About

Heidi Feldman is a professor associated with Pediatrics and Developmental Behavioral Pediatrics at Stanford University. Her role involves research and teaching within the Human Biology program, focusing on developmental and behavioral aspects. She is part of a faculty that contributes to the interdisciplinary exploration of human biology, integrating insights from pediatrics and developmental sciences to advance understanding in these fields.

Research topics

• Medicine
• Psychology
• Pediatrics
• Radiology
• Political Science
• Neuroscience
• Endocrinology
• Internal medicine
• Medical education
• Biology
• Anatomy
• Philosophy
• Geography
• Psychiatry

Selected publications

• Neonatal brain-age models in full- and preterm infants

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-01-23

  articleOpen access

  Prematurity affects brain development and increases risk for neurodevelopmental impairments. Yet reliable biomarkers for at-risk infants remain limited. The goals of this study are (1a) to construct a white matter neonatal brain-age model including full-term and preterm neonates from a large publicly-available data set, (1b) to evaluate the accuracy of this model for characterizing the preterm brain from the same data set; (2) to determine if a similar model can predict brain-age based on clinical MRI scans from high-risk neonates born preterm; and (3) to evaluate whether this predictive model provides information about the infant's health beyond conventional clinical and demographic measures. We developed brain-age prediction models using diffusion magnetic resonance imaging-derived white matter features from two datasets: (1) the developing Human Connectome Project (dHCP; 368 healthy infants) and (2) a clinical sample collected at the Lucile Packard Children's Hospital (LPCH; 162 high-risk preterm infants). White matter features demonstrated strong predictive performance in the dHCP dataset (within 3.9 days) and the LPCH clinical dataset (within 6.6 days). However, brain-age metrics (i.e., brain-age gap) showed no significant associations with health complications measured by a composite score of common prematurity complications. While tractometry-derived brain-age models accurately characterize brain maturation in the neonatal brain, their sensitivity to clinical complications in preterm infants appears limited. Global white matter maturation measures derived from clinical grade data may be insufficiently sensitive to capture the cumulative burden of prematurity-related morbidities, suggesting need for multimodal or longitudinal biomarkers.

  PublisherOA PDFDOI

• Autism Screening and Diagnosis in Children with Congenital Heart Disease

  2026-03-24

  articleOpen accessSenior author

  AimThis study determined the prevalence of positive autism screening at 18-30 months of age and autism diagnosis after age 2 years among children with congenital heart disease (CHD). We evaluated sociodemographic, clinical, and medical factors, and subgroups of heart disease associated with autism likelihood and/or diagnosis. MethodsSecondary analysis of data collected from the Stanford site of the California Perinatal Quality Care Collaborative and Lucile Packard Children's Hospital electronic health records. Participants (N = 94) were children born between 2016–2020 with CHD who required surgery before discharge from the neonatal intensive care unit and who had a High Risk Infant Follow-Up (HRIF) visit at 18-30 months of age. Heart disease was classified as cyanotic or acyanotic based on the need for oxygen supplementation. Scores on the Modified Checklist for Autism - Revised/Follow-up (M-CHAT R/F) at HRIF visit and evidence of an autism diagnosis were extracted. We compared sociodemographics, clinical, and medical factors across two heart disease subgroups, positive or negative screens, and autism diagnosis. ResultsThe prevalence of a positive screen was 14.6% and of autism diagnosis 11.7%. When considered independently, length of hospital stay was the only clinical factor that was significantly associated with positive screen results; children with longer stays in hospital had higher likelihood of positive screens. However, when sociodemographic and clinical factors were considered together, no factors were associated with screen results. Performance of the M-CHAT-R/F in relation to an eventual autism diagnosis showed sensitivity of 66% and positive predictive value of 57%.ConclusionIn this sample, children with CHD were more than 3 times more likely to have positive autism screens and to receive the diagnosis of autism than is reported for children in the general population. Early screening for autism is critical in children with CHD to promote early diagnosis and intervention. Future studies should consider other factors in relation to the development of autism in children with CHD to identify strategies for reducing autism prevalence.

