A, Electrodes included in analysis (EGI Inc). Pink shading represents standard 10-20 electrodes; blue shading, additional electrodes chosen. B and C, Association between maternal Perceived Stress Scale (PSS) scores and infant EEG power in the β (B) and γ (C) frequency bands. Shading indicates 95% CIs.
Latent profile analysis identified 2 unique electroencephalography (EEG)–based profiles from infant log10-transformed absolute EEG power (A) and relative EEG power (B).
eTable 1. Comparison of infants retained vs excluded from EEG analysis
eTable 2. Comparison of absolute and relative EEG power across sites
eTable 3. Regression to determine whether variables of interest vary as a function of site for EEG profile membership
eTable 4. Regression to determine whether variables of interest vary as a function of site for individual frequency bands
eTable 5. Correlations between demographic variables
eTable 6. Regression of absolute infant EEG power on maternal sensitivity residual scores
eTable 7. Regression of relative infant EEG power on maternal sensitivity residual scores
eTable 8. Regression of infant EEG profiles on maternal sensitivity residual scores: latent profile analysis
eTable 9. Tests of sex effects in absolute infant EEG power for maternal perceived stress scores
eTable 10. Tests of sex effects in absolute infant EEG power for maternal sensitivity residual scores
eTable 11. Tests of sex effects in latent profile analysis of relative EEG profiles for maternal perceived stress
eTable 12. Tests of sex effects in latent profile analysis of relative EEG profiles for maternal sensitivity residual scores
eFigure. Flowchart of participant inclusion and exclusion criteria
eMethods. Supplementary methods
Customize your JAMA Network experience by selecting one or more topics from the list below.
Pierce LJ, Thompson BL, Gharib A, et al. Association of Perceived Maternal Stress During the Perinatal Period With Electroencephalography Patterns in 2-Month-Old Infants. JAMA Pediatr. 2019;173(6):561–570. doi:10.1001/jamapediatrics.2019.0492
Is caregiver exposure to or perception of stress associated with resting electroencephalography power in 2-month-old infants?
In this cohort study of electroencephalographic data from 70 infants who were 2 months of age, perceived maternal stress was significantly and negatively associated with infant spectral power in the β and γ frequency bands and with a unique profile identified using latent profile analysis of the electroencephalographic data.
The findings suggest that infant electroencephalography can be used as an index of the association between caregiver stress and neurodevelopment and to identify indicators of risk and resilience in very young infants.
Variation in child responses to adversity creates a clinical challenge to identify children most resilient or susceptible to later risk for disturbances in cognition and health. Advances in establishing scalable biomarkers can lead to early identification and mechanistic understanding of the association of early adversity with neurodevelopment.
To examine whether maternal reports of stress are associated with patterns in resting electroencephalography at 2 months of age and whether unique electroencephalographic profiles associated with risk and resiliency factors can be identified.
Design, Setting, and Participants
For this cohort study, a population-based sample of 113 mother-infant dyads was recruited from January 1, 2016, to March 1, 2018, during regularly scheduled pediatric visits before infants were 2 months of age from 2 primary care clinics in Boston, Massachusetts, and Los Angeles, California, that predominantly serve families from low-income backgrounds. Data are reported from a single time point, when infants were aged 2 months, of an ongoing cohort study longitudinally following the mother-infant dyads.
Maternal reported exposure to stressful life events and perceived stress.
Main Outcomes and Measures
Spectral power (absolute and relative) in different frequency bands (Δ, θ, low and high α, β, and γ) from infant resting electroencephalography (EEG) and EEG profiles across frequency bands determined by latent profile analysis.
Of 113 enrolled infants, 70 (mean [SD] age, 2.42 [0.37] months; 35 girls [50%]) provided usable EEG data. In multivariable hierarchical linear regressions, maternal perceived stress was significantly and negatively associated with absolute β (β = −0.007; 95% CI, −0.01 to −0.001; semipartial r = −0.25) and γ power (β = −0.008; 95% CI, −0.01 to −0.002; semipartial r = −0.28). Maternal educational level was significantly and positively associated with power in high α, β, and γ bands after adjusting for covariates (high school: γ: β = 0.108; 95% CI, 0.014-0.203; semipartial r = −0.236; associate’s degree or higher: high α: β = 0.133; 95% CI, 0.018-0.248; semipartial r = 0.241; β: β = 0.167; 95% CI, 0.055-0.279; semipartial r = 0.309; and γ: β = 0.183; 95% CI, 0.066-0.299; semipartial r = 0.323). Latent profile analysis identified 2 unique profiles for absolute and relative power. Maternal perceived stress (β = 0.13; 95% CI, 0.01-0.25; adjusted odds ratio [AOR], 1.14; 95% CI, 1.01-1.28) and maternal educational level (high school: β = 3.00; 95% CI, 0.35-5.65; AOR, 20.09; 95% CI, 1.42-283.16; associate’s degree or higher: β = 4.12; 95% CI, 1.45-6.79; AOR, 61.56; 95% CI, 4.28-885.01) were each associated with unique profile membership.
