[Skip to Navigation]
Sign In
Figure.  Exposures Associated With Cognitive Test Scores
Exposures Associated With Cognitive Test Scores

Difference in cognitive test scores and 95% CIs for the 29 exposures from the independent exposure-specific regression analyses that were adjusted for a priori covariates (ie, maternal cognitive test score, child’s birth year, child’s sex and age, and maternal and paternal ages at child’s birth) and were retained after correction for false discovery rates. Black dots represent test scores, and black horizontal lines represent 95% CIs.

aCategorical exposure. For categorical variables, binary indicators for each category were created; the reference group included all other exposure categories. For example, for individuals with 3 or more previous pregnancies, the reference group included individuals with 0, 1, or 2 previous pregnancies. Because each indicator variable was evaluated in an independent multivariable regression analysis, all categories of a given variable (eg, maternal educational level) may not be represented in the figure if one of the categories did not meet the false discovery rate–corrected statistical threshold.

bContinuous exposure. All continuous variables were standardized using a mean of 0 and an SD of 1.0; thus, the reported difference in cognitive test scores corresponds with an increase of 1.0 SD in the continuous exposure. All exposures examined are described in eTable 1 in the Supplement. SoFAS indicates the solid fats and added sugars component of the Healthy Eating Index; WIC, Women, Infants, and Children program.