  PublisherDOI

• Opportunities and Alternative Approaches for Improving Language Outcomes in Young Children

  PEDIATRICS · 2026-03-16

  article1st authorCorresponding

  In this issue of Pediatrics, Wong et al1 used a speech auditory brain stem response (ABR) method, administered between 7 and 24 months of age, to predict those infants and toddlers who would later score more than 1 SD below the mean on the language subscale of the Bayley Scales of Infant Development. The ABR is a functional assessment of early neural responses, generated from electroencephalogram recordings acquired during the repeated presentations of auditory stimuli. The speech ABR substitutes speech stimuli for clicks or tones used in conventional ABR. In this study, machine learning techniques generated a set of neural features used for prediction. As the authors point out, the significance of the study is that if this early screening would allow accurate identification of children at high risk for language delay, it could lead to the provision of targeted early intervention, ideally mitigating the long-term consequences of language disorders in a cost-effective manner. This commentary has 2 aims: (1) to discuss limitations in the study’s methods that should be addressed in the further development of the speech ABR technique as a universal screening tool and (2) to offer an alternative approach to improving early language skills in infants and toddlers that can be applied in both high- and low-risk infant populations, thereby reducing the need for screening.Wong et al1 found that a set of features on the speech ABR test substantially improved prediction of later poor language outcomes compared with use of pediatric variables alone, including child sex, birth weight, and gestational age. In this manuscript, they did not label the features or provide a theoretical account of why the specific set might be useful in prediction. Nonetheless, the data are impressive.It is important to evaluate variables in this study to determine whether their selection impacted the effectiveness of the speech ABR as a potential screening tool. (1) Age of screening. The study conducted the speech ABR on children aged 1 to 24 months. Brain stem responses to auditory stimuli mature between birth and 24 months of age. The authors did not discuss whether the features had to be adjusted for child age or whether prediction accuracy changed across the age range tested. Determining an optimal age for screening is an important next step. (2) Age of outcomes assessment. The standardized language assessment was collected between 7 and 32 months of age. Prediction from infant to preschool language skills is generally limited, except in children with extreme delays.2 Even scores at 24 months of age are not excellent predictors of later language skills; approximately half of the children identified as delayed at age 2 have caught up to peers by age 3.3 Again, while Wong et al1 demonstrated impressive gains in predictive validity over traditional factors at 7 to 32 months of age, a rigorous evaluation requires outcomes testing at age 3 years or older. (3) Speech stimuli. Wong et al1 used 3 tones embedded in the same syllable (/ga/) as the speech stimuli. In Cantonese, varying the tone of a syllable can change word meaning. In English and many other languages, tone differentiates the speaker’s feelings (eg, angry vs pleased) or sentence meaning (eg, statement vs question) but not word meaning. The authors predict these stimuli will work regardless of the child’s language environment, but the stimuli must be tested in infants exposed to languages other than Cantonese. (4) Impact of socioeconomic status (SES). It is surprising that adding SES to the predictive models after neural factors were included did not improve model performance. Socioeconomic status is the most powerful predictor of language outcomes in many studies.4 Multiple methods are available for estimating SES, including income-to-needs ratios and levels of parent education. This study used an income variable. Replication efforts should ensure a wide range of SES in the assessed sample and should also include measures of parental education to assess the contribution of both income- and education-based measures of SES in relation to neural features in predicting outcomes. (5) Linkage to the universal newborn hearing screening (UNHS). An attractive feature about use of the speech ABR for early language screening is that the automated version of the ABR is currently used in the newborn period as part of UNHS. As the authors recognize, the addition of the speech ABR to current protocols could determine in a single session whether children can hear and their risk of language delay. Testing in this study began at age 7 months. Conducting this screening beyond the newborn period would require additional health care visits or additional procedures during scheduled infant health supervision visits, increasing expense and reducing the cost-effectiveness of this approach.A limitation of the use of the speech ABR for predicting language outcomes was the relatively low positive predictive value. This result means that many children who screened positive did not go on to display language delays. False-positive results risk worrying families or treating children unnecessarily. Additional research is necessary to understand what additional factors would increase the positive predictive value and other psychometric properties of the speech ABR.Importantly, the rationale behind early screening for language delay assumes that the provision for early intervention services would be extremely expensive, and therefore, only those children with the greatest need should receive such services. This rationale misses a key point about the impact of early intervention for language skills: All children benefit from early intervention services. Intervention methods have been shown to improve language outcomes in infants and toddlers similarly for children at high and low risk for language delays, including children with a medical complexity such as preterm birth.5,6The key ingredient for improving language outcomes is improving the child’s “language nutrition.”7 Language nutrition can be defined as the combination of environmental factors that is associated with favorable or healthy language outcomes. Specifically, good language nutrition includes a high quantity of adult speech directed to the child, a wide variety of sentence types, many sentences aligned with the child’s interests and attention, and provision of language input in warm, loving, and responsive caregiver-child interactions. Decades of research, including randomized clinical trials, have shown that improving language nutrition improves child language outcomes in the toddler-preschool period.8 The data are supportive of the same approach for children born preterm5,6 and children with intellectual disability and autism.5Thus, while early screening has value, a viable alternative to extensive screening is to provide high-quality coaching about language nutrition universally, to parents and caregivers of all newborns and infants. The method for disseminating the information need not be expensive and could include public service announcements, enhanced training of primary care clinicians and early care and education providers who connect routinely with parents, and direct individual or group parent education and training in community settings or online.9 It is critical that such programs operate in communities, such as low-resourced, low-SES neighborhoods, that may otherwise not receive these important messages. The benefit of universal design for this language intervention is that it holds the promise of raising language levels in all children regardless of early screening results.