Conclusions and Relevance
These findings suggest that unique contributions of caregiver stress and maternal educational level on infant neurodevelopment are detectable at 2 months; EEG might be a promising tool to identify infants most susceptible to parental stress and to reveal mechanisms by which neurodevelopment is associated with adversity. Additional studies validating subgroups across larger cohorts with different stressors and at different ages are required before use at the individual level in clinical settings.
Childhood exposure to adverse experiences increases later risk of learning, behavior, and health consequences.1,2 However, associations between early experiences and later outcomes are complex, arising from interactions between the type and timing of experiences, intrinsic (heritable) characteristics of the individual, and access to supports that can buffer early adverse exposures. Because of the multidimensional association among many variables, recent research has sought to describe specific mechanisms that link early experience and later outcomes and to discover biological markers that identify children who show resilience and those susceptible to long-term physical or mental health concerns and behavioral disruptions.
Evidence suggests that early adversity leads to negative outcomes in part through the association of excessive prenatal and early postnatal stress with developing neural and physiologic systems.2-8 These negative outcomes can vary, affecting a subset of infants who may be most vulnerable. Measures of neural function might therefore serve as early indexes of heightened risk for long-term negative outcomes. Given the plasticity of the developing brain, identifying neural indicators of risk early in development creates the opportunity to interrupt and mitigate downstream consequences of early-life stress.
The present study uses electroencephalography (EEG), which can be administered relatively quickly and easily within a developmental context. Perturbations in EEG features that reflect underlying neurodevelopmental alterations can be observed within the first months of life9 and are sensitive to exposure to adversity in different contexts.10-14 However, little work15 has examined how the nature of adversity might influence EEG findings during the earliest stages (ie, within the first 6 months) or whether protective factors (eg, maternal educational level) might buffer potential disturbances. The present study recorded baseline EEG from 2-month-old infants from predominantly low-income households whose mothers reported varying degrees of exposure to and perception of stress. Traditional and novel16 methods were used to characterize spectral profiles of the EEG with the hypothesis that the association of the profiles with stress and protective factors can be identified early in development.
Mother-infant dyads (n = 113) who were followed up longitudinally as part of a cohort study examining responses to early adversity participated when infants were approximately 2 months of age. Participants were recruited from Boston Children’s Hospital Primary Care Center, Boston, Massachusetts, and AltaMed Clinic, Children’s Hospital Los Angeles, Los Angeles, California, from January 1, 2016, to March 1, 2018. These clinics predominantly serve families from low-income backgrounds (ie, below the federal poverty level, qualify for public health insurance). Exclusion criteria were gestational age younger than 37 weeks; birth weight less than 2500 g; identified genetic, metabolic, or neurologic disorders; uncorrected vision difficulties; or birth-related complications (eg, extended stay in the neonatal intensive care unit). Parents provided written informed consent. The Boston Children's Hospital Institutional Review Board and the Children's Hospital Los Angeles Institutional Review Board approved all procedures. All data were deidentified.
Demographic information was collected from families (Table 1). Because not all participants reported income, zip code data were used to determine percentage of households below the federal poverty level within each participant’s neighborhood (American Community Survey 5-Year Data 2015 US Census tract); these data were used as a socioeconomic marker.17
Exposure to recent stressful life events was assessed using a modified 14-item Recent Life Events Questionnaire (RLEQ).18 Endorsed events occurred during pregnancy or since the child’s birth. Maternal appraisals of stress during the prior month were assessed using the Perceived Stress Scale (PSS).19,20 Both measures were administered during the 2-month study visit (eTable 1 and eFigure in the Supplement).
Identical protocols were used at each collection site. Infant recordings were completed using a 128-Channel Hydrocel Sensor Net System (EGI Inc) soaked in a 37°C salt water solution.
