Food environment index and preterm birth rate in the counties of the United States

Abstract

OBJECTIVES:

To determine the relationship between Food Environment Index (FEI) and Preterm Birth (PTB) rate at the county level of the United States of America (USA) (primary), while evaluating the interaction of multiple factors within a framework of sociodemographic, maternal health, maternal behavioral, and environmental factors.

METHODS:

This is a population-based retrospective cohort ecological study from 2015-2018. The study compares the characteristics of the population of the counties of the USA. All counties with complete data on their PTB rate and the independent variables were included in the study. Independent variables with greater than 20% missing data were excluded from the study. Purposive sampling technique was applied. A total of 2983/3142 counties were included in the study.

RESULTS:

The median PTB rate of all counties was 9.90%. The highest PTB rate (23.3%) was in Tallapoosa County, Alabama and the lowest (3.4%) in San Juan County, Washington State. After adjusting for variables, PTB rate had a significant association with FEI (coefficient of correlation – 0.36, p < 0.01, 95% CI – 0.19 to – 0.04). Increase in the rate of unemployment, African American race, adult smoking, obesity, uninsured rate, sexually transmitted diseases (STD), high school education and air pollution was associated with an increase in PTB rate, while an increase in FEI and alcohol abuse rates was associated with a decrease in PTB rate.

CONCLUSIONS:

FEI can predict the PTB rate in USA counties after adjusting for sociodemographic, health, behavioral and environmental factors. Future studies are needed to confirm these associations and consider them when making policies to reduce PTBs.

Keywords

Ecological study Food Environment Index premature birth preterm social determinants of health

1 Introduction and background

Preterm Birth (PTB) is defined by the World Health Organization (WHO) as any delivery that occurs at less than 37 completed weeks of gestation [1]. An estimated 15 million infants are born prematurely every year globally, of which 1 million dies from the complications of prematurity [1]. PTB is the number one cause of perinatal mortality in the United States of America (USA). The PTB rate had declined in the USA from 12.8% in 2006 to 9.5% in 2014 with an increasing trend to 9.9% since then, in 2018, per 100 live births [2]. The infant mortality rate (IMR) in the US was 5.42 deaths per 1000 live births in 2020 but the IMR in infants born at a gestational age of < 28 weeks was 363.39/1000 live births, 178 times higher [3]. PTB has significant mortality, morbidity, and financial implications but also has multiple modifiable risk factors, the reason for its importance as a public health issue. A study in the Ohio, USA found that prolonging each premature birth by one week could potentially reduce initial hospital expenditures by over $25 million [4]

The long-term effects of PTB can be severe, sometimes leading to lifelong disabilities such as cognitive impairment, cerebral palsy, intraventricular hemorrhage, and periventricular leukomalacia, leading to delayed development and adverse neurodevelopmental outcomes [3]. Nearly half of the infants born at 22–25 weeks gestation suffer from significant neurodevelopmental disabilities, requiring the early introduction of Special Education Programs (SEP) [4]. In the life of a prematurely born individual, hypertension, cardiac abnormalities, obstructive pulmonary disorders, abnormal blood sugar, and mental health problems are seen at a higher rate later in life compared to individuals born at term. Several sociodemographic factors are associated with a higher incidence of PTB. Behrman (2017) showed that young maternal age, pregnancy in unmarried women, ethnicity, educational attainment, household income, and occupational status are associated with the rate of PTB. In addition, maternal health conditions such as obesity and STD, and maternal behavioral factors such as smoking and alcohol abuse, have been associated with PTB [5]. This study utilized the differences in the population characteristics of geographically restricted areas such as the United States of America (USA) counties to predict PTB rate. Reduced maternal adherence to a healthy diet is associated with a higher percentage of premature birth and lower birth weights (LBW) [6]. Environmental factors, such as distance from supermarkets and retail of unhealthy foods and easy accessibility throughout communities, are increasingly being recognized as essential determinants of individuals’ choices that are directly related to diet-related health outcomes [7]. The Food Environment Index (FEI) is an indicator that could be related to maternal health outcomes and PTB rate in the population. This study aims to determine an association between the FEI and the rate of PTB in USA counties, while exploring and adjusting for the relationship with other risk factors using population-based secondary data sources, providing the basis for further research and policy making.

2 Methods

This was a population-based retrospective cohort ecological study. High-quality PTB rate and risk factors data gathered from a large population base with variability at the county level was obtained from publicly accessible platforms. A purposive sampling approach was followed. Only counties with available PTB rate data were included. Counties with > 20% missing independent variable data were excluded.

