Abstract
Aim:
To evaluate whether pregnant individuals with pregestational diabetes who live in a food-insecure community have worse glycemic control compared to those who do not live in a food-insecure community.
Methods:
A retrospective analysis of pregnant individuals with pregestational diabetes enrolled in a multidisciplinary prenatal and diabetes care program. The exposure was community-level food insecurity per the Food Access Research Atlas. The outcomes were hemoglobin A1c (A1c) < 6.0 % in early and late pregnancy, and an absolute decrease in A1c ≥ 2.0 % and mean change in A1c across pregnancy.
Results:
Among 418 assessed pregnant individuals with pregestational diabetes, those living in a food-insecure community were less likely to have an A1c < 6.0 % in early pregnancy compared to those living in a community without food insecurity [16 % vs. 30 %; adjusted risk ratio (aRR): 0.55; 95 % CI: 0.33–0.92]. Individuals living in a food-insecure community were more likely to achieve a decrease in A1c ≥ 2.0 % [35 % vs. 21 %; aRR: 1.55; 95 % CI: 1.06–2.28] and a larger mean decrease in A1c across pregnancy [mean: 1.46 vs. 1.00; adjusted beta: 0.47; 95 % CI: 0.06–0.87)].
Conclusions:
Pregnant individuals with pregestational diabetes who lived in a food-insecure community were less likely to enter pregnancy with glycemic control, but were more likely to have a reduction in A1c and achieve similar A1c status compared to those who lived in a community without food insecurity. Whether interventions that address food insecurity improve glycemic control and consequent perinatal outcomes remains to be studied.
Keywords: Diabetes, Pregestational diabetes, Pregnancy, Low food access, Food insecurity, Glycemic control
1. Introduction
Community-level food insecurity is defined as a lack of consistent and sustainable access to nutritionally adequate, culturally acceptable, and safe food in a specific geographic location [1], and is an adverse social determinant of health affecting 10 % of individuals in the U.S [2]. In pregnancy, food insecurity, whether defined at the individual- or community-level, affects nearly 20 % of people [3], and up to 33 % of single mothers postpartum [2,4,5]. Pregnant individuals with community-level food insecurity are at higher risk of adverse pregnancy outcomes [6], gestational diabetes and its severity [3,7,8], abnormal gestational weight gain [9], obesity [10], iron deficiency [11], and depressive symptoms and anxiety [12], compared to those without community-level food insecurity. In addition, food insecurity contributes to racial and ethnic health disparities in pregnancy, and is more frequently reported among non-Hispanic Black and Hispanic pregnant individuals compared to non-Hispanic White individuals [7,13].
Community-level food insecurity is frequent among people living with diabetes, and affects 20 % of these individuals [14]. Pregestational diabetes (type 1 and 2) affects nearly 1 in 20 pregnant persons [15,16], but the frequency and impact of community-level food insecurity in this high risk population during pregnancy remains to be defined. Glycemic control, such as that measured by glycosylated hemoglobin A1c (A1c), is affected by food insecurity [17], and food insecurity has increasingly been recognized as an intervention target to achieve health equity [18–23]. In pregnancy, individuals with diabetes and worse glycemic control are at higher risk of adverse maternal and infant outcomes [24–26]. Hence, guidelines recommend achieving a target A1c in pregnancy of at least < 6.5 %, and ideally < 6 %, to optimize outcomes [27,28]. The goal of achieving glycemic control has been shown to pose a unique and acute challenge for food-insecure pregnant individuals with pregestational diabetes [29,30].
We evaluated whether pregnant individuals with pregestational diabetes who lived in a food-insecure community had worse glycemic control, as measured by A1c, compared to those who did not live in a food-insecure community.
2. Methods
2.1. Study setting
This is a retrospective cohort analysis of pregnant individuals with pregestational diabetes (either type 1 or 2) and singleton gestations receiving prenatal and diabetes care as part of a multidisciplinary program from 2012 to 2016 at a tertiary care center. As has been previously described [26,31,32], individuals received integrated care with a team of maternal-fetal medicine specialists, diabetes nurse educators, and endocrinologists. Patients were encouraged to enroll in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) program, and consultations with social work were available but not universally a required part of the integrated care program. Individuals were referred to this care program regardless of gestational age from community-based prenatal care sites. Once enrolled, providers were encouraged to assess an A1c at least once per trimester. Socio-demographic and clinical data, including participant billing address at the delivery encounter, were manually abstracted from the electronic health record (EHR). This retrospective study was approved by the biomedical institutional review board at The Ohio State University (study ID: 2017H0145).
2.2. Study participants
We included individuals with a diagnosis of pregestational diabetes who were enrolled in the pregnancy and diabetes program and gave birth to a singleton, liveborn ≥ 22 weeks. For those with more than one pregnancy during the study, the first pregnancy was included in this analysis. Following an initial data query in the EHR to identify cases of pregestational diabetes, all diagnoses were then reviewed via manual chart review [25]. An individual was considered to have a diagnosis of pregestational diabetes if they met any of the following criteria: a prior documented diagnosis before pregnancy, A1c ≥ 6.5 %, a one-hour, 50-gram glucose challenge test ≥ 200 mg/dL at less than 20 weeks of gestation, or two elevated values on a 3-hour, 100-gram glucose tolerance test at less than 20 weeks of gestation [28]. Those diagnosed based on the latter two criteria also had a documented diagnosis of pregestational diabetes in the EHR during pregnancy. Of note, 88 % (275/313) of participants with an available A1c in the periconception period (4 weeks before last menstrual period through the first trimester or 14 weeks) met diagnostic criteria for pregestational diabetes based on A1c ≥ 6.5 % criteria alone.