Table 1.  Characteristics of Participants Included in and Excluded From the Analytic Sample
Characteristics of Participants Included in and Excluded From the Analytic Sample
Table 2.  Fully Adjusted Models of Target Exposures Associated With Childhood Cognitive Test Scores at Age 4 to 6 Years
Fully Adjusted Models of Target Exposures Associated With Childhood Cognitive Test Scores at Age 4 to 6 Years
Table 3.  Distribution of Indicators of Cognitive Test Scores by Race in the Analytic Sample
Distribution of Indicators of Cognitive Test Scores by Race in the Analytic Sample
1.
Martin  LT, Kubzansky  LD, LeWinn  KZ, Lipsitt  LP, Satz  P, Buka  SL.  Childhood cognitive performance and risk of generalized anxiety disorder.   Int J Epidemiol. 2007;36(4):769-775. doi:10.1093/ije/dym063 PubMedGoogle Scholar
2.
Batty  GD, Deary  IJ.  Early life intelligence and adult health.   BMJ. 2004;329(7466):585-586. doi:10.1136/bmj.329.7466.585 PubMedGoogle Scholar
3.
Lager  A, Bremberg  S, Vagero  D.  The association of early IQ and education with mortality: 65 year longitudinal study in Malmo, Sweden.   BMJ. 2009;339:b5282. doi:10.1136/bmj.b5282 PubMedGoogle Scholar
4.
Kendler  KS, Turkheimer  E, Ohlsson  H, Sundquist  J, Sundquist  K.  Family environment and the malleability of cognitive ability: a Swedish national home-reared and adopted-away cosibling control study.   Proc Natl Acad Sci U S A. 2015;112(15):4612-4617. doi:10.1073/pnas.1417106112 PubMedGoogle Scholar
5.
Nelson  CA  III, Zeanah  CH, Fox  NA, Marshall  PJ, Smyke  AT, Guthrie  D.  Cognitive recovery in socially deprived young children: the Bucharest Early Intervention Project.   Science. 2007;318(5858):1937-1940. doi:10.1126/science.1143921 PubMedGoogle Scholar
6.
Ritchie  SJ, Tucker-Drob  EM.  How much does education improve intelligence? a meta-analysis.   Psychol Sci. 2018;29(8):1358-1369. doi:10.1177/0956797618774253 PubMedGoogle Scholar
7.
Campbell  FA, Ramey  CT, Pungello  E, Sparling  J, Miller-Johnson  S.  Early childhood education: young adult outcomes from the Abecedarian Project.   Appl Dev Sci. 2002;6(1):42-57. doi:10.1207/S1532480XADS0601_05 Google Scholar
8.
Turkheimer  E, Haley  A, Waldron  M, D’Onofrio  B, Gottesman  II.  Socioeconomic status modifies heritability of IQ in young children.   Psychol Sci. 2003;14(6):623-628. doi:10.1046/j.0956-7976.2003.psci_1475.x PubMedGoogle Scholar
9.
Tucker-Drob  EM, Bates  TC.  Large cross-national differences in gene × socioeconomic status interaction on intelligence.   Psychol Sci. 2016;27(2):138-149. doi:10.1177/0956797615612727 PubMedGoogle Scholar
10.
Tucker-Drob  EM, Briley  DA, Harden  KP.  Genetic and environmental influences on cognition across development and context.   Curr Dir Psychol Sci. 2013;22(5):349-355. doi:10.1177/0963721413485087 PubMedGoogle Scholar
11.
Del Carmen Ruiz  J, Quackenboss  JJ, Tulve  NS.  Contributions of a child’s built, natural, and social environments to their general cognitive ability: a systematic scoping review.   PLoS One. 2016;11(2):e0147741. doi:10.1371/journal.pone.0147741 PubMedGoogle Scholar
12.
Lee  VE, Burkam  DT.  Inequality at the Starting Gate: Social Background Differences in Achievement as Children Begin School. Economic Policy Institute; 2002.
13.
Canfield  RL, Henderson  CR  Jr, Cory-Slechta  DA, Cox  C, Jusko  TA, Lanphear  BP.  Intellectual impairment in children with blood lead concentrations below 10 microg per deciliter.   N Engl J Med. 2003;348(16):1517-1526. doi:10.1056/NEJMoa022848 PubMedGoogle Scholar
14.
Perera  FP, Li  Z, Whyatt  R,  et al.  Prenatal airborne polycyclic aromatic hydrocarbon exposure and child IQ at age 5 years.   Pediatrics. 2009;124(2):e195-e202. doi:10.1542/peds.2008-3506 PubMedGoogle Scholar
15.
Liu  J, Lewis  G.  Environmental toxicity and poor cognitive outcomes in children and adults.   J Environ Health. 2014;76(6):130-138.PubMedGoogle Scholar
16.
Northstone  K, Joinson  C, Emmett  P, Ness  A, Paus  T.  Are dietary patterns in childhood associated with IQ at 8 years of age? a population-based cohort study.   J Epidemiol Community Health. 2012;66(7):624-628. doi:10.1136/jech.2010.111955 PubMedGoogle Scholar
17.
Freitas-Vilela  AA, Pearson  RM, Emmett  P,  et al.  Maternal dietary patterns during pregnancy and intelligence quotients in the offspring at 8 years of age: findings from the ALSPAC cohort.   Matern Child Nutr. 2018;14(1):e12431. doi:10.1111/mcn.12431 PubMedGoogle Scholar
18.
Lawlor  DA, Najman  JM, Batty  GD, O’Callaghan  MJ, Williams  GM, Bor  W.  Early life predictors of childhood intelligence: findings from the Mater-University study of pregnancy and its outcomes.   Paediatr Perinat Epidemiol. 2006;20(2):148-162. doi:10.1111/j.1365-3016.2006.00704.x PubMedGoogle Scholar
19.
Eriksen  H-LF, Kesmodel  US, Underbjerg  M, Kilburn  TR, Bertrand  J, Mortensen  EL.  Predictors of intelligence at the age of 5: family, pregnancy and birth characteristics, postnatal influences, and postnatal growth.   PLoS One. 2013;8(11):e79200. doi:10.1371/journal.pone.0079200 PubMedGoogle Scholar
20.
Koenen  KC, Moffitt  TE, Caspi  A, Taylor  A, Purcell  S.  Domestic violence is associated with environmental suppression of IQ in young children.   Dev Psychopathol. 2003;15(2):297-311. doi:10.1017/S0954579403000166 PubMedGoogle Scholar
21.
van Bakel  HJA, Riksen-Walraven  JM.  Parenting and development of one-year-olds: links with parental, contextual, and child characteristics.   Child Dev. 2002;73(1):256-273. doi:10.1111/1467-8624.00404 PubMedGoogle Scholar
22.
Patel  CJ, Rehkopf  DH, Leppert  JT,  et al.  Systematic evaluation of environmental and behavioural factors associated with all-cause mortality in the United States National Health and Nutrition Examination Survey.   Int J Epidemiol. 2013;42(6):1795-1810. doi:10.1093/ije/dyt208 PubMedGoogle Scholar
23.
Patel  CJ, Ioannidis  JPA, Cullen  MR, Rehkopf  DH.  Systematic assessment of the correlations of household income with infectious, biochemical, physiological, and environmental factors in the United States, 1999-2006.   Am J Epidemiol. 2015;181(3):171-179. doi:10.1093/aje/kwu277 PubMedGoogle Scholar
24.
Sontag-Padilla  L, Burns  RM, Shih  RA,  et al.  The Urban Child Institute CANDLE Study: Methodological Overview and Baseline Sample Description RAND Corporation; 2015. doi:10.7249/RR1336
25.
Wechsler  D.  Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II). NCS Pearson; 2011.
26.
Roid  GH.  Stanford-Binet Intelligence Scales, Fifth Edition. Riverside Publishing; 2003.
27.
Benjamini  Y, Hochberg  Y.  Controlling the false discovery rate: a practical and powerful approach to multiple testing.   J R Stat Soc Series B Stat Methodol. 1995;57(1):289–300. doi:10.1111/j.2517-6161.1995.tb02031.xGoogle Scholar
28.
Glickman  ME, Rao  SR, Schultz  MR.  False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies.   J Clin Epidemiol. 2014;67(8):850-857. doi:10.1016/j.jclinepi.2014.03.012 PubMedGoogle Scholar
29.
Tibshirani  R.  Regression shrinkage and selection via the lasso.   J R Stat Soc Series B Stat Methodol. 1996;58(1):267–288. doi:10.1111/j.2517-6161.1996.tb02080.xGoogle Scholar
30.
Van Houwelingen  JC.  Shrinkage and penalized likelihood as methods to improve predictive accuracy.   Stat Neerl. 2001;55(1):17–34. doi:10.1111/1467-9574.00154 Google Scholar
31.
Zhao  P, Yu  B.  On model selection consistency of lasso.   J Mach Learn Res. 2006;7:2541–2563.Google Scholar
32.
Burgette  LF, Reiter  JP.  Multiple imputation for missing data via sequential regression trees.   Am J Epidemiol. 2010;172(9):1070-1076. doi:10.1093/aje/kwq260 PubMedGoogle Scholar
33.
Stekhoven  DJ, Buhlmann  P.  MissForest—non-parametric missing value imputation for mixed-type data.   Bioinformatics. 2012;28(1):112-118. doi:10.1093/bioinformatics/btr597 PubMedGoogle Scholar
