5 vs. ~33 years), lower maternal college degree attainment (~20% vs. ~80%), higher rates of late or no prenatal care (~16% vs. 11%), and increased prevalence of smoking and hypertension (all P < 0.001). Community-level indicators showed lower median household incomes ($40,000 vs. ~$105,000), greater poverty (~16% vs. ~7% below $10,000/year), higher public assistance use (~32% vs. ~5%), and reduced healthcare access (greater distances to emergency and specialty care) in high-risk areas (all P < 0.001). Neighborhood deprivation indices were significantly elevated, commutes were longer, and population density was lower in these clusters. These findings highlight that adverse birth outcomes cluster in neighborhoods with pronounced socioeconomic and health disparities.

Conclusion

High-risk birth clusters highlight intertwined individual, socio-economic, and geographic. Addressing these requires comprehensive interventions focusing on social and structural determinants of health.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-025-08039-4.

Key Points

A. Why was this study conducted?

This study was conducted to examine how individual-level and neighborhood-level characteristics relate to adverse pregnancy outcomes in Washington, D.C., with the goal of identifying opportunities to improve care for those at highest risk.

B. What are the key findings?

The key findings show that both medical and non-medical factors contribute to increased pregnancy risks, particularly in the southern and eastern parts of Washington, D.C. The analysis points to geographic patterns in risk that align with differences in age, health status, and living conditions.

C. What does this study add to what is already known?

This study builds on previous research by demonstrating how local data and mapping techniques can be used to identify areas with higher rates of adverse pregnancy outcomes. These findings support more effective planning and delivery of healthcare services where they are needed most.

Background

Preterm birth (PTB) and small-for-gestational-age (SGA) births are persistent and pervasive public health crises in the United States, disparately associated with race and socioeconomic status [1]​. Social and structural determinants of health (SSDOH) have been linked to adverse maternal and neonatal outcomes [2–5] ​. Geospatial analysis can identify geographic health disparities, such as in adverse maternal and neonatal health outcomes, as well as inequities in access to health care services to highlight priority areas for public health intervention; these analyses are an underutilized and underexplored tool in public health, and specifically in maternal and infant health research [6] ​.

To explore utilization of geospatial analyses to better understand adverse pregnancy outcomes in the context of both associated individual- and community-level SSDOH, we consider Washington, D.C. (D.C.) as a case study. D.C. has a population of approximately 690,000 residents and is divided into eight wards, with the Anacostia River acting as both a natural and socio-economic barrier, associated with both racial and economic segregation. A visual representation of D.C.’s ward structure, including key landmarks and geographic ward boundaries, is provided in Supplementary Fig. 1. Wards 1–6 lie west of the river, and Wards 7 and 8 are situated to the east. Wards 2 and 3 are primarily populated by White residents, and Wards 7 and 8 are predominantly Black/African American [7] ​​. Historically, infrastructure and commercial investment have been disproportionately directed towards the western side of the river, leaving Wards 7 and 8 under-resourced and underserved. Residents in these areas face numerous SSDOH disparities (e.g., discrimination, violence, limited access to healthcare, poverty) linked to poor health outcomes [8]. To date, the specific SSDOH linked to adverse birth outcomes in racial and ethnic minority neighborhoods—particularly in Wards 7 and 8—remain understudied.

In this study, we analyzed a cohort of birth records from D.C. to examine the spatio-temporal patterns of PTB and SGA birthweight, alongside individual and community-level SSDOH factors. Our primary goal was to understand the relationships between these factors and to explore geospatial analysis as a tool for providing community-level insights into influences on adverse health outcomes. Analyzing PTB and SGA together is crucial because these outcomes are often interconnected, with preterm infants at a higher risk of being SGA. This combined analysis clarifies their relationship and identifies shared risk factors, such as SSDOH and maternal behaviors, contributing to a more comprehensive understanding of maternal and neonatal health. Such insights can inform targeted interventions to improve health outcomes and guide policymakers in developing effective strategies to reduce disparities and promote health equity. Furthermore, the demographics of D.C. provide valuable insights that may be generalizable to many diverse U.S. areas.

Methods

Data

We gathered individual-level birth data from 2010 to 2018 through the D.C. Health Vital Records Division (DCVRD). Residential addresses at the time of birth were geocoded using the D.C. Master Address Repository (DC MAR), which links standardized addresses to geographic coordinates and United States (U.S.) Census tracts. We ensured a match score of > 95% between the birth record address and the reference DC MAR address to assign births to specific U.S. Census tracts. This enabled to link individual birth records with specific U.S. Census tract community-level data, facilitating a comprehensive ecological analysis of contextual influences on health outcomes.

Exclusion criteria included non-D.C. residents, infants delivered before 20 weeks, multiple gestations, and records without a geocoded home address, estimated gestational age, birthweight, or maternal age. We defined PTB as delivery before 37 weeks of gestational age [9] ​; additionally, very PTB (VPTB) was further defined as delivery before 34 weeks of gestational age. SGA was classified as a birthweight below the 10^th^ percentile for gestational age, adjusted for infant sex [10]; additionally, we defined very SGA (VSGA) for infants with a birthweight below the 3^rd^ percentile for gestational age, adjusted for infant sex [10], as a marker of potential fetal growth restriction, which may indicate impaired placental function. Centiles for birthweight by gestational age and sex are available in Supplementary Table 1.

At the individual-level, we examined maternal demographics (i.e., age, race and ethnicity as a representation of social context rather than biological differences, and education), maternal prenatal care (i.e., no prenatal care or first prenatal care after 27^th^ week), maternal smoking (prior to and during pregnancy), and maternal comorbid conditions (pre-pregnancy and gestational diabetes and hypertension, preeclampsia and previous PTB); paternal demographics (i.e., race and ethnicity as a reflection of systemic social influences, and education); and infant’s birth weight, gestational age at delivery, and neonatal sex.

At the community level, we gathered yearly U.S. Census tract–level data from the Agency for Healthcare Research and Quality (AHRQ) Social Determinants of Health (SDOH) database [11] ​. This included demographic indicators (e.g., racial and ethnic distributions, educational attainment, foreign-born population, and households with limited English proficiency), which highlight the cultural diversity and social cohesion of communities; housing metrics (e.g., median home value, gross rent) and economic indicators (e.g., household income, unemployment rates, and government assistance) to assess neighborhood stability and material conditions; and transportation variables (e.g., average commute time and distance to healthcare facilities) to evaluate healthcare accessibility. Additionally, to comprehensively assess neighborhood-level socioeconomic disadvantage, we included two composite the Neighborhood Deprivation Index (NDI​) [12] and the Social Deprivation Index (SDI)​ [13]​. Both indices are based on principal components analysis (PCA) of variables derived from the American Community Survey (ACS), such as income, education, employment, housing, and occupation. We computed the NDI annually using the publicly available ndi R package [14], while the SDI was obtained directly from the project’s website ​ [13]​.