  PublisherDOI

• Neonatal brain-age models in full- and preterm infants

  SSRN Electronic Journal · 2026-01-01

  preprintOpen access
  PublisherDOI

• Care Across Contexts: Patterns of Caregiver-Infant Engagement in Spanish- and English-Speaking Families of Preterm Infants

  medRxiv · 2026-03-20

  articleOpen accessSenior authorCorresponding

  Objective: To examine group differences and continuity in caregiving environments of infants born preterm from Spanish- and English-speaking families. Study Design: = 23) families of infants born preterm (< 32 weeks gestation). Caregiver-infant engagement was assessed neonatally via hospital visitation and skin-to-skin (STS) care, and at home via child-directed adult word counts/hour (CD-AWC/hour) from all-day audio recordings. Result: No significant group differences were observed in family visitation, neonatal STS care, or in-home verbal engagement, although STS care rates varied considerably, especially within Spanish-speaking families. Across both groups, greater STS care was associated with higher CD-AWC/hour at home. Conclusion: Spanish- and English-speaking families showed comparable patterns of caregiver-infant engagement, as a group, however, many Spanish-speaking families engaged in less STS than English-speaking families. STS care predicted caregiver-infant verbal engagement at home, highlighting continuity from hospital to home.

  PublisherOA PDFDOI

• The Power of Early Experience: Neonatal Skin‐To‐Skin Care Mitigates SES‐Related Disparities in Developmental Outcomes

  Infancy · 2025-05-01 · 2 citations

  articleOpen access

  Infants from lower-socioeconomic (SES) backgrounds are at increased risk for compromised developmental outcomes compared to infants from higher-SES backgrounds. Features of caregiver-child interactions have been proposed as mechanisms through which SES-related factors are associated with child outcomes. This study assessed whether rates of tactile interactions between neonates and family members (skin-to-skin caregiving) served as a mechanism, that is, statistically mediated, SES-related developmental disparities in infants born preterm (n = 95). Infants from lower-SES backgrounds experienced less skin-to-skin care and scored lower on developmental assessments than infants from higher-SES backgrounds. Infants who experienced more skin-to-skin care had better outcomes than infants who experienced less skin-to-skin care. Critically, the direct association between SES and outcomes was significantly reduced after controlling for skin-to-skin care rates. Thus, SES-related disparities were linked to caregiving experiences as early as the neonatal period. Parallel analyses on non-skin-to-skin tactile care (swaddled holding, touch, and massage) revealed no associations, highlighting the specificity of skin-to-skin caregiving. These findings make substantial contributions to developmental theory and offer concrete and scalable recommendations for intervention.

