Racial Disparities in Blood Pressure Trajectories of Preterm Children: The Role of Family and Neighborhood Socioeconomic Status

Abstract

Racial disparities in cardiovascular disease mortality in the United States remain substantial. However, the childhood roots of these disparities are not well understood. In the current study, we examined racial differences in blood pressure trajectories across early childhood in a sample of African-American and European-American low-birth-weight preterm infants. Family and neighborhood socioeconomic status (SES), measured at baseline, were also examined as explanations for subsequent group disparities. Analyses focused on 407 African-American and 264 European-American children who participated in the Infant Health and Development Program, a US longitudinal study of preterm children born in 1985. Blood pressure was assessed on 6 occasions between the ages of 24 and 78 months, in 1987–1992. Across this age range, the average rate of change in both systolic and diastolic blood pressure was greater among African-American children than among European-American children. Neighborhood SES explained 29% and 24% of the racial difference in the average rate of change in systolic and diastolic blood pressure, respectively, whereas family SES did not account for group differences. The findings show that racial differences in blood pressure among preterm children emerge in early childhood and that neighborhood SES accounts for a portion of racial disparities.

Racial disparities in cardiovascular disease mortality in the United States remain substantial, with hypertension being the largest contributor to black-white disparities in mortality (1, 2). Understanding the childhood roots of these enduring differences is crucial, particularly within at-risk samples of low-birth-weight and premature children (3, 4). Premature birth is associated with elevated blood pressure beginning in infancy and early childhood (5, 6), which in turn is a risk factor for adult hypertension, heart disease, and early mortality (7, 8). Black/African-American (AA) infants are approximately twice as likely to be born premature as white/European-American (EA) infants (9) and consequently are at greater risk for a wide range of adult health problems (10). As such, racial disparities in health may be partially rooted in the perinatal and early childhood periods.

In samples of full-term, normal-birth-weight infants, some studies have shown that African Americans have higher levels of systolic blood pressure (SBP) and diastolic blood pressure (DBP) than European Americans by approximately 10 years of age (11, 12), and that African Americans experience greater increases in blood pressure across childhood and adolescence than do European Americans (12, 13). Although the higher prevalence of low birth weight and premature birth among African Americans is well documented (9), no studies (to our knowledge) have investigated racial differences in the blood pressure trajectories of preterm infants using longitudinal data. Furthermore, although cross-sectional work indicates that preterm birth may be more strongly associated with elevated blood pressure among AA children than among EA children (14, 15), research on this topic is scarce. One key objective of this study was therefore to examine racial differences in blood pressure between age 24 months and age 78 months in a sample of preterm infants. This topic is important in health disparities research because preterm infants are at elevated risk for subsequent health problems (10, 16) and because AA infants are disproportionately born preterm (9, 17). Identifying when racial disparities in blood pressure emerge among preterm children may be a key step toward reducing racial disparities.

Another important step towards reducing group disparities involves identifying environmental factors that account for group differences. Both family and neighborhood socioeconomic status (SES) are associated with child health outcomes, including inflammation, obesity, and aggregated risk factors for cardiovascular disease (18–21). In addition, due to a history of racist housing policies, labor market discrimination, and other varieties of unfair treatment (22–25), African Americans disproportionately live in low-SES neighborhoods and are raised by caregivers with lower income and educational levels (26). Thus, preterm AA infants are at greater risk of subsequent health problems than their generally more advantaged EA counterparts.

Family and neighborhood SES, although related, have independent health risks. Neighborhood SES represents aggregate characteristics of individuals within a particular spatial locality, and it can influence the health outcomes of children in various ways (27–31). For example, neighborhood SES can influence children through associations with parental stress levels (32, 33) and sleep patterns (34), which in turn influence family functioning and aspects of parenting linked to child health (34–36). Neighborhood SES may also influence the health of children through the availability of health-promoting resources such as parks and playgrounds (37), through the availability of healthy food choices (38), and through exposure to toxins that are present at higher levels in less advantaged neighborhoods (39–41). Family SES is an alternative conceptualization of SES linked with home environmental characteristics, family stress, and parenting (42).

Neighborhood and family SES have been linked to blood pressure (43–45), and researchers considering them simultaneously have found evidence for the independent influence of each (46, 47). However, these studies have not used longitudinal blood pressure data or focused on childhood samples. Moreover, no studies have considered the degree to which neighborhood or family SES accounts for racial disparities in blood pressure trajectories among children. We hypothesized that AA children would have greater increases in blood pressure across early childhood than EA children. Additionally, we predicted that lower family and neighborhood SES at baseline would be associated with greater subsequent increases in blood pressure and would attenuate expected racial differences in blood pressure trajectories.

