Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jul 30.
Published in final edited form as: Infant Behav Dev. 2025 Jul 30;80:102117. doi: 10.1016/j.infbeh.2025.102117

NIH Baby Toolbox® Methodology and Norms Development

Y Catherine Han 1, Elizabeth M Dworak 1, Maxwell Mansolf 1, Hubert Adam 1, Lihua Yao 1, Miriam A Novack 1, Sarah Pila 1, Rachel M Flynn 2, Amanda M Flagg 3, Vitali Ustsinovich 1, Kay Savio 4, Greg J Byrne 1, Richard C Gershon 1, Aaron J Kaat 1
PMCID: PMC12372496  NIHMSID: NIHMS2101250  PMID: 40743801

Abstract

The NIH Baby Toolbox is a comprehensive battery of cognitive, motor, and social-emotional assessments designed for infants aged 1–42 months. Although many standardized, normed early developmental assessments exist, these measures are time-intensive, costly, and require trained experts to administer, score, and interpret. The Baby Toolbox is intended to serve as a “common currency” across developmental fields and provide an extension to the NIH Toolbox for Assessment of Neurological and Behavioral Function® (NIHTB), which is appropriate for individuals from 3–90 years of age. The Baby Toolbox was normed using a nationally representative sample of infants and toddlers aged 16 days to 48 months old from the US population (N = 2,515; n = 2,042 English, n = 473 Spanish). The sample was weighted to match the US Census based on the demographic targets used to create the norming plan. After data collection, the demographic targets were updated to reflect the current figures. Cases were individually weighted to ensure representativeness. Unweighted and weighted demographics are reported herein. Scores were regressed on age in months (for ages where the tests or composites were appropriate) using generalized additive models for location, scale, and shape to obtain expected reference values and standard deviations across age. The Baby Toolbox norming study has been designed to minimize bias and maximize representativeness and precision of estimates. The availability of a “toolbox” of tests built using modern measurement theory to address developmental concerns will be an important foundation for addressing critical research questions in infant neurological and behavioral health.

Keywords: NIH Baby Toolbox, infancy, early childhood, developmental assessment, normed scores

1. Introduction

Infancy—defined here as the first 3 years of life—is a period of rapid neurological development and foundational skill acquisition (Rice & Barone, 2000). Monitoring neurodevelopmental changes during infancy and early childhood is critical for identifying typical versus atypical developmental trajectories, understanding underlying neurological conditions, and implementing timely interventions to optimize developmental outcomes (Blauw-Hospers & Hadders-Algra, 2005). The NIH Baby Toolbox (Baby Toolbox), developed under the NIH Neuroscience Blueprint, provides a downward extension of the NIH Toolbox® for Assessment of Neurological and Behavioral Function (NIHTB) to offer common standardized and normed measures for children from 16 days through 42 months (Gershon, et al., 2024).

The NIHTB was created as a comprehensive assessment for research and clinical use (Gershon et al., 2010). It provides normed measures of cognitive, emotional, sensory, and motor functioning for individuals from 3 years through 85 years and, since its release in 2013, has been extensively utilized in developmental research (Akshoomoff et al., 2014; Becker et al., 2023; Paolillo et al., 2020), clinical research (Fox et al., 2022), and large longitudinal studies spanning childhood to older adulthood (e.g., ECHO, Knapp et al., 2023; ABCD, Thompson, 2019). Although the NIHTB provides an impressive framework for lifespan assessments, it lacks tools for infants under 3 years of age.

Being able to compare a child’s scores on developmental measures with those of the same and different ages is a necessary component of determining the child’s development status and tracking milestones. As such, a major goal of the Baby Toolbox is to create standardized, norm-referenced values (normed scores) for measures within the primary domains of Cognition, Motor, and Social-Emotional Functioning. In doing so, researchers and clinicians have access to the most updated comparisons available between the child’s scores on these domains and other children in the general population based on age. Normed scores provide context for evaluating performance relative to one’s peer group—that is, by allowing for the identification of what is typical compared to atypical development and the implementation of early screening and intervention methods when an infant is not meeting expected developmental outcomes. In order to improve evidence for validity, ensuring test scores are compared to a representative reference group improves validity, especially for the consequences of testing related to decision-making (American Educational Research Association, American Psychological Association, & National Council on Measurement in Education, 2014).

The relevant population for the Baby Toolbox was infants and toddlers between the ages of 16 days and 42 months from English- or Spanish-speaking homes in the United States. Importantly, obtaining norms for this demographic can be challenging and requires a high level of methodological consideration (Oakes, 2017). Here, we describe the methodology that was used to develop normed scores for the Baby Toolbox, the process of obtaining data, and statistical considerations when deriving reference values.

2. Method

2.1. Baby Toolbox Measure Selection and Development

A challenge of developmental science is choosing assessments to fit research designs and have appropriate validity evidence for the proposed purpose of assessment. Assessments vary from standardized and normed measures (e.g., the Bayley Scales of Infant Development, Bayley & Aylward, 2019) to standardized laboratory-based paradigms (e.g., the LabTab, Goldsmith & Rothbart, 1996), to unstandardized lab-specific paradigms. The goal of the Baby Toolbox was to understand which measures or paradigms may be appropriate for monitoring infant and toddler development to ensure that the selected measures (1) reflected the most up-to-date research in the field and they (2) addressed known problems with existing measures. A two-pronged approach was implemented to achieve these goals. First, an expert survey of developmental researchers and clinicians was administered and reviewed. Second, a scoping review of available instruments was conducted to leverage existing science in developing a new standardized assessment battery for birth to 42-month-olds.

A structured survey with questions having fixed and open-ended short responses was sent to developmental listservs, professional organizations, and research groups. The survey was completed by 504 individuals. The majority of respondents reported multiple areas of expertise, including developmental psychology (71%), clinical psychology (21%), cognitive neuroscience (21%), neuropsychology (13%), pediatrics (8%), social neuroscience (8%), and psychometrics (7%). These experts’ initial survey evaluated six content areas (Language, Cognition, Motor, Self-Regulation, Social Functioning, Numeracy) and administrative details (e.g., maximum battery administration duration, battery material costs, etc.).

