To evaluate the effect of grandmother and mother educational attainment on low birth weight (LBW) in children and grandchildren.
The National Longitudinal Study of Adolescent to Adult Health is a multigenerational study that collected survey data from 1994 to 2018. Using this database, we constructed a cohort of 2867 non-Hispanic Black (NHB) and non-Hispanic White (NHW) grandmother-mother-grandchild triads to evaluate how education affects the likelihood of having LBW children and grandchildren, while adjusting for socioeconomic and maternal health factors using multivariable logistic regression.
Similar to previous studies, NHB women were more likely to have LBW descendants compared with NHW women in unadjusted and adjusted analyses. The prevalence of LBW descendants was lower in women with college education, regardless of race. Irrespective of race, mother and grandmother college education was associated with decreased odds of LBW children and grandchildren after adjusting for individual variables. When mother and grandmother education were examined together, and after adjusting for all individual, community, and health variables together, mother college education remained associated with lower odds of LBW (adjusted odds ratio, 0.58; 95% confidence interval, 0.44-0.77). There were no statistically significant differences in these effects between NHW and NHB populations.
Educational attainment in mothers is associated with decreased odds of LBW descendants after adjusting for multiple individual, community, and health covariates, regardless of race. Targeting improvements in education may ameliorate adverse pregnancy outcomes that disproportionately affect minority communities and cause significant lifelong consequences.
What’s Known on this Subject:
Compared with White women, Black women are 2 times more likely to have low birth weight infants. Increased educational attainment may lead to increased racial disparities in adverse pregnancy outcomes, including low birth weight; however, results of past research are inconsistent.
What This Study Adds:
In a subgroup of non-Hispanic Black and White mothers and grandmothers from a nationally representative, multigenerational database, increased educational attainment was associated with decreased rates of low birth weight in children and grandchildren regardless of race and without increasing disparities.
Racial disparities in adverse pregnancy outcomes are a critical problem in the United States. Non-Hispanic Black (NHB) women are 2 times more likely to deliver a low birth weight (LBW) infant than non-Hispanic White (NHW) women.1 LBW is associated with an increased risk of infant mortality, inferior health and development outcomes in childhood, and worse health in adulthood.2–11 Disparities in health outcomes not only lead to worse outcomes for disadvantaged individuals, but also worse population-level health measures.12
The social determinants of health are likely responsible for the majority of racial disparities in pregnancy outcomes, potentially mediated by epigenetic modifications and a complex array of biologic pathways.13–16 They can be broken down into individual factors, such as income, education, and discrimination, and community factors, such as health care access, crime, and pollution.17–23 Education is a key determinant in many health indicators in children and adults, but the extent to which it affects adverse pregnancy outcomes has been less well studied, with studies reporting inconsistent results, especially the differential impact of education on Black and White women.24,25 Several studies suggest increased educational attainment leads to worse disparities secondary to weaker improvements in adverse pregnancy outcomes for Black women,26–29 whereas other studies show improved pregnancy outcomes in both Black and White women with increased educational attainment.30,31 Additionally, there are challenges in applying research from past decades26,28–32 to the current era because of significant changes in fertility and family structure, including older and more educated mothers, more working mothers, fewer 2 parent households, and fewer households with married parents.33 These factors highlight the need for greater study of the relationship between education and birth weight using recent data.
The life-course framework posits that women’s pregnancy outcomes are influenced by early exposures and chronic stressors and therefore best studied with longitudinal and multigenerational datasets.34–39 However, most past outcomes research has drawn its conclusions from simpler datasets because longitudinal and multigenerational datasets are rare. To fill these gaps, we designed this study to assess the impact of transgenerational educational attainment on the risk of having an LBW infant in NHW and NHB families using the National Longitudinal Study of Adolescent to Adult Health (Add Health) database.