Among 534 identified pregnant individuals enrolled for pregnancy and diabetes care, 116 were excluded due to missing A1c data (n = 62), termination of pregnancy (n = 6), spontaneous abortion (n = 34), fetal demise (n = 6), or an address that could not be geocoded (n = 8), resulting in a final analytic sample of 418 individuals. As previously described [31], those without A1c data (n = 62) were more likely to have type 2 diabetes (89 % vs. 67 %; p < 0.001) compared to the final study sample, but did not differ by White diabetes classification, diabetes pharmacotherapy, self-reported race and ethnicity, age, body mass index (BMI), or insurance status (p > 0.05 for all).
2.3. Exposure, outcome, and covariates
The exposure was community-level food insecurity (yes/no) per criteria for both low income and low access defined per the Food Access Research Atlas of the Economic Research Service of the U.S. Department of Agriculture (USDA), which provides a spatial overview of food access indicators by income and access using measures of supermarket accessibility [33]. Residential addresses from the delivery encounter were geocoded using ArcGIS, and then linked at the census-tract level to the Food Access Research Atlas. Low-income and low-access status at the census level were measured separately, and then the overlap between the two measures was assessed [34,35]. Community-level low income was defined as a census tract in which at least 20 % of the population had a median family income at or below 80 % of the metropolitan area or state median income. Low access to food at the community level, as recommended by the USDA, was defined as the number (at least 500) and proportion (at least 33 %) of individuals at the census tract level who were more than 1 mile (urban) and 10 miles (rural) from the nearest food store.
In secondary analyses, we defined low access at a closer threshold for urban areas of 0.5 miles, but did not analyze a different threshold for rural areas to be consistent with prior analyses [6]. In addition, because another important factor affecting food access is access to motor vehicles, we also included a measure of low vehicle access in addition to low food access; this measure was defined as a census tract where ≥ 100 households do not have vehicle access and live more than half a mile from the nearest food store; or ≥ 500 individuals or 33 % of the population live more than 20 miles from the nearest food store, regardless of vehicle access [23].
The primary outcomes were glycemic control measured cross-sectionally as an A1c < 6.0 % in early and late pregnancy and longitudinally as a decrease in A1c ≥ 2.0 % across pregnancy or mean change in A1c percentage (continuous measure). We secondarily assessed an outcome of A1c < 6.5 % in early and late pregnancy. These threshold values of A1c were chosen to be consistent with current pregnancy recommendations from the American Diabetes Association (ADA) and American College of Obstetricians and Gynecologists [27,28]. A cut-off value of < 6.0 % was designated as the primary outcome as it has been considered the optimal target by the ADA [27], has the lowest risk of adverse pregnancy outcomes [36], and reflects the physiologic drop of A1c during pregnancy [37]. In secondary analyses, a cut-off value of 6.5 % was also assessed because it is associated with a lower risk of adverse pregnancy outcomes [38], and is considered a reasonable threshold [27].
The first pregnancy assessment was selected to be the value closest to the time of initial care with the diabetes and pregnancy care program, which is a time that may be most amenable to interventions aimed at addressing social determinants of health, including food insecurity [17, 39]. For those with repeated assessment of A1c, the late-in-pregnancy assessment was selected to be the measure that was closest to the delivery date to reflect glycemic control associated with adverse pregnancy outcomes at delivery. Throughout the study, the same standard assay was used for A1c assessment, which was measured using automated high pressure liquid chromatography (VARIANT II TURBO HgbA1c Kit, Bio-Rad laboratories, Hercules, CA), with a normal range 4.7–5.6 % and coefficient of variation < 3 % in non-pregnant individuals [40].
Additional covariates at delivery were assessed using data from the EHR including age, BMI, insurance status, self-reported race and ethnicity recognizing race as a social construct and social determinant of health, parity, chronic hypertension, White classification of diabetes (Class A: Diet alone, any duration or onset age; Class B: Onset age 20 years or older and duration less than 10 years; Class C: Onset age 10–19 years or duration 10–19 years; Class D: Onset age younger than 10 years, duration over 20 years, background retinopathy, or hypertension (not preeclampsia); Class R: Proliferative retinopathy or vitreous hemorrhage; Class F: Nephropathy with over 500 mg/d proteinuria; Class RF: Criteria for both Classes R and F coexist; Class H: Arteriosclerotic heart disease clinically evident; Class T: Prior renal transplantation) [41], and diabetes pharmacotherapy (insulin, oral agent, both, or none).
2.4. Statistical analysis
Sociodemographic and clinical characteristics were compared between those identified as living in a community with and without food insecurity. Categorical variables were compared with chi-square and continuous variables with Student t test. Modified Poisson regression with robust error variance was used to examine associations between community-level food insecurity and A1c as a categorical measure, and linear regression for A1c as a continuous measure. We calculated unadjusted and adjusted relative risks (RR, aRR) and beta coefficients with 95 % confidence intervals (95 % CI), respectively. A directed acyclic graph (DAG) was used to assess the potential for confounding a priori. We adjusted for individual-level socio-demographic (maternal age [continuous], insurance status (Medicaid, private), clinical (parity [0, 1, 2], BMI [continuous], hemoglobin at initiation of care as anemia may impact A1c [42] [continuous]), and diabetes-related characteristics (gestational age at A1c measurement or change in gestational age between measurements, respectively [continuous], White diabetes classification [B, C, D, R/F/RF] [41], and pharmacotherapy at delivery [insulin only, oral agents only, both insulin and oral agents, none]).
