Healthy Food Density is Not Associated with Diet Quality among Pregnant Women with Overweight/Obesity in South Carolina

Abstract

Objective:

Examine the association and moderating effect of residential location (urban/rural) on the relationship between neighborhood healthy food density and diet quality.

Design:

Cross-sectional analysis of baseline data from the Health in Pregnancy and Postpartum (HIPP) study, a randomized trial designed to prevent excessive gestational weight gain.

Participants:

Pregnant women in South Carolina with pre-pregnancy overweight/obesity (n=228).

Main outcome measures:

Healthy Eating Index (HEI)-2015 was used to measure diet quality from two 24-hour dietary recalls. HEI-2015 total scores and 11 binary HEI-2015 components (met standard for maximum component score vs. not) were calculated as dependent variables.

Analysis:

Multiple linear and logistic regression models were used to examine the association between healthy food density and HEI-2015 total scores and meeting the standards for maximum component scores. Healthy food density*residential location tested for moderation. P < .05 indicated significance.

Results:

Participants’ diet quality was suboptimal (M=52.0 ± 11.7, range 27–85). Healthy food density was not significantly related to HEI-2015 total scores or components and residential location was not a moderator.

Conclusions and Implications:

Diet quality was sub-optimal and there was no relationship between healthy food density and diet quality among HIPP participants. These data support examining behavioral factors that could influence diet quality.

Introduction

Poor diet quality during pregnancy is an important public health problem due to its widespread nature and adverse effects on maternal and offspring health.¹ Over half of U.S. women (55%) begin pregnancy overweight (body mass index [BMI] 25.0–29.9 kg/m²) or obese (BMI ≥ 30 kg/m²).² Additionally, almost one-half (47%) of all pregnant women in the U.S. exceed the Institute of Medicine’s 2009 weight gain recommendations during pregnancy, with higher rates of excessive gestational weight gain (GWG) (45%−64%) among women who begin pregnancy overweight or obese.³ Energy-dense, nutrient-poor diets may be an important factor contributing to excessive GWG⁴ and postpartum weight retention in women,⁵ and greater newborn adiposity⁶ and overweight in childhood.⁷ Diet quality among U.S. women during pregnancy falls short of national recommendations in the Dietary Guidelines for Americans (DGAs),¹ so there is a need to better understand the determinants of diet quality in pregnancy. Furthermore, the Scientific Report of the 2020 DGA Committee focuses on dietary patterns and highlights the importance of healthy dietary patterns during pregnancy for improving maternal and child health outcomes.⁸

Community nutrition environments, also commonly referred to as neighborhood food environments, encompass the distribution of food sources which includes the number, type, location, and accessibility of food retailers available to the general population.⁹ The neighborhood food environment has been increasingly investigated as a factor influencing dietary intake, overall diet quality, and obesity among the general U.S. population.¹⁰ Research has found that individuals living in the lowest-ranked food environments (based on supermarket density within 1 mile of participants’ homes, and participants’ and non-participants’ self-reported availability of healthy food) are 22–35% less likely to have healthy diet quality, compared to those in the best-ranked food environments among non-pregnant adults;¹¹ however, there is a paucity of research investigating the relationship between the neighborhood food environment and diet quality in pregnant women.¹² It is theorized that individuals are more likely to engage in healthier behaviors when they are in supportive environments,¹³ so poor access to nutrient-dense food may act as a barrier to improving diet quality during the important period of pregnancy.¹⁴ There is variability in the way neighborhood food access is defined, such as proximity or density of food outlets, which contributes to a largely inconclusive body of evidence on the relationship between food environments and dietary intake.¹⁵ Researchers have called for greater standardization in the methods used to define neighborhood food access. The Centers for Disease Control’s (CDC) Modified Retail Food Environment Index (mRFEI) is a standard way of assessing the food environment by calculating the percentage of food retailers considered “healthy” out of the total number of food retailers in a Census tract.¹⁶

It is important to consider how the relationship between the food environment and diet quality could differ between urban and rural areas. Individuals living in rural areas face additional barriers that could negatively impact diet quality such as traveling longer distances to buy groceries, having less independent access to a vehicle, having greater reliance on car ownership to get to stores, less access to public transportation, and going grocery shopping less frequently.^17,18 Variety, quality, and price of fresh produce have been found to be barriers to consuming fruits and vegetables in rural communities.¹⁹ There is typically greater variety of food in larger food retailers, which are more common in urban areas.¹⁹ Produce in smaller food stores, which are typical of rural environments, is often of lower quality (less fresh) than larger grocery stores more commonly located in urban areas.²⁰ Additionally, larger food retailers in urban areas are known to have better food prices²⁰ which may influence purchasing decisions and subsequently support better diet quality. For these reasons, it is important to examine how the relationship between healthy food density and diet quality may differ between individuals living in urban vs. rural areas.