METHODS

Data

This study drew on data from the Infant Health and Development Program (48). Participants were 985 infants born at 8 participating US medical institutions who weighed 2,500 g or less, had a gestational age of 37 weeks or less, and reached 40 weeks of postconceptional age between January and October 1985. An intervention program for 38% of the infants began at discharge from the neonatal nursery and continued until each child was a minimum of 36 months of age. The intervention included family support and an education curriculum focused on the healthy development of the child (48). Infants included in the intervention did not differ from control participants with regard to baseline sociodemographic or child health measures. Intervention status (treatment group) was also not associated with systolic or diastolic blood pressure at any time point, and sensitivity analyses carried out in children who did not receive the intervention showed the same pattern of findings as those reported for the full sample.

Participants

Analyses focused on 407 AA children and 264 EA children. The total sample size of 671 was consistent across all models presented below. Descriptive statistics for the sample are shown in Table 1. Hispanics were excluded due to an insufficient sample size (n = 82). Participants with missing data on person-level or neighborhood-level covariates (n = 106) and those without any blood pressure data (n = 126) were excluded. Relative to the analytical sample, children who were excluded because of missing data were more likely to have a parent who smoked (P < 0.001) and were less likely to be AA (P = 0.043). No other differences regarding variables shown in Table 1 were identified. Birth weights in the analytical sample ranged from 540 g to 2,500 g (mean = 1,793 g; standard deviation (SD), 458).

Table 1.

Characteristics of a Sample of African-American (n = 407) and European-American (n = 264) Low-Birth-Weight Preterm Infants, Infant Health and Development Program, 1987–1992

Level and Characteristic	African Americans		European Americans		Racial Differencea
	Mean (SD)	%	Mean (SD)	%	Cohen's d	P Value
Level 1b
Systolic blood pressure, mm Hg	100.2 (10.6)		101.2 (11.1)		0.09	0.33
Diastolic blood pressure, mm Hg	58.8 (8.9)		59.0 (10.2)		0.02	0.84
Body mass indexc	15.8 (1.4)		15.7 (1.3)		0.07	0.53
Level 2
Birth and demographic characteristics
Child sex (female)		52.0		50.0		0.58
Birth weight, g	1,750 (470)		1,860 (434)		0.23	0.003
Gestational age, weeks	33.1 (2.7)		33.0 (2.6)		0.05	0.52
Fetal growth restriction status (yes)d		21.9		14.1		0.003
Neonatal health indexe	103.7 (13.7)		96.9 (16.3)		0.46	<0.001
Maternal health and prenatal care
Gestational weight gain, poundsf	22.1 (13.3)		25.9 (11.8)		0.30	<0.001
Trimester in which prenatal care begang	0.61 (0.83)		0.26 (0.62)		0.48	<0.001
Smoked during pregnancy		31.5		34.0		0.50
Preeclampsia or eclampsia		16.9		15.5		0.61
Gestational diabetes		1.7		1.4		0.67
Hypertension or heart disease		5.6		3.8		0.27
Obesity (weight >200 pounds (>91 kg))		5.0		2.0		0.03
Adolescent pregnancy (age <19 years)		22.1		5.7		<0.001
Later-age pregnancy (age >30 years)		11.8		28.8		<0.001
Family socioeconomic status
Maternal education
Less than high school graduate		45.5		22.0		<0.001
High school graduate		25.4		33.2		0.03
Some college		17.4		23.5		0.06
College graduate		3.9		29.2		<0.001
Income-to-needs ratio	1.1 (1.0)		3.0 (2.0)		1.18	<0.001
Level 3
Mean neighborhood income, thousands of  dollars/year	15.8 (5.2)		24.5 (7.3)		1.37	<0.001

Abbreviation: SD, standard deviation.

^a Cohen's d statistic indicates the magnitude of the racial difference for continuous variables. Independent sample t tests and χ² tests were used to determine the statistical significance of group differences.

^b Level 1 variables were assessed at all 6 time points. Descriptive statistics are shown only for the 24-month assessments. No racial differences in body mass index were detected at any time.

^c Weight (kg)/height (m)².