Statistical analyses

Standardized incidence rates in smaller, less populated areas are prone to fluctuation due to minor changes in case numbers. To enhance stability, we applied Empirical Bayes (EB) [15]​ smoothing techniques, adjusting rates for each tract towards the overall average. This method provides more reliable measures of relative risk, enabling comparisons across diverse areas, even those with small populations or limited data. To visualize spatial variability in the outcome, we developed descriptive maps of EB smoothed rates, grouped into six categories based on breakpoints derived from the corresponding boxplot distribution. These categories included four interquartile-based intervals (< 25%, 25–50%, 50–75%, and > 75%) and two outlier groups, defined using Tukey’s method as values falling below Q1 − 1.5×IQR or above Q3 + 1.5×IQR, where Q1 and Q3 represent the first and third quartiles, respectively.

We employed Moran’s I Statistic [16] to assess spatial autocorrelation, i.e. whether neighboring spatial units have similar attribute values. Calculation involves constructing the weights matrix that defines the connections between units; we utilized a comprehensive queen contiguity spatial weights matrix (Supplementary Fig. 2), which accounts for both neighboring edges and corners, providing a robust assessment of how the 179 D.C. tracts are interconnected. On average, each tract is linked to 6 neighbors, highlighting substantial spatial continuity. Positive values of Moran’s I indicate clustering, negative values indicate dispersion, and zero suggests no spatial pattern. This analysis was conducted using spdep R package [17].

We utilized Kulldorff’s spatial and space–time scan statistics to detect clusters of Census tracts with significantly higher or lower rates of adverse outcomes than would be expected by random chance. Specifically, we assessed clustering of PTB and SGA cases using a retrospective Poisson spatial and space-time scan model with circular scanning windows. The analysis tested varying maximum cluster sizes set at 10%, 25%, and 50% of the total number of births per tract [18]. The circle with the highest likelihood ratio identified the most probable cluster, representing areas with significantly higher or lower observed cases relative to expected counts. For the space–time scan statistics, we applied similar spatial parameters while incorporating a temporal dimension to detect persistent clustering patterns. The temporal window ranged from a minimum aggregation of 1 year to a maximum covering the full study period. Statistical significance was assessed using Monte Carlo simulations, with clusters reported at a P-value threshold of < 0.05. To maintain spatial independence, secondary clusters were reported only if they did not overlap with the primary (most likely) cluster. We calculated relative risks (RRs) to compare the magnitude of risk within clusters to surrounding areas, with RR values greater than 1 indicating higher risk and values less than 1 indicating lower risk within the clusters.

Finally, ANOVA tests were conducted to evaluate individual- and community-level factors within the identified higher and lower risk clusters.

We used the tmap [19] and ggplot2 [20] R packages to create geospatial and statistical visualizations. All analysis were performed in R [21] ​​and SatScan [22] software.

Results

The study flow diagram is presented in Fig. 1**.** We initially obtained 84,788 birth records from 2010 to 2018. We excluded births to non-D.C. residents (n = 8,362), infants delivered before 20 weeks of gestation (n = 44), multiple-gestation births (n = 2,950), and those without a geocoded residential address (n = 2,221). Additional exclusions included records with missing or invalid gestational age, birthweight (n = 126), or maternal age (n = 2), resulting in a final analytic sample of 71,083 singleton births with complete data. Among these, 6,005 (8.4%) were classified as PTB, of which 2,018 (2.8%) were VPTB. A total of 4,031 (5.67%) infants were identified as SGA, and 1,119 (1.57%) met the criteria for VSGA. Of the 6,005 PTB infants, 352 (5.87%) were also classified as SGA, representing 8.73% of all SGA births.Fig. 1Study flow diagram illustrating the process of identifying dyads (mother-infant pairs) for inclusion in geospatial analysis. The final dataset comprised 71,083 complete birth records for 2010-2018, among which 6,005 were classified as preterm birth (PTB), 2,018 as very preterm birth (VPTB), 4,031 as small-for-gestational-age (SGA), and 1,119 as very SGA (VSGA). Centiles for birthweight by gestational age and sex are available in Supplementary Table 1

Cohort characteristics are shown in Table 1 which highlights similarities and differences across PTB, VPTB, SGA, and VSGA births relative to their respective comparison groups. PTB and SGA births are disproportionately associated with mothers of Black race, accounting for 67% and 69% of cases, respectively, compared to 51% in both non-PTB and non-SGA groups. This disparity is even more pronounced in VPTB and VSGA births, with Black mothers comprising 74% and 73% of cases, respectively.

Table 1Maternal, paternal, and infant cohort characteristics by preterm birth (PTB), small-for-gestational-age (SGA) birth, very preterm birth (VPTB), and very small-for-gestational-age (VSGA) including demographic details, prenatal care, smoking habits, comorbidities, year and residential address at time of delivery, paternal information, and infant birthweight and sex. The data is divided into non-outcome vs outcome groups with the sample sizes and means or percentages provided for each characteristicNon-PTB *N = 65,078PTBN *= 6,005Non-SGA N = 67,052SGA N= 4,031Non-VPTB *N *= 69,065VPTB *N = 2,018Non-VSGA N = 69,964VSGAN *= 1,119 * N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

* N *

Mean

(SD)