The PTB data were obtained from the March of dimes Organization which provided the rate of premature births in each county of the USA [8]. The socioeconomic and sociodemographic data were obtained from the University of Wisconsin Population Health Institute (UWPHI) website [9].

SPSS software was utilized for data analysis (IBM SPSS version 27, 2020). Mean and standard deviation were reported for all normal data, while for non-normal continuous data, median and interquartile range were reported. Pearson’s product moment correlation was used to measure the linear relationship between the continuous variables. Factors demonstrating significant multicollinearity were removed from the final stage of regression. Each independent factor (demographic or risk factor) was examined for univariate association with the outcome variable using linear regression method. A p-value of < 0.05 was considered as the cut-off for a significant association. In the second stage, all factors with a significant association with the outcome variable was included in multiple linear regression models using “enter” method.

3 Results

The number of PTBs in 2019 were 383,169, an average of 10.23% in the USA [10]. Out of 3006 counties, data for PTB were available for 2983 counties, which were included in the study. The median PTB rate of all Counties was 9.90%. The characteristics of the population are described in Table 1.

Table 1
Demographic characteristics of the Counties of USA

Characteristics Median Minimum Maximum Interquartile range

PTB (%) 9.90 3.4 23.30 8.70 – 11.40

2019 Population 25726 86 10039107 10900 – 68104

Hispanic (%) 4.10 0.50 96.30 2.2 – 9.4

NHW (%) 84 2.80 98 64.9 – 92.7

Unemployment (%) 3.90 1.30 19.90 3.10 – 4.80

African American (%) 2.22 0 85.41 0.70 – 10.23

Adult smoking (%) 16.95 5.90 41.49 14.93 – 19.73

Obesity (%) 33.10 12.40 57.70 29.20 – 36.50

Uninsured (%) 12.48 2.68 42.39 8.56 – 17.41

FEI County Value 7.60 0 10 6.90 – 8.20

Alcohol Abuse (%) 17% 0.09 0.29 0.15 – 0.20

STD (%) 302.20 38.70 2889.70 206.50 – 454.90

Teen Birth (per 1000 female population between 15–19) 33 3 113 22 – 45

High school education (%) 88 30 100 82 – 92

Air Pollution (micrograms per cubic meter) 9.20 4.20 15.40 7.90 – 10.00

Median household Income (in dollars) 47600 22000 134600 41100 – 55300

Residential segregation index 46 0 90 34 – 57

Characteristics	Median	Minimum	Maximum	Interquartile range
PTB (%)	9.90	3.4	23.30	8.70 – 11.40
2019 Population	25726	86	10039107	10900 – 68104
Hispanic (%)	4.10	0.50	96.30	2.2 – 9.4
NHW (%)	84	2.80	98	64.9 – 92.7
Unemployment (%)	3.90	1.30	19.90	3.10 – 4.80
African American (%)	2.22	0	85.41	0.70 – 10.23
Adult smoking (%)	16.95	5.90	41.49	14.93 – 19.73
Obesity (%)	33.10	12.40	57.70	29.20 – 36.50
Uninsured (%)	12.48	2.68	42.39	8.56 – 17.41
FEI County Value	7.60	0	10	6.90 – 8.20
Alcohol Abuse (%)	17%	0.09	0.29	0.15 – 0.20
STD (%)	302.20	38.70	2889.70	206.50 – 454.90
Teen Birth (per 1000 female population between 15–19)	33	3	113	22 – 45
High school education (%)	88	30	100	82 – 92
Air Pollution (micrograms per cubic meter)	9.20	4.20	15.40	7.90 – 10.00
Median household Income (in dollars)	47600	22000	134600	41100 – 55300
Residential segregation index	46	0	90	34 – 57

PTB: Preterm birth, NHW: Non-Hispanic white, FEI: Food Environment Index, STD: Sexually Transmitted Diseases.

On univariate linear regression, all the predictors had a significant association with PTB rate at the county level individually (Table 2). While a higher percentage or number of unemployment, African American race, smoking, obesity, uninsured, STD, teen birth, and air pollution was associated with an increase in the rate of PTB; a higher percentage or number of total population, Hispanic, Non-Hispanic White (NHW), FEI, alcohol abuse, high school education and median household income was associated with a decrease in PTB rate (Appendix 2). The strongest association was observed for teen births and preterm rate (0.492) while the weakest correlation was observed between Hispanic race and PTB (– 0.042). The highest R² was demonstrated by African American (0.23) race followed by teen birth rate (0.22). This means that 23% and 22% of the variation in PTB rate in counties can be accounted for by the African American race and teen birth respectively.