Methods
Sample
Add Health is a longitudinal, nationally representative, and multigenerational survey-based study designed to understand how health behaviors in adolescence shape health outcomes in adulthood. The study selected a stratified sample of schools and enrolled more than 20 000 adolescents in grades 7 to 12 to create a nationally representative cohort. Researchers interviewed participants during 5 waves of data collection from 1994 to 2018 and linked parent interviews and community statistics to participant files. There was 77% to 88% retention between waves of data collection. Participants were aged 24 to 32 years old by wave 4. A complete description of the Add Health study sample and design is reported elsewhere.40,41
Our study evaluated a subgroup of NHW and NHB biologically related grandmother-mother (primary study participant)-grandchild triads using data from waves 1, 3, and 4 (wave 5 data were not available when this study was completed). Exclusion criteria included participants missing a survey weight, participants that did not have children during the study period, participants that had children missing the primary outcome (birth weight), grandmothers not biologically related to participants, Hispanic ethnicity, and all races aside from White or Black. This left 2867 biologically related grandmother-mother-grandchild triads (Fig 1). To account for the survey design, survey weights were applied to ensure the validity of population-based conclusions.
Add Health received institutional review board (IRB) approval from the University of North Carolina for completion of the primary study. The Children’s Hospital of Philadelphia IRB deemed secondary analysis of deidentified Add Health data nonhuman subjects research; therefore, IRB approval was not necessary for this project.
Birth Outcome
The primary outcome of interest was the birth of any LBW grandchild, coded as a binary variable. LBW was defined as an infant born weighing less than 2500 g. Grandchild birth weight was obtained from wave 3 and wave 4 surveys administered to mothers from the questions, “What was [baby’s] birthweight?” and “How much did [baby] weigh at birth?”, respectively. To avoid duplicates, if a grandchild was included in wave 3, he or she was excluded from wave 4. Individuals missing the outcome of interest were excluded.
Variables
The survey questions and answers used as variables in this study are displayed in Table 1.
Individual and Community Variables
| Variables | Wave(s) | Survey Question | Survey Answer Choicesa | Recoded |
|---|---|---|---|---|
| Individual variables | ||||
| Race | ||||
| Grandmother | 1 | “What is your race?” “Which one best describes your racial background?” | White, Black, American Indian, Asian or Pacific Islander, Other | Excluded respondents that answered Hispanic to previous question (not shown) |
| Mother | 1 | Constructed variable from questions: “What is your race?” “Which one best describes your racial background?” | White, Black,Native American, Asian | N/A |
| Nativity status | ||||
| Grandmother | 1 | “Were you born in the US?” | Yes, no | N/A |
| Mother | 1 | “Were you born in the US?” | Yes, no | N/A |
| Education | ||||
| Grandmother | 1 | “How far in school did [the woman who served as your mother] go?”b Asked of primary study participants | 1 = 8th grade or less to 12 = respondent does not know if she went to school | Binary variable: Any college education or no college education |
| Mother | 4 | “What is the highest level of education you have achieved to date?” | 1 = 8th grade or less to 13 = Completed postbaccalaureate professional education | Binary variable: any college education or no college education |
| Marital status | ||||
| Grandmother | 1 | “Have you ever been married?” | Yes, no | N/A |
| Mother | 4 | Constructed variable from wave 4 question: “How many persons have you ever married? Be sure to include your current spouse if you are married now.” | Yes, no | N/A |
| Household income | ||||
| Grandmother | 1 | “About how much total income, before taxes, did your family receive in 1994?” | $0–$999 000 | 4 quantiles |
| Mother | 4 | “Thinking about your income and the income of everyone who lives in your household and contributes to the household budget, what was the total household income before taxes and deductions in [2006/2007/2008]? Include all sources of income, including nonlegal sources.” | 1 = <$5000 to 12 = $150 000 or more | 4 quantiles |