Maternal & delivery  Age at time of delivery65,07830 (6.5)6,00529 (6.6)67,05230 (6.4)4,03128 (6.8)69,06530 (6.5)2,01829 (6.6)69,96430 (6.5)1,11928 (6.9) Race White21,06832%1,12519%21,50932%68417%21,93032%26313%22,02331%17015% Black32,87151%4,00067%34,07951%2,79269%35,37351%1,49874%36,05052%82173% Asian & Pacific Islander3,0425%1783%3,0505%1704%3,1755%452%3,1895%313% American Indian/Alaskan Native1100%120%1170% <100%1160% <100%1190% <100% Other6,55510%5659%6,79710%3238%6,94910%1718%7,04310%777% Unknown1,4322%1252%1,5002%571%1,5222%352%1,5402%172% Ethnicity Hispanic9,84615%83014%10,20115%47512%10,43715%23912%10,55315%12311% Not Hispanic54,96784%5,14486%56,56884%3,54388%58,34584%1,76688%59,11884%99389% Unknown2650%311%1,1322%1063%1,1612%774%1,2042%343% Education level College degree and above29,61446%1,76229%30,21345%1,16329%30,87945%49725%31,08344%29326% No college degree and below34,81954%4,10168%36,10454%2,81670%37,48954%1,43171%38,10954%81172% Unknown6451%1422%7351%521%6971%904%7721%151% Prenatal care None or prenatal care after 27^th^ week8,66813%1,03417%9,08314%61915%9,35114%35117%9,50714%19517% Less than 10 prenatal care visits32,32450%4,74579%34,57452%2,49562%35,29151%1,77888%36,32252%74767% Smoking Prior to pregnancy6181%1292%6551%922%6941%533%7141%333% During pregnancy5941%1222%6271%892%6681%482%6831%333% Comorbidities Prepregnancy diabetes4441%1543%5731%251%5531%452%5911% <101% Gestational diabetes2,0253%3005%2,2383%872%2,2533%724%2,3003%252% Prepregnancy hypertension1,1252%3205%1,3332%1123%1,3222%1236%1,4072%383% Gestational hypertension2,6774%59410%2,9684%3038%3,0724%19910%3,1635%10810% Hypertension & eclampsia3331%1733%4521%541%4371%693%4891%172% Previous preterm births1,7913%72012%2,3383%1734%2,2163%29515%2,4554%565% Year at time of delivery 20106,85611%58710%7,03010%41310%7,23010%21311%7,33110%11210% 20116,96611%71512%7,21011%47112%7,44811%23312%7,53511%14613% 20127,16711%68211%7,40511%44411%7,62611%22311%7,74811%1019% 20137,10011%68711%7,40511%3829%7,53611%25112%7,69211%958% 20147,34311%61210%7,42711%52813%7,75311%20210%7,77711%17816% 20157,08911%59210%7,24211%43911%7,47411%20710%7,56311%11811% 20167,66312%74912%7,94812%46412%8,17512%23712%8,28912%12311% 20177,60712%71212%7,83812%48112%8,08412%23512%8,18412%13512% 20187,28711%66911%7,54711%40910%7,73911%21711%7,84511%11110% Residential address at delivery Ward 18,48913%62410%8,71713%39610%8,91313%20010%9,00213%11110% Ward 24,4357%2414%4,4707%2065%4,6217%553%4,6267%504% Ward 35,5659%3095%5,6868%1885%5,8058%693%5,8338%414% Ward 410,32016%82914%10,67416%47512%10,88116%26813%11,02616%12311% Ward 58,44013%81414%8,68013%57414%8,96513%28914%9,10213%15214% Ward 68,94314%65711%9,13514%46512%9,39714%20310%9,48014%12011% Ward 78,29813%1,12519%8,68113%74218%9,01213%41120%9,18913%23421% Ward 810,58816%1,40623%11,00916%98524%11,47117%52326%11,70617%28826% Paternal  Race White21,06632%1,10018%21,49532%67117%21,91632%25012%22,00231%16415% Black19,22130%2,05234%19,96830%1,30532%20,49530%77839%20,89930%37433% Asian & Pacific Islander2,1233%1232%2,0953%1514%2,2163%301%2,2153%313% American Indian/Alaskan Native1220%160%1300% <100%1290% <100%1370% <100% Other5,4868%4397%5,6598%2667%5,7898%1367%5,8598%666% Unknown17,06026%2,27538%17,70526%1,63040%18,52027%81540%18,85227%48343% Ethnicity Hispanic8,15213%64311%8,41013%38510%8,59612%19910%8,69312%1029% Not Hispanic41,12663%3,21153%42,27663%2,06151%43,28763%1,05052%43,79063%54749% Unknown15,80024%2,15136%16,36624%1,58539%17,18225%76938%17,48125%47042% Education level College degree and above27,20142%1,45624%27,66341%99425%28,29141%36618%28,43241%22520% No college degree and below21,99634%2,38540%22,92134%1,46036%23,52934%85242%23,94534%43639% Unknown15,88124%2,16436%16,46825%1,57739%17,24525%80040%17,58725%45841% Infant  Birthweight in grams65,0783,344 (464)6,0052,189 (760)67,0523,295 (567)4,0312,441 (336)69,0653,299 (499)2,0181,445 (651)69,9643,263 (578)1,1192,178 (339) Obstetric estimation of gestation65,07839 (1.1)6,00533 (3.5)67,05239 (2.2)4,03139 (1.6)69,06539 (1.4)2,01829 (3.5)69,96439 (2.2)1,11939 (1.6) Sex Female31,74649%2,85948%32,59149%2,01450%33,64049%96548%34,03049%57551% Male33,33251%3,14652%34,46151%2,01750%35,42551%1,05352%35,93451%54449%

Maternal education shows similar only 29% of both PTB and SGA births occurred among mothers with a college degree, compared to 46% and 45% in non-PTB and non-SGA births. The gap widens further for VPTB and VSGA, where only 25% and 26% of mothers held college degrees. In parallel, inadequate prenatal care is a consistent feature across all adverse maternal outcomes. A striking 79% of PTB and 62% of SGA mothers received fewer than 10 prenatal visits, compared to ~ 50% in their respective comparison groups; this increases to 88% for VPTB and 67% for VSGA births.

Paternal characteristics mirrored maternal trends. Among those with recorded data, Black fathers were overrepresented in adverse outcomes—comprising 34% of PTB and 32% of SGA births, compared to 30% in both non-PTB and non-SGA groups. Conversely, White fathers were underrepresented in these groups, accounting for only 18% of PTB and 17% of SGA births, compared to 32% in both non-PTB and non-SGA births.

Paternal educational attainment also only 24% of PTB and 25% of SGA births occurred among fathers with a college degree or higher, compared to 42% and 41% among non-PTB and non-SGA births, respectively. Conversely, 40% of PTB births were to fathers with no college degree, compared to 34% in non-PTB births. A similar trend was observed for SGA 36% had no college degree, compared to 34% in the non-SGA group.

Infant characteristics reflected expected clinical patterns associated with adverse birth outcomes. Infants classified as PTB had a mean birthweight of 2,189 g, while those in the VPTB category had a substantially lower mean birthweight of 1,445 g. In contrast, SGA infants had a mean birthweight of 2,441 g, which was markedly lower than the 3,295 g observed in non-SGA, and similarly those in the VSGA category had a substantially lower mean birthweight of 2,178 g. Mean gestational age also differed notably between PTB infants were born at an average of 33 weeks, and VPTB infants at 29 weeks, both significantly earlier than the 39 weeks observed in term infants.