Concurrently, a scoping review of the literature following PRISMA guidelines (Tricco et al., 2018) was conducted to identify common assessments in each of the six domains, excluding three commercial assessments frequently used in the field: Bayley Scales of Infant Development (Bayley & Aylward, 2019), Mullen Scales of Early Learning (Mullen, 1995), and Battelle Developmental Inventory (Berls & McEwen, 1999). As a result, 37,801 abstracts were eligible for title-and-abstract screening after removing duplication, ranging from 3000 (Self-Regulation) to over 9000 (Language) abstracts per domain. Approximately 10–15% of abstracts within each domain were eligible for full-text review ensuring content appropriateness. Finally, we extracted relevant information related to the tests utilized, whether the sample was clinically ascertained or from the general population, and any psychometric properties reported within the manuscript specific to that sample. After measure identification and measure approval by the NIH Steering Committee, the appropriate use and distribution rights for the chosen measures were secured, and the measures were integrated into an administration platform on the iPad; see Table 1 for list of measures. For further information on the development of specific Baby Toolbox measures see Novack, Dworak, Han, et al. (2025) for language measures, Flynn et al. (2025) for executive function/memory measures, Pila et al. (2025) for math measures, Blackwell et al. (2025) for social-emotional functioning measures, and Adolf et al. (2025) for motor measures.

Table 1.

List of NIH Baby Toolbox domains and measures

Domain Subdomain Measure Age range (months) Modality Time to administer (minutes)
Cognition Language Looking While Listening 6–24 Gaze 5
Language Picture Vocabulary MacArthur-Bates >=25 Touch 3
Language Communicative Development Inventories (CDI) Computerized Adaptive Tests (CAT) Production 9–30 Caregiver Report/Observation 2
Language MacArthur-Bates Communicative Development Inventories (CDI) Computerized Adaptive Tests (CAT) Comprehension 7–18 Caregiver Report/Observation 2
Language Mullen Receptive 1–42 Caregiver Report/Observation 3
Language Mullen Expressive 1–42 Caregiver Report/Observation 3
Executive Function/Memory Mullen Visual Reception 1–42 Caregiver Report/Observation & Touch 4
Executive Function/Memory Familiarization 6–21 Gaze 4
Executive Function/Memory Visual Delayed Response 22–42 Touch 5
Executive Function/Memory Delayed Memory 22–42 Touch 4
Early Math Who Has More 25–42 Touch 3
Early Math Subitizing 25–42 Observation 1
Early Math Verbal Arithmetic 37–42 Observation 2
Early Math Counting 25–42 Observation 2
Motor Gross Motor Get Up and Go 1–42 Caregiver Report/Observation 4.5
Fine Motor Reach to Eat 1–42 Caregiver Report/Observation 5.5
Postural Control (Gross Motor, supplementary) Sit and Stand 1–42 Caregiver Report/Observation 3.5
Social-Emotional Functioning Temperament Infant Behavior Questionnaire - Revised - Very Short Form (IBQ-R-VSF) 3–12 Caregiver Report 6
Temperament Early Childhood Behavior Questionnaire -Very Short Form (ECBQ-VSF) 13–36 Caregiver Report 6
Temperament Children’s Behavior Questionnaire - Very Short Form (CBQ-VSF) 37–42 Caregiver Report 6
Negative Affect PROMIS EC Parent-Report Bank v1.0 - Anger/Irritability 13–42 Caregiver Report/CAT 1
Negative Affect PROMIS EC Parent-Report Bank v1.0 - Anxiety 13–42 Caregiver Report/CAT 1
Negative Affect PROMIS EC Parent-Report Bank v1.0 - Depressive Symptoms 13–42 Caregiver Report/CAT 1
Psychological Well-Being PROMIS EC Parent-Report Bank v1.0 - Positive Affect 13–42 Caregiver Report/CAT 1
Self-Regulation PROMIS EC Parent-Report Scale v1.0 - Self-Reg - Flexibility 5a 13–42 Caregiver Report 1
Self-Regulation PROMIS EC Parent-Report Scale v1.0 - Self-Reg - Frustration Tolerance 6a 13–42 Caregiver Report 1
Social Relationships PROMIS EC Parent-Report SF v1.0 - Soc Rel - Child-Caregiver Interactions 13–42 Caregiver Report 1
Social Relationships PROMIS EC Parent-Report SF v1.0 - Soc Rel - Peer Relationships 4a 13–42 Caregiver Report 1
Social Relationships, Social Communication Caregiver Checklist 6–42 Caregiver Report 2
Social Relationships, Social Communication Social Observation 9–42 Observation 10
Open in a new tab

2.2. Sample Size Planning

2.2.1. Power analysis simulation study to obtain sample size

The development process for the Baby Toolbox normative scores followed a similar approach as the NIHTB Cognition Battery Version 3.0, using regression-based methods that are statistically more efficient than traditional norming approaches (LaForte et al., 2024; Oosterhuis et al., 2016; Ho et al., submitted). This involves using population-based deviations between the obtained scores and the expected scores for the reference group. To ensure adequate precision of the reference values, we conducted a simulation study prior to norming to estimate the necessary sample size for regression-based norming. For the simulation study, we followed a similar procedure as Oosterhuis et al. (2016).

The primary concern for reliable estimates during norming is the accurate quantification of the tails of the distribution. The simulation manipulated the sample size from 500 to 5000, assuming 75 items per test (though not all individuals would take all items given basal and ceiling rules) and manipulating the expected relationship with age. The primary outcome was the inter-percentile range (IPR), which represents the bootstrapped range from the 25th to 75th percentile and thus the accuracy of point-estimate scores. IPRs were formed around potential thresholds, or cut scores, and the population average, where accuracy is most important. As sample size increases, the IPR becomes very tight, but the more extreme potential cut scores (e.g., at −3 SD) have wider IPRs (Figure 1). Further details can be found in Supplementary Materials S1. Based on these results, the Baby Toolbox norming study proposed collecting a total planned sample size of up to 2,550 children from English-and Spanish-speaking households, which would provide sufficient precision to estimate the deviation-based percentiles.

Figure 1.

Figure 1.

Open in a new tab

Simulation-based power analyses for determining sample sizes for norming. IPR=interpercentile range. The 0.5, 0.75, and 0.9 labels represent the simulated correlational effects of age on scores.

2.2.2. Norming Statistical Plan

Although the Baby Toolbox is only intended to measure development through 42 months of age, we extended data collection through 48 months to allow us to observe scores at higher ranges of ability and therefore estimate scores for younger toddlers who perform higher than expected on certain measures with greater precision. Given developmental expectations and trends, high performing 36–42-month-olds may appear more like an average 48-month-old than like their same-age peers and thus observing higher scores (even if out of age range) would increase this precision.

The relevant norming statistics for deviation-based scores are an index of central tendency (mean or median ability at an age) and an estimate of dispersion (standard deviation of ability). The standardized residual between the obtained score and the expected score for one’s reference population can then be rescaled to an appropriate age-adjusted standard score with desirable characteristics such as a population mean of 100 and standard deviation of 15.