Table 2

The bivariate Pearson’s Coefficient of Correlation and Univariate Linear Regression of all the variables showing their relationship to preterm birth rate

Characteristics	Number of Counties	Coefficient of correlation	R²	P value (95% Confidence interval)
PTB	2984	1	N/A	N/A.
2019 Population	2983	– 0.058	0.004	0.002 (0.00 to 0.00)
Hispanic	3142	– 0.042	0.002	0.02 (– 1.20 to – 0.09)
NHW	3142	– 0.316	0.100	< 0.001 (– 3.70 to – 3.00)
Unemployment	2983	0.305	0.070	< 0.001 (0.35 to 0.45)
African American	2983	0.404	0.230	< 0.001 (0.06 to 0.07)
Adult smoking	2983	0.453	0.200	< 0.001 (0.24 to 0.28)
Obesity	2983	0.312	0.100	< 0.001 (0.11 to 0.14)
Uninsured	2983	0.301	0.060	< 0.001 (0.07 to 0.10)
FEI County Value	2964	– 0.366	0.150	< 0.001 (– 0.76 to – 0.64)
Alcohol Abuse	2983	– 0.433	0.180	< 0.001 (– 30.20 to – 25.90)
STD	2922	0.306	0.110	< 0.001 (0.00 to 0.00)
Teen Birth	2955	0.492	0.220	< 0.001 (0.05 to 0.06)
High school education	2636	– 0.076	0.007	< 0.001 (– 3.10 to – 1.10)
Air Pollution	2962	0.217	0.030	< 0.001 (0.18 to 0.28)
Median household Income (in dollars)	2983	– 0.413	0.130	< 0.001 (0.00 to 0.00)
Residential segregation index	2051	– 0.012	0	0.58 (– 0.01 to 0.01)

PTB: Preterm birth, NHW: Non-Hispanic white, FEI: Food Environment Index, STD: Sexually Transmitted Diseases.

PTB rate and FEI have a negative linear association (Fig. 1). The coefficient of correlation was – 0.36, which is a weak correlation, but the association was significant (p value < 0.001, CI – 0.76 to – 0.64). This indicates that with a certain degree of increase in FEI, the rate of PTB will decrease to a specific degree, and this relationship occurs more than by chance. Approximately, 41% of the variation in the PTB rate is explained by the independent variables and their correlation is positive indicating that when the rate of the combined variables increases, the PTB rate also increases (Fig. 2).

Fig. 1

A scatterplot diagram demonstrating the correlation between PTB rate and the FEI value.

Fig. 2

Scatterplot diagram showing the relationship between the outcome and the standardized predictors (R2 = 0.41).

Prior to proceeding with the second step of the regression, multicollinearity was identified, and the three covariates (NHW, teen birth and median household income) were removed from the multiple regression (step 2), removing all correlations between covariates with coefficients ≥ 0.5.

For the second step, all the variables that had a significant association with the outcome were included in a multiple regression model, except for Residential Segregation Index (RSI) (due to non-significant correlation), NHW, teen birth rate and median household income (due to multicollinearity). The final count of the covariates other than FEI was eleven: 2019 population, Hispanic population, unemployment rate, African American population, adult smoking, obesity, uninsured rate, alcohol abuse, STD rate, high school education, and air pollution (Table 3).

Table 3

Multiple linear regression analysis of the independent variables demonstrating their relationship with PTB rate adjusted for the effect of the covariates. The standardized beta coefficients have also been included

Characteristics	Standardized beta coefficient	P value	95% Conf. interval
PTB^**	–	< 0.01	5.96 to 8.31
2019 Population	0.03	0.10	0.00 to 0.00
Hispanic	– 0.03	0.11	– 0.11 to 1.10
Unemployment^**	0.05	< 0.01	0.02 to 0.12
African American^**	0.28	< 0.01	0.03 to 0.04
Adult smoking^**	0.19	< 0.01	0.09 to 0.14
Obesity^**	0.04	0.04	0.00 to 0.03
Uninsured^**	0.08	< 0.01	0.01 to 0.04
FEI County Value^**	– 0.07	< 0.01	– 0.19 to – 0.04
Alcohol Abuse^**	– 0.15	< 0.01	– 12.22 to – 7.11
STD^**	0.07	< 0.01	0.00 to 0.01
High school education^**	0.05	< 0.01	0.44 to 2.25
Air Pollution^**	0.05	< 0.01	0.02 to 0.12

PTB: Preterm birth, NHW: Non-Hispanic white, FEI: Food Environment Index, STD: Sexually Transmitted Diseases. **p < 0.05.