Geographic disparities were concentrated in Wards 7 and 8, which are historically underserved areas of Washington, D.C. Ward 8 accounted for a disproportionately high share of PTB (23%) and VPTB (26%) births, compared to 16% and 17% of non-PTB and non-VPTB births, respectively. Similarly, SGA (24%) and VSGA (26%) births were overrepresented in Ward 8, relative to 16% of non-SGA and 17% of non-VSGA births. Ward 7 showed a similar pattern, with 19% of PTB and 20% of VPTB births occurring in the ward, versus 13% of their respective comparison groups. For SGA and VSGA, 18% and 21% of births occurred in Ward 7, compared to 13% among both non-SGA and non-VSGA births.

The 2010–2018 EB smoothed rates for PTB, VPTB, SGA, and VSGA are shown in Fig. 2. Across all four outcomes, tracts shaded in blue represent areas with low rates (< 50%), primarily located in the north and west of D.C., encompassing neighborhoods like Kent, Cathedral Heights, Georgetown, and Tenleytown, which are typically more affluent based on indicators of differential health opportunities [8]. Conversely, tracts shaded in dark orange and red denote rates significantly above the average (> 75%), predominantly in the east and south of D.C., encompassing historically marginalized and disinvested areas such as Historic Anacostia. The spatial patterns for PTB and SGA significantly overlap in identifying high-risk areas (Wards 7 and 8) and low-risk areas (Ward 3), with VPTB and VSGA exhibiting similar distributions to their broader categories.Fig. 2Empirical Bayes (EB) smoothing rates maps for preterm birth (PTB; A), very preterm birth (VPTB; B), small-for-gestational-age (SGA; C) and very low birth weight (VSGA; D) from 2010–2018. The maps highlight outlying areas with significantly lower (blue) or higher (red) observed event rates compared to the expected rates, which are based on the reference rate of the population at risk

Consequently, the remainder of the study focuses on the broader PTB and SGA categories to ensure sufficient sample size, stable estimates, and because their spatial patterns closely mirror those of the more specific outcomes, VPTB and VSGA.

The global Moran’s I test to determine geospatial clustering of PTB and SGA within D.C., revealed significant clustering of EB-smoothed rates across the district (Moran’s I statistics = 0.257, P < 0.001 for PTB; 0.226, P < 0.001 for SGA).

To pinpoint cluster locations, we conducted spatial and space–time analyses. Results are presented for clusters detected using a maximum spatial cluster size of 25% of the population. Clusters identified using alternative maximum sizes of 50%, 25%, and 10% of the population are shown in Supplementary Fig. 3. The spatial analysis of adverse birth outcomes (Fig. 3A and B) revealed significant clusters with low values in northwest D.C. and high values in the southeast, indicating reduced likelihoods of PTB and SGA in the northwest and heightened risks in the southeast. All cluster details are reported in Supplementary Table 1**.** Specifically, spatial clustering analysis revealed that birthing individuals in clusters 1, 3, 5, 7, 10, and 12 had a 47% higher risk of PTB (RR = 1.467), while those in clusters 2, 4, and 5 faced a 56% higher risk of SGA (RR = 1.560), compared to individuals in census tracts outside these clusters. High-risk spatial clusters for both PTB and SGA were primarily located in Wards 7 and 8, as well as the southeastern portions of Ward 5. The space–time analysis (Fig. 3C and 3D) showed that high-risk clusters varied in persistence across outcomes; for PTB, cluster 2 (RR = 1.520) was present from January 2011 to December 2014, indicating a shorter duration. In contrast, SGA clusters 1 and 3 (RR = 1.675) persisted for a longer period, from January 2011 to December 2017. Ward 7 contained high-risk space–time clusters for both PTB and SGA, whereas in Ward 8, high-risk clusters were identified for SGA, but only a few adjacent tracts near the Ward 7 boundary showed elevated PTB risk.Fig. 3Spatial (A, B) and space-time (C,** D**) clusters of relative risks (RR) for preterm birth (PTB; A,** C**) and small-for-gestational age (SGA; B,** D**) during the study period from 2010 to 2018. RR represents the magnitude of risk within clusters compared to surrounding areas, with values greater than 1 indicating higher risk within clusters (red, RR: (1;1.8]) and values less than 1 indicating lower risk (blue, RR: (0.4;1]). All identified clusters are statistically significant (*P *< 0.05). The space-time analysis captures clustering patterns over time, with a minimum time aggregation of one year and a maximum spanning the entire study period. Results are presented for clusters detected using a maximum spatial cluster size of 25% of the population

ANOVA results in Table 2 demonstrate significant differences in individual- and community-level factors between high- and low-risk clusters for PTB and SGA presented in Fig. 3.