In secondary analyses, we repeated the above analyzes 1) with the outcome at a higher threshold value (A1c < 6.5 % rather than 6.0 %), 2) with the exposure at a more proximate urban cutoff (low access at 0.5 miles rather than at 1 mile from the nearest food store), and 3) with the exposure defined as low vehicle in addition to low food access. In sensitivity analyses, for the outcomes of A1c in late pregnancy and change in A1c, we adjusted for A1c in early pregnancy (continuous), but not in the primary analysis as early pregnancy A1c was theorized to be on the causal pathway between the exposure and outcome, which could result in collider bias. Missing data were minimal for covariates because of manual data abstraction (3 %), and imputation was not performed. All statistical analyses were performed using STATA (STATACORP, version 16.1, College Station, TX). A p-value < 0.05 was considered statistically significant.
3. Results
Among the study sample of 418 pregnant individuals with pregestational diabetes (33 % type 1 % and 67 % type 2), 31.6 years (SD: 6.32), 49 % had public insurance, 29 % identified as non-Hispanic Black, and the mean BMI at delivery was 35.3 kg/m2 (SD: 9.29). Most individuals were on insulin by delivery (77 % insulin only, 8 % insulin with oral agents). The mean gestational age at delivery was 37.0 weeks (SD: 2.18).
Eighteen percent of individuals lived in a community with food insecurity and 82 % lived in a community without food insecurity (Table 1). Those who lived in a community with food insecurity were more likely to be publicly insured (61 % vs. 47 %; p = 0.02), or self-identify as non-Hispanic Black (45 % vs. 25 %) or Hispanic (17 % vs. 11 %; overall p across groups<0.001) compared to those who lived in a community without food security. Age, parity, BMI, tobacco use, chronic hypertension, type of diabetes, gestational age delivery, diabetes pharmacotherapy, White diabetes classification, and hemoglobin did not vary by food-insecurity status. Similarly, gestational age at early and late pregnancy assessment as well as time between assessments did not vary by food insecurity status.
Table 1.
Individual and community-level characteristics overall and by community-level food insecurity (N = 418).
| Community-level food insecurity n (%) | ||||
|---|---|---|---|---|
| Variable | Overall, N = 418a | No N = 343a | Yes N = 75a | p-valuea |
| Individual-level measures | ||||
| Age, years, mean (SD) | 31.6 (6.32) | 31.7 (6.07) | 30.8 (7.34) | 0.2 |
| Parity, nulliparous (n = 416) | 154 (37.0) | 130 (38.1) | 24 (32.0) | 0.6 |
| Insurance status, public | 206 (49.3) | 160 (46.7) | 46 (61.3) | 0.02 |
| Self-reported race and ethnicity | ||||
| Non-Hispanic White | 222 (53.1) | 194 (56.6) | 28 (37.3) | < 0.001 |
| Non-Hispanic Black | 120 (28.7) | 86 (25.1) | 34 (45.3) | |
| Non-Hispanic Asian | 24 (5.7) | 24 (7.0) | 0 (-) | |
| Hispanic/Latina | 52 (12.4) | 39 (11.4) | 13 (17.3) | |
| Body mass index at delivery, kg/m2, mean (SD) (n = 416) | 35.3 (9.29) | 35.3 (9.64) | 35.4 (7.55) | 0.9 |
| Tobacco use (n = 415) | ||||
| Past use | 69 (16.6) | 50 (14.7) | 19 (25.3) | 0.07 |
| Current | 53 (12.8) | 43 (12.7) | 10 (13.3) | |
| Chronic hypertension (n = 415) | 131 (3.6) | 105 (30.9) | 26 (34.7) | 0.5 |
| Type of diabetes | 138 (33.0) | 120 (35.0) | 18 (24.0) | 0.06 |
| Type 1 | 280 (67.0) | 223 (65.0) | 57 (76.0) | |
| Type 2 | ||||
| Gestational age at delivery, weeks, mean (SD) | 37.0 (2.18) | 37.0 (2.24) | 36.8 (1.92) | 0.4 |
| Mediation use at delivery (n = 415) | ||||
| None | 17 (4.1) | 13 (3.8) | 4 (5.3) | 0.7 |
| Insulin only | 319 (76.9) | 264 (77.7) | 55 (73.3) | |
| Oral agents | 45 (10.8) | 37 (10.9) | 8 (10.7) | |
| Insulin and oral agents | 34 (8.2) | 26 (7.7) | 8 (10.7) | |
| White’s classification of diabetes in pregnancy (n = 414) | ||||
| B | 238 (57.5) | 195 (57.5) | 43 (57.3) | 0.8 |
| C | 94 (22.7) | 79 (23.3) | 15 (20.0) | |
| D | 45 (10.9) | 35 (10.3) | 10 (13.3) | |
| R/F/RF | 37 (8.9) | 30 (8.9) | 7 (9.3) | |
| Hemoglobin, g/dl, mean (SD) (n = 415) | 12.6 (1.63) | 12.6 (1.72) | 12.6 (1.14) | 0.8 |
| Gestational age at early pregnancy A1c, weeks, median (SD) | 8.6 (6.9, 10.7) | 8.6 (7.0, 10.7) | 7.7 (6.6, 10.9) | 0.9 |
| Gestational age at late pregnancy A1c, weeks, median (SD) | 31.1 (28.6, 34.0) | 31.1 (28.6, 34.1) | 31.0 (28.6, 33.7) | 0.4 |
| Gestational age between early and late pregnancy A1c, weeks, median (SD) | 21.0 (17.0, 25.0) | 21.2 (17.0, 25.0) | 20.0 (15.6, 24.8) | 0.3 |
| Community-level measures | ||||
| Family income of census tract, $, median (IQR) | 57,734 (41,893, 87,250) | 64,686 (42,661, 91,766) | 47,500 (34,898, 53,956) | < 0.001 |
| Census tract poverty rate, mean (SD) | 19.2 (13.50) | 17.9 (13.70) | 25.4 (10.64) | < 0.001 |
Abbreviations: standard deviation (SD), interquartile range (IQR).