The aims of the current study were to 1) examine the association between healthy food density (via the CDC’s mRFEI) and diet quality (via Healthy Eating Index (HEI)-2015 total scores and meeting the standards for maximum HEI-2015 component scores) and 2) examine residential location as a moderator in the relationship between healthy food density and diet quality. It was hypothesized that an increase in healthy food density would be associated with higher HEI-2015 total scores and higher odds of meeting the standards for maximum HEI-2015 component scores overall and for urban compared to rural women. Residential location was tested as a moderator.

Methods

The Health in Pregnancy and Postpartum (HIPP) study is a randomized controlled trial examining the efficacy of a theory-based behavioral lifestyle intervention to reduce excessive GWG among White and African-American (AA) pregnant women with overweight and obesity, as compared to a standard care intervention. This paper reports a cross-sectional analysis of demographic, food environment, and dietary data measured at baseline (n=228). Baseline assessments were conducted from January 2015 to January 2019. Participants had to complete all baseline measures before they were randomized, so this analysis includes women with complete baseline data.

Women were recruited to participate in the study primarily through 13 obstetrics and gynecology (OB/GYN) clinics in the greater Columbia, South Carolina (SC) area and adjacent counties, with some self-referrals in response to community and social media advertisements. Women were eligible if they: (a) were between 18–44 years of age, (b) self-identified as White or Black/AA (possible choices: White, Black or AA, Asian, Native Hawaiian or Pacific Islander, or American Indian or Alaska Native), (c) could read and speak English, (d) had no plans to move outside of the geographic area in the next 18 months, (e) were ≤ 16 weeks gestation, and (f) had a pre-pregnancy body mass index (BMI) ≥ 25 kg/m² and a pre-pregnancy weight ≤ 167.8 kg. Women were excluded if they had contraindications to physical activity during pregnancy.²¹ Institutional Review Boards at Palmetto Health, University of South Carolina, Lexington Medical Center, and the Medical University of South Carolina approved the study protocol. All participants provided written informed consent.

Measures

At the baseline visit, demographic data, home addresses, and anthropometric measures were collected. The demographic questionnaires and anthropometric measures were interviewer-administered, while the 24-hour dietary recalls were self-administered. Baseline demographic variables were categorized as follows: age (18–24 years, 25–29 years, 30–34 years, or 35–42 years), race (White or AA/Black), education (high school diploma/GED or less, some college, or college degree or higher), parity (nulliparous or multiparous), marital status (married or not married), enrollment in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) (yes or no), pre-pregnancy weight status (overweight or obese), and self-reported distance traveled in miles to buy groceries. Pre-pregnancy BMI was calculated from participants’ self-reported pre-pregnancy height and weight.

Neighborhood food environment

South Carolina food retailer addresses were acquired from ReferenceUSA, a commercial database of U.S. businesses, in December 2017.²² Retailers were categorized based on North American Industry Classification System (NAICS) codes. The categories of interest included: grocery stores/supermarkets (Group 445110), convenience stores (445120), gas stations with food marts (447110), drug stores (446110), discount merchandise stores (452319), and limited-service restaurants (722513). Limited-service restaurants are where customers order and pay before eating, the food is typically served quickly after ordering, and the food is kept cold, cooked in advance, and/or reheated (e.g., fast-food).²³ Food retailers and participants’ home addresses were geocoded to the point or street address level using the ArcGIS Online World Geocoding Service address locator in ArcGIS Pro, version 1.2 (Esri, Inc., Redlands, CA, 2016).²⁴

The neighborhood food environment was determined by calculating the 5-mile network distance from participants’ homes using the “Network Analyst” tool. While the location and frequency of participants’ grocery shopping was not collected, participants were asked how far they typically traveled in miles to buy groceries. The 5-mile distance was based on the average distance participants reported traveling to buy groceries across urban and rural areas. This average distance may capture a portion of retailers that participants frequent. Five-mile network buffers were created around each participant’s home. Food retailers that were contained in each buffer were clipped and summed for use in the mRFEI formula. Within the 5-mile network areas, there were grocery stores (n=182), convenience stores (n=457), drug stores (n=84), discount merchandise stores (n=150), and limited-service restaurants (n=580) that resulted in a total of 1,453 retailers that were included in the analyses.