34.
Chow  GC.  Tests of equality between sets of coefficients in two linear regressions.   Econometrica. 1960;28(3):591–605. doi:10.2307/1910133 Google Scholar
35.
Organisation for Economic Co-operation and Development.  Society at a Glance 2014: OECD Social Indicators. OECD Publishing; 2014.
36.
Buckley  JP, Doherty  BT, Keil  AP, Engel  SM.  Statistical approaches for estimating sex-specific effects in endocrine disruptors research.   Environ Health Perspect. 2017;125(6):067013. doi:10.1289/EHP334 PubMedGoogle Scholar
37.
Christian  K, Morrison  FJ, Bryant  FB.  Predicting kindergarten academic skills: interactions among child care, maternal education, and family literacy environments.   Early Child Res Q. 1998;13(3):501–521. doi:10.1016/S0885-2006(99)80054-4 Google Scholar
38.
Roberts  J, Jurgens  J, Burchinal  M.  The role of home literacy practices in preschool children’s language and emergent literacy skills.   J Speech Lang Hear Res. 2005;48(2):345-359. doi:10.1044/1092-4388(2005/024) PubMedGoogle Scholar
39.
Green  CM, Berkule  SB, Dreyer  BP,  et al.  Maternal literacy and associations between education and the cognitive home environment in low-income families.   Arch Pediatr Adolesc Med. 2009;163(9):832-837. doi:10.1001/archpediatrics.2009.136 PubMedGoogle Scholar
40.
Sayegh  P, Arentoft  A, Thaler  NS, Dean  AC, Thames  AD.  Quality of education predicts performance on the Wide Range Achievement Test–4th Edition Word Reading subtest.   Arch Clin Neuropsychol. 2014;29(8):731-736. doi:10.1093/arclin/acu059 PubMedGoogle Scholar
41.
Baer  J, Kutner  M, Sabatini  J, White  S. Basic Reading Skills and the Literacy of America’s Least Literate Adults: Results From the 2003 National Assessment of Adult Literacy (NAAL) Supplemental Studies. NCES 2009-481. National Center for Education Statistics; 2009. Accessed December 27, 2019. https://eric.ed.gov/?id=ED505187
42.
Ladson-Billings  G.  From the achievement gap to the education debt: understanding achievement in U.S. schools.   Educ Res. 2006;35(7):3-12. doi:10.3102/0013189X035007003 Google Scholar
43.
Margo  RA.  Race and Schooling in the South, 1880-1950: An Economic History. University of Chicago Press; 1990. doi:10.7208/chicago/9780226505015.001.0001
44.
Bornstein  MH.  How infant and mother jointly contribute to developing cognitive competence in the child.   Proc Natl Acad Sci U S A. 1985;82(21):7470-7473. doi:10.1073/pnas.82.21.7470 PubMedGoogle Scholar
45.
Fewell  RR, Deutscher  B.  Contributions of receptive vocabulary and maternal style: variables to later verbal ability and reading in low-birthweight children.   Top Early Child Special Educ. 2002;22(4):181–190. doi:10.1177/027112140202200401 Google Scholar
46.
Herbers  JE, Cutuli  JJ, Lafavor  TL,  et al.  Direct and indirect effects of parenting on the academic functioning of young homeless children.   Early Educ Dev. 2011;22(1):77–104. doi:10.1080/10409280903507261 Google Scholar
47.
Craig  F, Operto  FF, De Giacomo  A,  et al.  Parenting stress among parents of children with neurodevelopmental disorders.   Psychiatry Res. 2016;242:121-129. doi:10.1016/j.psychres.2016.05.016 PubMedGoogle Scholar
48.
Estes  A, Munson  J, Dawson  G, Koehler  E, Zhou  X-H, Abbott  R.  Parenting stress and psychological functioning among mothers of preschool children with autism and developmental delay.   Autism. 2009;13(4):375-387. doi:10.1177/1362361309105658 PubMedGoogle Scholar
49.
Baker  BL, Blacher  J, Crnic  KA, Edelbrock  C.  Behavior problems and parenting stress in families of three-year-old children with and without developmental delays.   Am J Ment Retard. 2002;107(6):433-444. doi:10.1352/0895-8017(2002)107<0433:BPAPSI>2.0.CO;2 PubMedGoogle Scholar
50.
Neece  CL, Green  SA, Baker  BL.  Parenting stress and child behavior problems: a transactional relationship across time.   Am J Intellect Dev Disabil. 2012;117(1):48-66. doi:10.1352/1944-7558-117.1.48 PubMedGoogle Scholar
51.
Tachibana  Y, Fukushima  A, Saito  H,  et al.  A new mother-child play activity program to decrease parenting stress and improve child cognitive abilities: a cluster randomized controlled trial.   PLoS One. 2012;7(7):e38238. doi:10.1371/journal.pone.0038238 PubMedGoogle Scholar
52.
Burgdorf  V, Szabo  M, Abbott  MJ.  The effect of mindfulness interventions for parents on parenting stress and youth psychological outcomes: a systematic review and meta-analysis.   Front Psychol. 2019;10:1336. doi:10.3389/fpsyg.2019.01336 PubMedGoogle Scholar
53.
Lee  JJ, Wedow  R, Okbay  A,  et al; 23andMe Research Team; COGENT (Cognitive Genomics Consortium); Social Science Genetic Association Consortium.  Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals.   Nat Genet. 2018;50(8):1112-1121. doi:10.1038/s41588-018-0147-3 PubMedGoogle Scholar
54.
Cheesman  R, Hunjan  A, Coleman  JRI,  et al.  Comparison of adopted and non-adopted individuals reveals gene-environment interplay for education in the UK Biobank.   Psychol Sci. 2020;31(5):582-591. doi:10.1177/0956797620904450PubMedGoogle Scholar
55.
Cortes  KE.  Do bans on affirmative action hurt minority students? evidence from the Texas Top 10% Plan.   Econ Educ Rev. 2010;29(6):1110-1124. doi:10.1016/j.econedurev.2010.06.004 Google Scholar
56.
Blume  GH, Long  MC.  Changes in levels of affirmative action in college admissions in response to statewide bans and judicial rulings.   Educ Eval Policy Anal. 2014;36(2):228-252. doi:10.3102/0162373713508810 Google Scholar
57.
Pitre  CC.  Improving African American student outcomes: understanding educational achievement and strategies to close opportunity gaps.   West J Black Stud. 2014;38(4):209-217.Google Scholar
58.
Berkner  L, Chavez  L. Access to Postsecondary Education for the 1992 High School Graduates. Postsecondary Education Descriptive Analysis Reports. Statistical Analysis Report. Office of Educational Research and Improvement, US Department of Education; 1997. Accessed May 21, 2020. https://eric.ed.gov/?id=ED413854
59.
Carnevale  AP, Strohl  J. Separate and Unequal: How Higher Education Reinforces the Intergenerational Reproduction of White Racial Privilege. Center on Education and the Workforce, Georgetown Public Policy Institute; July 2013. Accessed April 1, 2019. https://cew.georgetown.edu/wp-content/uploads/SeparateUnequal.FR_.pdf
60.
Kaufman  P, Chen  X. Projected postsecondary outcomes of 1992 high school graduates. National Center for Education Statistics working paper 1999-15. June 1999. Accessed April 15, 2019. https://nces.ed.gov/pubs99/199915.pdf
61.
Wagmiller  RL, Adelman  RM. Childhood and intergenerational poverty: the long-term consequences of growing up poor. National Center for Children in Poverty. Accessed April 15, 2019. http://www.nccp.org/publications/pub_909.html
62.
Anderson  JW, Johnstone  BM, Remley  DT.  Breast-feeding and cognitive development: a meta-analysis.   Am J Clin Nutr. 1999;70(4):525-535. doi:10.1093/ajcn/70.4.525 PubMedGoogle Scholar
63.
Kramer  MS, Aboud  F, Mironova  E,  et al; Promotion of Breastfeeding Intervention Trial (PROBIT) Study Group.  Breastfeeding and child cognitive development: new evidence from a large randomized trial.   Arch Gen Psychiatry. 2008;65(5):578-584. doi:10.1001/archpsyc.65.5.578 PubMedGoogle Scholar
64.
Deave  T, Heron  J, Evans  J, Emond  A.  The impact of maternal depression in pregnancy on early child development.   BJOG. 2008;115(8):1043-1051. doi:10.1111/j.1471-0528.2008.01752.x PubMedGoogle Scholar
65.
Cassidy-Bushrow  AE, Sitarik  AR, Havstad  S,  et al.  Burden of higher lead exposure in African-Americans starts in utero and persists into childhood.   Environ Int. 2017;108:221-227. doi:10.1016/j.envint.2017.08.021 PubMedGoogle Scholar
Original Investigation
September 21, 2020