Continuous variables were compared with student T tests and categorical variables were compared with chi square tests.
At the community-level, those with food insecurity were more likely to live in a community with a lower median family income (median: $47,500 vs. $64,686; p < 0.001) and to live in a census tract with a higher poverty rate (25 % vs. 18 %; p < 0.001).
In early pregnancy (median gestational age [IQR]: 9.7 weeks [7.4, 14.1]), 27 % (112/418) of individuals had a A1c < 6.0 % and 38 % (160/418) had A1c < 6.5 %. Among a subset of 376 individuals with a follow-up assessment in late pregnancy (median gestational age [IQR]: 30.4 weeks [27.7, 33.6]), 46 % (173/376) of individuals had a A1c < 6.0 % and 59 % (222/376) had a A1c < 6.5 %. From early to late pregnancy, 24 % (90/376) of individuals had a decrease in A1c of ≥ 2.0 %.
Individuals living in a community with food insecurity were less likely to achieve a A1c < 6.0 % in early pregnancy compared to those living in a community without food insecurity (16 % vs. 29 %; aRR: 0.55; 95 % CI: 0.33–0.92) (Table 2). However, A1c did not differ by food insecurity status in late pregnancy. Across pregnancy individuals living in a community with food insecurity were more likely to achieve a decrease in A1c of ≥ 2.0 % (35 % vs. 21 %; aRR: 1.55; 95 % CI: 1.06–2.28) and have a higher mean decrease in A1c (mean [SD]: 1.46 [1.65] vs. 1.00 [1.51]; adj. β: 0.47; 95 % CI: 0.06–0.87) compared to those living in a community without food insecurity.
Table 2.
Association between community level food insecurity and glycemic control among pregnant individuals with pregestational diabetes.
| Community level food insecurity | ||||
|---|---|---|---|---|
| Yes n (%) or mean (SD) | No n (%) or mean (SD) | Unadjusted analysisa | Adjusted analysisa,b | |
| Risk ratio (95 % CI) | Adjusted risk ratio (95 % CI) | |||
| Cross-sectional analysis: | ||||
| Early pregnancy A1c < 6.0 %c | 12/75 (16.0) | 100/343 (29.2) | 0.55 (0.32–0.95) | 0.55 (0.33–0.92) |
| Late pregnancy A1c < 6.0 %d | 28/71 (39.4) | 145/305 (47.5) | 0.83 (0.61–1.13) | 0.90 (0.67–1.20) |
| Longitudinal analysis: | ||||
| Decrease in A1c of ≥ 2 %d | 25/71 (35.2) | 65/305 (21.3) | 1.65 (1.13–2.42) | 1.55 (1.06–2.28) |
| Beta coefficient (95 % CI) | Adjusted beta coefficient (95 % CI) | |||
| Mean change in A1cd | 1.46 (1.65) | 1.00 (1.51) | 0.46 (0.06–0.86) | 0.47 (0.06–0.87) |
Abbreviations: CI: confidence interval.
Unadjusted and adjusted risk ratios calculated using modified Poisson regression with robust error variance.
Model adjusted for age (continuous), insurance status (Medicaid, private), body mass index (continuous), race and ethnicity (White, Black, Hispanic, other), diabetes pharmacotherapy (none, insulin, oral agents only, both insulin and oral agents), White Diabetes Classification (B, C, D, R/F/RF), baseline hemoglobin (continuous), and gestational age of hemoglobin A1C measurement (continuous) or change in gestational age between first and last measurement (continuous).
N = 418 in the unadjusted adjusted model, and N = 413 in the adjusted model.
N = 376 in the unadjusted adjusted model, and N = 373 in the adjusted model.
In secondary analyses, the primary results were similar (Table 3). First, when the outcome was defined at a higher threshold of A1c < 6.5 %, individuals living in a community with food insecurity were less likely to achieve glycemic control in early pregnancy compared to those living in a community without food insecurity (24 % vs. 41 %; aRR: 0.58; 95 % CI: 0.39–0.87). Second, when the exposure of low food access was lowered in an urban setting, individuals living in a community with food insecurity were less likely to achieve a A1c < 6.0 % in early pregnancy (24 % vs. 29 %; aRR: 0.73; 95 % CI: 0.53–1.00). Third, when the exposure was defined as low vehicle access in addition to low food access, individuals living in a community with food insecurity were less likely to achieve a A1c < 6.0 % in early pregnancy (18 % vs. 32 %; aRR: 0.62; 95 % CI: 0.43–0.90). In sensitivity analysis, after adjustment for early pregnancy A1c, individuals living in a community with food insecurity were not more likely to achieve glycemic control across pregnancy compared to those living in a community without food insecurity.
Table 3.