The mRFEI combines the concepts of food deserts (i.e., areas with poor access to supermarkets) with the concept of food swamps (i.e., areas with a high amount of unhealthy food) into a single score at the census-tract level.¹⁶ The current study calculated healthy food density scores at the individual-level based on HIPP participants’ home addresses. Healthy food density scores were calculated by dividing the total number of healthy food retailers by the total number of “healthy” and “less healthy” food retailers, and then multiplying by 100 to get a percentage. mRFEI scores range from 0 (no food retailers that typically sell healthy food) to 100 (only food retailers that sell healthy food). According to the CDC, “healthy” retailers include supermarkets, larger grocery stores, supercenters, and produce stores, while “less healthy” retailers include establishments such as convenience stores and limited-service restaurants.¹⁶ These classifications are based off of typical food offerings in these types of retailers, with “healthy” retailers offering a variety of nutrient-dense foods such as fruits and vegetables, low-fat dairy items, meat products, and whole grain foods.²⁵

Urban and rural areas were determined by the Census Bureau’s 2017 Urban Areas Boundary file.²⁶ Rural areas include all populations and areas not included within an urban area. Participants’ addresses were spatially joined to associated urban area boundaries. Participants’ addresses that fell within urban areas were categorized as urban and those outside urban areas were categorized as rural.

Diet quality

Participants completed 2 unannounced 24-hour dietary recalls (1 weekday and 1 weekend day, which included Fridays) at baseline through the National Cancer Institute (NCI)’s Automated Self-Administered 24-hour Dietary Recall (ASA24) online system.²⁷ For the first recall, measurement staff explained to participants how to complete the recall using the computer on-site and were there to answer any questions. For the second recall, participants completed the recall using their computer or mobile device. If participants did not have access to the Internet, measurement staff conducted an interviewer-administered recall over the phone and entered their recall into ASA24. The ASA24 is a web-based dietary assessment tool that provides complete nutrient analysis of all foods and beverages reported during the data collection timeframe.²⁷ The 2 dietary recalls per participant were averaged and then scored using the simple HEI-2015 scoring algorithm method via SAS code provided by the NCI²⁸ to generate HEI-2015 scores, which measure adherence to the 2015 DGAs.²⁹

The HEI-2015 includes 13 components, including 9 adequacy components (i.e., Total Fruits, Whole Fruits, Total Vegetables, Greens and Beans, Whole Grains, Dairy, Total Protein Foods, Seafood and Plant Proteins, and Fatty Acids), which are dietary components that need to be increased. There are 4 moderation components (i.e., Refined Grains, Sodium, Added Sugars, and Saturated Fats), which are dietary components that need to be reduced. All components are scored on a density basis out of 1,000 calories, with the exception of Fatty Acids, which is a ratio of unsaturated to saturated fatty acids.³⁰ For each component, higher scores reflect greater adherence to the DGAs. Component scores are summed to create a total score that can reach a maximum of 100 points, with higher scores indicating better diet quality. HEI-2015 total scores were analyzed as a continuous variable while HEI-2015 components were analyzed as dichotomous outcomes based on meeting the standards for maximum HEI-2015 component scores or not.

Statistical Analyses

Descriptive statistics (i.e., means, standard deviations (SD), and percentages) were used to summarize participants’ sociodemographic characteristics, food environment variables (proximity to food retailers, self-reported distance for grocery shopping, and healthy food density scores), and diet quality (i.e., HEI-2015 total scores and components) at baseline. Independent samples t-tests were used to test for mean differences in continuous variables (e.g., age, parity, gestational age, healthy food density scores, HEI-2015 total scores, HEI-2015 component scores) by residential location. The Pearson’s χ² test was used to examine differences in the proportion of categorical characteristics (e.g., marital status, education level, and pre-pregnancy weight status) by residential location.