| Employment | ||||
| Grandmother | 1 | “Does [your mom/dad] work for pay?” Asked of primary study participants | Yes, no | N/A |
| Mother | 4 | “Are you currently working for pay at least 10 hours a week?” | Yes, no | N/A |
| Community variablesc | Wave(s) | Variable | Range of answersd | Data source |
| Education | 1, 4 | Proportion of individuals in the census tract > 25 y old with a college degree | W1: 0-1 W4: 0-1 | Census tract |
| Income | 1, 4 | Median household income | W1: $4999–$125 053 W4: $4800–$171 600 | Census tract |
| Unemployment | 1, 4 | Proportion unemployed | W1: 0–0.66 W4: 0–0.70 | Census tract |
| Poverty | 1, 4 | Proportion of households living below federal poverty level | W1: 0–1 W4: 0–1 | Census tract |
| Total arrests | 1, 4 | Total adult arrests/100 000 population | W1: 0–9404 W4: 0–13 000 | County data |
| Violent arrests | 1, 4 | Total violent arrests/100 000 population | W1: 0–5535 W4: 0–500 | County data |
| Variables | Wave(s) | Survey Question | Survey Answer Choicesa | Recoded |
|---|---|---|---|---|
| Individual variables | ||||
| Race | ||||
| Grandmother | 1 | “What is your race?” “Which one best describes your racial background?” | White, Black, American Indian, Asian or Pacific Islander, Other | Excluded respondents that answered Hispanic to previous question (not shown) |
| Mother | 1 | Constructed variable from questions: “What is your race?” “Which one best describes your racial background?” | White, Black,Native American, Asian | N/A |
| Nativity status | ||||
| Grandmother | 1 | “Were you born in the US?” | Yes, no | N/A |
| Mother | 1 | “Were you born in the US?” | Yes, no | N/A |
| Education | ||||
| Grandmother | 1 | “How far in school did [the woman who served as your mother] go?”b Asked of primary study participants | 1 = 8th grade or less to 12 = respondent does not know if she went to school | Binary variable: Any college education or no college education |
| Mother | 4 | “What is the highest level of education you have achieved to date?” | 1 = 8th grade or less to 13 = Completed postbaccalaureate professional education | Binary variable: any college education or no college education |
| Marital status | ||||
| Grandmother | 1 | “Have you ever been married?” | Yes, no | N/A |
| Mother | 4 | Constructed variable from wave 4 question: “How many persons have you ever married? Be sure to include your current spouse if you are married now.” | Yes, no | N/A |
| Household income | ||||
| Grandmother | 1 | “About how much total income, before taxes, did your family receive in 1994?” | $0–$999 000 | 4 quantiles |
| Mother | 4 | “Thinking about your income and the income of everyone who lives in your household and contributes to the household budget, what was the total household income before taxes and deductions in [2006/2007/2008]? Include all sources of income, including nonlegal sources.” | 1 = <$5000 to 12 = $150 000 or more | 4 quantiles |
| Employment | ||||
| Grandmother | 1 | “Does [your mom/dad] work for pay?” Asked of primary study participants | Yes, no | N/A |
| Mother | 4 | “Are you currently working for pay at least 10 hours a week?” | Yes, no | N/A |
| Community variablesc | Wave(s) | Variable | Range of answersd | Data source |
| Education | 1, 4 | Proportion of individuals in the census tract > 25 y old with a college degree | W1: 0-1 W4: 0-1 | Census tract |
| Income | 1, 4 | Median household income | W1: $4999–$125 053 W4: $4800–$171 600 | Census tract |
| Unemployment | 1, 4 | Proportion unemployed | W1: 0–0.66 W4: 0–0.70 | Census tract |
| Poverty | 1, 4 | Proportion of households living below federal poverty level | W1: 0–1 W4: 0–1 | Census tract |
| Total arrests | 1, 4 | Total adult arrests/100 000 population | W1: 0–9404 W4: 0–13 000 | County data |
| Violent arrests | 1, 4 | Total violent arrests/100 000 population | W1: 0–5535 W4: 0–500 | County data |
N/A, not applicable; W, wave.
Where applicable, all choices included options for “I don't know,” “Refused,” “Legitimate Skip,” and “Missing.”
Excluded biologically unrelated grandparents.
Community variables relevant to grandmothers were taken from wave 1 data and community variables relevant to mothers were taken from wave 4 data.
Where applicable, all choices included options for “I don't know,” “Refused,” “Legitimate Skip,” and “Missing.” Range of answers for community variables education, unemployment, poverty represent proportions of the population.