The R of the model was 0.62, which shows high correlation between the predicted value and observed value of FEI. The adjusted R2 was 0.385, indicating that 38.5% of the variation in the dependent variable is explained by the variables in PTB rate of the counties. The adjusted R² of the model is higher than the univariate prediction between FEI and PTB (0.15), signifying that adjusting for other variables increased the strength of the association between them. The model showed significant association of the FEI with PTB rate when adjusted for all other covariates. The F-ratio was 138.46 which is statistically significant at p < 0.001, implying that the regression model predicted the PTB rate significantly well.

When adjusted for all other independent variables, an increase in the rates of unemployment, African American race, adult smoking, obesity, uninsured rate, STD, high school education and air pollution was associated with an increase in PTB rate, while an increase in FEI and alcohol abuse rates was associated with a decrease in PTB rate. Total 2019 county population and the Hispanic population rate were not associated with changes in the PTB rate. The standardized coefficient beta (SCB) (showing the strength of association) was strongest for the African American race (0.28), followed by adult smoking (0.19). The adjusted SCB for FEI was – 0.07. The lowest SCB was for 2019 county population and Hispanic race (– 0.03).

Being unemployed, African American, uninsured, and teen were associated with increased PTB rate; being Hispanic, NHW, high school graduate, and higher median income were associated with lower PTB rate. RSI had a negative correlation with PTB rate which was not significant.

Teen birth rate had the highest positive correlation coefficient with PTB rate (0.49), followed by African American population (0.40). The highest negative correlation coefficient of (– 0.41) was with median household income, followed by NHW population (– 0.31). After adjusting for multiple variables, 2019 county population and Hispanic race were not significant predictors of PTB rate. However, African American race, high school education, unemployment rate and uninsured rate were found to be able to significantly predict PTB rate. Therefore, the strongest predictor of outcome when considering demographics was African American race (0.28), while uninsured rate was the most significant socioeconomic predictor of the outcome (0.08).

Obesity and STD had a positive correlation with PTB rate (0.3) which was significant. Both obesity and STD were shown to be able to predict the outcome (PTB rate) more accurately than the mean, after adjusting for other variables (p 0.04 for obesity and p < 0.01 for STD). STD was a stronger predictor of PTB compared to obesity (0.07 vs 0.04).

While Adult smoking had a significant positive correlation with PTB rate (0.45), alcohol abuse had a significant negative correlation (– 0.43). After adjusting for other variables, both of these variables were significantly associated with PTB rate (p < 0.01). Smoking was the stronger predictor of the two (0.19 vs – 0.15).

FEI (calculated from 0–10) is the primary independent variable which showed a significant negative correlation with PTB rate, indicating that the lower the FEI value, the higher the PTB rate (– 0.36). Approximately 15% of the variability of the PTB rate can be explained by FEI with a significant correlation (p < 0.001). Air pollution is the second environmental variable with a significant positive correlation, indicating that the higher the pollution in a county, the greater the chance of PTB (0.22) (p < 0.001). Both FEI and air pollution were significant predictors of PTB after adjusting for multiple other predictors, with FEI being the stronger predictor (both with p value < 0.01) (FEI – 0.07 vs air pollution 0.05).

4 Discussion

To the best of our knowledge, this is the first study that attempted to demonstrate an association between Food Environment Index and Preterm Birth rate at the county level in the USA. This study attempted to untangle the complexity of the relationship between PTB and multiple other factors, determine the predictors of PTB, as well as the strength of the associations between them. The result showed that FEI is a predictor of PTB after adjusting for the confounders.

The findings of this study showed a significant variability in PTB rate between the counties. This is consistent with a study conducted by Riley (2019) which showed that pregnant women residing in counties with higher well-being (using Gallup-Sharecare Well-Being Index) have lower PTB rate [11].

The National Center for Health Statistics (NCHS) has classified USA counties into six-levels based on urban/rural communities and population size (Large central metro counties to Micropolitan and Non-core counties) [12]. The type of county and their population size may affect the medical facilities present there and the outcome of pregnancies. No association was found between county population size and PTB rate after adjusting for all other variables.