^d Infants with a birth weight in the bottom 10th percentile for gestational age (using Lubchenco Growth Standards) were coded as having fetal growth restriction (50).

^e The Neonatal Health Index (51) was standardized to have mean score of 100 in the full study sample (SD, 16), with higher scores indicating a shorter-than-average discharge time for a given birth weight.

^f 1 pound = 0.45 kg.

^g 0 = first trimester, 1 = second trimester, 2 = third trimester, and 3 = no prenatal care.

Measures

Outcome: blood pressures

Resting SBP and DBP were calculated at ages 24, 30, 36, 48, 60, and 78 months as the average of 3 readings taken at each time point. Readings were taken on the child's right arm with the child in a seated position. Consistent with previous research (49), biologically implausible values were set to missing and extreme values were censored at 3 standard deviations from the mean. Dinamap automated oscillometric devices (Critikon, Tampa, Florida) were used for the majority of readings (n = 2,646), and other readings were assessed using auscultation (n = 369). Method of blood pressure measurement (0 = Dinamap, 1 = auscultation) and child's behavioral state during readings (0 = fully or somewhat cooperative, 1 = not very cooperative) were included as controls. On average, children included in analyses had blood pressure data for 4.5 of the 6 time points (SD, 1.4), and children were cooperative in 93% (n = 2,815) of readings.

Birth variables and demographic factors (base model controls)

Data on birth weight, gestational age, and neonatal health were collected from medical records. Birth weight (grams) and gestational age (weeks) were continuous variables. Infants with a birth weight in the bottom 10th percentile for gestational age (using Lubchenco Growth Standards) were coded as having fetal growth restriction (50). Neonatal Health Index scores were calculated on the basis of length of stay in the neonatal nursery, adjusted for birth weight (51). Information on child race and sex was obtained from medical records. Body mass index, assessed at all 6 time points and calculated as weight in kilograms divided by squared height in meters, was also included as a time-varying covariate in all models (52).

Prenatal care and maternal health

Several measures relating to infant prenatal experience were also included as controls. Gestational diabetes and preeclampsia or eclampsia were coded from medical records, and gestational weight gain was calculated as the difference between maternal reports of weight before conception and at delivery. Each of these variables has been associated with child blood pressure (53, 54). Smoking status during pregnancy (1 = smoked, 0 = did not smoke), trimester in which mother began receiving prenatal care (0 = first trimester; 3 = no prenatal care), adolescent pregnancy (<19 years), and later-age pregnancy (>30 years) were also included due to known associations with child blood pressure (55, 56), as were 2 additional maternal health risk factors assessed from obstetrical medical records: hypertension or heart disease (1 = had condition; 0 = did not have condition) (54) and obesity status (1 = >200 pounds (>91 kg); 0 = ≤200 pounds (≤91 kg)) (57).

Baseline family SES

Income-to-needs ratio was calculated by dividing each family's income, assessed at ages 12 and 24 months, by its corresponding poverty threshold and averaging the 2 assessments together. Three dichotomously coded variables (less than high school graduation, some college, and college degree or more) were used to assess maternal education, with high school graduates being the reference category.

Baseline neighborhood SES

Neighborhood characteristics were constructed by matching family addresses at birth to 1980 US Census data. Mean neighborhood income was used to assess neighborhood socioeconomic characteristics. Income at the neighborhood level is an established indicator of neighborhood deprivation (58, 59). Study participants were distributed across 429 census tracts. The mean neighborhood income was $19,189 per year (SD, $7,415).

Analyses

We fitted a series of multilevel models to examine racial differences in trajectories of SBP and DBP and to consider reasons for these differences. Three-level models were used, in which measures across time (level 1) were nested within individuals (level 2) and neighborhood census tracts (level 3). Equations for each of our models are shown in the Web Appendix (available at http://aje.oxfordjournals.org/). Linear and quadratic growth terms were included in all models. Time was coded as age in months and was centered at the midpoint of the trajectory (51 months).

In a model with linear and quadratic growth terms, centering at the midpoint allows the linear slope parameter to be interpreted not only as the rate of instantaneous growth at the midpoint of the trajectory but also as the “average rate of growth” across the entire trajectory (60). The average rate of growth is a meaningful parameter because it is an index of overall change in blood pressure across the data collection period. Therefore, in the models shown, the interpretation of the racial association with the linear slope represents the difference between AA children and EA children in the average rate of growth across the trajectory.