Multiple linear regression models determined the association between healthy food density scores (independent variable) and HEI-2015 total scores (dependent variable), and the association between healthy food density scores (independent variable) and meeting the standards for maximum HEI-2015 component scores (dependent variable) as secondary outcomes. A multiplicative interaction term (healthy food density*residential location) was used to test healthy food density as a moderator. Beta coefficients and standard errors are presented along with estimated odds ratios (ORs) and 95% confidence intervals (CIs).

The Whole Grains and Sodium components could not be analyzed due to the small cell size of participants who met the standards for maximum HEI-2015 component scores. Models adjusted for maternal race, educational attainment, age, marital status, parity, WIC enrollment, and pre-pregnancy BMI. These potential confounders were chosen a priori based on existing literature.^12,14 A P-value <0.05 indicated statistical significance. Statistical analyses were performed using SAS^® software, version 9.4 (SAS Institute, Inc., Cary, NC, 2013).³¹

Results

Study population

Participants (n=228) were racially-diverse (56% White, 44% AA), primarily married (67%), more than a third were 30–34 years old (37%), and almost a quarter of women (24%) were enrolled in WIC (Table 1). The sample was well-educated because most women (59%) earned a college degree or higher. More than half of women had at least one child (57%) and just over half (52%) had a BMI ≥ 30 prior to pregnancy. The mean gestational age was 12.5 weeks (±2.4 weeks). Additionally, most women (85%) lived in urban areas at baseline. In terms of demographic differences by race, AA women were slightly younger (M=28.6 ± 5.5. years vs. 30.5 ± 4.5 years) and had a higher gestational age at baseline (M=13.0 ± 2.6 weeks vs. 12.2 ± 2.2 weeks) compared to White women. Over half of AA women were not married (55.4% vs. 15.8%) and just under half had a college degree or higher (49.5% vs. 66.9%), and more AA women were enrolled in the WIC program (40.6% vs. 10.2%) compared to White women (not pictured). There were no significant differences in demographic characteristics by residential location; however, there were significant urban vs. rural differences in food environment characteristics (Table 2).

Table 1.

HIPP^a participants’(n=228) baseline demographic and psychosocial characteristics in early pregnancy by residential location.

Characteristic	Total (n=228), n	%	Urban (n=193; 84.6%), %	Rural (n=35; 15.3%), %	p-value
Race, %					0.10
White	127	55.7	53.4	68.6
African-Americanb	101	44.3	46.6	31.4
Age, %					0.92
18–24 years	39	17.1	17.1	17.1
25–29 years	63	27.6	26.9	31.5
30–34 years	85	37.3	37.3	37.1
35–42 years	41	18.0	18.7	14.3
Marital Status, %					0.34
Married	152	66.7	65.3	74.3
Not married	76	33.3	34.7	25.7
Education level, %					0.86
High school or less	28	12.3	11.9	14.3
Some college	65	28.5	29.0	25.7
College degree/higher	135	59.2	59.1	60.0
WICc, %					0.67
Enrolled (parent and/or child receives food)	54	23.7	24.3	20.0
Not enrolled	174	76.3	75.7	80.0
Parity					0.27
Nulliparous	98	43.0	44.6	34.3
Multiparous	130	57.0	55.4	65.7
Pre-pregnancy weight statusd, %					0.27
Overweight (BMIe 25.0–29.9 kg/m2)	110	48.2	46.6	57.1
Obese (BMI ≥ 30 kg/m2)	118	51.8	53.4	42.9
Group randomization, %					0.71
Intervention	114	50.0	50.8	45.7
Standard Care	114	50.0	49.2	54.3
Characteristic, Mean ± SD		Total	Urban	Rural	p-value
Age (years), range 18–42		29.7 ± 5.1	29.8±5.1	29.1±4.8	0.49
Gestational age (weeks), range 7–20		12.5 ± 2.4	12.4±2.3	13.2±2.7	0.09

The χ² test was used to examine differences in the proportion of categorical characteristics by residential location. Independent samples t tests were used to test for mean differences in continuous demographic characteristics by residential location.

p < 0.05 indicated statistical significance.

^aHealth in Pregnancy and Postpartum

^bIncludes two participants who indicated both African-American and White as their race.

^cSpecial Supplemental Nutrition Program for Women, Infants, and Children

^dBased upon self-reported pre-pregnancy height and weight.

^eBody mass index

Table 2.