Secondary and sensitivity analyses of the association between community level food insecurity and glycemic control among pregnant individuals with pregestational diabetes.
| Community level food insecurity | ||||
|---|---|---|---|---|
| Yes n (%) or mean (SD) | No n (%) or mean (SD) | Unadjusted analysisa | Adjusted analysisb,2 | |
| SECONDARY ANALYSIS #1: Outcome defined as A1c < 6.5 % | ||||
| Cross-sectional analysis: | ||||
| Early pregnancy A1c < 6.5 %c | 18/75 (24.0) | 142/343 (41.4) | 0.58 (0.38–0.88)* | 0.58 (0.39–0.87) |
| Late pregnancy A1c < 6.5 %d | 35/71 (49.3) | 187/305 (61.3) | 0.80 (0.62–1.03) | 0.84 (0.66–1.07) |
| SECONDARY ANALYSIS #2: Exposure defined as low food access at 0.5 miles (urban), 10 miles (rural) | ||||
| Cross-sectional analysis: | ||||
| Early pregnancy A1c < 6.0 %c | 45/188 (23.9) | 67/230 (29.1) | 0.82 (0.59–1.14) | 0.73 (0.53–1.00) |
| Late pregnancy A1c < 6.0 %d | 72/169 (42.6) | 101/207 (48.8) | 0.87 (0.70–1.09) | 0.92 (0.73–1.16) |
| Longitudinal analysis: | ||||
| Decrease in A1c of ≥ 2 %d | 46/169 (27.2) | 44/207 (21.3) | 1.28 (0.89–1.84) | 1.09 (0.72–1.65) |
| Mean change in A1cd | 1.17 (1.59) | 1.01 (1.51) | 0.16 (− 0.16 to 0.47) | 0.15 (− 0.20 to 0.49) |
| SECONDARY ANALYSIS #3: Exposure defined as low vehicle and food access | ||||
| Cross-sectional analysis: | ||||
| Early pregnancy A1c < 6.0 %c | 27/48 (18.2) | 85/270 (31.5) | 0.57 (0.39, 0.83) | 0.62 (0.43, 0.90) |
| Late pregnancy HbA1c < 6.0 %d | 54/134 (40.3) | 119/242 (49.2) | 0.82 (0.64, 1.04) | 0.92 (0.73, 1.17) |
| Longitudinal analysis: | ||||
| Decrease in HbA1c of ≥ 2 %d | 42 (31.3) | 48 (19.8) | 1.58 (1.11–2.26) | 1.39 (0.95–2.04) |
| Mean change in HbA1cd | 1.24 (1.68) | 1.00 (1.46) | 0.24 (− 0.09 to 0.57) | 0.22 (− 0.13 to 0.56) |
| SENSITIVITY ANALYSIS: Adjusted for early pregnancy A1c | ||||
| Late pregnancy A1c < 6.0 %d | – | – | – | 1.06 (0.81–1.40) |
| Longitudinal analysis: | ||||
| Decrease in HbA1c of ≥ 2 %d | – | – | – | 1.15 (0.81–1.65) |
| Mean change in HbA1cd | – | – | – | 0.02 (− 0.24 to 0.27) |
Abbreviations: CI: confidence interval.
Unadjusted and adjusted risk ratios calculated using modified Poisson regression with robust error variance.
Model adjusted for age (continuous), insurance status (Medicaid, private), BMI (continuous), race and ethnicity (White, Black, Hispanic, other), diabetes pharmacotherapy (none, insulin, oral agents only, both insulin and oral agents), White Diabetes Classification (B, C, D, R/F/RF), baseline hemoglobin (continuous), and gestational age of hemoglobin A1C measurement (continuous).
N = 418 in the unadjusted adjusted model, and N = 413 in the adjusted model.
N = 376 in the unadjusted adjusted model, and N = 373 in the adjusted model.
4. Discussion
Pregnant individuals with pregestational diabetes living in a community with food insecurity were less likely to enter pregnancy with glycemic control as measured by A1c, but were more likely to have a reduction in A1c compared to those living in a community without food insecurity. A1c measured late in pregnancy—after multidisciplinary and longitudinal prenatal care had occurred—was similar regardless of food insecurity.
In the current study, nearly 1 in 5 pregnant individuals with pregestational diabetes were classified as living in a community with food insecurity. The inverse association between community-level food insecurity and glycemic control was present at different A1c thresholds as well as when using different criteria to define community-level food insecurity. In a previous analysis of the current study population, higher community-level social vulnerability, defined using the U.S. Centers for Disease Control and Prevention Social Vulnerability Index, also was associated with worse glycemic control [31]. Prior studies have demonstrated an association between community-level food insecurity or food deserts and adverse pregnancy outcomes in individuals without diabetes [4,29]. Tipton et al. found that living in a food desert was associated with at least one of the assessed pregnancy co-morbid conditions (e.g., gestational hypertension, preeclampsia, gestational diabetes) in a low-risk population of pregnant individuals [6]. Fonge et al. demonstrated that pregnant individuals living in a better food environment were less likely to have gestational diabetes requiring medication to achieve glycemic control as well as type 2 diabetes [8]. Li et al. found that individuals with gestational diabetes who reported household food insecurity per the U.S. Food Security Survey Module were more likely to develop type 2 diabetes in the National Health and Nutrition Examination Survey (NHANES) from 2007 to 2018 [43].
In the current study, differences in glycemic control by living in a food-insecure area were not apparent by the end of pregnancy. Guidelines currently do not prescribe the frequency of A1c assessment in pregnancy [27]. Hence, we determined the association between food insecurity and A1c in early pregnancy, which may be a time that is more amenable to interventions aimed at addressing food insecurity or nutrition behaviors to improve glycemic control, and late pregnancy, which is temporally proximate to adverse pregnancy outcomes at delivery. Furthermore, it is likely that enrollment in a multidisciplinary diabetes and pregnancy program led to improvements in glycemic control through improved access to pharmacotherapy, increased glycemic monitoring, frequent counseling from clinicians, and WIC referral for eligible participants [26]. In a previous analysis of this study population, non-Hispanic Black individuals had worse glycemic control in early pregnancy compared to non-Hispanic White individuals, which was likely due to higher community-level social vulnerability however. By late pregnancy, however, there were no differences in glycemic control between both groups [32].