Recordings were completed in dimly lit rooms with a low-electrical-signal background. Infants were 60 cm from a computer monitor (Dell model P2314Ht, 49 × 29 cm) held backward over their mother’s shoulder or seated and facing forward on her lap. NetAmps 300 Amplifier and NetStation, version 4.5.4 (Boston Children’s Hospital) and NetAmps 400 Amplifier and NetStation, version 5 (Children’s Hospital Los Angeles) were used to record EEG data. Data were amplified and sampled at 500 Hz while infants watched a video of infant toys for up to 5 minutes (mean impedance, <100 kΩ). A trained researcher engaged minimally with infants using a toy and bubbles to redirect attention toward the screen only when necessary. Electroencephalographic data were collected from 110 infants (in 3 cases, infants would not tolerate the EEG net being placed). After processing and data quality checks, EEG recordings from 70 participants were suitable for analysis (eFigure in the Supplement).
Raw EEG files were exported from NetStation, version 4.5.4 in MATLAB format (MathWorks Inc) and preprocessed using the Harvard Automated Processing Pipeline for Electroencephalography,21 optimized for use with infant EEG and used with MATLAB, version 2014b and EEGLAB, version 14.0.0b.22 The Batch EEG Automated Processing Platform23 was used for subsequent spectral analysis. Within Harvard Automated Processing Pipeline for Electroencephalography, data were band-pass filtered at 1 to 249 Hz. A subset of 46 channels was selected for further processing,21 including standard 10-20 electrodes and additional electrodes based on previous infant reports14,24 (Figure 1A). Independent component analysis and advanced processing techniques were used to correct artifact (eMethods in the Supplement), and 2-second segments were extracted. Segments with remaining artifact were removed before re-referencing to the average. Power spectra for each channel were extracted using a fast Fourier transform with a multitaper window.22 For each participant, mean power across all segments was calculated. Log10-transformed absolute power was calculated for each frequency band of interest: Δ (2-4 Hz), θ (4-6 Hz), low α (6-9 Hz), high α (9-13 Hz), β (13-30 Hz), and γ (30-50 Hz). These numbers were then averaged across all included electrodes. Relative power was calculated as power in each frequency band divided by total power.25 Site differences in absolute and relative power were explored and statistically accounted for as needed (eMethods and eTables 2-4 in the Supplement).
To examine associations between reported maternal stress and infant EEG findings, multivariable hierarchical linear regressions were performed in SPSS statistical software, version 24.0 (IBM Inc). A hierarchical approach was used to examine how much additional variance in infant EEG power (absolute or relative in each of the 6 frequency bands) can be explained by maternal stress measures (RLEQ and PSS scores) after adjusting for demographic factors (infant age, neighborhood poverty, and maternal educational level) entered into the first step of the model. Sample size calculation using G*Power, version 3.126,27 indicated that a sample size of 70 could identify an effect size of 0.2, with 80% power using 5 indicators. Significance was set at P < .05, 2-sided, for all measures. The 95% CIs and semipartial correlations are presented for all analyses. Correlations among all demographic variables are reported in eTable 5 in the Supplement.
Latent profile analysis (LPA) is a statistically driven technique that was used to isolate unique profiles within the EEG, providing a powerful data reduction strategy while determining whether maternal characteristics were associated with infant membership in separable EEG profiles. Mplus, version 7.3128 was used to fit latent class models to whole-brain EEG power in each frequency band (log10-transformed absolute or relative). Each analysis began with a 2-class model. Successive models added more classes until model fit no longer increased. Model fit was assessed with goodness-of-fit statistics: smallest Bayes information criteria (BIC), smallest Akaike information criteria (AIC), entropy that approaches 1, and Lo-Mendell-Rubin (LMR) P < .05.
To explore whether subgroups of infants had different EEG profiles, variables associated with membership in latent profiles were identified using multinomial logistic regression in Mplus (eMethods in the Supplement), with demographic factors and maternal stress as indicators and profile membership (absolute or relative EEG profiles) as the outcome variable. Unstandardized β coefficients, 95% CIs, and adjusted odds ratios (AORs) are presented for all analyses. Because factors associated with profile membership (eg, maternal stress levels) can be explored, this novel use of latent profile analysis15 could ultimately be used to isolate subgroups of infants with neurodevelopmental profiles that indicate higher or lower relative risk and resilience.