Identification of Modifiable Social and Behavioral Factors Associated With Childhood Cognitive Performance

Author Affiliations
  • 1Weill Institute for Neurosciences, Department of Psychiatry and Behavioral Sciences, University of California, San Francisco, San Francisco
  • 2Department of Pediatrics, Division of Developmental Medicine, University of California, San Francisco, San Francisco
  • 3University of California, San Francisco, San Francisco
  • 4Preventive Medicine, University of Tennessee Health Science Center, Memphis
  • 5Division of Primary Care and Population Health, Department of Medicine, Stanford University School of Medicine, Stanford, California
JAMA Pediatr. 2020;174(11):1063-1072. doi:10.1001/jamapediatrics.2020.2904
Key Points

Question  Are specific exposures associated with childhood cognitive performance and are inequities in these exposures associated with racial disparities in cognitive test scores?

Findings  In this cohort study of 1055 mother-child dyads, 24 of 155 prenatal and postnatal exposures were associated with childhood cognitive performance; models that included all of the exposures fully accounted for the racial disparities in cognitive test scores. Modifiable exposures included breastfeeding, parental educational levels, fostering of cognitive growth during mother-child interactions, parenting stress, and maternal reading ability, which together were associated with 0.5% of a standard deviation difference in cognitive test scores.

Meaning  The study’s results indicated that addressing inequities in the early environment could help to reduce racial disparities in childhood cognitive performance.

Abstract

Importance  Inequities in social environments are likely associated with a large portion of racial disparities in childhood cognitive performance. Identification of the specific exposures associated with cognitive development is needed to inform prevention efforts.

Objective  To identify modifiable factors associated with childhood cognitive performance.

Design, Setting, and Participants  This longitudinal pregnancy cohort study included 1503 mother-child dyads who were enrolled in the University of Tennessee Health Science Center–Conditions Affecting Neurodevelopment and Learning in Early Life study between December 1, 2006, and July 31, 2011, and assessed annually until the children were aged 4 to 6 years. The analytic sample comprised 1055 mother-child dyads. A total of 155 prenatal, perinatal, and postnatal exposures were included to evaluate environment-wide associations. Participants comprised a community-based sample of pregnant women who were recruited between 16 weeks and 28 weeks of gestation from 4 hospitals in Shelby County, Tennessee. Women with high-risk pregnancies were excluded. Data were analyzed from June 1, 2018, to April 15, 2019.

Exposures  Individual and neighborhood socioeconomic position, family structure, maternal mental health, nutrition, delivery complications, birth outcomes, and parenting behaviors.

Main Outcomes and Measures  Child’s full-scale IQ measured by the Stanford-Binet Intelligence Scales, Fifth Edition, at age 4 to 6 years.

Results  Of 1055 children included in the analytic sample, 532 (50.4%) were female. Among mothers, the mean (SD) age was 26.0 (5.6) years; 676 mothers (64.1%) were Black, and 623 mothers (59.0%) had an educational level of high school or less. Twenty-four factors were retained in the least absolute shrinkage and selection operator regression analysis and full models adjusted for potential confounding. Associations were noted between child cognitive performance and parental education and breastfeeding; for each increase of 1.0 SD in exposure, positive associations were found with cognitive growth fostering from observed parent-child interactions (β = 1.12; 95% CI, 0.24-2.00) and maternal reading ability (β = 1.42; 95% CI, 0.16-2.68), and negative associations were found with parenting stress (β = −1.04; 95% CI, −1.86 to −0.21). A moderate increase in these beneficial exposures was associated with a notable improvement in estimated cognitive test scores using marginal means (0.5% of an SD). Black children experienced fewer beneficial cognitive performance exposures; in a model including all 24 exposures and covariates, no racial disparity was observed in cognitive performance (95% CIs for race included the null).

Conclusions and Relevance  The prospective analysis identified multiple beneficial and modifiable cognitive performance exposures that were associated with mean differences in cognitive performance by race. The findings from this observational study may help guide experimental studies focused on reducing racial disparities in childhood cognitive performance.

Introduction

Childhood cognitive performance is associated with future mortality and morbidity across a number of conditions, including heart disease, cancer, and anxiety disorders.1-3 Although once considered a primarily inherited characteristic, childhood cognitive performance has been reported to be sensitive to environmental conditions in multiple studies.4-7 This sensitivity is especially relevant among children growing up in households with low socioeconomic conditions, in which environmental circumstances account for a larger proportion of the variance in childhood cognitive function than do genetic predispositions.8-10 The preponderance of data suggests that early environmental exposures are important for early cognitive development.