Individual Variables
The focus of this study was to evaluate the impact of educational attainment, coded as binary variable “any college education,” on the risk of the birth of an LBW infant. Other variables included were based on known or plausible demographic and socioeconomic factors that affect birth weight: race, nativity, marital status, household income, and employment. Grandmother variables represent her socioeconomic and sociodemographic status before the birth of her grandchild (ren). Mother educational attainment and other demographic variables were selected from the final wave of data collection. Because the primary outcome was the birth of any LBW infant, time-varying covariates such as mother’s education, which may differ for each pregnancy, could have occurred before or after the birth of her child (ren). The percent of missing variables is reported in Table 2.
Grandmother and Mother Characteristics
| Grandmother | Mother | |
|---|---|---|
| Race | ||
| White | 78.5% | 76.4% |
| Black | 21.5% | 23.2% |
| Any college education | ||
| No | 54.9% | 27.1% |
| Yes | 44.9% | 72.9% |
| Missing | 0.1% | 0.0% |
| Household income quantile | ||
| 1 | 25.8% | 40.4% |
| 2 | 24.0% | 32.6% |
| 3 | 24.1% | 12.2% |
| 4 | 14.2% | 8.9% |
| Missing | 12.0% | 5.9% |
| US born | ||
| No | 2.5% | 1.2% |
| Yes | 97.5% | 98.8% |
| Employed | ||
| No | 21.5% | 31.3% |
| Yes | 66.7% | 56.0% |
| Missing | 11.8% | 12.7% |
| Married | ||
| No | 5.9% | 30.7% |
| Yes | 93.8% | 69.3% |
| Missing | 0.3% | 0.0% |
| Grandmother | Mother | |
|---|---|---|
| Race | ||
| White | 78.5% | 76.4% |
| Black | 21.5% | 23.2% |
| Any college education | ||
| No | 54.9% | 27.1% |
| Yes | 44.9% | 72.9% |
| Missing | 0.1% | 0.0% |
| Household income quantile | ||
| 1 | 25.8% | 40.4% |
| 2 | 24.0% | 32.6% |
| 3 | 24.1% | 12.2% |
| 4 | 14.2% | 8.9% |
| Missing | 12.0% | 5.9% |
| US born | ||
| No | 2.5% | 1.2% |
| Yes | 97.5% | 98.8% |
| Employed | ||
| No | 21.5% | 31.3% |
| Yes | 66.7% | 56.0% |
| Missing | 11.8% | 12.7% |
| Married | ||
| No | 5.9% | 30.7% |
| Yes | 93.8% | 69.3% |
| Missing | 0.3% | 0.0% |
Community Variables
Community variables, which provided contextual information on Add Health participants, were treated as potential confounders in this study because past research has shown that community attributes may have a significant impact on birth outcomes.3 Add Health researchers linked census tract data (except for arrests, which were reported by county) to each respondent based on address global positioning system coordinates at each wave of data collection.40 Variables were chosen based on factors known or suspected to have an association with adverse health outcomes: education, income, unemployment, poverty, and arrests. There were no missing values for community variables.
Pregnancy and Health Covariates
The pregnancy and mother health covariates in this study were selected based on factors that have been shown to have an impact on pregnancy outcomes in past research. They were selected from wave 3 and 4 questions on mothers’ insurance status, mothers’ personal history of LBW, overweight/obese (body mass index ⩾ 25), hypertension, diabetes, oral disease, depression, cigarette use during pregnancy, alcohol use during pregnancy, and prenatal care established in the first trimester. The proportion of variables missing for these factors did not differ between the LBW and non-LBW groups (Supplemental Table 4).
Analyses
We used StataSE 16 to complete our analysis. To account for the complex sampling structure of the survey and loss to follow-up through the 4 waves of data collection, Add Health researchers created sampling weights,42 which for our study design was a longitudinal analysis of data from waves 1, 3, and 4. We used these weights in our analyses as per the Stata Survey Data Reference Manual instructions.43
We used χ2 tests to assess general participant characteristics in subgroups stratified by race and educational attainment. t tests are not validated for use in survey analysis of continuous data (community-level variables in this study). As a result, we assessed these variables for normal distribution, obtained means, and then performed regressions to assess statistically significant differences between these continuous variables. Missing variables resulting from skipped questions during survey administration are reported in the results as missing.