It is anticipated that food insecurity may increase in the near future as food prices have increased by 10 % in 2022 alone [44]. The USDA Food Access Research Atlas is a publicly available resource and readily accessible online. It could be used to risk stratify pregnant individuals with regards to their likelihood of community-level food insecurity, and identified individuals could then undergo further screening and referral to community-based resources [4,29]. For example, nearly half of the study population was enrolled in Medicaid, and Medicaid enrollees have the option to enroll in two programs that address food insecurity, WIC as well as the Supplemental Nutrition Assistance Program (SNAP). Of concern, enrollment in WIC has decreased from 2011 to 2017 from 79 % to 68 % regardless of diabetes status [45]. Hence, screening for food insecurity may provide an additional opportunity to emphasize enrollment in these programs. Whether these programs adequately address individual- or community-level food insecurity and improve glycemic control for pregnant individuals with diabetes remains to be studied. Prior interventions have been developed to address food insecurity in pregnancy, including the use of food vouchers [46], but interventions have generally not focused on the unique needs of individuals with pregestational diabetes [7].
Further studies are needed to identify whether interventions that aim to address community-level food insecurity in addition to other adverse social determinants of health, such as transportation barriers and built environment not conducive to exercise, can improve glycemic control and diabetes-related pregnancy outcomes [30]. The relative contributions of individual- and community-level food insecurity to glycemic control in pregnancy and the associated impact of targeted interventions at both levels also need to be assessed further.
Strengths of this study include extending the use of the USDA Food Access Research Atlas to understand the association of community-level food insecurity with glycemic control among pregnant individuals with pregestational diabetes. The USDA updates the Food Access Research Atlas using up-to-date census tract data with additional measures, such as distance to stores, income, and vehicle access [35]. This tool employs an ecologic or community approach to understand food insecurity and associated social vulnerability as opposed to individual risk factors. Hence, this approach may provide a broader economic and social context in which individuals experience food insecurity. Finally, this analysis utilized data from a diverse cohort in which nearly half the assessed pregnant individuals with pregestational diabetes self-identified as being from minoritized populations who are more likely experience community-level food insecurity.
There are study limitations to note. First, given that some of the diagnoses were made in pregnancy, we may have included some individuals with early-onset gestational diabetes who were diagnosed based on an abnormal three-hour 100-gram glucose tolerance test. However, given an early diagnosis within pregnancy, and the fact that misdiagnosis would likely bias our observed results toward the null, we do not believe this could account for the observed association. Second, exposure misclassification is possible because addresses were retrospectively abstracted from the EHR. It is possible participants may have changed locations during pregnancy or lived at another location other than their listed address. Since this misascertainment, if it occurred, should not be systematic, this also should bias toward the null. Third, the USDA Food Access Research Atlas was not specifically designed for the clinical setting or population of this analysis. Fourth, glycemic control and community-level food insecurity may be associated with additional factors not assessed in this study, such as WIC and SNAP enrollment. Fifth, this analysis excluded about 20 % of the original study population, primarily due to the lack of A1c data. While those who were excluded were more likely to have type 2 diabetes, they did not differ by other assessed socio-demographic characteristics. Sixth, the current study occurred at a single institution among pregnant individuals enrolled in a multidisciplinary diabetes and pregnancy program, and these findings may not be generalizable to all settings.
In conclusion, pregnant individuals with pregestational diabetes living in a community with food insecurity were less likely to enter pregnancy with glycemic control, but were more likely to have a reduction in A1c across pregnancy compared to living in a community without food insecurity. Whether interventions that address food insecurity improve glycemic control and consequent pregnancy outcomes in pregnant individuals with pregestational diabetes remains to be studied.
Funding
Dr. Venkatesh was supported by the Care Innovation and Community Improvement Program at The Ohio State University. Drs. Venkatesh and Joseph were supported by AHRQ Grant #R01HS028822.
Footnotes
Conflict of interest
None of the authors report a conflict of interest.