Random effects were estimated for the level 2 intercept and linear slope and the level 3 intercept. The quadratic slope was estimated as a fixed effect because significant variance was not detected. The magnitude of racial differences in blood pressure levels at the beginning (age 24 months) and end (age 78 months) of the trajectory was also assessed. P < 0.05 was selected to indicate statistical significance, using 2-sided tests in multilevel regression models.

Variance components in null models showed that, for SBP, 11% of the total variance in the data existed across neighborhoods (level 3), 18% across people within neighborhoods (level 2), and 71% within individuals over time (level 1). For DBP, the corresponding variance components at levels 3, 2, and 1 were 5%, 17%, and 77%.

A series of 4 models was fitted for SBP and DBP. All level 2 and level 3 covariates were entered as predictors of the intercept and the linear rate of change. Model 1 and all subsequent models included blood pressure measurement technique, cooperation during blood pressure reading, and body mass index as time-varying covariates at level 1, as well as the following level 2 variables: race, sex, treatment group, birth weight, gestational age, fetal growth restriction, and Neonatal Health Index. The quadratic growth term was also allowed to vary as a function of race. Model 2 added the following level 2 variables relating to prenatal care and maternal health: gestational weight gain, trimester in which prenatal care began, smoking during pregnancy, obesity (>200 pounds (>91 kg)), heart condition or hypertension, gestational diabetes, preeclampsia or eclampsia, and maternal age. Model 3 adjusted for all of the above covariates and added family SES variables at level 2: income-to-needs ratio and maternal education. In model 4, we added neighborhood income at level 3 to consider its association with blood pressure trajectories after accounting for family SES variables and all other covariates. We conducted sensitivity analyses to consider racial disparities separately for males and females (results not shown). No race × sex interactions were found.

RESULTS

Estimated SBP and DBP trajectories for African Americans and European Americans are shown in Figures 1 and 2. We show separate sets of trajectories for models 2–4 in each figure in order to help the reader visualize the degree to which racial differences in blood pressure were explained by family and neighborhood SES. Tables 2 and 3 show parameter estimates from growth models for SBP and DBP, respectively.

Figure 1.

Estimated systolic blood pressure trajectories for African-American (AA) and European-American (EA) children (models 2–4), Infant Health and Development Program, 1987–1992. Black lines represent AA children, and gray lines represent EA children. Solid lines represent fitted plots from model 2, which included the following variables: level 1—linear slope, quadratic slope, blood pressure measurement technique, child's cooperation during blood pressure reading, and body mass index; level 2—race, sex, treatment group, birth weight, gestational age, fetal growth restriction, Neonatal Health Index, gestational weight gain, trimester in which prenatal care began, smoking during pregnancy, obesity, heart condition or hypertension, gestational diabetes, preeclampsia or eclampsia, and maternal age. Dashed lines represent fitted plots from model 3, which added income-to-needs ratio and maternal education at level 2. Dotted lines represent fitted plots from model 4, which added neighborhood mean income at level 3.

Figure 2.

Estimated diastolic blood pressure trajectories for African-American (AA) and European-American (EA) children (models 2–4), Infant Health and Development Program, 1987–1992. Black lines represent AA children, and gray lines represent EA children. Solid lines represent fitted plots from model 2, which included the following variables: level 1—linear slope, quadratic slope, blood pressure measurement technique, child's cooperation during blood pressure reading, and body mass index; level 2—race, sex, treatment group, birth weight, gestational age, fetal growth restriction, Neonatal Health Index, gestational weight gain, trimester in which prenatal care began, smoking during pregnancy, obesity, heart condition or hypertension, gestational diabetes, preeclampsia or eclampsia, and maternal age. Dashed lines represent fitted plots from model 3, which added income-to-needs ratio and maternal education at level 2. Dotted lines represent fitted plots from model 4, which added neighborhood mean income at level 3.

Table 2.