Summary of food environment characteristics by residential location among HIPP^a participants (n=228).

	Total (n=228)	Urban (n=193)	Rural (n=35)	p-valueb
	Mean±SD (Range)	Mean±SD	Mean±SD
Distance to nearest grocery store, miles	1.3±1.2 (0.1–7.2)	1.0±0.7	3.0±1.9	<0.0001
Distance to nearest convenience store, miles	0.8±0.7 (0.0–3.3)	0.7±0.5	1.6±0.9	<0.0001
Distance to nearest limited-service restaurant, miles	1.2±1.3 (0.0–8.5)	0.9±0.7	3.2±1.9	<0.0001
Self-reported distance for grocery shopping, miles	5.8±5.2 (1.0–37.0)	4.9±3.8	10.9±8.2	0.0001
5-mile healthy food densityc, %	12.2±5.9 (0.0–43.0)	12.3±4.8	11.5±10.0	0.63

^aHealth in Pregnancy and Postpartum

^bIndependent samples t tests were used to test for mean differences in food environment characteristics by residential location.

^cHealthy food density represents the percentage of healthy food retailers out of the total number of retailers (both healthy and less healthy).

Healthy food density and HEI-2015 Scores

Overall, HIPP participants’ diet quality was sub-optimal (M=52.0 ± 11.7, range 27–85). Healthy food density was not significantly related to HEI-2015 total scores [adjusted β (SE): −0.22 (0.14), p=0.11] (Table 3).

Table 3.

Adjusted linear regression models of baseline associations between 5-mile healthy food density and Healthy Eating Index-2015 (HEI) total scores, HIPP^a study (n=228)

	Model 1	Model 2	Model 3
	β (SE)	β (SE)	β (SE)
5-mile healthy food densityb	−0.24 (0.13)	−0.22 (0.14)	−0.29 (0.20)
Urban (refc: Rural)		3.51 (2.17)	2.06 (3.86)
Healthy food density*urban			0.12 (0.27)
Control variables
Education level
High school or less		−2.21 (2.71)	−2.18 (2.72)
Some college (ref: College degree or higher)		−1.82 (1.94)	−1.79 (1.94)
Race
Black (ref: White)		2.80 (1.77)	2.78 (1.78)
Age
18–24 years		−2.21 (3.12)	−2.22 (3.12)
25–29 years		−1.93 (2.46)	1.87 (2.46)
30–34 years (ref: 35–42 years)		−0.83 (2.24)	−0.86 (2.25)
Marital Status
Not married (ref: Married)		−2.66 (2.08)	−2.63 (2.08)
Parity
Multiparous (ref: Nulliparous)		0.98 (1.71)	0.97 (1.71)
Proxy for income
Enrolled in WICd (ref: Not enrolled)		0.55 (2.24)	0.38 (2.27)
Pre-pregnancy weight status
Obese (ref: Overweight)		0.45 (1.60)	0.46 (1.60)

^aHealth in Pregnancy and Postpartum

^bHealthy food density represents the percentage of healthy food retailers out of the total number of retailers (both healthy and less healthy).

^cReference group

^dSpecial Supplemental Nutrition Program for Women, Infants, and Children

Healthy food density scores are continuous, ranging from 0–50.

Linear regression models were used to examine the relationship between 5-mile healthy food density and HEI total scores.

Model 1: crude model examining the relationship between healthy food density and HEI total scores.

Model 2: adjusted model including covariates.

Model 3: adjusted model including healthy food density*urban interaction term and other covariates.

^*= p < 0.05

In terms of HEI-2015 components, less than 10% of participants met the standard for a maximum HEI-2015 component score for Whole Grains (3%), Fatty Acids (10%), Sodium (0.4%), and Saturated Fats (8%). Most participants (63%) met the standard for a maximum Total Protein foods component score. Approximately a third of participants met the standard for a maximum component score for Total Vegetables (31%), Greens and Beans (35%), and Seafood and Plant Proteins (31%). About a quarter of participants met the standard for a maximum component score for Total Fruits (25.0%) and Added Sugars (24%). Approximately 40% of participants met the standard for a maximum component score for Whole Fruits. Additionally, healthy food density was not significantly related to the odds of meeting the standards for maximum HEI-2015 component scores (Table 4). Residential location did not moderate these relationships (P-values all > 0.05).

Table 4.