This study found that the strongest positive demographic predictor of PTB was African American race. The findings are consistent with previous studies. The rate of PTB was twice in African American women compared to NHW race in one study, while another study found that 38% of PTB and 31% of the very PTB can be attributed to the disparity between African Americans and NHWs [13]. The lower socioeconomic status of Africa American women, maternal education, differences in biomarkers, genetic variation, microbiome, neighborhood deprivation, and short interpregnancy interval are believed to be the reason behind the difference [14]. The larger the size of the NHW population in a county, the lower the PTB rate. This is consistent with previous studies that showed that NHW have reduced rate of PTB compared to other races [14].

As the percentage of Hispanic people in counties increased, the PTB rate decreased. Previous studies have verified this association [15]. Although Hispanic race seems to have a lower incidence of PTB compared to other races, one of the reasons for increasing incidence of PTB in the USA in the past several years is increase in the rate of PTB in the Hispanic population [14].

With increase in teen birth in the county, the PTB rate also increased. Teen birth rate also increased with the rate of smoking, alcohol abuse, median household income and uninsured. In this study, teen birth could be predicted accurately by multiple other covariates, an example of multicollinearity. To mitigate the effect of teen birth rate on all the other variables (multicollinearity) in the final model, teen birth rate was not considered in the final step of the multiple regression model. Although the rate of teen pregnancy has been decreasing in the USA, Kawakita (2016) found a 36% increase in PTB rate in adolescents compared to the older counterparts [16]. There are several socioeconomic and demographic factors that are associated with teen birth that also affected the rate of PTB, such as poor education, African American race, rural residence, and poverty, including the uninsured, the unemployed, and the homeless [17].

This study found that the higher the rate of completion of high school, the lower the PTB rate, which is consistent with previous studies. Carmichael (2018) demonstrated higher rate of PTB in peri-viable gestation is associated with lower education and uninsured status [18]. Other studies have also shown an association of lower levels of maternal education and PTB. Liu (2019) in a USA based study found that the rate of PTB was 8.9% in less than high school educated mothers compared to 8.3% and 6.7% in mothers who were educated up to high school or beyond respectively [19].

This study found that higher the uninsured and unemployment rate in the county, the higher the PTB rate. Margerison (2019), in their study based in Michigan in the USA, found that a 1% increase in state unemployment in the first trimester of pregnancy was associated with a 3% increase in PTB odds [20]. Additionally, uninsured and unemployed pregnant women end up with poor prenatal care, which has been proven to be associated with PTB [21].

The results also indicate that the higher the median household income, the lower the rate of PTB in the county. Like teen births, median household income was also found to be a confounder with strong correlation (> ± 0.5) with other covariates: alcohol abuse, teen birth rate, and adult smoking. Median household income was removed from stage 2 of the regression model to minimize the effect of multicollinearity. Median household income has been known to be an indicator of income and poverty. In a study from North Carolina, Coley (2015) showed that mothers who lived in neighborhoods with lower income had a higher chance of delivering a preterm child [22].

Residential segregation pertains to structural, individual and institutional racism, which is measured by the degree to which two or more races (in this case African Americans and White) live separately from each other in a geographic area [23]. This study did not find a relationship between the magnitude of residential segregation and PTB rate. Since the index ranges from 0–100 and gives higher scores to the predominance of one race over another in a geographic area, both NHW and African American predominant areas had higher scores in this study. This may have led to offsetting of the higher PTB rate of African American predominant neighborhoods by the low PTB rate of the predominantly NHW neighborhoods, resulting in a net lack of effect of segregation on PTB. Previous studies, such as by Anthopolos (2011) and colleagues, found that infants born to African American and white parents living in isolated neighborhoods with predominantly African Americans had, on average, reduced birthweight and increased odds of PTB compared to their counterparts in areas with lesser isolation [24].

Counties with a higher rate of obesity also had a higher rate of PTB. Liu (2019), in a large USA population based study, concluded that maternal pre-pregnancy obesity was significantly associated with PTB (18% higher in higher BMI), but the risk is different based on maternal age, race, or ethnicity [19]. The higher the number of new cases of STDs in a county, the higher the rate of PTB. This was consistent with the findings of the study by Baer (2019) showing that higher rate of Gonorrhea and Syphilis during pregnancy, in a population from California, increased the odds of PTB significantly [25]. STDs are mostly asymptomatic, but when left untreated, women have a 3.3 times higher risk for preterm delivery compared to women who have received treatment according to a study [26].