Estimation of Racial Differences in Systolic Blood Pressure in a Series of Hierarchical Growth Models, With Consideration of Family and Neighborhood Socioeconomic Indicators as Explanations for Racial Differences in Growth Parameters, Infant Health and Development Program, 1987–1992

Predictor Variable	Unstandardized Parameter Estimate (B (SE))
	Model 1a	Model 2b	Model 3c	Model 4d
Intercept (level at age 51 months), mm Hg	99.67e (0.50)	99.66e (0.52)	99.74e (0.55)	99.67e (0.56)
AA race (referent: EA)	1.45f (0.64)	1.44f (0.66)	1.33 (0.72)	1.48f (0.75)
Household income-to-needs ratio			0.21 (0.67)	0.08 (0.69)
Maternal education
Less than high school graduate			−0.62 (0.63)	−0.58 (0.63)
High school graduate (referent)
Some college			−1.33 (0.70)	−1.32 (0.70)
College graduate			−1.37 (0.94)	−1.39 (0.94)
Neighborhood mean income, thousands of dollars per year				0.03 (0.05)
Average linear rate of change, mm Hg/month	0.047g (0.015)	0.048g (0.016)	0.037f (0.017)	0.045g (0.017)
AA race (referent: EA)	0.053g (0.019)	0.051f (0.020)	0.070g (0.023)	0.050f (0.024)
Household income-to-needs ratio			0.023 (0.025)	0.040 (0.026)
Maternal education
Less than high school graduate			−0.001 (0.024)	−0.006 (0.024)
High school graduate (referent)
Some college			−0.018 (0.026)	−0.019 (0.026)
College graduate			0.051 (0.035)	0.054 (0.034)
Neighborhood mean income, thousands of dollars per year				−0.004f (0.002)
Quadratic rate of change, mm Hg/month2	0.0032e (0.0008)	0.0032e (0.0008)	0.0033e (0.0008)	0.0034e (0.0008)
AA race (referent: EA)	−0.0022f (0.0010)	−0.0024f (0.0010)	−0.0025f (0.0010)	−0.0025f (0.0010)

Abbreviations: AA, African-American; EA, European-American; SE, standard error.

^a Model 1 (base model) variables: level 1—linear slope, quadratic slope, blood pressure measurement technique, child's cooperation during blood pressure reading, and body mass index; level 2—race, sex, treatment group, birth weight, gestational age, fetal growth restriction, and Neonatal Health Index.

^b Model 2 added gestational weight gain, trimester in which prenatal care began, smoking during pregnancy, obesity, heart condition or hypertension, gestational diabetes, preeclampsia or eclampsia, and maternal age (level 2).

^c Model 3 added income-to-needs ratio and maternal education (level 2).

^d Model 4 added neighborhood mean income (level 3).

^eP < 0.001.

^fP < 0.05.

^gP < 0.01.

Table 3.

Estimation of Racial Differences in Diastolic Blood Pressure in a Series of Hierarchical Growth Models, With Consideration of Family and Neighborhood Socioeconomic Indicators as Explanations for Racial Differences in Growth Parameters, Infant Health and Development Program, 1987–1992

Predictor Variable	Unstandardized Parameter Estimate (B (SE))
	Model 1a	Model 2b	Model 3c	Model 4d
Intercept (level at 51 months), mm Hg	55.72e (0.44)	55.92e (0.45)	56.13e (0.47)	56.23e (0.48)
AA race (referent: EA)	3.23e (0.55)	2.92e (0.57)	2.59e (0.62)	2.36e (0.66)
Household income-to-needs ratio			−0.35 (0.57)	−0.19 (0.59)
Maternal education
Less than high school graduate			−0.24 (0.55)	−0.29 (0.55)
High school graduate (referent)
Some college			−0.70 (0.60)	−0.69 (0.60)
College graduate			−1.08 (0.81)	−1.02 (0.81)
Neighborhood mean income, thousands of  dollars per year				−0.05 (0.04)
Average linear rate of change, mm Hg/month	−0.037f (0.014)	−0.031g (0.014)	−0.041f (0.015)	−0.034g (0.015)
AA race (referent: EA)	0.061e (0.017)	0.051f (0.018)	0.068e (0.021)	0.052g (0.022)
Household income-to-needs ratio			0.033 (0.023)	0.046 (0.023)
Maternal education
Less than high school graduate			−0.004 (0.022)	−0.008 (0.022)
High school graduate (referent)
Some college			−0.017 (0.024)	−0.018 (0.024)
College graduate			0.003 (0.032)	0.005 (0.031)
Neighborhood mean income, thousands of  dollars per year				−0.003g (0.001)
Quadratic rate of change, mm Hg/month2	0.0032e (0.0008)	0.0032e (0.0008)	0.0032e (0.0008)	0.0032e (0.0008)
AA race (referent: EA)	−0.0023g (0.0010)	−0.0024g (0.0010)	−0.0024g (0.0010)	−0.0024g (0.0010)

Abbreviations: AA, African-American; EA, European-American; SE, standard error.