Adjusted logistic regression models of baseline associations between 5-mile healthy food density and achieving maximum Healthy Eating Index-2015 (HEI) component scores, Health in Pregnancy and Postpartum study (n=228)

	Total Vegetables	Greens & Beans	Total Fruits	Whole Fruits	Dairy	Total Protein Foods	Seafood & Plant Proteins	Fatty Acids	Refined Grains	Saturated Fats	Added Sugars
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
5-mile healthy food density	1.00 (0.93, 1.08)	0.98 (0.90, 1.06)	0.98 (0.90, 1.06)	0.97 (0.90, 1.05)	1.00 (0.92, 1.09)	1.02 (0.95, 1.09)	1.02 (0.94, 1.10)	1.01 (0.82, 1.24)	1.03 (0.94, 1.13)	0.98 (0.83, 1.16)	1.05 (0.94, 1.18)
Urban (refb: Rural)	4.34 (0.84, 22.26)	1.08 (0.26, 4.50)	1.06 (0.23, 4.84)	0.79 (0.19, 3.31)	0.44 (0.09, 2.22)	2.16 (0.56, 8.31)	2.53 (0.52, 12.41)	15.62 (0.45, 541.51)	2.76 (0.40, 18.95)	1.09 (0.08, 14.91)	5.79 (0.55, 61.08)
Healthy food density*urban	0.89 (0.78, 1.00)	1.04 (0.94, 1.16)	1.01 (0.90, 1.12)	1.03 (0.93, 1.13)	1.03 (0.92, 1.15)	0.98 (0.89, 1.08)	0.95 (0.84, 1.06)	0.88 (0.68, 1.14)	0.93 (0.81, 1.06)	1.04 (0.86, 1.26)	0.97 (0.84, 1.11)
Control variables
Education level
High school or less	0.75 (0.26, 2.19)	1.02 (0.38, 2.73)	0.62 (0.21, 1.82)	2.92 (1.06, 8.07)	2.29 (0.76, 6.91)	1.94 (0.69, 5.48)	1.38 (0.48, 3.93)	0.94 (0.18, 4.84)	0.85 (0.25, 2.94)	0.37 (0.04, 3.55)	0.43 (0.12, 1.61)
Some college (ref: College degree or higher)	0.84 (0.40, 1.74)	0.79 (0.38, 1.61)	0.70 (0.32, 1.52)	1.14 (0.56, 2.31)	0.77 (0.29, 2.01)	0.99 (0.50, 1.97)	0.66 (0.31, 1.44)	1.24 (0.39, 3.95)	0.94 (0.39, 2.23)	1.82 (0.59, 5.65)	1.18 (0.54, 2.58)
Race
Black (ref: White)	1.47 (0.75, 2.90)	1.47 (0.77, 2.78)	1.13 (0.56, 2.27)	0.90 (0.48, 1.72)	0.77 (0.33, 1.78)	2.00 (1.05, 3.83)	1.75 (0.89, 3.46)	4.08 (1.30, 12.82)*	2.69 (1.21, 5.97)	0.99 (0.31, 3.19)	1.88 (0.92, 3.84)
Age
18–24 years	0.39 (0.12, 1.31)	1.52 (0.50, 4.63)	1.59 (0.46, 5.48)	0.29 (0.09, 0.93)*	0.55 (0.13, 2.40)	0.94 (0.31, 2.89)	0.23 (0.07, 0.82)*	1.11 (0.20, 6.17)	0.65 (0.16, 2.70)	3.12 (0.23, 43.17)	0.37 (0.10, 1.38)
25–29 years	0.64 (0.26, 1.59)	1.36 (0.57, 3.24)	1.49 (0.54, 4.08)	0.71 (0.30, 1.65)	0.66 (0.21, 2.13)	0.87 (0.37, 2.07)	0.44 (0.18, 1.09)	0.51 (0.11, 2.36)	0.51 (0.16, 1.59)	4.82 (0.53, 43.99)	0.64 (0.24, 1.72)
30–34 years (ref: 35–42 years)	0.84 (0.37, 1.91)	0.71 (0.32, 1.59)	1.24 (0.48, 3.18)	0.65 (0.30, 1.42)	0.96 (0.35, 2.64)	1.38 (0.62, 3.08)	0.67 (0.31, 1.49)	0.73 (0.20, 2.66)	1.37 (0.54, 3.48)	3.94 (0.46, 33.83)	0.67 (0.27, 1.63)
Marital Status
Not married (ref: Married)	1.14 (0.52, 2.50)	0.57 (0.26, 1.24)	1.07 (0.47, 2.40)	0.50 (0.23, 1.09)	0.86 (0.32, 2.33)	0.91 (0.43, 1.96)	0.45 (0.20, 1.03)	0.66 (0.20, 2.16)	0.93 (0.38, 2.25)	0.81 (0.23, 2.80)	0.78 (0.34, 1.79)
Parity
Multiparous (ref: Nulliparous)	0.74 (0.39, 1.40)	1.43 (0.77, 2.67)	1.02 (0.52, 2.02)	1.23 (0.66, 2.28)	0.75 (0.34, 1.68)	1.02 (0.56, 1.89)	0.71 (0.36, 1.36)	1.54 (0.53, 4.53)	1.54 (0.71, 3.34)	1.46 (0.48, 4.46)	0.57 (0.28, 1.15)
Proxy for income
Enrolled in WICc (ref: Not enrolled)	1.24 (0.52, 2.97)	0.85 (0.37, 1.94)	2.02 (0.85, 4.83)	0.80 (0.34, 1.88)	1.77 (0.62, 5.09)	0.90 (0.39, 2.07)	1.11 (0.44, 2.81)	1.66 (0.47, 5.80)	1.08 (0.40, 2.90)	2.22 (0.57, 8.59)	2.13 (0.84, 5.45)
Pre-pregnancy weight status
Obese (ref: Overweight)	0.97 (0.54, 1.77)	0.87 (0.49, 1.55)	1.18 (0.63, 2.22)	1.49 (0.83, 2.65)	0.95 (0.45, 2.00)	1.00 (0.56, 1.77)	1.46 (0.79, 2.70)	0.80 (0.31, 2.10)	0.55 (0.27, 1.12)	0.38 (0.13, 1.10)	0.93 (0.48, 1.80)