The higher the rate of adult smoking in the county, the higher the PTB rate. Previous research has shown a significant association between smoking and PTB, with the earliest mention in 1957 [27]. The mechanism of this association seems to be multifactorial, such as vasoconstriction, effects of carbon monoxide and cadmium, altered steroid hormone production, and other factors [28]. This study showed that the higher the rate of alcohol abuse in a county, the lower the PTB rate. This finding was contrary to what was expected based on previous research, which demonstrated a positive association of binge drinking (excessive drinking) with PTB rate [29] [30]. In national population-based studies, binge drinking (4 or more drinks on one occasion for women and 5 for men) and heavy drinking (average of > 1 drink per day on an average for women and > 2 drinks per day for men) were associated with white males, single individuals (women or men), younger age, and those who have higher income compared to individuals who did not report these drinking behaviors [31, 32]. Some of these factors, such as NHW population and higher income, are related to decrease in PTB rate. Furthermore, since the survey was conducted over the phone, it is likely that there were more male respondents who have more access to phones, reporting that they consumed alcohol excessively, which may confound the data for the county with the lack of inclusion of the targeted population: women in the childbearing age.

This study demonstrated that when FEI is lower, the PTB rate is higher in a county. Dietary imbalance in the mother and resulting deficiency in certain nutrients (Vitamin E and lipids) have been shown to be associated with PTB [33]. This study aimed at investigating the association between food environment and PTB, since both obesity and uncontrolled gestational diabetes, which are influenced by the food environment, have been shown to be associated with increased risk for PTB [34, 35]. Food environment of an individual has been shown to affect the requirement for medication in gestational diabetes, which has led to adverse maternal outcomes [36]. Therefore, it was essential to explore the association between maternal food environment and PTB after adjusting for all other risk factors that may be considered confounders. Minaker (2013) showed that the distance between the home and the nearest convenience store had a strong association with female BMI and waist circumference in that neighborhood [37]. A systematic review with metanalysis by Torlani (2009) showed that BMI > 35 is associated with an increased risk for PTB [38]. This study showed that the higher degree of air pollution, the higher the PTB rate at the county level. In a USA based systematic review, Bekker (2019) and colleagues showed that exposure to particulate matter 2.5 (PM 2.5) was associated with an increased risk of PTB in 79% of the studies considered [40].

4.1 Limitations

The major limitation of the study is the measurement of the exposure (independent factors) as an average of the entire population and not specific to the target group (women in the reproductive age group who can give birth). This may lead to “Ecological Fallacy,” which results from the inability of linking the outcome with exposure of disease in the same individuals [41]. Responder’s bias may also have overestimated or underestimated the magnitude of the variables. There is also a potential for differences in measurement of exposure in different geographic areas as state laws vary geographically. Individual physical condition (co-morbidities/habits/environments/mental health) were not considered which may interact with the exposure acting as confounders. The study was restricted to include whichever variables were available and may be missing confounders that are relevant for study analysis such as: obstetric facilities, trained staff, home births etc.

5 Conclusions

This ecological study confirms the multifactorial etiology with a complex relationship of several factors with the outcome of PTB rate. The PTB rate had a significant association with FEI after accounting for potential confounders. Sociodemographic factors, maternal behavior, and environmental factors affected PTB rate at the county level. Future quantitative studies are required to examine the relationship between FEI and PTB rate with a large number of human subjects to confirm the findings of this research.

Footnotes

Acknowledgments

The March of Dimes and University of Wisconsin Population Health Institute for giving us permission to use their data.

Disclosure statement

The authors have no potential conflicts of interest to disclose, no funding was received for this study and consent was not required for the conduct of this study.

Appendix 1

The data source and the year data were collected

Variable	Primary data source	Year of data collection
2019 Population	USCB	2019
Unemployment	Bureau of Labour statistics	2018
African American population	USCB	2019
Adult smoking	Behavioural Risk Factor Surveillance System: Self-reported annual survey	2018
Obesity	United States Diabetes Surveillance System	2017
Uninsured	US Census Bureau’s Small Area Health Insurance Estimates (SAHIE) program	2018
FEI County Value	United States Department of Agriculture (USDA) Food Environment Atlas, Map the Meal Gap from Feeding America	2015 and 2018
Alcohol Abuse	Telephone based survey conducted annually by Behavioral Risk Factor Surveillance System	2018
STD	National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention	2018
Teen Birth	National Center for Health Statistics - Natality files	2013–2019
High school education	American Community Survey, 5-year estimates	2015–2019
Air Pollution	CDC’s National Environmental Public Health Tracking Network website	2016
Median household Income (in dollars)	The US Census Bureau’s Small Area Income and Poverty Estimates (SAIPE) program	2019

USCB: United States Census Bureau; PTB: Preterm birth, NHW: Non-Hispanic white, FEI: Food Environment Index, STD: Sexually Transmitted Diseases, SAHIE: Small Area Health Insurance Estimates, CDC: Centers for Disease Control and Prevention, SAIPE: Small Area Income and Poverty Estimates, USDA: United States Department of Agriculture.