2.3. Sampling Plan

The sample accrual targets were based on the 5-year estimates from the 2019 American Community Survey (ACS) Public Use Microdata Sample (PUMS; US Census Bureau, 2019), which allowed for unique demographic combinations (e.g., racial identity by age within geographic region) that are not generally reported by the US Census Bureau. We developed targets for primary language spoken to the child by race and ethnicity nested within geographic region, age, sex assigned at birth, and maternal educational attainment (from women who gave birth within the previous 12 months, which is the extent of available Census data). Maternal educational attainment categories for Census targets consisted of High School Graduate, general educational diploma (GED), or Less; Some College (including 2-year Associate’s Degree and Technical School); and Bachelor’s Degree, or Higher.

The primary demographic variable of interest was either chronological age or adjusted age in months (CAAMs). Age was rounded to the nearest integer month (≥16 days were rounded up, ≤15 days were rounded down). Adjusted age in months was used for children who were born prematurely if they were 36 months or younger in chronological age and whose expected due date was at least three weeks prior to their actual date of birth. Strata were in 1-month increments from 1 to 6 CAAMs (6 strata), 2-month increments from 7 to 12 CAAMs (3 strata), 3-month increments from 13 to 30 CAAMs (6 strata), and 6-month increments from 31 to 48 CAAMs (3 strata). Due to practical constraints recruiting the youngest infants, a sample size of 100 for 1–3 CAAMs and 150 per strata for 4 CAAMs and above was used. Age strata were nested within geographic regions to account for population differences. Other relevant demographic information for norming included race and ethnicity by language exposure, sex assigned at birth, and caregiver household educational attainment. Targets for sex assigned at birth were nested within age strata, while caregiver education and the interaction between race/ethnicity and language were nested within broader age categories to prevent small, and potentially unattainable, target sample sizes.

For educational attainment, we had difficulty recruiting mothers with less than a high school education, reflecting a trend of increasing educational attainment for women and higher education of women who gave birth prior to when data collection began in 2023 in particular (Bailey et al., 2022; Livingston & Cohn, 2013). To avoid over-weighting the small proportion of children whose mothers had less than a high school education, we re-categorized some children (n = 39 of 2,515; 1.55%) into this group based on other caregiver information (e.g., father); and this new “caregiver education” variable was used for weight calculations based on the census proportions and for reporting here (Table 2) and in the other papers within this special issue. See Supplemental Materials S2 for more information on this reclassification and potential impacts.

Table 2.

Demographics of the full sample that complete the NIH Baby Toolbox

Demographic Category N Unweighted % Weighted % Weighted difference
Geographic Region Northeast 291 11.6% 15.9% 0.0%
Midwest 814 32.4% 21.4% 0.0%
South 981 39.0% 39.0% 0.0%
West 429 17.1% 23.7% 0.0%
Primary English 2042 81.2% 76.5% 0.0%
Language Spanish 473 18.8% 23.5% 0.0%
Sex Assigned at Birth Female 1236 49.1% 50.4% 1.5%
Male 1279 50.9% 49.6% 1.5%
Race / Ethnicity Hispanic (any race) 856 34.0% 26.0% 0.0%
Non-Hispanic White 835 33.2% 47.5% 0.0%
Non-Hispanic Black 536 21.3% 12.9% 0.0%
Non-Hispanic Asian 55 2.2% 5.0% 0.0%
Non-Hispanic Other 233 9.3% 8.6% 0.0%
Caregiver Education Less Than High School Graduate 114 4.5% 10.4% 0.0%
High School Graduate or GED 515 20.5% 23.2% 0.0%
Some College (including Associates or Technical Degree) 759 30.2% 29.6% 0.0%
Bachelor’s Degree 700 27.8% 22.6% 0.0%
Advanced Degree 427 17.0% 14.2% 0.0%
Open in a new tab

Among Spanish-speaking households, targets were based on CAAM groups for individuals of Hispanic or Latino ethnicity, which included targets for 1–6 CAAMs, 7–12 CAAMs, 13–24 CAAMs, 25–36 CAAMs, and 37–48 CAAMs. For Spanish-speaking households that were non-Hispanic, targets were based on any age, geographic region, and racial identity. For English-speaking households, targets were based on Hispanic or Latino ethnicity regardless of racial identity or Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Asian, and Non-Hispanic Other (including multiracial) for the same CAAM age strata as Spanish-speaking Hispanic individuals. These strata were nested within geographic region, except for Non-Hispanic Asian, who could be recruited regardless of geographic region given the census proportions of primarily English or Spanish-speaking households in that target age range. Maternal/caregiver education targets were also based on these broader age strata nested within geographic region.

2.4. Collection of the Baby Toolbox data

To establish psychometric properties of the Baby Toolbox, we conducted a national study of infants and toddlers with sampling targets based on our simulations. The final obtained sample size was 2,515 infants and toddlers, with n = 2,042 administered the English version and n = 473 administered the Spanish version (Table 2). In total, age was adjusted for 173 children (6.9% of total sample), with an average adjusted age difference of 1.12 months (SD = 0.85).

2.4.1. Participant inclusion/exclusion criteria.

Caregivers or legal guardians were required to provide consent and be able to read and respond to written materials in either English or Spanish, both for the purpose of consent and for caregiver-completed measures. Children with known genetic diagnoses were excluded.

2.4.2. Language classification.

Children were classified into English or Spanish testing based on the caregiver-reported “primary” language to which they were exposed. Children were eligible to participate if they were exposed to multiple languages, as long as English or Spanish was the primary language to which they were exposed.

2.4.3. Recruitment strategies.

Participants were tested in partnership with a marketing research firm that had existing locations around the U.S. in all four census regions, and across 16 unique sites, which included: Appleton (WI), Atlanta (GA), Baltimore (MD), Boston (MA), Chicago (IL), Dallas (TX), Houston (TX), Los Angeles (CA), Minneapolis (MN), Nashville (TN), Iselin (NJ), Orlando (FL), Philadelphia (PA), Phoenix (AZ), Seattle (WA), and St. Louis (MO). Data collection occurred between June 2023 and March 2024. Harder-to-reach participants (e.g., those with lower education or who were non-English speakers) were recruited through a variety of channels, including digital outreach via social media and in-person methods such as through local religious organizations and other non-profits. Potential participants were pre-screened for eligibility via a phone call. If eligible, a testing appointment was scheduled, and caregivers or legal guardians were provided with detailed information about expectations for the testing appointment. Caregivers of participants were compensated for their participation.