Children develop in complex learning environments that are characterized by myriad social, physical, and chemical exposures, many of which have been associated with cognitive performance and are distributed inequitably by race and socioeconomic position.11,12 Exposures include those modifiable by prevention and intervention efforts, such as prenatal and postnatal exposure to environmental toxins13-15; prenatal and postnatal nutrition16,17; parental educational level and household income18; smoking during pregnancy18; breastfeeding, birth weight, and maternal body mass index during pregnancy19; exposure to domestic violence20; and parenting behaviors.21 However, these exposures are often examined independently, reflecting disciplinary perspectives. Concurrent testing of multiple aspects of the early environment may reveal targets for intervention that are most likely to be associated with reductions in cognitive performance disparities.

We used an environment-wide association approach22,23 to examine 155 factors associated with childhood cognitive performance in a prospective pregnancy cohort study of mother-child dyads living in an urban area in the southern US. Measurements included detailed multilevel and multimodal phenotyping in the prenatal and postnatal periods; the examined variables encompassed individual and neighborhood measures of socioeconomic position, family structure, maternal mental health, nutrition, birth outcomes, and observed and self-reported measures of parenting behaviors.24 Our goal was to identify modifiable exposures that may be targeted in future interventions designed to achieve equity in early learning environments for all children.

Methods

Quiz Ref IDThe Conditions Affecting Neurodevelopment and Learning in Early Life (CANDLE) study conducted through the University of Tennessee Health Sciences Center is a prospective pregnancy cohort study in Shelby County, Tennessee, that was originally initiated to identify early-life factors of neurocognitive development. From December 1, 2006, to July 31, 2011, the CANDLE study enrolled 1503 pregnant women between 16 weeks and 28 weeks of gestation who were planning to deliver at 1 of 4 participating hospitals. Women with known chronic conditions (eg, hypertension and diabetes) were excluded. Extensive longitudinal data collection included clinical visits conducted twice during pregnancy and at annual intervals throughout childhood.24 The analytic sample included all children with a recorded full-scale IQ measurement from the visit conducted at age 4 to 6 years. All research activities were approved by the institutional review board of the University of Tennessee Health Sciences Center, and written informed consent was obtained from all mothers. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Based on previous literature and theory,4-20 we evaluated the full CANDLE data set to identify factors that may be associated with childhood cognitive performance. We excluded measures of the same construct that had high Spearman correlation coefficients (defined as >0.85); if subscale scores and overall scores had high correlation coefficients, only overall scores were included. We also calculated the mean of the same measure at multiple points, keeping prenatal and postnatal measurements separate if they did not have high correlation coefficients. Continuous variables were standardized to a mean of 0 and an SD of 1.0 for all analyses. In the first stage of analysis, indicator variables were created from categorical variables, with indicators assigned a value of −1.0 or 1.0, and the reference group included all other categories. This approach allowed us to examine each category as a separate exposure. In stages 2 and 3, all categories of an exposure selected in stage 1 were included in the final models. Our analysis included multiple distinct measures of the following multilevel broad domains of exposure measured across the prenatal and postnatal periods: socioeconomic position; family structure; maternal cognition, mental health, physical health, and stress; mother and infant nutrition; labor and delivery complications; birth outcomes; infant and child health; parenting behaviors; enrollment in supplementary programs (eg, the Women, Infants, and Children program); and maternal perceptions of neighborhood quality.

A total of 161 variables were identified through this process described in eTable 1 in the Supplement). From these variables, we identified a minimal set of a priori–defined confounding variables to include in all models that (1) may have been associated with cognitive performance, (2) were present at the time of child’s birth, or (3) captured important study design outcomes. These covariates comprised child’s age at assessment, child’s year of birth, maternal cognitive performance on the Wechsler Abbreviated Scales of Intelligence,25 child’s sex, and paternal and maternal ages at child’s birth. The remaining 155 variables were considered target exposures in our analyses.

Quiz Ref IDEach child’s standardized full-scale IQ was assessed by trained psychologists using the Stanford-Binet Intelligence Scales, Fifth Edition (the mean [SD] score for the test is 100 [15] points, with higher scores indicating higher cognitive performance), which has been validated and normed in large diverse populations and includes tasks that assess working memory, processing speed, visual-spatial skills, vocabulary, and language comprehension.26

Statistical Analysis

The analysis occurred in 3 stages. During stage 1, we selected those exposures (from the 155 total exposures) that were associated with cognitive performance, adjusting for the a priori–selected covariates that were identified as potential confounding variables. Thus, using 155 independent multivariable regression models (which were adjusted for the child’s age and sex, maternal and paternal ages at the child’s birth, and maternal cognitive performance), we estimated associations between each target exposure and the child’s cognitive performance. Target exposures were considered statistically significant at P = .05 after multiple-comparison adjustment for the false discovery rate.27,28

In stage 2, we applied the least absolute shrinkage and selection operator (LASSO) method to a model that included all target exposures that had a statistically significant association with the outcome after correction for false discovery rate. The LASSO method decreases unstable effect estimates toward 0 and excludes colinear covariates.29-31

In stage 3, factors retained in the LASSO regression model were incorporated into multiple parsimonious models to evaluate associations with adjustment for potential confounding variables of the specific exposures associated with cognitive performance that were being examined. For example, the final models for target exposures did not include variables that occurred after the target exposure (eg, no postnatal covariates were included in models of prenatal exposures).

Quiz Ref IDWe evaluated the distributions of exposures retained in stage 3 by race (White vs Black or other race). We also calculated the proportion of the mean difference in cognitive test scores by race through inclusion of these exposures plus covariates in a single fully adjusted model.

We used the missForrest package in R software (R Foundation for Statistical Computing) to impute the values of missing indicator and control variables in the analytic sample. This method does not rely on distributional assumptions, accommodates nonlinear associations and interactions,32 and outperforms or is similar to multivariate imputation by chained equations for contexts in which many exposures are examined.33

Global Chow tests34 were used to examine whether the final models from stage 3 differed by household income (dichotomized at <50% of the median of disposable income for the US in 2010, or $14 550)35 and child sex or race. If the results of the global Chow test were statistically significant, we performed fully interactive models using the modifier of interest.36 All data were analyzed from June 1, 2018, to April 15, 2019.

Results

Of 1503 participants enrolled, 11 fetuses and 5 children died; 71 participants withdrew from the study, and 260 participants did not attend the visit at age 4 to 6 years. Among those who attended the visit, 101 participants did not have a completed outcome measure. The analytic sample included 1055 mother-child dyads with cognitive test scores recorded at age 4 to 6 years (mean [SD] score, 99.8 [14.9] points) (Table 1). Among mothers in the analytic sample, the mean (SD) age was 26.0 (5.6) years; 676 mothers (64.1%) were Black, 311 mothers (29.5%) were White, and 68 mothers (6.4%) were of another race (most of whom were of multiple races and identified as Black). A total of 623 mothers (59.0%) had an educational level of high school or less. A total of 630 mothers (59.7%) had Medicaid or TennCare insurance, and 440 mothers (41.7%) had household incomes of less than $25 000. Among children in the analytic sample, the mean (SD) age was 4.4 (0.5) years, and 532 children (50.4%) were female. A comparison of participants included in the analytic sample with those excluded from the analysis indicated that the samples had similar demographic characteristics. Compared with participants excluded from the analytic sample, mothers included in the analysis were slightly older (mean [SD] age, 25.3 [5.1] years vs 26.0 [5.6] years, respectively) and more children were born in 2009 or later (72.5% vs 60.8%, respectively). Descriptive data on all variables are available in eTable 1 in the Supplement.