Appendix 2

References

WHO. Preterm birth [Internet]. 2018 [cited 2021 Apr 9]. Available from: https://www.who.int/news-room/fact-sheets/detail/preterm-birth.

Rita

, Malosso

, Saccone

, Simonetti

, Squillante

, Berghella

US trends in abortion and preterm birth. The Journal of Maternal-Fetal & Neonatal Medicine. 2018;31:2463–7.

Patel

Short- and Long-Term Outcomes for Extremely Preterm Infants. Vol. 33, American Journal of Perinatology. Thieme Medical Publishers, Inc.; 2016. pp. 318–28.

Jarjour

Neurodevelopmental outcome after extreme prematurity: A review of the literature. Vol. 52, Pediatric Neurology. Elsevier Inc.; 2015;143–52.

Behrman

, Butler

, Outcomes I of M (US) C on UPB and AH. Sociodemographic and community factors contributing to preterm birth. 2007;

Chia

, Chen

, Lai

, Wong

, Neelakantan

, Van Dam

, et al. Maternal dietary patterns and birth outcomes: A systematic review and meta-analysis. Vol. 10, Advances in Nutrition. Oxford University Press; 2019. pp. 685–95.

Rose

, Hutchinson

, Bodor

, Swalm

, Farley

, Cohen

, et al. Neighborhood food environments and body mass index. The importance of in-store contents. Am J Prev Med. 2009;37(3):214–9.

March of Dimes. Preterm United States | PeriStats | March Of Dimes [Internet]. 2018 [cited 2020 Aug 11]. Available from: https://www.marchofdimes.org/Peristats/ViewSubtopic.aspx?reg=99&top=3&stop=60&lev=1&slev=1&obj=1.

UWPHI. The University of Wisconsin Population Health Institute. County Health Rankings & Roadmaps, 2020 [Internet]. countyhealthrankings.org. 2020 [cited 2020 Jun 26]. Available from: https://www.countyhealthrankings.org/explore-health-rankings.

10.

Hamilton

, Martin

, Osterman

MJK

Vital statistics rapid release births: Provisional data for 2019. 2019.

11.

Riley

, Roy

, Herrin

, Spatz

, Silvestri

, Arora

, et al. Do pregnant women living in higher well-being populations in the USA experience lower risk of preterm delivery A cross-sectional study. BMJ Open. 2019;9(4).

12.

Mohamoud

, Kirby

, Ehrenthal

Poverty, urban-rural classification and term infant mortality: A population-based multilevel analysis. BMC Pregnancy Childbirth. 2019;19(1).

13.

Schaaf

, Liem

SMS

, Mol

BWJ

, Abu-Hanna

, Ravelli

ACJ

Ethnic and racial disparities in the risk of preterm birth: A systematic review and meta-analysis. Vol. 30, Am J Perinatol. 2013;30:433–50.

14.

Manuck

. Racial and ethnic differences in preterm birth: A complex, multifactorial problem. Vol. 41, Seminars in Perinatology. W.B. Saunders; 2017;511–8.

15.

Bediako

, BeLue

, Hillemeier

A Comparison of Birth Outcomes Among Black, Hispanic, and Black Hispanic Women. J Racial Ethn Health Disparities. 2015;2(4):573–82.

16.

Kawakita

, Wilson

, Grantz

, Landy

, Huang

, Gomez-Lobo

Adverse Maternal and Neonatal Outcomes in Adolescent Pregnancy. J Pediatr Adolesc Gynecol. 2016;29(2):130–6.

17.

Kappeler

. Adolescent health and teen pregnancy in the United States: A progress report. Vol. 130, Public Health Reports. Association of Schools of Public Health; 2015;196–8.

18.

Carmichael

, Blumenfeld

, Mayo

, Profit

, Shaw

, Hintz

, et al. Stillbirth and Live Birth at Periviable Gestational Age: A Comparison of Prevalence and Risk Factors. Am J Perinatol. 2019;36(5):537–44.

19.

Liu

, Xu

, Sun

, Du

, Gao

, Snetselaar

, et al. Association between maternal pre-pregnancy obesity and preterm birth according to maternal age and race or ethnicity: a population-based study. Lancet Diabetes Endocrinol. 2019;7(9):707–14.