2.4.4. Trainer certification and fidelity.

Market research firm staff were trained to administer the Baby Toolbox using a Train-the-Trainer model. Thirty staff members (approximately two per site) attended a week-long training workshop to learn how to administer the Baby Toolbox. Following the workshop and sufficient practice with adults and children in the target age range, attendees submitted two recorded administrations, one with an infant under 24 months old and one with a toddler aged 24–48 months old. These research staff received comprehensive and detailed feedback on all aspects of administration from the Baby Toolbox training team and were asked to submit additional videos if they had many errors (more than 75% of administration steps). Staff also completed observational scoring of eight videos and were required to complete additional scoring if their accuracy was lower than 75% on scoring. If they met both administration and scoring certification requirements, they were considered certified trainers.

Certified trainers then used the same process to train their own staff (5–15 per certified trainer, 90 in total), who then completed the same requirements (i.e., two video-recorded administration sessions and observational coding). Certified trainers completed the first level of certification by using checklists and writing summaries of trainee administration videos, which they then used to provide feedback to the examiners. Baby Toolbox training staff reviewed the feedback and videos for all 90 trainee examiners, provided additional comments, and determined whether a given examiner could be considered a certified examiner. In addition to these requirements, examiners were also required to pass scoring certification. Examiners who did not meet requirements were removed from the project. Site visits at each of the sites were conducted by core Baby Toolbox members within the first few weeks of accrual, with ongoing visits as needed by the market research firm leadership for initial and ongoing fidelity checks.

2.4.5. Testing Procedure.

All children were tested at one of the testing sites with their caregivers present. Following a warm-up period that lasted at least 5–10 minutes, caregivers were given an iPad to complete caregiver-report questionnaires while the child completed the performance-based tests of the Baby Toolbox battery, tailored to their age, with another iPad. See Table 3 for battery order based on age band. Note that examiners were able to delay a measure in the battery, and, depending on the measure-specific rules and the location within the test for the delay, resume it later within the battery. However, infants can be fussy or refuse to participate at any time during administration of the battery. Attempts were made to gather as much data as possible from an infant, including bringing infants back on a following day, if necessary, but missing data occurred throughout the battery when a child did not want to participate or comply with testing procedures. Caregivers were also asked to complete a sociodemographic questionnaire that included questions about social determinants of health, child and family health characteristics, and the child’s media use. Participants aged 21 months and older were administered a touch-based measure (e.g., Visual Delayed Response, Memory Task, Picture Vocabulary, Who Has More) after first being administered the Baby Toolbox Touch Screen Tutorial to become acquainted with responding to the iPad via touch. Children between 6 and 24 months were administered gaze-based versions of measures that used built-in eye tracking (Novack, Han, Pila, et al., 2025 for methodological details).

Table 3.

Battery order of the NIH Baby Toolbox

Order Age in Months
1–5 months 6–8 months 9–15 months 16–21 months 22–24 months 25–36 months 37+ months
1 Sit and Stand Mullen Visual Reception Social Observation Social Observation Social Observation Social Observation Social Observation
2 Get Up and Go Sit and Stand Familiarization Familiarization Mullen Receptive Touch Screen Tutorial Touch Screen Tutorial
3 Mullen Receptive Get Up and Go Mullen Receptive Mullen Receptive Mullen Expressive (Prompted) Delayed Memory (Learning) Delayed Memory (Learning)
4 Mullen Expressive (Prompted) Familiarization Looking While Listening Looking While Listening Looking While Listening Visual Delayed Response Visual Delayed Response
5 Mullen Visual Reception Mullen Receptive Mullen Expressive (Prompted) Mullen Expressive (Prompted) Sit and Stand Delayed Memory (Test) Delayed Memory (Test)
6 Reach to Eat Looking While Listening Sit and Stand Sit and Stand Get Up and Go Mullen Receptive Mullen Receptive
7 Mullen Expressive (Observational) Mullen Expressive (Prompted) Get Up and Go Get Up and Go Touch Screen Tutorial Mullen Expressive (Prompted) Mullen Expressive (Prompted)
8 Reach to Eat Mullen Visual Reception Mullen Visual Reception Delayed Memory (Learning) Picture Vocabulary Picture Vocabulary
9 Mullen Expressive (Observational) Reach to Eat Reach to Eat Visual Delayed Response Sit and Stand Sit and Stand
10 Mullen Expressive (Observational) Mullen Expressive (Observational) Delayed Memory (Test) Get Up and Go Get Up and Go
11 Mullen Visual Reception Counting (Verbal) Counting (Verbal)
12 Reach to Eat Counting (Object) Counting (Object)
13 Mullen Expressive (Observational) Subitizing Subitizing
14 Who Has More Who Has More
15 Mullen Visual Reception Verbal Arithmetic
16 Reach to Eat Mullen Visual Reception
17 Mullen Expressive (Observationl) Reach to Eat
18 Mullen Expressive (Observational)
Open in a new tab

2.4.6. Data management.

Measure specific data were directly entered into the Baby Toolbox iPad app, which were then transferred using a secure protocol to a centralized repository each evening. Data extractions were examined thrice weekly by the Baby Toolbox data team to ensure data transfer was successful, and to review and identify any administration problems. The Baby Toolbox app is set up to export data to a secure web server. Media files (image or video format) require usage of a secure web server, as file size limitations prevent using other methods. Reports were generated and reviewed after each data collection per participant to confirm that the intended data was received. Additional screening data collected by the marketing research firm were provided separately as a flat file and were stored on a secure server.

2.5. Validation measures

To validate the Baby Toolbox measures against existing commercial tests, a subset of infants (n = 120 recruited) returned to complete the Bayley Scales of Infant and Toddler Development 4th edition (Bayley-4; Bayley & Aylward, 2019), the current version of a widely-used norm-referenced test in both clinical and developmental research settings. The Bayley-4 assesses Cognition, Expressive Language, Receptive Language, Gross Motor, and Fine Motor scales. Caregivers additionally completed a Social-Emotional scale and Adaptive Behavior scales assessing Receptive, Expressive, Personal, Interpersonal Relationships, and Play and Leisure subscales. The Bayley-4 was given to children by trained, certified administrators. Nearly all caregivers of the full sample also completed the Ages and Stages Questionnaire 3rd edition (ASQ-3; Squires & Bricker, 2009), which is a caregiver-report form assessing developmental milestones for Communication, Gross Motor, Fine Motor, Personal Social, and Problem-Solving domains. The ASQ-3 has been validated in medical literature and has a sensitivity of about 75% to detect delayed development (Squires et al., 1997).