References
- [1].Hamm M, Bellows AC, Community food security: background and future directions, J. Nutr. Educ. Behav 35 (1) (2003) 37–43. [DOI] [PubMed] [Google Scholar]
- [2].Coleman-Jensen A, Rabbitt MP Gregory CA, Singh A, Household food security in the United States in 2019, Economic Research Service, US Department of Agriculture. Accessed 8 June 2022, 〈https://www.ers.usda.gov/publications/pub-details/?pubid=99281〉. [Google Scholar]
- [3].Richards M, Weigel M, Li M, Rosenberg M, Ludema C, Food insecurity, gestational weight gain and gestational diabetes in the National Children’s Study, 2009–2014, J. Public Health 43 (3) (2021) 558–566. [DOI] [PubMed] [Google Scholar]
- [4].Dolin C, Compher CC, Oh JK, Durnwald CP, Pregnant and hungry: addressing food insecurity in pregnant women during the COVID-19 pandemic in the United States, Am. J. Obstet. Gynecol. MFM 3 (4) (2021), 100378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Cukrowska-Torzewska E, Matysiak A, The motherhood wage penalty: a meta-analysis, Soc. Sci. Res 88–89 (2020), 102416. [DOI] [PubMed] [Google Scholar]
- [6].Tipton M, Wagner SA, Dixon A, Westbay L, Darji H, Graziano S, Association of living in a food desert with pregnancy morbidity, Obstet. Gynecol 136 (1) (2020) 140–145. [DOI] [PubMed] [Google Scholar]
- [7].Banner G, Weaver KN, Rushovich T, York SL, Yee LM, Association between food deserts and gestational diabetes mellitus in a large metropolitan area, Am. J. Perinatol 38 (S 01) (2021) e39–e45. [DOI] [PubMed] [Google Scholar]
- [8].Fonge Y, Jain VD, Harrison C, Brooks M, Sciscione AC, Examining the relationship between food environment and gestational diabetes, Am. J. Obstet. Gynecol. MFM 2 (4) (2020), 100204. [DOI] [PubMed] [Google Scholar]
- [9].Cheu L, Yee LM, Kominiarek MA, Food insecurity during pregnancy and gestational weight gain, Am. J. Obstet. Gynecol. Matern. Fetal Med 2 (1) (2020), 100068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Laraia B, Gundersen C, Household food insecurity is associated with self-reported pregravid weight status, gestational weight gain, and pregnancy complications, J. Am. Diet. Assoc 110 (2010) 692–701. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Park C, Eicher-Miller HA, Iron deficiency is associated with food insecurity in pregnant females in the United States: National Health and Nutrition Examination Survey 1999–2010, J. Acad. Nutr. Diet 114 (12) (2014) 1967–1973. [DOI] [PubMed] [Google Scholar]
- [12].Tarasuk V, Gundersen C, Wang X, Roth DE, Urquia ML, Maternal food insecurity is positively associated with postpartum mental disorders in Ontario, Canada, J. Nutr 150 (11) (2020) 3033–3040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Hromi-Fiedler AB-MA, Segura-Perez S, Perez-Escamilla R, Household food insecurity is associated with depressive symptoms among low-income pregnant Latinas, Matern. Child Nutr 7 (2011) 421–430. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Berkowitz S, Baggett TP, Wexler DJ, Huskey KW, Wee CC, Food insecurity and metabolic control among U.S. adults with diabetes, Diabetes Care 36 (2013) 3093–3099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Shah N, Wang MC, Freaney PM, Perak AM, Carnethon MR, Kandula NR, Gunderson EP, Bullard KM, Grobman WA, O’Brien MJ, Khan SS, Trends in gestational diabetes at first live birth by race and ethnicity in the US, 2011–2019, J. Am. Med. Assoc 326 (7) (2021) 660–669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Feig D, Hwee J, Shah BR, Booth GL, Bierman AS, Lipscombe LL, Trends in incidence of diabetes in pregnancy and serious perinatal outcomes: a large, population-based study in Ontario, Canada, 1996–2010, Diabetes Care 37 (6) (2014) 1590–1596. [DOI] [PubMed] [Google Scholar]
- [17].Haire-Joshu D, Hill-Briggs F, The next generation of diabetes translation: a path to health equity, Annu. Rev. Public Health 40 (2019) 391–410. [DOI] [PubMed] [Google Scholar]
- [18].Pantell M, Baer RJ, Torres JM, Felder JN, Gomez AM, Chambers BD, Dunn J, Parikh NI, Pacheco-Werner T, Rogers EE, Feuer SK, Ryckman KK, Novak NL, Tabb KM, Fuchs J, Rand L, Jelliffe-Pawlowski LL, Associations between unstable housing, obstetric outcomes, and perinatal health care utilization, Am. J. Obstet. Gynecol. MFM 1 (4) (2019), 100053. [DOI] [PubMed] [Google Scholar]
- [19].Amjad S, MacDonald I, Chambers T, Osornio-Vargas A, Chandra S, Voaklander D, Ospina MB, Social determinants of health and adverse maternal and birth outcomes in adolescent pregnancies: a systematic review and meta-analysis, Pediatr. Perinat. Epidemiol 33 (1) (2019) 8–99. [DOI] [PubMed] [Google Scholar]
- [20].Marmot M, Allen JJ, Social determinants of health equity, Am. J. Public Health 104 (S4) (2014) S517–S519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Walker R, Garacci E, Ozieh M, Egede LE, Food insecurity and glycemic control in individuals with diagnosed and undiagnosed diabetes in the United States, Prim. Care Diabetes 15 (5) (2021) 813–818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Lyles C, Wolf MS, Schillinger D, Davis TC, Dewalt D, Dahlke AR, Curtis L, Seligman HK, Food insecurity in relation to changes in hemoglobin A1c, self-efficacy, and fruit/vegetable intake during a diabetes educational intervention, Diabetes Care 36 (6) (2013) 1448–1453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Berkowitz S, Karter AJ, Corbie-Smith G, Seligman HK, Ackroyd SA, Barnard LS, Atlas SJ, Wexler DJ, Food insecurity, food “deserts,” and glycemic control in patients with diabetes: a longitudinal analysis, Diabetes Care 41 (6) (2018) 1188–1195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Davidson A, Park AL, Berger H, Aoyama K, Harel Z, Cook JL, Ray JG, Risk of severe maternal morbidity or death in relation to elevated hemoglobin A1c preconception, and in early pregnancy: a population-based cohort study, PLoS Med. 17 (5) (2020), e1003104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Finneran M, Kiefer MK, Ware CA, Buschur EO, Thung SF, Landon MB, Gabbe SG, The use of longitudinal hemoglobin A1c values to predict adverse obstetric and neonatal outcomes in pregnancies complicated by pregestational diabetes, Am. J. Obstet. Gynecol. MFM 2 (1) (2020), 100069. [DOI] [PubMed] [Google Scholar]