Univariable and multivariable analysis were completed with logistic regression. To determine the impact of education on the odds of having an LBW descendant, we controlled for a variety of mother and grandmother individual, community, and health variables in several models of variable breadth. These models were developed based on the outcomes of univariable analysis and included variables that were associated with LBW transgenerationally. For each group of factors, models included all significant variables listed in Table 1. We examined whether the effects of our primary risk factor, education, varied by race/ethnicity and found no statistically significant interaction. This finding was confirmed by an additional multivariable analysis on separate NHB and NHW subgroups, which revealed the similar findings to those described in the following section.
Results
Overall, 14.7% of study participants reported ever giving birth to an LBW infant. In both grandmothers and mothers, ∼20% reported NHB race (Table 2). A higher proportion of mothers (72.9%) reported college education compared with grandmothers (44.9%).
The prevalence of LBW children and grandchildren stratified by grandmother and mother educational attainment and race is displayed in Figs 2 and 3. Black race and no college education were associated with a significantly higher prevalence of LBW for both grandmothers and mothers in unadjusted analysis (refer to Fig 2 legend for rates and P values). College education did not increase the disparity in LBW between racial subgroups. This interaction between race and education on LBW was not statistically significant for mothers (P = .26) or grandmothers (P = .61), suggesting that educational benefits for LBW applied similarly to NHB and NHW women.
Educational attainment in grandmothers and mothers and prevalence of LBW. The vertical axis corresponds to percent of infants born classified as LBW and the horizontal axis corresponds to degree of educational attainment stratified by White race (dark gray) and Black race (light gray). Rates of LBW grandchildren in NHW and NHB grandmothers without college education were 14.85% and 24.31% (P = .004), respectively, and rates in those with college education were 10.25% and 20.03% (P = .002), respectively. Rates of LBW children in NHW and NHB mothers without college education were 19.38% and 27.53% (P = .046), respectively, and rates in those with college education were 10.13% and 19.29% (P < .001), respectively.
Educational attainment in grandmothers and mothers and prevalence of LBW. The vertical axis corresponds to percent of infants born classified as LBW and the horizontal axis corresponds to degree of educational attainment stratified by White race (dark gray) and Black race (light gray). Rates of LBW grandchildren in NHW and NHB grandmothers without college education were 14.85% and 24.31% (P = .004), respectively, and rates in those with college education were 10.25% and 20.03% (P = .002), respectively. Rates of LBW children in NHW and NHB mothers without college education were 19.38% and 27.53% (P = .046), respectively, and rates in those with college education were 10.13% and 19.29% (P < .001), respectively.
Forest plot of univariable analysis of grandmother and mother individual variables and odds of LBW in grandchildren and children, respectively. Individual variables, listed on the left, are in separate rows. Black squares represent ORs and horizontal lines represent 95% CIs. Variables with CI crossing 1, demarcated with a vertical line, had no statistically significant effect on odds of LBW. Black race in grandmothers and mothers was associated with an increased odds of LBW descendants (OR, 2.00; 95% CI, 1.46-2.74 and OR, 1.94; 95% CI, 1.48-2.54, respectively). College education in grandmothers and mothers was associated with a decreased odds of LBW descendants (OR, 0.66; 95% CI, 0.50-0.88 and OR, 0.50; 95% CI, 0.38-0.66, respectively).
Forest plot of univariable analysis of grandmother and mother individual variables and odds of LBW in grandchildren and children, respectively. Individual variables, listed on the left, are in separate rows. Black squares represent ORs and horizontal lines represent 95% CIs. Variables with CI crossing 1, demarcated with a vertical line, had no statistically significant effect on odds of LBW. Black race in grandmothers and mothers was associated with an increased odds of LBW descendants (OR, 2.00; 95% CI, 1.46-2.74 and OR, 1.94; 95% CI, 1.48-2.54, respectively). College education in grandmothers and mothers was associated with a decreased odds of LBW descendants (OR, 0.66; 95% CI, 0.50-0.88 and OR, 0.50; 95% CI, 0.38-0.66, respectively).