  PublisherOA PDFDOI

• Positive Autism Screening in Children Born Preterm

  Journal of Autism and Developmental Disorders · 2025-06-04 · 1 citations

  articleSenior author
  PublisherDOI

• Positive Autism Screening in Children Born Preterm

  medRxiv · 2025-04-06

  preprintOpen accessSenior authorCorresponding

  Abstract Purpose Autism is more common among children born preterm than term. This study determined the prevalence of positive autism screening among 18-30 month old preterm-born children and evaluated sociodemographic, clinical, and neurodevelopmental factors associated with positive screens. Methods Secondary analyses of Stanford data from California Perinatal Quality Care Collaborative. Infants born &lt; 32 weeks gestation between 2016-2020, who attended High-Risk Infant Follow-Up at 18-30 months, were classified into two groups based on results of Modified Checklist for Autism in Toddlers-Revised with Follow-Up (M-CHAT-R/F): positive-screen (score &gt; 2) and negative-screen (≤ 2). We compared sociodemographics, clinical factors, and language development across groups. Results Prevalence of positive-screens was 12.2% (41/337). Children in the positive-screen group had lower gestational age, birthweight and longer hospital stays than children in the negative-screen group (all p &lt; .05); gestational age was the primary factor associated with a positive screen (OR 0.75, 0.56-0.99). We found no group differences in sociodemographics or medical complications; children in the positive-screen group had lower language scores than children in the negative-screen group ( p &lt; .001). All children in the positive-screen group and ~⅓ of children in the negative-screen group scored low in language. Conclusion The high prevalence of positive-screens reinforces the importance of early screening for autism in very preterm children. Gestational age at birth was the only factor associated with positive-screens. Language difficulties were not specific to children with positive-screens, highlighting the need for autism screening and routine developmental assessments for preterm children.

  PublisherOA PDFDOI

• Listening to mom in the neonatal intensive care unit: a randomized trial of increased maternal speech exposure on white matter connectivity in infants born preterm

  Frontiers in Human Neuroscience · 2025-10-14

  articleOpen accessSenior author

  Objective Early speech experiences are presumed to contribute to the development of brain structures involved in processing speech. Previous research has been limited to correlational studies. Here, we conducted a randomized trial with neonates born preterm to determine whether increased exposure to maternal speech during NICU hospitalization is causally linked to structural white matter maturation. Study design We enrolled 46 neonates born preterm (24–31 weeks gestational age). Participants were randomly assigned to receive increased (T: n = 21) or routine (C: n = 25) exposure to mother’s speech. The T-group heard 10-min audio recordings of their mothers reading a children’s story two times/hour between 10pm and 6am, increasing speech exposure by 2.67 h/day. The C-group did not hear recorded speech. At near-term-equivalent age, we obtained two high-angular resolution diffusion MRI (scan 1: b = 700, scan 2: b = 1500) and T1 relaxometry scans. We assessed mean diffusivity (MD), pre-registered primary outcome (NCT02847689), of the left and right arcuate fasciculus, tracts implicated in language processing. Secondary outcomes included fractional anisotropy (FA) and R1 (1/T1). We hypothesized that neonates randomized to the T-group would show evidence for increased maturation within the arcuate, indexed as decreased MD and increased FA and R1, compared to neonates in the C-group. Results Groups were equivalent on medical and demographic variables. Linear mixed models demonstrated that compared to the C-group, the T-group demonstrated significantly lower MD in the left (scan 1: β = −0.11, Marginal R 2 = 0.27; scan 2: β = −0.12, Marginal R 2 = 0.33) but not right arcuate (scan 1: β = −0.06, Marginal R 2 = 0.09; scan 2: β = −0.03, Marginal R 2 = 0.01). The T-group also demonstrated significantly higher FA (scan 1 β = 0.02, Marginal R 2 = 0.20; scan 2: β = 0.03, Marginal R 2 = 0.31) and R1 (β = 0.02, Marginal R 2 = 0.39) in the left but not right arcuate. Conclusion Preterm neonates with increased maternal speech exposure showed more mature left arcuate microstructure, supporting a causal role of exposure to speech in brain development. Enhancing speech exposure in the NICU may benefit preterm children’s language outcomes.

  PublisherOA PDFDOI

• Family Navigation Engagement and Outcomes for Children With a New Autism Diagnosis

  Journal of Autism and Developmental Disorders · 2025-11-28

  articleSenior author
  PublisherDOI

Recent grants

• NIH Grant R01HD046500

  NIH · $1.8M · 2012

• NIH Grant R01HD069162

  NIH · $1.8M · 2017

• Predicting language processing efficiency in preterm children: Social-environmental and neuro-biological factors

  NIH · $5.8M · 2011–2026

Frequent coauthors

• Irene M. Loe

  Stanford University

  71 shared

• Janine E. Janosky

  69 shared

• Jack L. Paradise

  University of Pittsburgh

  65 shared

• Thomas F. Campbell

  61 shared

• Lynne C. Huffman

  Research Network (United States)

  50 shared

• Benjamin L. Handen

  University of Pittsburgh

  49 shared

• Katherine E. Travis

  Stanford University

  48 shared

• Marcia Kurs‐Lasky

  Children's Hospital of Pittsburgh

  45 shared

Education

• Ph.D., Human Biology

  Stanford University

  2000

• B.A., Human Biology

  University of California, Berkeley

  1994