We identified 29 exposures that were associated with cognitive test scores after correcting for false discovery rate (Figure). These exposures spanned multiple domains and comprised both modifiable and nonmodifiable characteristics, including family structure (previous pregnancies, first birth order, and adult-to-child ratio); maternal psychosocial factors (social support, social ties in their neighborhood, and parenting stress); neighborhood characteristics (maternal perceptions of neighborhood quality, mean total score on the Child Opportunity Index [score range, 1-100, with higher scores indicating more childhood opportunity at the neighborhood level], and score on the education domain of the Child Opportunity Index [measured in quintiles, with higher quintiles indicating more opportunity]); socioeconomic characteristics (maternal and paternal educational levels; income adjusted for household size; enrollment in the Women, Infants, and Children program and reduced-price meal programs; receipt of food stamps; and health insurance status); parenting behaviors and cognitive enrichment (parental involvement [measured using the Parenting Relationship Questionnaire], child abuse potential, maternal reading ability, and fostering of cognitive growth [assessed by trained behavioral coders who observed mothers and children completing interactive tasks]); and maternal nutrition during pregnancy and infant feeding (plasma folate in the second trimester, total score on the Healthy Eating Index [score range, 0-100, with higher scores indicating higher dietary quality], score on the solid fats and added sugars component of the Healthy Eating Index [score range, 0-20, with higher scores indicating lower consumption of solid fats and added sugars], and breastfeeding). Positive factors associated with cognitive performance at this stage included higher educational levels (ie, mother and father completed college or graduate school), high educational opportunity in the neighborhood, and high adjusted household income. Negative factors included 3 or more pregnancies, enrollment in a reduced-price meal program, and public health insurance, the latter 2 of which are also factors associated with low-income status.

A correlation matrix (eFigure in the Supplement) indicated associations between several of the 29 exposures; LASSO regression analysis excluded 5 exposures (public health insurance; receipt of food stamps; enrollment in the Women, Infants, and Children program; child abuse potential; and total score on the Child Opportunity Index). Adjusted associations for the remaining 24 exposures were assessed using 5 separate models. We examined exposures during childhood in a single model, which was adjusted for the a priori covariates included in stage 1 as well as prenatal nutrition exposures, breastfeeding, birth weight, and maternal race (model 1). Prenatal nutrition exposures were examined in separate models and included the a priori covariates as well as maternal and paternal educational levels, adjusted household income, neighborhood educational opportunity, and maternal race (models 2-4). Breastfeeding was adjusted for the a priori covariates as well as prenatal nutrition variables, maternal and paternal educational levels, adjusted household income, neighborhood educational opportunity, birth weight, and maternal race (model 5) (Table 2).

Quiz Ref IDIn the final models, 7 of the 24 exposures were associated with cognitive test scores and had 95% CIs that did not include the null. Cognitive test scores were positively associated with parental educational level. Compared with children of parents who had a high school education or less, children with mothers and fathers who completed graduate school had higher scores of 1.70 points (95% CI, 0.33-3.07 points) and 2.36 points (95% CI, 0.70-4.02 points), respectively. An increase of 1.0 SD in parenting stress was associated with a decrease of 1.04 points (95% CI, −1.86 to −0.21 points), while maternal reading ability and parenting behaviors that fostered cognitive growth were associated with increases of 1.42 points (95% CI, 0.16-2.68 points) and 1.12 points (95% CI, 0.24-2.00 points), respectively. Compared with the reference group (<2 months of breastfeeding), all categories of breastfeeding (2-4 months, 5-6 months, and ≥6 months) were associated with higher scores; 6 months or more of breastfeeding was associated with a higher score of 1.66 points (95% CI, 0.51-2.80 points). Using a heuristic approach to describe the potential impact of intervention for these exposures, we calculated the estimated marginal mean score for a child with reference values for these characteristics (ie, children with a score of 0 for continuous variables, children who were breastfed for <2 months, and children with parents who had a high school education or less) compared with a child who was breastfed for 6 or more months, had parents with college degrees, and had a mother whose reading level and observed fostering of cognitive growth scores were 1.0 SD higher than the mean and whose parenting stress level was 1.0 SD lower than the mean. Based on coefficients from the fully adjusted model (eTable 3 in the Supplement), a child with more beneficial exposures would have an estimated score of 116.2 points (95% CI, 111.2-121.2 points) compared with a reference child’s score of 108.8 points (95% CI, 105.5-112.1 points), which represents a mean (SD) difference of 7.4 (0.5) points.

We observed no modification of associations by race or socioeconomic position, but greater statistical significance was found in the associations between cognitive test scores and paternal technical school education among girls (for girls, β = 3.13; 95% CI, 1.06-5.20; for boys, β = −0.92; 95% CI, −3.17 to 1.34; P = .03) and maternal graduate school education among boys (for boys, β = −3.09; 95% CI, −0.89 to 5.30; for girls, β = 0.60; 95% CI, −1.19 to 2.38; P = .05) (eTable 2 in the Supplement).

Quiz Ref IDFor all but 2 of the 24 exposures identified, Black children were less likely to be exposed to factors that are beneficial for cognitive performance, with notable differences in prenatal nutrition, maternal reading ability, social support and social ties, neighborhood educational opportunity, breastfeeding, and multiple measures of SEP, including maternal educational attainment (Table 3). The unadjusted mean (SD) difference in test scores by race was 13.3 (11.6) points (mean [SD], 95.9 [14.1] points for Black participants and participants of other races vs 109.2 [12.3] points for White participants). In a mutually adjusted model that included all 24 exposures and covariates, we observed no difference in cognitive test scores by race (ie, the 95% CI for the race coefficient included the null) (eTable 3 in the Supplement).

Discussion

Our goal was to identify the most salient factors associated with childhood cognitive performance at age 4 to 6 years and the extent to which inequities in these exposures were associated with racial disparities. Of 155 prenatal and postnatal target exposures evaluated, 24 exposures were selected for inclusion in fully adjusted models; with the exception of birth order, all of these exposures could be considered modifiable. Exposures associated with cognitive performance included well-known factors, such as breastfeeding and parental educational levels, as well as factors that are understudied in epidemiological cohorts, such as parenting stress, maternal reading ability, and a measure of observed maternal fostering of cognitive growth. When comparing estimated mean scores between a child with a moderate increase in these exposures and a child with reference values, we observed a mean difference of 0.5% of an SD. In a mutually adjusted model that included all target exposures and covariates, we observed no association between race and cognitive test scores.

Maternal reading ability was positively associated with childhood cognitive performance in a model adjusted for multiple confounders, such as maternal cognitive performance and educational level. Maternal reading ability is associated with the number of words spoken in the home and the literacy environment (eg, the number of books in the home)37-39 and reflects the quality of the mother’s early educational environment.40 In the US, 14% of adults are illiterate, and 15% of adults graduate high school with only basic reading skills.41 The quality of early educational environments varies substantially by socioeconomic position and race, which some have called the education debt to refer to the legacy of disparities in education access and quality42 that is particularly prevalent in the southern US.43 Addressing these disparities in educational quality is necessary to reduce the consequences of this legacy for the next generation of US children and to minimize inequitable downward trajectories in cognitive performance among children whose parents are in low-quality educational environments.

The fostering of cognitive growth, which was assessed by trained behavioral coders who observed mothers and children completing interactive tasks, was also positively associated with childhood cognitive performance. Psychological studies have reported similar associations between maternal behaviors that encourage cognitive growth and improved cognitive outcomes.21,44 However, most parenting research using observed behavioral coding has relied on small samples of primarily White mother-child dyads without robust adjustment for potential confounders, such as socioeconomic position or breastfeeding.45 We observed an association that was independent of family and neighborhood socioeconomic position, maternal cognitive performance and reading ability, birth weight, and breastfeeding. Notably, even in environments with high levels of adversity, parenting behaviors have been observed to change in response to intervention, with subsequent improvements in childhood cognitive development46; therefore, expanding accessibility to interventions aimed at improving parent-child interactions during cognitive tasks may help to improve childhood cognitive development.