All Bayley-4 and ASQ-3 data were entered by administrators in a pen and paper format, and scanned copies of the reports were uploaded to a secure web server. A double-entry data entry process was undertaken to convert responses from the Bayley-4 and ASQ-3 scanned paper forms to a digital format. A third independent reviewer compared and completed reconciliation for mismatching entries. The Bayley-4 and ASQ-3 data were then scored according to their respective technical and user manuals (Bayley & Aylward, 2019; Squires et al., 2009).

3. Computing normative values

3.1. Scoring model generation

The primary objective of the norming study was to generate scoring models for the Baby Toolbox at the measure level, domain level, and composite level, and to explore the psychometric properties (validity, reliability) of those scores. The Baby Toolbox approach to scoring was based on three ideals: that a measurement system should be (1) conceptually comprehensive and methodologically sound, (2) produce clinically actionable measure scores, and (3) be sensitive to individualized unique perception while allowing for meaningful comparisons across developmental stages. We started from the aspiration that individuals at different ages can be measured on the same conceptual latent variable, and their status can be monitored at different time points without compromising measurement properties. This is not always achievable, but it can often be accomplished with careful conceptual work and the aid of item response theory (IRT) models.

Various algorithms can yield scores amenable to norming, including summed, total, or average (ad hoc) scores. Here, where possible, we used IRT (Embretson & Reise, 2013), which uses item-level responses to yield continuous scores representing an individual’s relative standing on the domain of interest. IRT has several advantages over ad hoc scores in terms of accounting for the informativeness and difficulty of each item and thereby providing more accurate scores. IRT also yields individual-level standard errors of measurement (Wang et al., 1997), allowing quantification of uncertainty in score estimation, construct measures of statistically significant change over time (Hays et al., 2021), and account for measurement error at the individual level (Mansolf, 2023). For these reasons, as many measures as possible were scored using IRT, although some measures utilize non-IRT scoring where the statistical assumptions for IRT were not obtained (e.g., when local item dependence assumptions were violated) which was evaluated on a case-by-case basis. Factor analysis methods were used to fit composites to the domains and overall, see Han, Dworak, et al. (2025). Due to time constraints, for score model development, we used two-thirds of the dataset to explore the item responses and select the appropriate model. Once all data were collected, we then applied these model specifications to the full dataset to finalize parameter estimation and maximize the precision of item and person estimates for any subsequent analyses.

A variety of models were used across Baby Toolbox measures, which were tailored to the specific measure, item structure, and responses (for the measure-level scoring model information, see Novack, Dworak, Han, et al., 2025; Flynn et al., 2025; Pila et al., 2025; Adolf et al., 2025; Blackwell et al. 2025). An iterative modeling approach was used in some cases to determine the most appropriate unidimensional or multidimensional model via standard fit indices (e.g., the M2 statistic, and M2-based root mean square error of approximation [RMSEA] and comparative fit indices [CFI]; Maydeu-Olivares, 2014). In all cases, scoring model generation was both data and theory driven. The score estimation procedures also varied depending on the final scoring model and item type. For unidimensional models with dichotomous items, a maximum likelihood (ML) estimation method was used. For unidimensional models with polytomous items or measures that utilized a Computerized Adaptive Test, an expected a posteriori (EAP) estimation method was used. For multidimensional models or unidimensional models with mixed item types (dichotomous and polytomous), a maximum a posteriori (MAP) estimation procedure was used.

3.2. Obtaining sample weights

For data collection and the original sampling plan, we utilized the 2019 American ACS 5-year estimates with a 10% margin, but for data analysis we updated and weighed the sample (DeBell & Krosnick, 2009). The demographic targets were updated to the 5-year estimates for the ACS using the 2022 values (U.S. Census Bureau, 2022), with the proportional fitting reweighting the sample any time a single marginal proportion was outside of a 5% limit on the target. In this way, each participant was weighted to better approximate the U.S. Census demographics. Weighting factors included race and ethnicity, sex assigned at birth, and caregiver education categories nested within Census regions. By definition, the mean sample weight was 1.0. Weights were truncated to a maximum value of 5.0. Figure 2 provides a density plot for the sample weights by broad age group.

Figure 2.

Figure 2.

Open in a new tab

Distribution of sampling weights from the Baby Toolbox norming study, reproduced with permission from NIH Baby Toolbox Technical Manual (Han, Blackwell, et al., 2025).

3.3. Obtaining reference values

The relevant norming statistics for deviation-based scores are an index of central tendency (mean or median ability at an age), and an estimate of dispersion (standard deviation of ability). Then the standardized residual between the obtained score and the expected score for one’s reference population can be rescaled to an appropriate age-adjusted standard score with desirable characteristics such as a population mean of 100 and standard deviation of 15.

Theta ability estimates and corresponding standard errors, or raw scores for other measures where IRT was not available, were used to generate Change Sensitive Scores (CSS). These CSSs were a linear conversion of the IRT theta score to positive integers. The raw composite scores, which were the weighted sum of specific measure-level scores as derived from factor analyses (c.f., Han, Dworak, et al., 2025, this issue), for overall Cognition and its subcomponents, Executive Function/Memory, Math, and Language, as well as composites for Motor, Self-Regulation, Negative Affect, and Social Communication, were also converted to CSSs prior to norming procedures (Figure 3). Along with composites, a select number of measure scores were also normed.

Figure 3.

Figure 3.

Open in a new tab

Procedure for norming measure and composite scores for the NIH Baby Toolbox, reproduced with permission from the NIH Baby Toolbox Technical Manual (Han, Blackwell, et al., 2025) (A) Norming procedure for individual measure level scores. (B) Norming procedure for composite scores.

Bootstrap resamples of the CSS were drawn with replacement for the selected cases with 1000 bootstraps. Within each bootstrap resample, plausible value imputation was used to sample an expected score for each examinee based on their obtained CSS and its associated standard error or multiple imputation was utilized for infants within the age range of the test but for whom the measure was not successfully completed. Within each imputation, expected CSS were regressed on CAAM using Generalized Additive Models for Location, Scale, and Shape (GAMLSS; Stasinopoulos et al., 2018; Timmerman et al., 2021) to develop age-adjusted reference values for each measure. Figure 4 shows the age-adjusted reference values across chronological age in months for each of the eight composites. For the Cognition-related, Motor, and Social Communication composites, all the reference value CSSs scaled monotonically as age increased. In contrast, there were less but expected changes in reference values with age for the Self-Regulation and Negative Affect composites, which is consistent with prior work indicating that these constructs are trait-like and have more within person stability and is further discussed in the Social-Emotional Functioning measure development manuscript in this current issue (Blackwell et al., 2025).

Figure 4.