For participants who contributed EEG data, 17 PSS scores (24%) and 1 neighborhood poverty value (1%) were missing (late introduction of measure, address not disclosed) (eTable 1 in the Supplement). Multiple imputation with 10 completed data sets was used to replace missing indicator values.29
Of 113 enrolled infants, 70 (mean [SD] age, 2.42 [0.37] months; range, 1.93-3.38 months; 35 girls [50%]) provided usable EEG data. Mean (SD) maternal age was 29 (5.7) years. A total of 51 mothers (73%) were high school educated or below, and 20 (29%) reported a family income of less than $16 000 in the past year. Race/ethnicity was white in 35 (50%), African American in 23 (33%), and Hispanic, Latino, or Spanish origin in 40 (57%). Mean (SD) neighborhood poverty levels were 24.6% (11.2%). Included and excluded families did not differ on demographic variables reported (eTable 1 in the Supplement). Table 1 summarizes demographic characteristics.
Hierarchical linear regression was used to estimate the association between maternal stress and log10-transformed absolute EEG power, adjusting for demographic factors (Table 2). Inclusion of maternal stress measures (PSS and RLEQ) was associated with more variance in absolute infant EEG power than demographic factors alone (8.5%, γ band). Maternal PSS scores were associated with lower infant EEG power in the β and γ bands (Figure 1B and C). Semipartial correlations indicated a small to medium association between PSS scores and variance in EEG power, with demographic variables held constant. Maternal RLEQ scores were not associated with absolute EEG power in any frequency band. Other significant factors included infant age (positive, all frequency bands) and mothers having an associate’s degree or higher (higher high α, β, and γ power) or high school education (higher γ power) compared with no high school. Neighborhood poverty level was not significantly associated with EEG power in any frequency band.
For relative EEG power (Table 3), no additional variance was explained by maternal stress measures after adjusting for demographic variables. Infant age was associated with lower relative θ and lower relative β power. No other associations were found.
Maternal perceived stress, educational level, and infant age were all associated with variation in EEG power. Different patterns emerged for each variable as well as for absolute vs relative power. Whereas maternal educational level was positively associated with higher absolute spectral power across multiple frequency bands, maternal perceived stress was uniquely associated with decreased absolute power in high-frequency bands.
For log10-transformed, whole-brain absolute EEG power, 3 profiles were tested, and a 2-profile solution was selected as the best model (2-profile: BIC = −461.43, AIC = −504.16, entropy = 0.966, LMR P = .007; 3-profile: BIC = −596.16, AIC = −654.62, entropy = 0.968, and LMR P = .10). Mean posterior probabilities were 44 (99.7%) in profile 1 and 26 (98.5%) in profile 2. Profile 1 (low power) displayed lower power than profile 2 (high power) across all frequency bands (Figure 2).
Neither maternal stress measure was associated with profile membership (PSS: β = −0.15; 95% CI, −0.33 to 0.03; AOR, 0.86; 95% CI, 0.72-1.03; RLEQ: β = 0.39; 95% CI, −0.47 to 1.05; AOR, 1.34; 95% CI, 0.62-2.87); however, older infant age (β = 4.47; 95% CI, 218.00-6.76; AOR, 87.36; 95% CI, 8.81-865.41) and higher maternal educational level (high school: β = 3.00; 95% CI, 0.35-5.65; AOR, 20.09; 95% CI, 1.42-283.16; associate’s degree or higher: β = 4.12; 95% CI, 1.45-6.79; AOR, 61.56; 95% CI, 4.28-885.01) were associated with membership in the high-power profile. Poverty level was not associated with profile membership (β = −2.34; 95% CI, −8.34 to 3.66; AOR, 0.10; 95% CI, 0.00-38.77).
For whole-brain relative EEG power, 3 profiles were tested, and a 2-profile solution was selected as the best model (2-profile: BIC = −2271.50, AIC = −2314.22, entropy = 0.92, and LMR P = .001; 3-profile: BIC = −2308.21, AIC = −2366.67, entropy = 0.907, and LMR P = .12). Mean posterior probabilities were 36 (97.3%) for profile 1 and 34 (99.0%) for profile 2. Profile 1 (relative high) displayed higher power than profile 2 in high-frequency bands (ie, β and γ) but lower power than profile 2 (relative low) in low-frequency bands (ie, Δ and θ) (Figure 2).