^aAdequacy components- dietary components that should be increased.

^bReference group

^cSpecial Supplemental Nutrition Program for Women, Infants, and Children

^*p < 0.05

^dModeration components- dietary components that should be consumed in moderation.

Whole Grains and Sodium component could not be analyzed due to the small cell size of participants who met the sodium intake recommendation.

Discussion

This study found that HIPP participants had sub-optimal diet quality overall and healthy food density was not associated with overall diet quality. Previous studies that have examined the relationship between food retailer density and diet quality show conflicting findings across pregnant and non-pregnant samples.^10,11,14

The current study’s results are consistent with the null findings of Laraia et al. (2004), who examined the relationship between the density of multiple food outlets (i.e., supermarkets, grocery stores, and convenience stores) and diet quality measured by the Diet Quality Index for Pregnancy in a sample (n=918) of pregnant women.¹⁴ They also found no significant association between food outlet density and diet quality scores. Conversely, previous research that examined supermarket density and fast-food restaurant density independently found more favorable results in non-pregnant individuals.¹¹ Given the conflicting findings across pregnant vs. non-pregnant samples and differences in how food density was measured, further research could help clarify relationships for pregnant women.

It is possible that a significant association between healthy food density and overall diet quality was not observed due to limited variability in healthy food density scores overall. Additionally, there was no difference in healthy food density scores between participants living in urban vs. rural areas, meaning access to healthy food within a 5-mile radius was comparable across participants. Another potential contributing factor could be how neighborhoods were conceptualized, which varies widely in the literature.¹⁵ This study used a tailored approach and based neighborhood size on participants’ average self-reported distance traveled for grocery shopping across urban and rural areas.

In terms of HEI-2015 components, residential location did not moderate the relationships between healthy food density and meeting the standards for maximum HEI-2015 component scores. Previous studies that have examined the relationship between healthy food density and vegetable intake have shown mixed results.^32–34 Thornton and colleagues found that a higher density of supermarkets and produce stores is associated with more frequent vegetable consumption among non-pregnant Australian women (n=1,399) living in urban areas.³² Similarly, Powell and Han (2011) found that higher supermarket density was significantly associated with slightly higher weekly vegetable consumption among low-income non-pregnant adolescents (n=1,134).³⁴ Alternatively, previous studies have found no association between grocery store density and vegetable consumption among non-pregnant Japanese women.³³ It is worth noting that pregnant women were not included in any of these previous samples; highlighting the need for additional research.