^a Model 1 (base model) variables: level 1—linear slope, quadratic slope, blood pressure measurement technique, child's cooperation during blood pressure reading, and body mass index; level 2—race, sex, treatment group, birth weight, gestational age, fetal growth restriction, and Neonatal Health Index.

^b Model 2 added gestational weight gain, trimester in which prenatal care began, smoking during pregnancy, obesity, heart condition or hypertension, gestational diabetes, preeclampsia or eclampsia, and maternal age (level 2).

^c Model 3 added income-to-needs ratio and maternal education (level 2).

^d Model 4 added neighborhood mean income (level 3).

^eP < 0.001.

^fP < 0.01.

^gP < 0.05.

Systolic blood pressure

Results from models 1–4 for SBP are shown in Table 2, and results from models 2–4 are depicted in Figure 1. The trajectory for EA children had a concave curvature characterized by small decreases in SBP between the ages of approximately 24 and 42 months, followed by increases. The trajectory for AA children showed less curvature (B = −0.002, 95% confidence interval (CI): −0.004, 0.000; P = 0.034) and thus more consistent increases across the data collection period. Based on model 1, a significant racial difference in SBP level was also present at age 51 months (B = 1.45 mm Hg, 95% CI: 0.20, 2.69; P = 0.023) and in the average rate of change across the trajectory (B = 0.053 mm Hg per month, 95% CI: 0.016, 0.090; P = 0.006), such that AA children had elevated SBP and a 0.64-mm Hg greater average increase in SBP per year than EA children. Overall, race accounted for 9% of the variance in the linear slope and 0% of the total intercept variance. Post hoc estimates indicated that AA children had a marginally significantly lower SBP level at age 24 months (B = −1.57, 95% CI: −0.01, 3.15; P = 0.052) and a more positive linear rate of change at this time point (B = 0.17, 95% CI: 0.057, 0.285; P = 0.004). At age 78 months, race was not a significant predictor of SBP level (B = 1.29, P = 0.14) or linear rate of change (B = −0.065, P = 0.28).

Racial differences in SBP level at age 51 months and average rate of change were robust to the inclusion of prenatal care and maternal health risk factors (model 2). In model 3, maternal education and family income were added but were not significantly associated with SBP level at age 51 months or average linear rate of change in SBP. Neighborhood income was added in model 4 and was inversely associated with the average SBP rate of change (B = −0.004, 95% CI: −0.008, 0.000; P = 0.011); for each standard-deviation ($7,400) lower mean neighborhood income, children experienced a 0.36-mm Hg greater increase in SBP per year. Neighborhood income also accounted for 16% of the variance in the average rate of growth and 29% of the racial difference in this parameter. As depicted in Figure 1, although some of the racial difference in the slope was explained, a relatively small portion of racial differences in the overall trajectory was accounted for by covariates.

Diastolic blood pressure

Results from models 1–4 for DBP are shown in Table 3, and results from models 2–4 are depicted in Figure 2. The trajectory for EA children had a concave curvature characterized by decreases in SBP between the ages of approximately 24 and 56 months, followed by increases. For AA children, the trajectory showed less curvature (B = −0.002, 95% CI: −0.004, 0.000; P = 0.018) and was characterized by stability between the ages of approximately 24 and 48 months, followed by gradual increases. In model 1, racial differences were also present in DBP levels at age 51 months (B = 3.23, 95% CI: 2.15, 4.31; P < 0.001) and in the average rate of change (B = 0.061, 95% CI: 0.028, 0.094; P < 0.001), such that AA children had elevated DBP levels and experienced, on average, a 0.73-mm Hg greater increase in DBP per year than EA children. Overall, race accounted for 25% of the variance in the linear slope and 6% of the total intercept variance at age 51 months (at levels 2 and 3). Post hoc estimates indicated that racial differences in DBP level at age 24 months were not significant (B = −0.07, P = 0.92), but racial differences in initial linear rate of change were evident (B = 0.18, 95% CI: 0.08, 0.29; P < 0.001). Furthermore, estimates of DBP levels at age 78 months indicated significant racial differences (B = 3.23, 95% CI: 1.67, 4.79; P < 0.001), with AA children having higher DBP levels than EA children.