At the individual-level, birthing individuals in high-risk clusters were significantly younger than those in low-risk clusters (PTB: 26.67 ± 1.89 vs. 33.09 ± 1.72 years; SGA: 26.46 ± 1.78 vs. 32.64 ± 2.05 years; all P < 0.001). Racial and ethnic distributions differed markedly. High-risk PTB and SGA clusters had substantially higher proportions of non-Hispanic Black mothers (PTB: 0.881 ± 0.188 vs. 0.110 ± 0.117; SGA: 0.898 ± 0.171 vs. 0.173 ± 0.173) and fewer non-Hispanic White (PTB: 0.079 ± 0.174 vs. 0.705 ± 0.198; SGA: 0.067 ± 0.153 vs. 0.632 ± 0.249) and Hispanic individuals (PTB: 0.039 ± 0.037 vs. 0.127 ± 0.131; SGA: 0.035 ± 0.036 vs. 0.148 ± 0.154; all P < 0.001). Educational attainment was also substantially lower in high-risk clusters, with only 20.3% ± 0.183 (PTB) and 17.9% ± 0.169 (SGA) of mothers holding a college degree or higher, compared to 84.0% ± 0.193 (PTB) and 76.4% ± 0.244 (SGA) in low-risk areas (P < 0.001). Also, high-risk clusters exhibited more adverse maternal health behaviors and limited care access. Rates of no or late prenatal care (after 27 weeks) were significantly higher (PTB: 15.5% ± 5.2 vs. 10.9% ± 5.6; SGA: 16.0% ± 5.6 vs. 10.8% ± 4.9), as were rates of fewer than 10 prenatal visits (PTB: 64.5% ± 10.8 vs. 35.3% ± 10.6; SGA: 66.1% ± 10.3 vs. 37.7% ± 11.8). Smoking prior to pregnancy was more common in high-risk clusters (PTB: 2.0% ± 1.6 vs. 0.2% ± 0.5; SGA: 2.2% ± 1.8 vs. 0.2% ± 0.7), as was smoking during pregnancy (PTB: 1.9% ± 1.6 vs. 0.1% ± 0.4; SGA: 2.1% ± 1.8 vs. 0.2% ± 0.6; all P < 0.001). Clinical conditions were more prevalent in high-risk areas. Pre-pregnancy diabetes rates were 1.0% ± 1.1 in PTB and 1.1% ± 1.1 in SGA clusters, compared to 0.4% ± 0.9 and 0.5% ± 0.9 in low-risk areas, respectively. Pre-pregnancy hypertension was also elevated (PTB: 3.1% ± 2.0 vs. 1.0% ± 1.4; SGA: 3.1% ± 2.0 vs. 1.1% ± 1.3), along with gestational hypertension (PTB: 5.4% ± 2.8 vs. 3.7% ± 2.5; SGA: 5.2% ± 2.7 vs. 4.0% ± 2.4) and history of previous PTB (PTB: 5.3% ± 3.3 vs. 1.6% ± 1.9; SGA: 5.6% ± 3.4 vs. 2.0% ± 2.2; all P < 0.001). Gestational diabetes showed no significant difference for PTB (2.6% ± 1.7 vs. 2.9% ± 2.3) but was marginally lower in SGA clusters (2.6% ± 1.7 vs. 3.2% ± 2.3; P < 0.05). Paternal characteristics reflected similar disparities. Fathers in high-risk PTB and SGA clusters were less likely to be White (PTB: 7.7% ± 17.7 vs. 71.7% ± 20.1; SGA: 6.4% ± 15.6 vs. 63.9% ± 25.7) and more likely to be Black (PTB and SGA: 45.8% ± 11.2 vs. 8.6% ± 8.5 and 45.8% ± 10.5 vs. 12.8% ± 12.0, respectively). Paternal educational attainment was similarly lower (college degree or PTB: 14.4% ± 18.4 vs. 81.6% ± 19.3; SGA: 12.6% ± 16.4 vs. 73.9% ± 24.6; all P < 0.001). No significant differences in infant sex distribution were found between high- and low-risk clusters for either outcome.

Table 2Comparison of individual- (A) and community-level (B) factors between high-risk and low-risk spatial and space-time clusters for preterm birth (PTB) and small-for-gestational age (SGA) from 2010–2018. ANOVA is used to assess the significance of differences in the means of individual- and community-level factors (tract level) between high-risk and low-risk clusters for the spatial and the space-time clustering providing insights into the characteristics and determinants associated with areas of elevated or reduced risk for PTB and SGA(A)PTB 2010-2018SGA 2010-2018Relative risk classHighLowHighLow Spatial clustering  CLUSTERS1, 3, 5, 7, 10, 122, 4, 6, 8, 9, 112, 4, 51, 3, 6, 7, 8 AVERAGE RELATIVE RISK (SD)1.467 (0.106)0.623 (0.050)1.560 (0.046)0.670 (0.132) START DATEJanuary 1, 2010January 1, 2010January 1, 2010January 1, 2010 END DATEDecember 31, 2018December 31, 2018December 31, 2018December 31, 2018 Space-time clustering  CLUSTERS21,31,32,4 AVERAGE RELATIVE RISK (SD)1.520 (-)0.560 (0.014)1.675 (0.064)0.430 (0.156) START DATEJanuary 1, 2011January 1, 2012January 1, 2011January 1, 2015 END DATEDecember 31, 2014December 31, 2017December 31, 2017December 31, 2018 Individual-level factors

Maternal & delivery

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Age at time of delivery26.672 (1.892)33.089 (1.715)26.464 (1.776)32.636 (2.048) White race0.079 (0.174)0.705 (0.198)0.067 (0.153)0.632 (0.249) Black race0.881 (0.188)0.110 (0.117)0.898 (0.171)0.173 (0.173) Hispanic Ethnicity0.039 (0.037)0.127 (0.131)0.035 (0.036)0.148 (0.154) Education level ≥ college degree0.203 (0.183)0.840 (0.193)0.179 (0.169)0.764 (0.244) No prenatal care or after 27^th^ week0.155 (0.052)0.109 (0.056)0.160 (0.056)0.108 (0.049) Less than 10 prenatal care visits0.645 (0.108)0.353 (0.106)0.661 (0.103)0.377 (0.118) Smoking prior to pregnancy0.020 (0.016)0.002 (0.005)0.022 (0.018)0.002 (0.007) Smoking during pregnancy0.019 (0.016)0.001 (0.004)0.021 (0.018)0.002 (0.006) Comorbidities: Pre-pregnancy diabetes0.010 (0.011)0.004 (0.009)0.011 (0.011)0.005 (0.009) Comorbidities: Gestational diabetes0.026 (0.017)0.029 (0.023)0.026 (0.017)0.032 (0.023)**** Comorbidities: Pre-pregnancy hypertension0.031 (0.020)0.010 (0.014)0.031 (0.020)0.011 (0.013) Comorbidities: Gestational hypertension0.054 (0.028)0.037 (0.025)0.052 (0.027)0.040 (0.024) Comorbidities: Hypertension & preeclampsia0.010 (0.011)0.002 (0.009)0.010 (0.011)0.003 (0.006) Comorbidities: Previous preterm births0.053 (0.033)0.016 (0.019)0.056 (0.034)0.020 (0.022) Paternal

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  White race0.077 (0.177)0.717 (0.201)0.064 (0.156)0.639 (0.257) Black race0.458 (0.112)0.086 (0.085)0.458 (0.105)0.128 (0.120) Hispanic Ethnicity0.031 (0.032)0.105 (0.108)0.027 (0.035)0.124 (0.127) Education level ≥ college degree0.144 (0.184)0.816 (0.193)0.126 (0.164)0.739 (0.246) Infant