Greater PSS scores (β = 0.13; 95% CI, 0.01-0.25; AOR, 1.14; 95% CI, 1.01-1.28) were associated with membership in the relative low profile. Increases of 1 point on the PSS were associated with 1.14 times greater odds of membership in the relative low profile, adjusting for demographic variables. The RLEQ scores were not associated with profile membership (RLEQ: β = −0.25; 95% CI, −0.70 to 0.20; AOR, 0.79; 95% CI, 0.50-1.22). Neither maternal educational level (high school: β = −0.60; 95% CI, −2.15 to 0.95; AOR, 0.55; 95% CI, 0.12-2.58; college: β = −0.59; 95% CI, −2.31 to 1.13; AOR, 0.55; 95% CI, 0.10-3.11) nor poverty level (β = −0.50; 95% CI, −5.69 to 4.69; AOR, 0.61; 95% CI, 0.00-109.30) were associated with profile membership. Infant age was positively associated with membership in the relative low profile (β = 2.38; 95% CI, 0.36-4.40; AOR, 10.80; 95% CI, 1.44-81.35).
To test whether maternal perceived stress remained associated with infant EEG findings in the context of exposure to stressful life events, standardized residual scores were created by regressing PSS on RLEQ scores. Higher residuals reflected mothers who reported greater perceptions of stress than predicted based on reported exposure to stressful life events. These maternal sensitivity scores may reflect higher sensitivity to stressors as opposed to acute expected increases attributable to isolated stressful exposures. All EEG outcomes were regressed on maternal sensitivity, and in line with patterns observed with PSS scores, every point increase in maternal sensitivity was associated with a small to moderate reduction in absolute β and γ power and 2.92 times greater odds of membership in the relative low profile, adjusting for demographic variables (eTables 6-8 in the Supplement).
Because responses to stress have been found to differ based on sex, all significant associations between maternal stress and infant EEG findings were explored. No associations with sex were observed for these outcomes (eTables 9-12 in the Supplement).
High levels of maternal perceived stress during early infancy have been correlated with neural development in ways that prior research has suggested might influence learning, behavior, and health.2,5,6,30 Our results suggest that infants whose mothers reported higher levels of perceived stress during the early postnatal period had lower high-frequency spectral power than infants whose mothers reported lower levels of perceived stress. This pattern was evident in both traditional regression analyses and LPA, whereby a unique spectral profile was associated with higher maternal perceived stress and sensitivity scores. Higher maternal educational level was associated with increased spectral power across multiple frequency bands. Together, these patterns suggest that (1) stress-based alterations to neurophysiologic maturation represent a pathway by which early adversity affects early child development, (2) EEG patterns might indicate risk of atypical developmental trajectories correlated with early exposure to caregiver stress, and (3) protective factors, such as maternal educational level, may mitigate effects of early adverse factors.
The correlation of high-frequency power with caregiver stress is consistent with prior studies11,14,15,31 of children at high risk for stress exposure. Reduced baseline γ power,14,15 sometimes with concomitantly elevated low-frequency power,11 occurs from 6 months of age in infants reared in low compared with high socioeconomic environments and in infants exposed to extreme psychosocial deprivation through institutionalization.14,31 Evidence suggests that altered patterns can persist into childhood.15 Although the association of caregiver or infant stress with EEG power in those studies was not explicitly tested, the present results suggest that, by 2 months of age, perceived maternal stress may be associated with EEG perturbations.
Higher levels of maternal education were associated with a high-power EEG profile and higher EEG power across multiple frequency bands, controlling for other factors, such as neighborhood poverty. Higher spectral power (eg, γ) during infancy has been associated with better cognitive outcomes into childhood,15,32 suggesting that experiences that accompany opportunities for achieving higher levels of maternal education may buffer against the association of early-life stress with neurodevelopment. In future research, we hope to elucidate precisely what it is about opportunities to attain more education that mediates a more positive outcome. Data on typical developmental trajectories of infant EEG are currently limited. Future efforts will characterize typical EEG profiles during the first postnatal year and relate specific profile patterns to short- and long-term behavioral, cognitive, and health outcomes.
Infant age and maternal perceived stress were associated with the same EEG profile, consistent with early-life stress exposure being associated with disruptions in the timing of neurodevelopmental processes, including accelerated maturation of certain brain structures. Such findings have been observed in animal models33-35 and human neonates.36 Perceived maternal stress was arguably associated with a more mature EEG pattern in terms of relative spectral power. This finding is consistent with the hypothesis of accelerated neurodevelopment and an earlier closing of sensitive periods of developmental plasticity. Critically, precocious maturation that arises from early-life stress may be associated with disrupted as opposed to enhanced long-term cognitive development.37 Of note, differences in absolute and relative EEG power were evident as a function of age differences on the order of weeks. This finding is consistent with prior studies38-40 and illustrates the rapid nature of neural development during the first months of life.