The univariable relationship between community and health factors and LBW are shown in Supplemental Tables 5 and 6. Community covariates showed an association between favorable community conditions and decreased odds of LBW (Supplemental Table 5). For health factors, hypertension was the only covariate associated with an increased risk of LBW (odds ratio [OR], 2.21; 95% confidence interval [CI], 1.52-3.22; Supplemental Table 6).
Results of the multivariable analysis examining the association between either grandmother or mother educational attainment and LBW grandchild or child, respectively, are displayed in Table 3. After controlling for income and race, having any college education was associated with a lower odds of having an LBW child or grandchild (adjusted odds ratio [aOR], 0.54; 95% CI, 0.41-0.72; and aOR, 0.74; 95% CI, 0.56-0.98, respectively). For grandmothers, this effect size remained stable when community and mother health variables were added to the model but lost statistical significance. The effect of college education in mothers remained significant after adjusting for individual, community, and health covariates (aOR, 0.55; 95% CI, 0.41-0.74). There were no significant differences in these effects between NHB and NHW women.
Multivariable Logistic Regression Educational Attainment and LBW
| aOR | P value | 95% CI | |
|---|---|---|---|
| Grandmother Education and Adjusted Odds LBW | |||
| Individual variablesa | 0.74 | .04 | 0.56–0.98 |
| Community variablesb | 0.74 | .05 | 0.55–1.00 |
| Individual and communityc | 0.78 | .10 | 0.58–1.05 |
| Individual, community, and mother healthd | 0.77 | .08 | 0.57–1.04 |
| Mother education and adjusted odds LBW | |||
| Individual variablesa | 0.54 | <.001 | 0.41–0.72 |
| Community variablesb | 0.56 | <.001 | 0.43–0.74 |
| Individual and communityc | 0.57 | <.001 | 0.44–0.75 |
| Individual, community, and mother healthd | 0.55 | <.001 | 0.41–0.74 |
| Grandmother and mother education and adjusted odds LBW | |||
| Grandmother | |||
| Individual variablesa | 0.79 | .08 | 0.61–1.03 |
| Community variablesb | 0.84 | .24 | 0.63–1.12 |
| Individual and communityc | 0.86 | .30 | 0.64–1.15 |
| Individual, community, and mother healthd | 0.85 | .26 | 0.64–1.13 |
| Mother | |||
| Individual variablesa | 0.57 | <.001 | 0.43–0.76 |
| Community variablesb | 0.59 | <.001 | 0.45–0.77 |
| Individual and communityc | 0.60 | <.001 | 0.46–0.79 |
| Individual, community, and mother healthd | 0.58 | <.001 | 0.44–0.77 |
| aOR | P value | 95% CI | |
|---|---|---|---|
| Grandmother Education and Adjusted Odds LBW | |||
| Individual variablesa | 0.74 | .04 | 0.56–0.98 |
| Community variablesb | 0.74 | .05 | 0.55–1.00 |
| Individual and communityc | 0.78 | .10 | 0.58–1.05 |
| Individual, community, and mother healthd | 0.77 | .08 | 0.57–1.04 |
| Mother education and adjusted odds LBW | |||
| Individual variablesa | 0.54 | <.001 | 0.41–0.72 |
| Community variablesb | 0.56 | <.001 | 0.43–0.74 |
| Individual and communityc | 0.57 | <.001 | 0.44–0.75 |
| Individual, community, and mother healthd | 0.55 | <.001 | 0.41–0.74 |
| Grandmother and mother education and adjusted odds LBW | |||
| Grandmother | |||
| Individual variablesa | 0.79 | .08 | 0.61–1.03 |
| Community variablesb | 0.84 | .24 | 0.63–1.12 |
| Individual and communityc | 0.86 | .30 | 0.64–1.15 |
| Individual, community, and mother healthd | 0.85 | .26 | 0.64–1.13 |
| Mother | |||
| Individual variablesa | 0.57 | <.001 | 0.43–0.76 |
| Community variablesb | 0.59 | <.001 | 0.45–0.77 |
| Individual and communityc | 0.60 | <.001 | 0.46–0.79 |
| Individual, community, and mother healthd | 0.58 | <.001 | 0.44–0.77 |
Race, education, and income variables.