Parenting stress was negatively associated with childhood cognitive test scores. Although raising a child with developmental delays is associated with increases in parenting stress,47-49 longitudinal research describes these associations in children with typical development as bidirectional, such that parenting stress is associated with increased risk of worse outcomes for children.50 Interventions to reduce parenting stress have been associated with improvements in the cognitive outcomes of children; although these interventions may have the most impact for parents of children with developmental delays, benefits have been observed for parents of all children.51,52

We also replicate previous studies that have reported positive associations between childhood cognitive performance and parental educational levels.18,19 Although the coefficients for maternal educational level described in the present study are smaller than those of similar studies,18,19 they are consistent with coefficients found in a study that accounted for shared genetic variance between siblings.4 Our inclusion of multiple salient factors associated with childhood cognitive performance in the final models may produce more accurate coefficients within a cohort of unrelated children. While recent genome-wide association studies have developed polygenic risk scores to estimate educational attainment, the study with the most statistical power53 explained only 11% to 13% of the variance. Among adopted children for whom the association between genetic factors and environmental conditions were addressed, the explained variance was further reduced.54 In contrast, multiple overarching social factors associated with racial and socioeconomic disparities in educational attainment have been elucidated, including bans on affirmative action,55,56 impoverished early-learning environments that inadequately prepare marginalized groups for higher education,57,58 and attendance at undercapitalized universities in which dropout rates are higher.59,60 In our sample, 71.1% of White mothers had a college education or more compared with 35.5% of Black mothers; our findings highlight the potential intergenerational impact of these socially driven patterns of educational attainment for childhood cognitive performance.

We adjusted for directly assessed maternal cognitive performance in all models; this adjustment is a strength of our study and provides evidence that the associations between exposures and childhood cognitive performance are not confounded by maternal cognitive ability. However, it is also important to note the likely inequities in the early learning environments of mothers in our study. Among adults from families with low socioeconomic positions, substantial variance in cognitive performance is associated with environmental factors.8 Thus, our adjustment for maternal cognitive performance may yield conservative estimates of the associations between potentially modifiable socioeconomic factors and childhood cognitive performance given the intergenerational persistence of poverty in the US, which is more likely among minority groups.61

Although correlation coefficients in the present study are consistent with those of previous studies,62,63 we did not observe a dose-response association between breastfeeding and cognitive test scores, which may be explained by our reliance on maternal retrospective reports. Children with mothers who had 3 or more previous pregnancies had lower cognitive test scores (although 95% CIs included the null), which is consistent with previous studies.18,19 Despite the use of widely used, reliable survey instruments, maternal prenatal depression,64 maternal experiences of interpersonal violence during pregnancy,20 and birth weight and gestational age18,19 did not meet the threshold for statistical significance in the independent models with minimal covariate adjustment (stage 1). We did not observe modification of associations by race or socioeconomic position.

Limitations

This study has several limitations. First, the estimated correlation coefficients from these models may be biased by unmeasured confounding. The intended use of our findings is to guide intervention studies and quasi-experimental research to further assess the associations between the identified factors and childhood cognitive performance. Second, important environmental exposures, such as lead, that are known to be associated with cognitive development and may be inequitably distributed by race were not measured in the CANDLE study.65 Residual confounding from this type of exposure and measurement errors in the exposures that were included57 may have produced an underestimation of the proportion of racial disparity in cognitive test scores associated with modifiable exposures. Third, differences in exposure assessment and measurement errors reduced our ability to directly compare exposure types. However, for most exposures, measurements were similar or had better reliability and validity compared with other longitudinal epidemiological cohorts. The CANDLE study excluded women with high-risk pregnancies at enrollment, and mothers had lower rates of preterm birth and low birth weight compared with the general population of Shelby County,24 which may have limited our ability to detect associations for these exposures. Fourth, the results of the CANDLE study are largely representative of Shelby County, Tennessee, limiting their generalizability to the US population. However, our participants share characteristics with other primarily Black communities with high levels of socioeconomic disadvantage in urban areas of the southern US.

Conclusions

In an analysis that was well adjusted for potential confounding variables, we identified multiple modifiable factors associated with childhood cognitive performance that, if addressed, could help improve cognitive test scores. We conducted this study among a sample of children in the southern urban US, which comprises a population that is often underrepresented in large studies of childhood cognitive development.18,19 In our study population, Black children experienced notably lower levels of most of the beneficial cognitive performance exposures, and inclusion of these exposures in a fully adjusted model accounted for the racial disparity in cognitive test scores. Our findings suggest that increasing equity in early social environments may help to reduce disparities in childhood cognitive performance.

Back to top
Article Information

Accepted for Publication: June 2, 2020.

Corresponding Author: Kaja Z. LeWinn, ScD, MS, Weill Institute for Neurosciences, Department of Psychiatry, University of California, San Francisco, 401 Parnassus Ave, San Francisco, CA 94143 (kaja.lewinn@ucsf.edu).

Published Online: September 21, 2020. doi:10.1001/jamapediatrics.2020.2904

Author Contributions: Dr LeWinn and Ms Batra had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: LeWinn, Bush, Tylavsky, Rehkopf.

Acquisition, analysis, or interpretation of data: LeWinn, Bush, Batra, Tylavsky.

Drafting of the manuscript: LeWinn, Batra, Rehkopf.

Critical revision of the manuscript for important intellectual content: LeWinn, Bush, Tylavsky, Rehkopf.

Statistical analysis: LeWinn, Batra.

Obtained funding: LeWinn, Bush, Tylavsky.

Administrative, technical, or material support: LeWinn, Bush, Tylavsky.

Supervision: LeWinn, Rehkopf.

Conflict of Interest Disclosures: Dr LeWinn reported receiving grants from the Urban Child Institute during the conduct of the study. Dr Bush reported receiving grants from the University of California, San Francisco, and the Urban Child Institute during the conduct of the study. Dr Tylavsky reported receiving grants from the National Institutes of Health and the Urban Child Institute and during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was funded by the Urban Child Institute in Memphis, Tennessee (Drs LeWinn, Bush, and Tylavsky); the Urban Child Institute also funded the Conditions Affecting Neurodevelopment and Learning in Early Life (CANDLE) study.

Role of the Funder/Sponsor: The Urban Child Institute had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Robert Davis, MD, of the Center for Biomedical Informatics, University of Tennessee Health Science Center, and Nancy Adler, MD, of the Center for Health and Community, University of California, San Francisco, provided insightful comments and suggestions that greatly improved our manuscript. Neither contributor received compensation. We are grateful for the participation of families enrolled in the Conditions Affecting Neurodevelopment and Learning in Early Life (CANDLE), as well as the dedication of CANDLE research staff and investigators.