Figure 4.

Open in a new tab

Reference values obtained for the composite scores. The solid black line indicates the reference value for that composite and age in months, and the dotted lines above and below the are the reference values for ±1 standard deviation, respectively.

Results were combined using Rubin’s Rules for the imputations (Rubin, 1976), and the empirical sampling distribution of reference values and standard deviations were aggregated across bootstraps after the combined imputations. Given the rapid growth of skills during initial months of life, followed by still rapid but slower development, GAMLSS are more appropriate than a linear model. For norming, only location and scale modeling were included, both of which were regressed on cubic splines representing participant-level chronological age in months. Following procedures used in the NIH Toolbox V3 (Ho et al., under review), a normal distribution was selected, with the identity and log links for the location and scale parameters, respectively. After converting the GAMLSS residual to an age-based standard score (Mean = 100, SD =15), the scores greater than 146 or less than 54 were winsorized to these values, respectively, which were one point more extreme than ±3 SD from the mean. See domain specific papers within this issue (Novack, Dworak, Han, et al., 2025; Flynn et al., 2025; Pila et al., 2025; Adolf et al., 2025; Blackwell et al., 2025), for measure level score validity and reliability results and Han, Dworak, et al. (2025) for composite level score validity and reliability results.

4. Discussion

This study presents the norming methodology for the Baby Toolbox, a comprehensive battery designed to assess early cognitive, motor, and socioemotional development. Establishing robust normative data is essential for interpreting individual performance relative to a representative sample, thereby enhancing the clinical and research utility of the Baby Toolbox. Our approach incorporated rigorous sampling, diverse participant recruitment, and state-of-the-art statistical techniques to ensure the generalizability and accuracy of normative scores.

There are several strengths of the Baby Toolbox norming study. One of the primary strengths is the use of a large and demographically representative sample. Our simulation approach, sample targets, and recruitment strategy ensured adequate representation across key sociodemographic variables, including race, ethnicity, socioeconomic status, and geographic region, in alignment with national census benchmarks. This diversity enhances applicability of the Baby Toolbox norms across various populations. A second key strength is the use of advanced statistical techniques in the development of scoring models and norms, such as placing as many measures as possible on an IRT metric.

Several limitations regarding this norming study must be acknowledged. First, the data were collected from infants and toddlers, many of whom were born during the global COVID-19 pandemic. It is unclear how this experience may have impacted development, and thus normative scores. Recent, smaller, cross-sectional studies have indicated that greater social distancing was associated with worse general development scores (Ferrari et al., 2022) and worse gross motor skills in infants (de Abreu et al., 2024). However, the impact of COVID-19 on child development is multifaceted (Ashikkali et al., 2024; Shah et al., 2020), and the long-term effects have yet to be well understood. Second, the battery was administered based on age in month groupings. Though this was important for consistency of data collection across sites, this was not necessarily ideal for validation. For example, infants were more likely to “fuss out” of measures placed towards the end of the battery, thus leading to attrition and less data available for scoring on those measures. Finally, although a small subsample participated in a test-retest reliability study (see Novack, Dworak, Han, et al., 2025; Flynn et al., 2025; Pila et al., 2025; Adolf et al., 2025; and Blackwell et al. 2025), the short period of time and small number of individuals limits our ability to capture within-individual change; the majority of the norming study was cross-sectional in nature. Longitudinal studies will be necessary to further validate these norms and assess individual growth trajectories over time.

Due to difficulties recruiting mothers with less than a high school education, we opted to re-categorize some education data, using this derived “caregiver education” to derive sampling weights. This avoids over-weighting the few mothers we did collect in this age range, which improves the statistical stability of the weights and all statistical quantities derived using them. However, the drawback of this approach is that we no longer have a perfect match between our “caregiver” education variable and the census “maternal” education variable, such that a small number of children we treat as having a mother with less than a high school education when calculating case weights instead have a mother with a higher education level. The effects of this recategorization, which are most prominent for measures with expected performance differences based on mother’s/caregiver’s education, are likely to be minimal given this only affected 1.5% of the sample, but should be acknowledged. Composite model parameter estimation may be slightly affected, but mainly via a (small) shift in mean factor loading and/or intercept parameters; while these affect CSS calculations, the CSSs are not referenced to any population, and therefore these differences have no practical impact. Normative scores which utilized these sampling weights are likely to be slightly higher, as mothers with at least a high school education were effectively oversampled; again, due to the small proportion of affected cases, we expect this inflation to be minimal.

The Baby Toolbox is adaptable to many contexts, including clinical, educational, and research settings. The Baby Toolbox will also facilitate having common measures and domains across studies, allowing for increased data sharing, bolstering the capability for secondary data analysis, and improving cross study comparisons. As an example, the NIH-wide Healthy Brain and Child Development study (HBCD; Nelson et al., 2024), is the first large scale research, study to use the Baby Toolbox even prior to public release to standardize measures and reduce time spent in assessments across all sites.

An important consideration for using the Baby Toolbox on a larger scale is weighing the benefits and harm of detection of developmental delays in infants. While the utility of this tool as a screening method still requires research, given the growing ethnic and linguistic diversity of the United States as well as major technological advancements occurring in the last twenty years, there is a need for appropriate, comprehensive norms that represent this diversity and modern advancements. This is especially important considering the growing research examining the effects of social determinants of health and their impact on development (Duncan et al., 2012; Ursache & Noble, 2016; Decker et al., 2020).

In conclusion, this study establishes a robust foundation for the application of the Baby Toolbox in both research and clinical practice. By leveraging a diverse, representative sample and sophisticated modeling approaches, we have developed normative metrics that enhance the interpretability and utility of the toolbox. Continued research efforts aimed at refining and expanding these norms will be essential for maximizing the impact of this assessment battery in early developmental screening and research.

Supplementary Material

1

Highlights.

  • The NIH Baby Toolbox is a suite of infant developmental assessments (1–42 months).

  • Cognition, Motor, and Social-Emotional Functioning measures are offered through a tablet-assisted administration and scoring platform.

  • A norming study assessed the reliability and validity of the measures.

  • Norm referenced scores were created for composites and select measures.

  • Evidence for reliability and validity of the tests scores is necessary and foundational for research and clinical use.

Acknowledgements:

We thank the study participants and their family members for their participation, along with research staff across all sites who administered this study. The authors would like to gratefully acknowledge the contributions of numerous individuals who participated in data collection efforts for the NIH Baby Toolbox norming study, including the site staff and trained examiners across the 16 data collection sites, the technical team at Northwestern and Helium Foot supporting implementation of measures in the iPad platform, and many scientists who advised on the norming and analytic strategy. In particular, we would like to thank the contributions of Maria Varela Diaz, Rachel Hanrahan, Paula Hom Sievert, Cecily Frederick, Christian Clare, Steven Glinberg, and Pranav Rajiv. We would also like to thank Callie M. Jones and Gabriella M. Juarez for their assistance with data entry.