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Sex: Female0.492 (0.053)0.488 (0.069)0.491 (0.058)0.486 (0.062) Sex: Male0.508 (0.053)0.512 (0.069)0.509 (0.058)0.514 (0.062)(B)PTB 2010-2018SGA 2010-2018Relative risk classHighLowHighLow Spatial clustering  CLUSTERS1, 3, 5, 7, 10, 122, 4, 6, 8, 9, 112, 4, 51, 3, 6, 7, 8 AVERAGE RELATIVE RISK (SD)1.467 (0.106)0.623 (0.050)1.560 (0.046)0.670 (0.132) START DATEJanuary 1, 2010January 1, 2010January 1, 2010January 1, 2010 END DATEDecember 31, 2018December 31, 2018December 31, 2018December 31, 2018 Space-time clustering  CLUSTERS21,31,32,4 AVERAGE RELATIVE RISK (SD)1.520 (-)0.560 (0.014)1.675 (0.064)0.430 (0.156) START DATEJanuary 1, 2011January 1, 2012January 1, 2011January 1, 2015 END DATEDecember 31, 2014December 31, 2017December 31, 2017December 31, 2018 Community-level factors (U.S. Census Tract-level)

Population Demographics

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  White race (%)6.110 (12.767)73.653 (14.502)5.255 (12.210)63.898 (23.305) Black race (%)90.669 (14.371)13.634 (13.480)91.723 (13.858)23.477 (21.977) Asian race (%)0.525 (0.933)6.209 (3.238)0.437 (0.893)5.154 (3.224) Hispanic ethnicity (%)2.588 (2.437)11.209 (7.850)2.530 (2.593)11.780 (8.465) Foreign-born (%)4.062 (2.706)20.482 (7.791)3.848 (2.722)19.534 (8.602) Limited English speaking HH (%)0.985 (1.291)3.061 (3.358)0.934 (1.217)3.541 (4.387) Education

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Less than high school education (%) ▴▴▴16.777 (5.834)5.447 (6.205)18.051 (6.187)7.128 (7.124) High school diploma (%) ▴▴▴38.290 (10.028)6.020 (4.816)38.834 (9.686)9.172 (7.572) Some college or associate's degree (%) ▴▴▴25.991 (5.602)9.417 (4.885)26.289 (5.585)11.547 (6.288) Bachelor's degree (%) ▴▴▴10.574 (6.258)31.283 (6.016)9.769 (5.770)29.156 (6.994) Master's or professional school degree or doctorate (%) ▴▴▴8.369 (8.429)47.833 (10.647)7.057 (6.969)42.997 (14.180) Any postsecondary education (%) ▴▴▴44.933 (13.613)88.532 (10.321)43.115 (13.008)83.700 (13.907) Housing Characteristics

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Average HH size2.473 (0.325)1.977 (0.379)2.537 (0.358)2.146 (0.402) Median home value of owner-occupied housing units (dollars)287,550 (110,924)684,609 (254,293)280,942 (104,515)673,421 (246,559) Median gross rent (dollars)960 (237)1,709 (328)948 (272)1,609 (407) Housing units built before 1979 (%)84.547 (11.664)84.612 (13.254)81.212 (14.094)85.230 (16.141)*** Owner-occupied housing units (%)37.066 (19.761)45.303 (18.784)31.834 (19.818)48.671 (19.664) HH with only one occupant (%)40.555 (7.951)48.228 (13.631)*39.045 (9.133)44.056 (12.241) Families with children that are single-parent families (%)77.115 (20.355)21.069 (21.645)79.188 (20.154)25.206 (21.655) HH that received food stamps/SNAP, past 12 months (%)29.655 (11.021)3.264 (4.527)32.456 (12.168)5.233 (6.090) HH with public assistance income or food stamps/SNAP (%)30.968 (11.294)3.789 (4.738)33.803 (12.486)5.868 (6.299) Economic Factors

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Gini index of income inequality0.484 (0.072)0.465 (0.055)0.486 (0.074)0.464 (0.057) Median HH income (dollars, inflation-adjusted)43,535 (23,413)108,040 (39,783)39,604 (21,170)103,979 (40,331) HH income less than $10,000 ($10,000 and $14,999 ($15,000 and $24,999 ($100,000 (%)15.467 (13.274)50.707 (15.467)13.503 (12.103)48.854 (15.892) Employment

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Armed forces (%)▴0.426 (2.876)0.410 (1.123)*0.985 (6.376)0.556 (1.929) Government (%)▴28.998 (8.041)24.198 (6.477)28.929 (8.578)24.710 (6.646) Arts, recreation, accommodation and food services (%)▴9.910 (5.070)8.436 (4.038)10.512 (5.141)8.679 (4.249) Construction (%)▴4.126 (2.667)1.780 (1.883)3.878 (2.631)2.321 (2.436) Finance, insurance, real estate, rental and leasing (%)▴4.622 (2.327)6.938 (2.741)4.533 (2.250)6.383 (2.761) Manufacturing (%)▴1.167 (1.091)1.298 (0.815)1.020 (1.023)1.241 (0.864) Agriculture, forestry, fishing, hunting and mining (%)▴0.056 (0.245)0.136 (0.298)***0.085 (0.371)0.131 (0.297) Not in labor force (%)▴40.397 (10.749)24.285 (12.253)41.569 (11.333)25.832 (10.707) Neighborhood Characteristics

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Neighborhood Deprivation Index0.033 (0.025)-0.028 (0.014)0.039 (0.025)-0.024 (0.017) Social Deprivation Index71.015 (21.203)36.474 (19.547)73.261 (21.805)38.143 (21.209) Transportation and Commuting among Workers

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  No vehicle available (%)▴26.624 (10.664)28.487 (17.286)28.963 (11.440)23.442 (14.894)*** Walking to work (%)▴3.831 (8.850)20.940 (16.292)4.220 (9.214)14.584 (13.777) < 15-minute commute time (%)▴9.803 (10.297)16.375 (10.450)*10.769 (11.655)14.227 (9.366) 15- to 29-minute commute time (%)▴29.596 (7.986)40.458 (7.436)28.219 (8.402)37.858 (7.713) 30- to 59-minute commute time (%)▴43.718 (8.028)38.040 (9.255)******42.942 (8.889)41.643 (9.654) At least 60-minute commute time (%)▴16.882 (7.074)5.128 (2.586)18.070 (7.592)6.272 (3.522) Public transport with < 15-minute commute time (%)▴1.706 (2.645)1.808 (2.231)1.655 (2.543)1.672 (2.232) Public transport with 15- to 29-minute commute time (%)▴16.186 (10.716)31.234 (11.446)15.007 (9.570)28.430 (12.021) Public transport with 30- to 59-minute commute time (%)▴52.735 (10.937)58.415 (11.400)52.708 (10.997)60.077 (11.321) Public transport with at least 60-minute commute time (%)▴29.372 (11.950)8.542 (5.300)30.630 (12.260)9.821 (6.726) Healthcare Access