Addition of controls relating to prenatal care and maternal health (model 2) attenuated racial differences in the average DBP linear rate of change and level at age 51 months by 16% and 10%, respectively, yet racial differences remained significant. In model 3, maternal education and family income were not significantly associated with DBP linear slope or level at age 51 months, and identified racial differences were not attenuated. However, higher neighborhood income (model 4) was associated with a flatter average DBP linear rate of change (B = −0.003, 95% CI: −0.001, −0.005; P = 0.024) and accounted for 16% of the variance in the linear rate of change. Neighborhood income also attenuated the racial difference in the average DBP linear rate of change by 24%.

DISCUSSION

In this study, we examined blood pressure trajectories from age 24 months to age 78 months among low-birth-weight, preterm AA and EA children and the degree to which racial differences emerged across this period. After controlling for birth characteristics, time-varying body mass index, and maternal health risk factors, AA children showed steeper increases in SBP and DBP over time than did EA children. These findings are congruent with studies showing a divergence in health outcomes across racial groups in early childhood (61) and provide support for the notion that racial disparities among adults, in part, have their origin in the first decade of life (4, 62). The results also extend previous work focusing on racial differences in blood pressure trajectories among full-term children of normal birth weight (11, 12) and suggest that in samples of preterm, low-birth-weight children, the divergence in blood pressure may occur earlier than in samples of normal-birth-weight children, among whom disparities have been shown to emerge between the ages of 9 and 15 years (11–13, 63). These findings are important given links between blood pressure in early childhood and subsequent hypertension and cardiovascular disease (7, 64), and they demonstrate the need for further research on racial disparities in the health trajectories of preterm infants across the life course.

An additional focus of this study was consideration of family- and neighborhood-level SES at birth as explanations for racial differences in subsequent blood pressure trajectories. Results of these analyses indicated that neighborhood SES accounted for a portion of the racial difference in blood pressure trajectories, particularly for DBP, while family-level SES did not. This finding is consistent with recent research showing that neighborhood conditions explain a larger portion of racial disparities in hypertension than do socioeconomic factors at the individual level (45–47). Although additional research is needed to determine specific mechanisms for possible effects of neighborhood SES on child blood pressure, current conceptualizations of neighborhood socioeconomic disadvantage suggest that multiple pathways of influence are likely at play (27–31). Elevated stress exposure, weaker support structures, and diminished social capital are all viable candidate mechanisms for the apparent impacts of disadvantaged neighborhoods (32, 33, 65). Such factors influence parental well-being, family functioning, and in turn, the quality of caregiving provided to a child (34–36). Another major pathway relates to the physical environment of neighborhoods, such as the availability of green spaces, playgrounds, and other child resources (37). Lastly, neighborhood SES may also influence child blood pressure through exposure to toxins and air pollutants (39–41).

Some limitations of this study and future research directions should be noted. First, the analyses presented considered family and neighborhood characteristics at the time of birth as predictors of subsequent blood pressure trajectories. Thus, although both family SES and neighborhood of residence tend to be relatively stable over short periods of time, studies that consider the importance of variability in time scales of exposure to neighborhood and family SES across childhood are needed. Because AA families are more likely to experience drops in income and downward mobility into areas with higher levels of poverty (66, 67), accounting for residential mobility and changes in family SES may explain additional variance in group disparities. Furthermore, because neighborhood SES may in part capture unmeasured aspects of family SES, such as home value, fuller assessment of family SES may more clearly elucidate family and neighborhood influences. Second, since only a few explanations for racial differences were assessed, further research should identify additional social determinants of group differences. Examples include stressors that influence parental behaviors and health, such as racial discrimination and harsh workplace environments, and children's health behaviors, including diet and exercise.

A third limitation relates to the length of time since the data were collected. Because data collection was completed in the early 1990s, sociocultural influences on blood pressure may have changed since that time. However, several findings suggest that racial disparities in blood pressure among children may be just as salient today as they were 2 decades ago. National trend data indicate that child blood pressure levels have been rising since 1988, and racial disparities in child blood pressure may also have increased (68). Furthermore, racial differences in neighborhood SES remain substantial (69–71). Since (to our knowledge) no previous studies have considered racial disparities in the blood pressure trajectories of low-birth-weight, preterm children, this study can become a building block for future research.

The results of this study show that racial disparities in blood pressure among low-birth-weight, preterm AA and EA children become evident across the early childhood years and that neighborhood SES explains a portion of racial differences in blood pressure trajectories. Additional research aimed at understanding contextual risk factors for preterm AA children is warranted and may help to address the early roots of cardiovascular health disparities.