Funding information:

This project has been funded in whole or in part with Federal funds from the NICHD, National Institutes of Health, Department of Health and Human Services, under Contract No. 75N94019D00005. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Footnotes

Author statement

The authors have no conflicts of interest to disclose. YCH: data curation, methodology, formal analysis, software, validation, writing (original draft). EMD: data curation, methodology, formal analysis, software, writing (original draft). MM: formal analysis, methodology, software, writing (review and editing). HA: data curation, project administration, validation, writing (review and editing). LY: formal analysis, methodology, writing (review and editing). MAN: conceptualization, investigation, resources, writing (review and editing). SP: conceptualization, investigation, resources, writing (review and editing). RMF: conceptualization, investigation, resources, writing (review and editing). AMF: conceptualization, investigation, writing (review and editing). VU: project administration, supervision, validation, writing (review and editing). KS: data curation, investigation, project administration, writing (review and editing). GJB: data curation, project administration, writing (review and editing). RCG: conceptualization, funding acquisition, investigation, methodology, project administration, supervision, validation, writing (review and editing). AJK: conceptualization, data curation, formal analysis, investigation, methodology, software, supervision, validation, writing (review and editing). All authors read and approved the manuscript.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

5. References

  1. Adolf K, Cole WG, Novack MA, Han YC, Kaat AJ, Rachwani J, Clark J, Dusing S, Heathcock J, Prosser LA, Lockman JJ, Whitall J, Gershon RC (2025). Development and validation of the NIH Baby Toolbox® motor measures. Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. Washington, DC: AERA. [Google Scholar]
  3. Ashikkali L, Seggie AJR, & Johnson C (2024). The long-term indirect impact of Covid-19 on child health. Paediatrics and Child Health, 34(5), 160–165. DOI: 10.1016/j.paed.2024.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bailey MJ, Currie J, & Schwandt H (2022). The Covid-19 baby bump: the unexpected increase in US fertility rates in response to the pandemic (No. w30569, p. 20220483642). Cambridge, MA: National Bureau of economic research. [Google Scholar]
  5. Bayley N, & Aylward GP (2019). Bayley-4: Scales of infant and toddler development, Technical manual (4th ed.). Pearson. [Google Scholar]
  6. Becker L, Condy E, Kaat A, & Thurm A (2023). How do 3-year-olds do on the NIH Toolbox Cognitive Battery? Child Neuropsychology, 29(4), 521–542. DOI: 10.1080/09297049.2022.2100337 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Berls AT, & McEwen IR (1999). Battelle developmental inventory. Physical Therapy, 79(8), 776–783. DOI: 10.1093/ptj/79.8.776 [DOI] [PubMed] [Google Scholar]
  8. Blackwell CK, Nowinski C, Wetherby A, Sommerville J, Bethell C, Wakschlag L, Hooker JL, Mansolf M, Han YC, Dworak EM, Ustsinovich V, Kaat AJ, Gershon RC (2025). Development and validation of the NIH Baby Toolbox® social-emotional functioning measures. Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Blauw-Hospers CH, & Hadders-Algra M (2005). A systematic review of the effects of early intervention on motor development. Developmental Medicine and Child Neurology, 47(6), 421–432. DOI: 10.1017/S0012162205000824 [DOI] [PubMed] [Google Scholar]
  10. de Abreu RWF, Lima CRG, Verdério BN, Dos Santos MM, Dos Santos AN, Brugnaro BH, & Rocha NACF (2024). Association between environmental factors during the COVID-19 pandemic and functioning of infants with biological risk in the first year of life: Cross-sectional exploratory study. Early Human Development, 191, 105987. DOI: 10.1016/j.earlhumdev.2024.105987 [DOI] [PubMed] [Google Scholar]
  11. DeBell M, & Krosnick JA (2009). Computing weights for American national election study survey data. ANES Technical Report Series. https://electionstudies.org/wp-content/uploads/2018/04/nes012427.pdf [Google Scholar]
  12. Decker AL, Duncan K, Finn AS, & Mabbott DJ (2020). Children’s family income is associated with cognitive function and volume of anterior not posterior hippocampus. Nature Communications, 11(1), 4040. DOI: 10.1038/s41467-020-17854-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Duncan GJ, & Magnuson K (2012). Socioeconomic status and cognitive functioning: moving from correlation to causation. Cognitive Science, 3(3), 377–386. DOI: 10.1002/wcs.1176 [DOI] [PubMed] [Google Scholar]
  14. Embretson SE, & Reise SP (2013). Item response theory. Psychology Press. [Google Scholar]
  15. Ferrari E, Palandri L, Lucaccioni L, Talucci G, Passini E, Trevisani V, & Righi E (2022). The Kids Are Alright (?). Infants’ Development and COVID-19 Pandemic: A Cross-Sectional Study. International Journal of Public Health, 67, 1604804. DOI: 10.3389/ijph.2022.1604804 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Flynn RM, Han YC, Kaat AJ, Mansolf M, Dworak EM, Pila S, Yao L, Carlson SM, Oakes LM, Gershon RC, Zelazo PD (2025). Development and validation of the NIH Baby Toolbox® executive function and memory measures. Infant Behavior and Development. [DOI] [PubMed] [Google Scholar]
  17. Gershon R, Novack MA, & Kaat AJ (2024). The Infant and Toddler Toolbox: A new standardized tool for assessing neurodevelopment in children ages 1–42 months. Child Development, 95(6), 2252–2254. DOI: 10.1111/cdev.14135 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Goldsmith HH, and Rothbart MK (1996). Prelocomotor and Locomotor Laboratory Temperament Assessment Battery, Lab-TAB; version 3.0. Technical Manual, Department of Psychology, University of Wisconsin, Madison, WI. [Google Scholar]
  19. Han YC, Blackwell CK, Dworak EM, Flynn RM, Mansolf MA, Novack MA, Pila S, Kaat AJ (2025). National Institutes of Health (NIH) Baby Toolbox® Technical Manual. [Google Scholar]