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Distance to the nearest emergency department (miles)†1.686 (0.738)1.055 (0.633)*1.482 (0.692)1.247 (0.746) Distance to the nearest medical-surgical ICU (miles) †1.571 (0.722)0.918 (0.568)*1.388 (0.674)1.120 (0.710) Distance to the nearest obstetrics department (miles) †1.601 (0.700)0.964 (0.564)1.468 (0.627)1.163 (0.703) Distance to the nearest pediatric ICU (miles) †4.086 (1.595)1.799 (0.917)4.154 (1.450)2.233 (1.215) Distance to the nearest designated trauma center (miles) †1.826 (0.821)1.464 (0.883)*2.126 (0.936)1.621 (0.921) Geographical Characteristics

Mean (SD)

Mean (SD)

P S

P ST

Mean (SD)

Mean (SD)

P S

P ST  Land area (square miles)0.306 (0.237)0.300 (0.281)0.344 (0.335)0.348 (0.324) Population density13,037 (5,827)23,190 (15,093)13,377 (6,808)19,066 (13,243)Abbreviations: P~S~ P-value for spatial clustering, P~ST~ P-value for space-time clustering, HH HouseholdLegend: ▴ages 16+; ▴▴ages 18+; ▴▴▴ages 25+; † calculated using weighted tract centroidsSignificance ‘’ 0.001; ‘’ 0.01; ‘*’ 0.05; ‘.’ 0.1; ‘ ’ 1

At the community-level, high-risk clusters for both PTB and SGA were located in census tracts with markedly different sociodemographic profiles compared to low-risk clusters. These tracts had significantly higher proportions of Black residents (PTB: 90.7% ± 14.4 vs. 13.6% ± 13.5; SGA: 91.7% ± 13.9 vs. 23.5% ± 22.0) and correspondingly lower proportions of White (PTB: 6.1% ± 12.8 vs. 73.7% ± 14.5; SGA: 5.3% ± 12.2 vs. 63.9% ± 23.3), Hispanic (PTB: 2.6% ± 2.4 vs. 11.2% ± 7.9; SGA: 2.5% ± 2.6 vs. 11.8% ± 8.5), and foreign-born populations (PTB: 4.1% ± 2.7 vs. 20.5% ± 7.8; SGA: 3.8% ± 2.7 vs. 19.5% ± 8.6; all P < 0.001). Educational attainment in high-risk tracts was substantially lower. The share of residents with a bachelor’s degree was significantly reduced (PTB: 10.6% ± 6.3 vs. 31.3% ± 6.0; SGA: 9.8% ± 5.8 vs. 29.2% ± 7.0), as was attainment of graduate or professional degrees (PTB: 8.4% ± 8.4 vs. 47.8% ± 10.6; SGA: 7.1% ± 7.0 vs. 43.0% ± 14.2). Overall, postsecondary education rates were nearly halved in high-risk tracts (PTB: 44.9% ± 13.6 vs. 88.5% ± 10.3; SGA: 43.1% ± 13.0 vs. 83.7% ± 13.9; all P < 0.001). Economic hardship was more pronounced in high-risk clusters. Median household incomes were less than half of those in low-risk areas (PTB: $43,535 ±$23,414 vs. $108,041 ±$39,783; SGA: $39,605 ±$21,171 vs. $103,979 ±$40,332). These communities had higher rates of extreme poverty, with more households earning below $10,000 annually (PTB: 15.9$100,000 (PTB: 15.5% ± 13.3 vs. 50.7% ± 15.5; SGA: 13.5% ± 12.1 vs. 48.9% ± 15.9). Reliance on public assistance was significantly greater, with 31.0% ± 11.3 (PTB) and 33.8% ± 12.5 (SGA) of households receiving SNAP or public assistance compared to just 3.8% ± 4.7 and 5.9% ± 6.3, respectively (P < 0.001). Housing in high-risk tracts reflected lower wealth accumulation and stability. Median home values were substantially lower (PTB: $287,551 ±$110,925 vs. $684,609 ±$254,293; SGA: $280,942 ±$104,515 vs. $673,422 ±$246,559), and rents followed a similar pattern (PTB: $960 ±$237 vs. $1,709 ±$329; SGA: $948 ±$273 vs. $1,609 ±$408). Fewer homes were owner-occupied (PTB: 37.1% ± 19.8 vs. 45.3% ± 18.8; SGA: 31.8% ± 19.8 vs. 48.7% ± 19.7), and household sizes were slightly larger on average (PTB: 2.47 ± 0.33 vs. 1.98 ± 0.38; SGA: 2.54 ± 0.36 vs. 2.15 ± 0.40). High-risk areas had significantly more single-parent households among families with children (PTB: 77.1% ± 20.4 vs. 21.1% ± 21.6; SGA: 79.2% ± 20.2 vs. 25.2% ± 21.7; all P < 0.001). Employment patterns diverged notably between high- and low-risk clusters. High-risk areas had greater employment in government (PTB: 29.0% ± 8.0 vs. 24.2% ± 6.5), construction (PTB: 4.1% ± 2.7 vs. 1.8% ± 1.9), and arts/accommodation sectors, but fewer jobs in finance and real estate (PTB: 4.6% ± 2.3 vs. 6.9% ± 2.7). Labor force nonparticipation was substantially higher (PTB: 40.4% ± 10.7 vs. 24.3% ± 12.3; SGA: 41.6% ± 11.3 vs. 25.8% ± 10.7; all P < 0.001). Neighborhood deprivation was notably greater in high-risk areas. Neighborhood Deprivation Index scores were positive in high-risk clusters (PTB: 0.033 ± 0.025 vs. − 0.028 ± 0.014; SGA: 0.039 ± 0.025 vs. − 0.024 ± 0.017), and Social Deprivation Index scores were nearly twice as high (PTB: 71.0 ± 21.2 vs. 36.5 ± 19.5; SGA: 73.3 ± 21.8 vs. 38.1 ± 21.2; all P < 0.001). Commuting characteristics also differed. Residents in high-risk tracts were less likely to walk to work (PTB: 3.8% ± 8.9 vs. 20.9% ± 16.3; SGA: 4.2% ± 9.2 vs. 14.6% ± 13.8), had fewer short commutes (< 15 PTB: 9.8% ± 10.3 vs. 16.4% ± 10.5), and a much higher proportion of long commutes (≥ 60 PTB: 16.9% ± 7.1 vs. 5.1% ± 2.6; SGA: 18.1% ± 7.6 vs. 6.3% ± 3.5; all P < 0.001). Finally, access to healthcare was poorer in high-risk clusters. For PTB, the mean distance to the nearest emergency department was 1.69 ± 0.74 miles in high-risk areas vs. 1.06 ± 0.63 miles in low-risk areas. For SGA, distances to specialized care such as pediatric ICUs (4.15 ± 1.45 vs. 2.23 ± 1.22 miles) and obstetric departments (1.47 ± 0.63 vs. 1.16 ± 0.70 miles) were significantly greater (all P < 0.01). Despite the concentration of high-risk births in dense urban areas, population density was lower in high-risk tracts (PTB: 13,037 ± 5,828 vs. 23,190 ± 15,093 persons/sq mi; SGA: 13,377 ± 6,809 vs. 19,067 ± 13,244; P < 0.01).