References
1.
Martin  LT, Kubzansky  LD, LeWinn  KZ, Lipsitt  LP, Satz  P, Buka  SL.  Childhood cognitive performance and risk of generalized anxiety disorder.   Int J Epidemiol. 2007;36(4):769-775. doi:10.1093/ije/dym063 PubMedGoogle Scholar
2.
Batty  GD, Deary  IJ.  Early life intelligence and adult health.   BMJ. 2004;329(7466):585-586. doi:10.1136/bmj.329.7466.585 PubMedGoogle Scholar
3.
Lager  A, Bremberg  S, Vagero  D.  The association of early IQ and education with mortality: 65 year longitudinal study in Malmo, Sweden.   BMJ. 2009;339:b5282. doi:10.1136/bmj.b5282 PubMedGoogle Scholar
4.
Kendler  KS, Turkheimer  E, Ohlsson  H, Sundquist  J, Sundquist  K.  Family environment and the malleability of cognitive ability: a Swedish national home-reared and adopted-away cosibling control study.   Proc Natl Acad Sci U S A. 2015;112(15):4612-4617. doi:10.1073/pnas.1417106112 PubMedGoogle Scholar
5.
Nelson  CA  III, Zeanah  CH, Fox  NA, Marshall  PJ, Smyke  AT, Guthrie  D.  Cognitive recovery in socially deprived young children: the Bucharest Early Intervention Project.   Science. 2007;318(5858):1937-1940. doi:10.1126/science.1143921 PubMedGoogle Scholar
6.
Ritchie  SJ, Tucker-Drob  EM.  How much does education improve intelligence? a meta-analysis.   Psychol Sci. 2018;29(8):1358-1369. doi:10.1177/0956797618774253 PubMedGoogle Scholar
7.
Campbell  FA, Ramey  CT, Pungello  E, Sparling  J, Miller-Johnson  S.  Early childhood education: young adult outcomes from the Abecedarian Project.   Appl Dev Sci. 2002;6(1):42-57. doi:10.1207/S1532480XADS0601_05 Google Scholar
8.
Turkheimer  E, Haley  A, Waldron  M, D’Onofrio  B, Gottesman  II.  Socioeconomic status modifies heritability of IQ in young children.   Psychol Sci. 2003;14(6):623-628. doi:10.1046/j.0956-7976.2003.psci_1475.x PubMedGoogle Scholar
9.
Tucker-Drob  EM, Bates  TC.  Large cross-national differences in gene × socioeconomic status interaction on intelligence.   Psychol Sci. 2016;27(2):138-149. doi:10.1177/0956797615612727 PubMedGoogle Scholar
10.
Tucker-Drob  EM, Briley  DA, Harden  KP.  Genetic and environmental influences on cognition across development and context.   Curr Dir Psychol Sci. 2013;22(5):349-355. doi:10.1177/0963721413485087 PubMedGoogle Scholar
11.
Del Carmen Ruiz  J, Quackenboss  JJ, Tulve  NS.  Contributions of a child’s built, natural, and social environments to their general cognitive ability: a systematic scoping review.   PLoS One. 2016;11(2):e0147741. doi:10.1371/journal.pone.0147741 PubMedGoogle Scholar
12.
Lee  VE, Burkam  DT.  Inequality at the Starting Gate: Social Background Differences in Achievement as Children Begin School. Economic Policy Institute; 2002.
13.
Canfield  RL, Henderson  CR  Jr, Cory-Slechta  DA, Cox  C, Jusko  TA, Lanphear  BP.  Intellectual impairment in children with blood lead concentrations below 10 microg per deciliter.   N Engl J Med. 2003;348(16):1517-1526. doi:10.1056/NEJMoa022848 PubMedGoogle Scholar
14.
Perera  FP, Li  Z, Whyatt  R,  et al.  Prenatal airborne polycyclic aromatic hydrocarbon exposure and child IQ at age 5 years.   Pediatrics. 2009;124(2):e195-e202. doi:10.1542/peds.2008-3506 PubMedGoogle Scholar
15.
Liu  J, Lewis  G.  Environmental toxicity and poor cognitive outcomes in children and adults.   J Environ Health. 2014;76(6):130-138.PubMedGoogle Scholar
16.
Northstone  K, Joinson  C, Emmett  P, Ness  A, Paus  T.  Are dietary patterns in childhood associated with IQ at 8 years of age? a population-based cohort study.   J Epidemiol Community Health. 2012;66(7):624-628. doi:10.1136/jech.2010.111955 PubMedGoogle Scholar
17.
Freitas-Vilela  AA, Pearson  RM, Emmett  P,  et al.  Maternal dietary patterns during pregnancy and intelligence quotients in the offspring at 8 years of age: findings from the ALSPAC cohort.   Matern Child Nutr. 2018;14(1):e12431. doi:10.1111/mcn.12431 PubMedGoogle Scholar
18.
Lawlor  DA, Najman  JM, Batty  GD, O’Callaghan  MJ, Williams  GM, Bor  W.  Early life predictors of childhood intelligence: findings from the Mater-University study of pregnancy and its outcomes.   Paediatr Perinat Epidemiol. 2006;20(2):148-162. doi:10.1111/j.1365-3016.2006.00704.x PubMedGoogle Scholar
19.
Eriksen  H-LF, Kesmodel  US, Underbjerg  M, Kilburn  TR, Bertrand  J, Mortensen  EL.  Predictors of intelligence at the age of 5: family, pregnancy and birth characteristics, postnatal influences, and postnatal growth.   PLoS One. 2013;8(11):e79200. doi:10.1371/journal.pone.0079200 PubMedGoogle Scholar
20.
Koenen  KC, Moffitt  TE, Caspi  A, Taylor  A, Purcell  S.  Domestic violence is associated with environmental suppression of IQ in young children.   Dev Psychopathol. 2003;15(2):297-311. doi:10.1017/S0954579403000166 PubMedGoogle Scholar
21.
van Bakel  HJA, Riksen-Walraven  JM.  Parenting and development of one-year-olds: links with parental, contextual, and child characteristics.   Child Dev. 2002;73(1):256-273. doi:10.1111/1467-8624.00404 PubMedGoogle Scholar
22.
Patel  CJ, Rehkopf  DH, Leppert  JT,  et al.  Systematic evaluation of environmental and behavioural factors associated with all-cause mortality in the United States National Health and Nutrition Examination Survey.   Int J Epidemiol. 2013;42(6):1795-1810. doi:10.1093/ije/dyt208 PubMedGoogle Scholar
23.
Patel  CJ, Ioannidis  JPA, Cullen  MR, Rehkopf  DH.  Systematic assessment of the correlations of household income with infectious, biochemical, physiological, and environmental factors in the United States, 1999-2006.   Am J Epidemiol. 2015;181(3):171-179. doi:10.1093/aje/kwu277 PubMedGoogle Scholar
24.
Sontag-Padilla  L, Burns  RM, Shih  RA,  et al.  The Urban Child Institute CANDLE Study: Methodological Overview and Baseline Sample Description RAND Corporation; 2015. doi:10.7249/RR1336
25.
Wechsler  D.  Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II). NCS Pearson; 2011.
26.
Roid  GH.  Stanford-Binet Intelligence Scales, Fifth Edition. Riverside Publishing; 2003.
27.
Benjamini  Y, Hochberg  Y.  Controlling the false discovery rate: a practical and powerful approach to multiple testing.   J R Stat Soc Series B Stat Methodol. 1995;57(1):289–300. doi:10.1111/j.2517-6161.1995.tb02031.xGoogle Scholar
28.
Glickman  ME, Rao  SR, Schultz  MR.  False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies.   J Clin Epidemiol. 2014;67(8):850-857. doi:10.1016/j.jclinepi.2014.03.012 PubMedGoogle Scholar
29.
Tibshirani  R.  Regression shrinkage and selection via the lasso.   J R Stat Soc Series B Stat Methodol. 1996;58(1):267–288. doi:10.1111/j.2517-6161.1996.tb02080.xGoogle Scholar
30.
Van Houwelingen  JC.  Shrinkage and penalized likelihood as methods to improve predictive accuracy.   Stat Neerl. 2001;55(1):17–34. doi:10.1111/1467-9574.00154 Google Scholar
31.
Zhao  P, Yu  B.  On model selection consistency of lasso.   J Mach Learn Res. 2006;7:2541–2563.Google Scholar
32.
Burgette  LF, Reiter  JP.  Multiple imputation for missing data via sequential regression trees.   Am J Epidemiol. 2010;172(9):1070-1076. doi:10.1093/aje/kwq260 PubMedGoogle Scholar
33.
Stekhoven  DJ, Buhlmann  P.  MissForest—non-parametric missing value imputation for mixed-type data.   Bioinformatics. 2012;28(1):112-118. doi:10.1093/bioinformatics/btr597 PubMedGoogle Scholar
34.
Chow  GC.  Tests of equality between sets of coefficients in two linear regressions.   Econometrica. 1960;28(3):591–605. doi:10.2307/1910133 Google Scholar
35.
Organisation for Economic Co-operation and Development.  Society at a Glance 2014: OECD Social Indicators. OECD Publishing; 2014.
36.
Buckley  JP, Doherty  BT, Keil  AP, Engel  SM.  Statistical approaches for estimating sex-specific effects in endocrine disruptors research.   Environ Health Perspect. 2017;125(6):067013. doi:10.1289/EHP334 PubMedGoogle Scholar
37.
Christian  K, Morrison  FJ, Bryant  FB.  Predicting kindergarten academic skills: interactions among child care, maternal education, and family literacy environments.   Early Child Res Q. 1998;13(3):501–521. doi:10.1016/S0885-2006(99)80054-4 Google Scholar
38.
Roberts  J, Jurgens  J, Burchinal  M.  The role of home literacy practices in preschool children’s language and emergent literacy skills.   J Speech Lang Hear Res. 2005;48(2):345-359. doi:10.1044/1092-4388(2005/024) PubMedGoogle Scholar
39.
Green  CM, Berkule  SB, Dreyer  BP,  et al.  Maternal literacy and associations between education and the cognitive home environment in low-income families.   Arch Pediatr Adolesc Med. 2009;163(9):832-837. doi:10.1001/archpediatrics.2009.136 PubMedGoogle Scholar
40.
Sayegh  P, Arentoft  A, Thaler  NS, Dean  AC, Thames  AD.  Quality of education predicts performance on the Wide Range Achievement Test–4th Edition Word Reading subtest.   Arch Clin Neuropsychol. 2014;29(8):731-736. doi:10.1093/arclin/acu059 PubMedGoogle Scholar
41.
Baer  J, Kutner  M, Sabatini  J, White  S. Basic Reading Skills and the Literacy of America’s Least Literate Adults: Results From the 2003 National Assessment of Adult Literacy (NAAL) Supplemental Studies. NCES 2009-481. National Center for Education Statistics; 2009. Accessed December 27, 2019. https://eric.ed.gov/?id=ED505187
42.
Ladson-Billings  G.  From the achievement gap to the education debt: understanding achievement in U.S. schools.   Educ Res. 2006;35(7):3-12. doi:10.3102/0013189X035007003 Google Scholar
43.
Margo  RA.  Race and Schooling in the South, 1880-1950: An Economic History. University of Chicago Press; 1990. doi:10.7208/chicago/9780226505015.001.0001
44.
Bornstein  MH.  How infant and mother jointly contribute to developing cognitive competence in the child.   Proc Natl Acad Sci U S A. 1985;82(21):7470-7473. doi:10.1073/pnas.82.21.7470 PubMedGoogle Scholar
45.
Fewell  RR, Deutscher  B.  Contributions of receptive vocabulary and maternal style: variables to later verbal ability and reading in low-birthweight children.   Top Early Child Special Educ. 2002;22(4):181–190. doi:10.1177/027112140202200401 Google Scholar
46.
Herbers  JE, Cutuli  JJ, Lafavor  TL,  et al.  Direct and indirect effects of parenting on the academic functioning of young homeless children.   Early Educ Dev. 2011;22(1):77–104. doi:10.1080/10409280903507261 Google Scholar
47.
Craig  F, Operto  FF, De Giacomo  A,  et al.  Parenting stress among parents of children with neurodevelopmental disorders.   Psychiatry Res. 2016;242:121-129. doi:10.1016/j.psychres.2016.05.016 PubMedGoogle Scholar
48.
Estes  A, Munson  J, Dawson  G, Koehler  E, Zhou  X-H, Abbott  R.  Parenting stress and psychological functioning among mothers of preschool children with autism and developmental delay.   Autism. 2009;13(4):375-387. doi:10.1177/1362361309105658 PubMedGoogle Scholar
49.
Baker  BL, Blacher  J, Crnic  KA, Edelbrock  C.  Behavior problems and parenting stress in families of three-year-old children with and without developmental delays.   Am J Ment Retard. 2002;107(6):433-444. doi:10.1352/0895-8017(2002)107<0433:BPAPSI>2.0.CO;2 PubMedGoogle Scholar
50.
Neece  CL, Green  SA, Baker  BL.  Parenting stress and child behavior problems: a transactional relationship across time.   Am J Intellect Dev Disabil. 2012;117(1):48-66. doi:10.1352/1944-7558-117.1.48 PubMedGoogle Scholar
51.
Tachibana  Y, Fukushima  A, Saito  H,  et al.  A new mother-child play activity program to decrease parenting stress and improve child cognitive abilities: a cluster randomized controlled trial.   PLoS One. 2012;7(7):e38238. doi:10.1371/journal.pone.0038238 PubMedGoogle Scholar
52.
Burgdorf  V, Szabo  M, Abbott  MJ.  The effect of mindfulness interventions for parents on parenting stress and youth psychological outcomes: a systematic review and meta-analysis.   Front Psychol. 2019;10:1336. doi:10.3389/fpsyg.2019.01336 PubMedGoogle Scholar
53.
Lee  JJ, Wedow  R, Okbay  A,  et al; 23andMe Research Team; COGENT (Cognitive Genomics Consortium); Social Science Genetic Association Consortium.  Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals.   Nat Genet. 2018;50(8):1112-1121. doi:10.1038/s41588-018-0147-3 PubMedGoogle Scholar
54.
Cheesman  R, Hunjan  A, Coleman  JRI,  et al.  Comparison of adopted and non-adopted individuals reveals gene-environment interplay for education in the UK Biobank.   Psychol Sci. 2020;31(5):582-591. doi:10.1177/0956797620904450PubMedGoogle Scholar
55.
Cortes  KE.  Do bans on affirmative action hurt minority students? evidence from the Texas Top 10% Plan.   Econ Educ Rev. 2010;29(6):1110-1124. doi:10.1016/j.econedurev.2010.06.004 Google Scholar
56.
Blume  GH, Long  MC.  Changes in levels of affirmative action in college admissions in response to statewide bans and judicial rulings.   Educ Eval Policy Anal. 2014;36(2):228-252. doi:10.3102/0162373713508810 Google Scholar
57.
Pitre  CC.  Improving African American student outcomes: understanding educational achievement and strategies to close opportunity gaps.   West J Black Stud. 2014;38(4):209-217.Google Scholar
58.
Berkner  L, Chavez  L. Access to Postsecondary Education for the 1992 High School Graduates. Postsecondary Education Descriptive Analysis Reports. Statistical Analysis Report. Office of Educational Research and Improvement, US Department of Education; 1997. Accessed May 21, 2020. https://eric.ed.gov/?id=ED413854
59.
Carnevale  AP, Strohl  J. Separate and Unequal: How Higher Education Reinforces the Intergenerational Reproduction of White Racial Privilege. Center on Education and the Workforce, Georgetown Public Policy Institute; July 2013. Accessed April 1, 2019. https://cew.georgetown.edu/wp-content/uploads/SeparateUnequal.FR_.pdf
60.
Kaufman  P, Chen  X. Projected postsecondary outcomes of 1992 high school graduates. National Center for Education Statistics working paper 1999-15. June 1999. Accessed April 15, 2019. https://nces.ed.gov/pubs99/199915.pdf
61.
Wagmiller  RL, Adelman  RM. Childhood and intergenerational poverty: the long-term consequences of growing up poor. National Center for Children in Poverty. Accessed April 15, 2019. http://www.nccp.org/publications/pub_909.html
62.
Anderson  JW, Johnstone  BM, Remley  DT.  Breast-feeding and cognitive development: a meta-analysis.   Am J Clin Nutr. 1999;70(4):525-535. doi:10.1093/ajcn/70.4.525 PubMedGoogle Scholar
63.
Kramer  MS, Aboud  F, Mironova  E,  et al; Promotion of Breastfeeding Intervention Trial (PROBIT) Study Group.  Breastfeeding and child cognitive development: new evidence from a large randomized trial.   Arch Gen Psychiatry. 2008;65(5):578-584. doi:10.1001/archpsyc.65.5.578 PubMedGoogle Scholar
64.
Deave  T, Heron  J, Evans  J, Emond  A.  The impact of maternal depression in pregnancy on early child development.   BJOG. 2008;115(8):1043-1051. doi:10.1111/j.1471-0528.2008.01752.x PubMedGoogle Scholar
65.
Cassidy-Bushrow  AE, Sitarik  AR, Havstad  S,  et al.  Burden of higher lead exposure in African-Americans starts in utero and persists into childhood.   Environ Int. 2017;108:221-227. doi:10.1016/j.envint.2017.08.021 PubMedGoogle Scholar
×