Education, income, unemployment, and poverty variables.
Individual race, education, income and community education, income, unemployment, and poverty variables.
Individual race, education, income, community education, income, unemployment, poverty, mother insurance, hypertension, diabetes, overweight/obese, depression, oral disease, adequate prenatal care, personal history of LBW, smoking during pregnancy, and alcohol use during pregnancy variables.
The final models included both mother and grandmother factors. Controlling for mother educational attainment, the decreased likelihood of grandmothers with college education having an LBW grandchild lost statistical significance. However, mother educational attainment continued to be associated with a significantly lower odds of having an LBW child (aOR, 0.58; 95% CI, 0.44-0.77). NHB race continued to be associated with increased adjusted odds of LBW in both grandmother and mother models after adjusting for individual and community factors.
Multiple secondary analyses were performed. One analysis looked at a larger sample that included all racial subgroups to evaluate how increased power may affect the results of the study (Supplemental Table 7). Here, increased educational attainment in mothers continued to be associated with a decreased odds (aOR, 0.58; 95% CI, 0.46-0.75) of an LBW child after adjusting for all individual, community, and health factors. Grandmother educational attainment was associated with a decreased odds of an LBW grandchild after adjusting for individual mother and grandmother variables (aOR, 0.76; 95% CI, 0.59-0.98). Another analysis evaluated the impact of age and parity on LBW. Age was obtained at the time of participant interview (age at time of pregnancy was not available in this dataset). In univariate analysis, there was no significant association between age and LBW (wave 3 OR, 1.02; 95% CI, 0.94-1.11 and wave 4 OR, 1.02; 95% CI, 0.94-1.11). Parity also was not associated with LBW (OR, 1.00; 95% CI, 1.00-1.02). Including these variables did not change the effect of education on LBW. The final analysis looked at the impact of grandmothers and mothers with different levels of college education on the odds of LBW (Supplemental Table 8). There were no differences in the rates of LBW between concordant pairs, in which mother and grandmother both had a college education, and discordant pairs, in which only mother or grandmother had any college education, regardless of race.
Discussion
This study evaluated the impact of multigenerational educational attainment on the likelihood of LBW using the Add Health dataset. We showed that college education in NHB and NHW mothers is associated with decreased odds of LBW after controlling for a wide variety of potential confounders. College education in grandmothers is also associated with a decreased odds of LBW that was largely mediated by mothers’ educational attainment. Although racial disparities in LBW persisted regardless of educational attainment and after adjusting for multiple individual and community variables, they did not increase with increased education, in contrast to past studies.26–29
This study is unique in its evaluation of mother and grandmother educational attainment impact on LBW. Several studies have shown that the maternal intrauterine environment, shaped by complex biologic, social, and environmental factors, may affect disease states later in life and health across generations39,44–46 ; however, research on the impact of transgenerational educational attainment on birth outcomes continues to be sparse. One of the only other such studies in the United States32 used national hospital survey data to find that increased education in grandfathers, but not grandmothers, led to improved rates of LBW, without a differential effect by race. The different results may reflect the household socioeconomic status of that generation (born ∼1948). Research has shown that the benefit education confers on health outcomes has increased over the past several decades,47 which supports our more contemporaneous results.