Discussion

Principal findings

Our study identified significant spatial and space-time clusters of adverse pregnancy outcomes—specifically PTB and SGA infants—within D.C., from 2010 to 2018. High-risk clusters for both PTB and SGA were predominantly located in the southern and eastern tracts. Individuals within these clusters experienced approximately 47% higher risk for PTB and 56% higher risk for SGA compared with those outside these areas. Space-time analysis further revealed periods of intensified risk (PTB: January 2011 to December 2014; SGA: January 2011 to December 2017), reflecting temporal fluctuations in adverse outcomes. These persistent high-risk clusters underscore the enduring impact of structural factors, including racial segregation, economic marginalization, and limited access to resources, despite ongoing public health efforts. At the individual level, high-risk areas were characterized by younger maternal age, a higher proportion of Black birthing individuals, lower educational attainment, increased maternal comorbidities, and adverse health behaviors such as smoking and reduced prenatal care utilization. Community-level factors included socioeconomic disadvantages, such as elevated unemployment rates, lower median household incomes, and neighborhood deprivation. Together, these findings emphasize the critical need for targeted, multi-level interventions to address structural inequities and reduce risks to maternal and infant health in vulnerable populations.

Results in the context of what is known

Our findings contribute to the growing body of literature demonstrating the profound influence of socioeconomic and structural factors on adverse birth outcomes [23] ​, emphasizing the role of geospatial analysis in identifying high-risk areas and socio-structural contributors to disparities [24]. Similar spatial epidemiologic investigations conducted in urban centers—including New York [18]​​, Ohio ​ [25]​, Michigan [26], and Massachusetts [27] —have also highlighted racial and economic disparities in maternal and infant health.

By situating our findings within this broader context, our study reinforces the importance of a socio-structural approach to maternal and child health, addressing the interplay of socio-economic factors and barriers to care that contribute to adverse birth outcomes in underserved communities.

Clinical implications

Poor maternal circumstances during pregnancy have lasting effects, underscoring the need for targeted antenatal care that addresses the specific needs of birthing individuals. Our identification of high-risk spatial and space-time clusters for PTB and SGA infants in D.C., highlights the importance of enhanced clinical vigilance in these areas. Healthcare providers should prioritize screening for maternal comorbidities, substance use, and engagement with prenatal care to improve early detection of high-risk pregnancies. Local community initiatives and public health services play a critical role in ensuring effective care delivery. A systematic approach that integrates both medical and non-medical risk factors, combined with continuity of care, can improve maternal and infant outcomes. Addressing non-medical factors such as poverty, education, and access to resources—including food and healthcare—is essential to optimizing prenatal care and mitigating adverse outcomes [28]. Although reaching birthing individuals in underserved neighborhoods poses challenges, targeted recruitment strategies, trust-building efforts, and sensitive outreach have demonstrated success in improving engagement and outcomes [29]​​. While this study focuses on D.C., the approach and lessons learned are applicable to other urban settings facing similar maternal and infant health disparities.

Research implications

Tackling systemic inequities is essential to improving healthcare access and outcomes. This study demonstrates the utility of geospatial and space-time analysis in uncovering persistent and localized disparities in adverse birth outcomes across D.C. The identification of high-risk clusters—particularly in areas marked by socioeconomic disadvantage, limited healthcare access, and higher prevalence of maternal risk factors—confirms that these outcomes are both spatially and socially patterned. These findings support the continued use of small-area surveillance to monitor PTB and SGA trends and to potentially inform resource allocation and public health planning. Future research should examine the causal mechanisms linking structural factors— such as neighborhood-level disadvantage—with adverse maternal and infant outcomes and evaluate the impact of targeted interventions—such as enhanced prenatal care, social services integration, or neighborhood-based programs—on reducing these disparities. Extending this analytic approach to other urban settings can help inform national strategies to improve maternal and child health outcomes.

Strengths and limitations

A major strength of this study is its integrated analysis of multiple adverse maternal and neonatal outcomes—PTB, SGA, VPTB, and VSGA—within a unified geospatial framework. This approach ensures methodological consistency across outcomes while allowing for outcome-specific insights. The 9-year timeframe adds robustness, revealing long-term trends in maternal and neonatal outcomes in D.C. By focusing on small geographic units (Census tract–level), the study provides locally relevant findings for targeted strategies. Additionally, incorporating various SSDOH factors offers a comprehensive view of barriers to maternal-child health.

However, several limitations should be noted. The study relies on birth certificate data, which may introduce misclassification or reporting bias, particularly for maternal health behaviors (e.g., smoking) and comorbidities. Birth records are also known to contain inconsistencies in reporting maternal weight, chronic conditions, and prior pregnancy history. In addition, while the findings highlight associations between geography and outcomes, they do not establish causality. Ecological limitations also apply. Although clusters were identified at the census tract-level, risks within those areas can vary due to individual-level factors such as health status, behaviors, or genetics. The analysis used maternal residence at delivery, which may not reflect location throughout pregnancy. While most moves likely occur within the same or adjacent tracts, spatial misclassification remains possible [23, 25]​. Additionally, records with ungeocoded addresses were excluded, potentially omitting individuals experiencing housing instability—who may be at elevated risk. Finally, this study did not incorporate environmental exposures such as air pollution, heat, or toxicants, which are known contributors to adverse birth outcomes. Future studies should integrate environmental factors [30] to better understand how structural and environmental factors jointly shape maternal and infant health risks.

Conclusions

Our study, which examines individual- and neighborhood-level factors, highlights the importance of integrating clinical care with non-medical interventions to address adverse maternal outcomes. These findings can inform targeted policies and programs aimed at supporting birthing individuals in communities with concentrated medical and social risks. Expanding similar analyses to other U.S. regions may enhance understanding of geographic disparities in maternal health and inform broader evidence-based strategies.

Supplementary Information

Supplementary Figures and Tables.