  20. Han YC*, Dworak EM*, Mansolf M, Gershon RC, Kaat AJ (2025). Composite scores for the NIH Baby Toolbox®. (* = contributed equally). Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hays RD, Spritzer KL, & Reise SP (2021). Using Item Response Theory to Identify Responders to Treatment: Examples with the Patient-Reported Outcomes Measurement Information System (PROMIS®) Physical Function Scale and Emotional Distress Composite. Psychometrika, 86(3), 781–792. DOI: 10.1007/s11336-021-09774-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Ho EH, Han YC, LaForte EM, Kaat AJ, Dworak EM, Ece B, Yao L, Olsen JB, Gershon RC (Submitted). Norming of the NIH Toolbox Cognition Battery Version 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Knapp EA, Kress AM, Parker CB, Page GP, McArthur K, Gachigi KK, ... & Influences on Child Health Outcomes, O. B. O. P. C. F. E. (2023). The environmental influences on child health outcomes (ECHO)-wide cohort. American Journal of Epidemiology, 192(8), 1249–1263. DOI: 10.1093/aje/kwad071 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. LaForte EM, Hook JN, & Giella AK (2024). National Institutes of Health (NIH) Toolbox® Version 3 Technical Manual. [Google Scholar]
  25. Livingston G, & Cohn DV (2013, May 10). Record share of new mothers are college educated. Pew Research Center. https://www.pewresearch.org/social-trends/2013/05/10/record-share-of-new-mothers-are-college-educated/ [Google Scholar]
  26. Mansolf M (2023). A true score imputation method to account for psychometric measurement error. Psychological Methods. DOI: 10.1037/met0000578 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Maydeu-Olivares A (2014). Evaluating the fit of IRT models. In Handbook of item response theory modeling (pp. 111–127). Routledge. [Google Scholar]
  28. Mullen EM (1995). Mullen scales of early learning (pp. 58–64). Circle Pines, MN: AGS. [Google Scholar]
  29. Nelson CA, Frankeberger J, & Chambers CD (2024). An introduction to the HEALthy Brain and Child Development Study (HBCD) study. Developmental Cognitive Neuroscience, 69, 101441. DOI: 10.1016/j.dcn.2024.101441 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Novack MA, Han YC, Pila S, Kaat AJ, Flynn RM, Bedjeti K, Varela Diaz M, Hanrahan RT, Glinberg S, Sievert PH, Frederick C, Rajiv P, Clare C, Ustsinovich V, Gershon RC (2025). Automated iPad-based gaze detection using the NIH Baby Toolbox®. Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Novack MA, Dworak EM, Han YC, Kaat AJ, Ustsinovich V, Saffran J, Frank MC, Waxman S, Gershon RC (2025). Development and validation of the NIH Baby Toolbox® language measures. Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Oakes LM (2017). Sample size, statistical power, and false conclusions in infant looking-time research. Infancy, 22(4), 436–469. DOI: 10.1111/infa.12186 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Oosterhuis HEM, Van Der Ark LA, & Sijtsma K (2016). Sample size requirements for traditional and regression-based norms. Assessment, 23(2), 191–202. DOI: 10.1177/1073191115580638 [DOI] [PubMed] [Google Scholar]
  34. Paolillo EW, McKenna BS, Nowinski CJ, Thomas ML, Malcarne VL, & Heaton RK (2020). NIH Toolbox® emotion batteries for children: Factor-based composites and norms. Assessment, 27(3), 607–620. DOI: 10.1177/1073191118766396 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Pila S, Han YC, Adam H, Kaat AJ, Sarama J, Clements DH, Gershon RC (2025). Development and validation of the NIH Baby Toolbox® math measures. Infant Behavior and Development. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Rice D, & Barone S (2000). Critical periods of vulnerability for the developing nervous system: Evidence from humans and animal models. Environmental Health Perspectives, 108(suppl 3), 511–533. DOI: 10.1289/ehp.00108s3511 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Rubin DB (1976). Inference and missing data. Biometrika, 63(3), 581–592. DOI: 10.1093/biomet/63.3.581 [DOI] [Google Scholar]
  38. Shah K, Mann S, Singh R, Bangar R, & Kulkarni R (2020). Impact of COVID-19 on the mental health of children and adolescents. Cureus, 12(8). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7520396/ [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Squires J, Bricker D, & Potter L (1997). Revision of a parent-completed developmental screening tool: Ages and Stages Questionnaires. Journal of Pediatric Psychology, 22(3), 313–328. DOI: 10.1093/jpepsy/22.3.313 [DOI] [PubMed] [Google Scholar]
  40. Squires J, & Bricker D (2009). Ages & stages questionnaires. Baltimore: Paul H. Brookes. [Google Scholar]
  41. Squires J, Twombly E, Bricker D, Potter L (2009). ASQ®−3 user’s guide. Baltimore: Paul H. Brookes. [Google Scholar]
  42. Stasinopoulos MD, Rigby RA, & Bastiani FD (2018). GAMLSS: A distributional regression approach. Statistical Modelling, 18(3–4), 248–273. DOI: 10.1177/1471082X18759144 [DOI] [Google Scholar]
  43. Thompson WK, Barch DM, Bjork JM, Gonzalez R, Nagel BJ, Nixon SJ, & Luciana M (2019). The structure of cognition in 9 and 10 year-old children and associations with problem behaviors: Findings from the ABCD study’s baseline neurocognitive battery. Developmental Cognitive Neuroscience, 36, 100606. DOI: 10.1016/j.dcn.2018.12.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Timmerman ME, Voncken L, & Albers CJ (2021). A tutorial on regression-based norming of psychological tests with GAMLSS. Psychological methods, 26(3), 357. DOI: 10.1037/met0000348 [DOI] [PubMed] [Google Scholar]
  45. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, Moher D, Peters MDJ, Horsley T, Weeks L, Hempel S, Akl EA, Chang C, McGowan J, Stewart L, Hartling L, Aldcroft A, Wilson MG, Garritty C, … Straus SE (2018). PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Annals of Internal Medicine, 169(7), 467–473. DOI: 10.7326/M18-0850 [DOI] [PubMed] [Google Scholar]
  46. Ursache A, & Noble KG (2016). Neurocognitive development in socioeconomic context: Multiple mechanisms and implications for measuring socioeconomic status. Psychophysiology, 53(1), 71–82. DOI: 10.1111/psyp.12547 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. U.S. Census Bureau. (2019). American Community Survey 5-Year Estimates. [Google Scholar]
  48. U.S. Census Bureau. (2022). American Community Survey 5-Year Estimates. [Google Scholar]
  49. Wang T, Kolen MJ, & Harris DJ (1997). Conditional standard errors, reliability and decision consistency of performance levels using polytomous IRT. Reliability Issues with Performance Assessments: A collection of papers, 13. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS2101250-supplement-1.pdf (130.7KB, pdf)

RESOURCES