The underlying reason for these effects between mother and grandmother education and child health is less certain. Education leads to improved health outcomes and greater health equity,48 possibly through changes in an individual’s health choices and behaviors, job prospects and earning potential, and altered social exposures and toxic stress.49 A systematic review of parent literacy and childhood health outcomes showed similar changes in health behaviors, with higher literacy resulting in improved child outcomes.50
Our results are similar to work from Virginia and Connecticut that found rates of LBW in NHB and NHW women improved with increased mother educational attainment,30 yet differ from prior work suggesting that increased educational attainment is associated with increased racial disparities in LBW,31 small for gestational age,51 mortality,28 and preterm birth.26,27,29 These studies suggest that the effect of education on pregnancy outcomes differs by maternal racial/ethnic status. There are several potential explanations for these differences. First, it is possible that education has more of an effect on birth outcomes in areas with higher amounts of poverty and degrees of racial disparity,52 although this is not supported by a study from North Carolina.28 The majority of earlier studies used single-state data,27,28,31,51 and we must consider whether generalizable results can be drawn from such limited and unique subsets of the US population. Second, the choice of birth outcome measure may also explain differences in results. LBW, used in this study, is primarily the result of preterm birth and is the second most common cause of infant mortality.53–55 However, these variables are not entirely collinear, and it is possible that education affects LBW differently than the way it affects preterm birth, small for gestational age, and infant mortality. Other considerations are data collection timeframes and data sources. Much of the past research used birth data from the 1980s and 1990s, which may not reflect the current birth climate.26,28,29,31,51 Additionally, several of the studies used birth certificate data, which could result in systematic errors because of a higher likelihood of missing data for young, unwed mothers with less than a high school education.26,28,29,31,33,51 Although survey data can also be prone to certain errors, when compared with medical records, it has been shown to be valid for birth weight even many years after a child’s birth regardless of mother’s educational attainment. Here, survey data have the additional benefit of including longitudinal information on the mother’s health and socioeconomic status that is not available in birth certificates.56
There are several limitations to this study. The decision to use a specific subgroup from the Add Health cohort decreased the population size and therefore the study’s power. To evaluate the potential impact, we performed a secondary analysis on a larger sample, which confirmed our main results. Missing data could have magnified or diminished the effect size of educational attainment on odds of LBW. Several factors were missing at greater than 5% but excluding patients with these covariates could have resulted in a biased sample. Including these data with a variable for missing data did not significantly change the result while maintaining the patients in the study. We also used LBW, rather than preterm birth, as the primary outcome of interest. Although LBW is most often the result of prematurity, the causes for and outcomes of these 2 diagnoses can be distinct.53,55 Our final analysis was adjusted for a number of variables that contribute to LBW, including hypertension and smoking, to address some of the limitations of this outcome measure. Another limitation was the use of the birth of any LBW infant to construct our triads, which we chose so that we could capture all families affected by this diagnosis at least once. This may have contributed to the different prevalence of LBW in our subgroup (14.7%), compared with reported prevalence of LBW in the United States population (8.3%), which is calculated on a per-birth, rather than a per-person rate.1 Importantly, the disparity in LBW between NHB and NHW families in our subgroup is comparable to the United States population (1.7 vs 2.0 times more likely).1 Survey data are subject to recall bias, self-report bias, nonresponse bias, and errors in data entry or recall.57–59 The Add Health researchers found that the effect of nonresponse by study participants on conclusions drawn from the dataset is largely trivial.60,61 A large, multigenerational, and prospective database would be superior for research in this field; however, to our knowledge there are none currently available in the United States.
Conclusions
We found that increased educational attainment in NHW and NHB mothers is associated with a decreased likelihood of LBW after controlling for a wide variety of individual, community, and health covariates. These results suggest that initiatives and policies that support educational advancement and improve access to quality education may have the potential to decrease adverse pregnancy outcomes that disproportionately affect underserved and minority patient populations. Prospective, multigenerational studies are needed to further elucidate the direct and indirect mechanisms by which educational attainment affects infant and child health.
ACKNOWLEDGMENTS
This research uses data from Add Health, a program directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill. Add Health is funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Information on how to obtain the Add Health data files is available on the Add Health Web site (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis.
Dr Echevarria conceptualized and designed the study, carried out and interpreted data analysis, drafted the initial manuscript, and reviewed and revised the manuscript. Dr Lorch supervised the conceptualization and design of the study and the data analysis, assisted with interpretation of data analysis, and reviewed and revised the manuscript. Both authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: N o external funding.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no potential conflicts of interest to disclose.
Comments