Previous literature shows that individuals living in rural areas travel farther to do their grocery shopping compared to those in urban areas.²⁰ This pattern was also observed among HIPP participants, with participants in rural areas traveling twice as far (10.9 vs. 4.9 miles) as participants in urban areas to buy groceries. Traveling a farther distance may result in less frequent grocery shopping for individuals in rural areas compared to those in urban areas. Additionally, there are differences in the quality of fresh produce sold in rural grocery stores that can have a detrimental impact on produce purchasing and subsequent consumption.³⁵

Participants who were enrolled in the WIC program had diet quality scores that were comparable to participants who were not enrolled in WIC. This may be due to participants’ high educational attainment and average household income levels overall. The majority of participants (59%) earned a college degree or higher. Additionally, the majority of participants (58%) had a household income greater than $50,000/year. The 2018 median household income for SC residents was $51,015,³⁶ indicating that most participants were financially in a better position than the typical SC resident. These factors could have contributed to diet quality scores being comparable among participants regardless of their participation in the WIC program.

The current study is not without limitations. First, the exact location of where participants shopped for groceries was not collected. Since participants reported grocery shopping an average of 5.5 miles away from home, it was estimated they might do their grocery shopping at some of the included stores; however, it is possible that participants do their grocery shopping elsewhere. Researchers have found that taste, cost, and convenience influence diet quality,³⁷ so it may be beneficial to investigate these factors in future studies. Future studies could compare women’s perceived access to healthy food with their daily travel patterns measured via Global Position System (GPS) devices to better understand how women interact with their food environments. Additionally, HIPP participants’ sub-optimal diet quality indicates that women are not purchasing and consuming foods that contribute to optimal diet quality, regardless of the healthy food density in their areas. This suggests that it may be valuable for future studies to examine other factors that influence food purchasing behaviors outside of objectively measuring neighborhood access.

The present study obtained food retailer data from a single database while using multiple sources could improve the accuracy/completeness of the data.³⁸ The cross-sectional design does not allow for the examination of healthy food density and diet quality at multiple time-points during pregnancy; therefore, the direction of the association cannot be determined. Geocoding can be inaccurate due to the inherent error in the geo-referencing process;³⁹ however, all of the food outlets in the current study were matched to the point- or street-address level. The locations of the food retailers were obtained in December 2017, while the HIPP baseline assessments were conducted from January 2015 to January 2019; therefore, some included retailers could have closed, or new ones could have opened. Additionally, the study used data from a randomized trial, so the sample may not be representative of all pregnant women in SC with overweight or obesity. Furthermore, inaccurate reporting is present in all self-report dietary assessment tools,⁴⁰ so there is a possibility that participants under- or over-reported their dietary intake.

This study’s use of GIS-based methods allowed for an objective and tailored measure of the availability of multiple food retailers in relation to participants’ homes. Few studies have examined the ratio of healthy food retailers to less healthy retailers in relation to diet quality at the individual-level.⁴¹ This study’s use of the HEI-2015 is a strength since it captures overall diet quality, scores diets based on adherence to federal dietary recommendations, allows for a variety of ethnic and cultural eating patterns, and is reliable for all segments of the population for which the United States Department of Agriculture’s Food Patterns are appropriate, including pregnant women.⁴² Furthermore, this analysis was conducted in a racially-diverse sample of pregnant women with overweight or obesity, which is a high-priority sample of women who have not been included in much research to date.^12,14

Implications for Research and Practice

Overall, HIPP participants had sub-optimal diet quality; however, healthy food density was not related to diet quality. Given the adverse effects of poor diet quality on maternal and child health outcomes,^4,6 additional research is needed to better understand how women interact with their food environments during the important period of pregnancy to inform future diet quality interventions. Practitioners working with similar populations of pregnant women can be aware that diet quality is likely sub-optimal. While the food environment does not seem to be a primary driver of a low diet quality, other modifiable behavioral factors may be important targets in order to improve diet quality among pregnant women with overweight and obesity. Some potential strategies could include building nutrition education classes into prenatal visits, providing women with group classes to teach healthy cooking and shopping skills, or providing women with dietary self-monitoring technology and practitioners could review records and provide feedback at appointments.