Determining how early, elevated levels of maternal stress may exert widespread and long-lasting functional effects on the infant remains an area of active investigation. Excessive early stress is associated with synapse formation and myelination41,42 and likely occurs through gene × environment interactions, which can accelerate development,33,35 disrupting temporally appropriate, circuit-based maturation that is essential for healthy brain development.43 The high sensitivity to powerful experiences, both positive and negative,11,12,44,45 is likely associated with the developmental timing and organization of neural circuits.46,47 Future studies will need to capture direct measures of stress in the infant, complementing the reporting here of a marker that is proximal to such early neurobiological changes. Such measures may ultimately help to identify infants at risk for negative consequences of early-life stress and factors that can mitigate early disruption.
Maternally reported number of recent stressful life events was not associated with EEG outcomes, which suggests that the number of reported events alone does not objectively reflect the varied association with different events or that stressful events may need to exceed a threshold to detect correlations in the infant.1 Moreover, very young infants may be less responsive to, or more protected from, exposure to stressful life events as a result of maternal buffering behaviors—an idea that is consistent with the observed associations among maternal educational level, perceived stress, sensitivity scores, and infant EEG. Of note, other variables not measured here could also contribute to the current findings.
Results from the present study suggest that greater levels of perceived maternal stress during the earliest stages of postnatal development are correlated with alterations in developing EEG. The EEG-based markers can be identified early in development, providing a potentially promising framework to understand how early stress exposure may be associated with neurodevelopment and how such markers might ultimately be used to facilitate prevention or limit long-term adverse effects.
Accepted for Publication: January 2, 2019.
Corresponding Author: Charles A. Nelson, PhD, Laboratories of Cognitive Neuroscience, Boston Children’s Hospital, 1 Autumn St, Sixth Floor, Boston, MA 02215 (email@example.com).
Published Online: April 8, 2019. doi:10.1001/jamapediatrics.2019.0492
Author Contributions: Dr Pierce had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Pierce, Thompson, Valdes, Levitt, Nelson.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Pierce, Thompson, Reilly, Nelson.
Critical revision of the manuscript for important intellectual content: Pierce, Thompson, Gharib, Schlueter, Valdes, Roberts, Conroy, Levitt, Nelson.
Statistical analysis: Pierce, Thompson.
Obtained funding: Levitt, Nelson.
Administrative, technical, or material support: All authors.
Supervision: Levitt, Nelson.
Conflict of Interest Disclosures: Drs Levitt and Nelson reported receiving grants from JPB Foundation during the conduct of the study. No other disclosures were reported.
Funding/Support: Support for this research was provided by grant 1025 from the JPB Foundation–funded Research Network on Toxic Stress and Health (Drs Levitt and Nelson) and Simms/Mann Family Foundation (Dr Levitt).
Role of the Funder/Sponsor: The funding source 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: Perry Dinardo, BA (research assistant), Emma Satterthwaite-Muresianu, BA (undergraduate thesis student), Saul Urbina-Johanson (undergraduate thesis student), and Marissa Rodriguez (undergraduate thesis student), Boston Children’s Hospital, Boston, Massachusetts, and Juan Ramirez, BS, Children’s Hospital Los Angeles, Los Angeles, California, assisted with data collection. Mark Wade, PhD (postdoctoral fellow), Boston Children’s Hospital, Sarah Jensen, PhD (postdoctoral fellow), Boston Children’s Hospital, and Kate McLaughlin, PhD, Harvard Graduate School of Education, provided input on statistical analyses. Sandra Figueroa, BS, Juan Ramirez, BS, Michele Castro, CMA, Dulce Angel Aceves, CMA, and Sara Palencia, BS, MA, CCC-SLP, Children’s Hospital Los Angeles, Los Angeles, California, and Jenny Chan, BA (research coordinator), Gabriela Kovacikova, BA (research assistant), Anna Hoover, BA (research assistant), and Hannah Durant, BA (research assistant), Boston Children’s Hospital Primary Care Center, Boston, Massachusetts, made essential contributions to family recruitment, engagement, and database administration. Mona Patel, MD, and Heydeh Khalili, MA, facilitated the study in Los Angeles with AltaMed Clinic, Children’s Hospital Los Angeles. We thank the children and families who participated in this research.