- [26].Kiefer M, Finneran MM, Ware CA, Fareed N, Joseph J, Thung SF, Costantine MM, Landon MB, Gabbe SG, Venkatesh KK, Association of change in haemoglobin A1c with adverse perinatal outcomes in women with pregestational diabetes, Diabetes Med. 39 (7) (2022), e14822. [DOI] [PubMed] [Google Scholar]
- [27].American Diabetes Association, 14. Management of diabetes in pregnancy: standards of medical care in diabetes—2020, Diabetes Care 43 (S1) (2020) S183–S192. [DOI] [PubMed] [Google Scholar]
- [28].Bulletins—Obstetrics. ACoOaGCoP, ACOG practice bulletin no. 201: pregestational diabetes mellitus, Obstet. Gynecol 132 (6) (2018) e228–e248. [DOI] [PubMed] [Google Scholar]
- [29].Gomez H, DiTosto JD, Niznik CM, Yee LM, Understanding food security as a social determinant of diabetes-related health during pregnancy, Am. J. Perinatol 10 (2021) 1055. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Yee L, McGuire JM, Taylor SM, Niznik CM, Simon MA, Social and environmental barriers to nutrition therapy for diabetes management among underserved pregnant women: a qualitative analysis, J. Nutr. Educ. Behav 48 (3) (2016) 170–180. [DOI] [PubMed] [Google Scholar]
- [31].Venkatesh K, Germann K, Joseph J, Kiefer M, Buschur E, Thung S, Costantine MM, Gabbe S, Grobman WA, Fareed N, Association between social vulnerability and achieving glycemic control among pregnant individuals with pregestational diabetes, Obstet. Gynecol 139 (6) (2022) 1051–1060. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Venkatesh K, Fareed N, Kiefer MK, Ware CA, Buschur E, Landon MB, Thung SF, Costantine MM, Gabbe SG, Joseph JJ, Differences in hemoglobin A1c during pregnancy between non-hispanic black versus white women with prepregnancy diabetes, Am. J. Perinatol (2022). [DOI] [PubMed] [Google Scholar]
- [33].Rhone A, Ver Ploeg M, Dicken C, Williams R, Breneman V, Low-income and low-supermarket-access Census tracts, 2010–2015, Econ. Inf. Bull 165 (2015) 1–15. [Google Scholar]
- [34].USDA, Introduction to the Food Research Atlas. Accessed 8 June 2022, 〈https://gisportal.ers.usda.gov/portal/apps/experiencebuilder/experience/?id=a53ebd7396cd4ac3a3ed09137676fd40&page=page_0〉.
- [35].Economic Research Service (ERS), U.S. Department of Agriculture (USDA), Food Access Research Atlas Accessed 8 June 2022, 〈https://www.ers.usda.gov/data-products/food-access-research-atlas/〉.
- [36].Cyganek K, Skupien J, Katra B, et al. , Risk of macrosomia remains glucose-dependent in a cohort of women with pregestational type 1 diabetes and good glycemic control, Endocrine 55 (2017) 447–455. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Nielsen L, Ekbom P, Damm P, et al. , HbA1c levels are significantly lower in early and late pregnancy, Diabetes Care 27 (2004) 1200–1201. [DOI] [PubMed] [Google Scholar]
- [38].M.B. HAPO Study Cooperative Research Group, Lowe LP, et al. , Hyperglycemia and adverse pregnancy outcomes, N. Engl. J. Med 358 (19) (2008) 1991–2002. [DOI] [PubMed] [Google Scholar]
- [39].Leziak K, Birch E, Jackson J, Strohbach A, Niznik C, Yee LM, Identifying mobile health technology experiences and preferences of low-income pregnant women with diabetes, J. Diabetes Sci. Technol 15 (5) (2021) 1018–1026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].Bio-Rad laboratories I, Variant II Turbo HbA1c Kit – 2.0. U. S Accessed 11 November 2021, 〈https://www.accessdata.fda.gov/cdrh_docs/reviews/K090699.pdf〉.
- [41].Bennett S, Tita A, Owen J, Biggio JR, Harper LM, Assessing White’s classification of pregestational diabetes in a contemporary diabetic population, Obstet. Gynecol 125 (5) (2015) 1217–1223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [42].Ahmad J, Rafat D, HbA1c and iron deficiency: a review, Diabetes Metab. Syndr 7 (2) (2013) 118–122. [DOI] [PubMed] [Google Scholar]
- [43].Li L, Ji J, Li Y, Huang YJ, Moon JY, Kim RS, Gestational diabetes, subsequent type 2 diabetes, and food security status: national health and nutrition examination survey, 2007–2018, Prev. Chronic Dis 19 (2022), E42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [44].FAO I, UNICEF, WFP, WHO, The State of Food Security and Nutrition in the World 2021. Transforming food systems for food security, improved nutrition and affordable healthy diets for all Accessed 6 June 2022, 〈 10.4060/cb4474en〉. [DOI]
- [45].Soneji S, Beltrán-Sánchez H, Association of special supplemental nutrition program for women, infants, and children with preterm birth and infant mortality, JAMA Netw. Open 2 (12) (2019), e1916722. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [46].Ridberg R, Marpadga S, Akers MA, Bell JF, Seligman HK, Fruit and vegetable vouchers in pregnancy: preliminary impact on diet & food security, J. Hunger Environ. Nutr 16 (S1) (2021) 149–163. [Google Scholar]