20.

Margerison

, Luo

, Li

Economic conditions during pregnancy and preterm birth: A maternal fixed-effects analysis. Paediatr Perinat Epidemiol. 2019;33(2):154–61.

21.

Maupin

, Lyman

, Fatsis

, Prystowiski

, Nguyen

, Wright

, et al. Characteristics of women who deliver with no prenatal care. Journal of Maternal-Fetal and Neonatal Medicine. 2004;16(1):45–50.

22.

Coley

, Nichols

, Rulison

, Aronson

, Brown-Jeffy

, Morrison

Race, socioeconomic status, and age: Exploring intersections in preterm birth disparities among teen mothers. Int J Popul Res. 2015;1–10.

23.

Williams

Racial Residential Segregation: A fundamental cause of racial disparities in health. Public Health Rep. 2001;116(5):404–16.

24.

Anthopolos

, James

, Gelfand

, Miranda

A spatial measure of neighborhood level racial isolation applied to low birthweight, preterm birth, and birthweight in North Carolina. Spat Spatiotemporal Epidemiol. 2011;2(4):235–46.

25.

Baer

, Chambers

, Ryckman

, Oltman

, Rand

, Jelliffe-Pawlowski

An evaluation of sexually transmitted infection and odds of preterm or early-term birth using propensity score matching. Sex Transm Dis. 2019;46(6):389–94.

26.

Olaleye

, Babah

, Osuagwu

, Ogunsola

, Afolabi

Sexually transmitted infections in pregnancy-An update on Chlamydia trachomatis and Neisseria gonorrhoeae. Eur J Obstet Gynecol. 2020;255:1–12.

27.

Simpson

A preliminary report on cigarette smoking and the incidence of prematurity - PubMed. 1957.

28.

Ion

, Bernal

Smoking and preterm birth. Reprod Sci. 2015;22(8):918–26.

29.

O’Leary

, Nassar

, Kurinczuk

, Bower

The effect of maternal alcohol consumption on fetal growth and preterm birth. BJOG. 2009;116(3):390–400.

30.

Ikehara

, Kimura

, Kakigano

, Sato

, Iso

, Saito

, et al. Association between maternal alcohol consumption during pregnancy and risk of preterm delivery: the Japan Environment and Children’s Study. BJOG. 2019;126(12):1448–54.

31.

Naimi

, Brewer

, Mokdad

, Denny

, Serdula

, Marks

Binge drinking among US adults. J Am Med Assoc. 2003;289(1):70–5.

32.

Serdula

, Brewer

, Gillespie

, Denny

, Mokdad

Trends in alcohol use and binge drinking, -Results of a multi-state survey. Am J Prev Med. 2004;26(4):294–8.

33.

Zhang

, Zhou

, Perkins

, Wang

, Sun

Maternal dietary nutrient intake and its association with preterm birth: A case-control study in Beijing, China. Nutrients. 2017;9(3).

34.

Marchi

, Berg

, Dencker

, Olander

, Begley

Risks associated with obesity in pregnancy, for the mother and baby: A systematic review of reviews. Obesity Reviews. 2015;16(8):621–38.

35.

Billionnet

, Mitanchez

, Weill

, Nizard

, Alla

, Hartemann

, et al. Gestational diabetes and adverse perinatal outcomes from 716,152 births in France in Diabetologia. 2017;60(4):636–44.

36.

Fonge

, Jain

, Harrison

, Brooks

, Sciscione

Examining the relationship between food environment and gestational diabetes. Am J Obstet Gynecol MFM. 2020;2(4):100204.

37.

Minaker

, Raine

, Wild

, Nykiforuk

CIJ

, Thompson

, Frank

Objective food environments and health outcomes. Am J Prev Med. 2013;45(3):289–96.

38.

Torloni

, Betrán

, Daher

, Widmer

, Dolan

, Menon

, et al. Maternal BMI and preterm birth: A systematic review of the literature with meta-analysis. Journal of Maternal-Fetal and Neonatal Medicine. 2009;22(11):957–70.

39.

Bekkar

, Pacheco

, Basu

, Denicola

Association of Air Pollution and Heat Exposure with Preterm Birth, Low Birth Weight, and Stillbirth in the US: A Systematic Review. JAMA Netw Open. 2020;3(6).

40.

Neumark

What can ecological studies tell us about death? Vol. 6, Israel Journal of Health Policy Research. BioMed Central Ltd.; 2017.