Reading Aloud, Play, and Social-Emotional Development

We performed a factorial, single-blind RCT at an urban public hospital serving low-income families (Bellevue Hospital Center [BHC]), with enrollment and first random assignment during the postpartum period, second random assignment at age 3 years, and 5 assessments through 4.5 years. The trial was approved by the New York University School of Medicine Institutional Review Board and BHC Research Review Committee and was registered with clinicaltrials.gov (NCT00212576). Separate informed consent was obtained for enrollment and first random assignment and for second random assignment.

Consecutive mother and infant dyads meeting inclusion criteria and providing informed consent were enrolled and randomly assigned postpartum from November 2005 through October 2008. Inclusion criteria were as follows: planned pediatric care at BHC, full-term, no significant medical complications or early intervention eligibility at birth, and mother primary caregiver was 18 years or older, English- or Spanish-speaking, with a working telephone or pager. Consent for the second random assignment took place at ∼3 years by research assistants masked to group assignment.

There were 2 phases of random assignment. Dyads were randomly assigned postpartum to VIP 0-3, control 0 to 3 years, or a third group (Building Blocks [BB])¹⁹ by using a random number generated by the project director using Microsoft Excel (Microsoft Corp, Redmond, WA). The second random assignment took place by using the same methodology at the time of the 3-year assessment. All families assigned at postpartum random assignment to VIP 0-3 or control 0 to 3 years were eligible at 3 years for second random assignment to VIP 3-5 or control 3 to 5 years. Families assigned to BB received no follow-up beyond 24 months, did not participate in the 3-year random assignment or 3- and 4.5-year assessments, and did not have data for the present analyses. Group assignments were concealed from staff and study participants until consent for each random assignment was completed.

1. VIP 0-3: As previously described, VIP 0-3 was provided by a bilingual facilitator over 15 possible 1-on-1 30-minute sessions from 2 weeks to 3 years.¹⁹ Parent and child dyads were video recorded during ∼5 minutes of play or shared reading by using a developmentally appropriate learning material (toy or book). Videos were immediately reviewed with the parent to identify and reinforce responsive interactions and promote self-reflection. Age-specific pamphlets were used to provide suggestions for positive parenting and opportunities for parents to describe their observations and develop goals for interacting and were taken home together with videos. After a 3-year implementation, VIP 0-3's cost is estimated at $175 to $200 per child per year, including staff, equipment, supplies, rent, and utilities, with 1 interventionist providing the VIP for ∼400 to 500 families.

2. VIP 3-5: VIP 3-5²¹ took place during 9 30- to 45-minute sessions from 3 to 5 years with the same structure as VIP 0-3 along with enhancements designed to promote interactions in the context of rapidly emerging developmental capacities during the preschool period. VIP 3-5 was developed in consultation with preschool education experts who suggested integration of strategies from research-based curricula,^24,–26 including the following: (1) building sessions around themes (eg, grocery store, birthday party); (2) video recording both story book reading and play, with play planned on the basis of the story; (3) integration of writing within play (eg, shopping list, party invitations); and (4) focusing on characters’ feelings.

Control 0 to 3 years and control 3 to 5 years families received standard pediatric care during the birth- to 3-year and the 3- to 5-year periods, respectively, including recommended anticipatory guidance and monitoring. All groups received ROR.

Bilingual research assistants masked to group assignment performed assessments.

As previously described,^19,23 we assessed baseline sociodemographic characteristics and socioeconomic status (SES; Hollingshead 4 Factor Index²⁷). Maternal literacy was assessed at 6 months in the mother’s preferred language by using the Woodcock-Johnson III and/or the Woodcock and M‎uñoz-Sandoval²⁸ Batería III Tests of Achievement, Letter-Word Identification Test; for families who had not received the literacy assessment, educational level was used as a proxy.²³ As in previous studies, mothers were categorized as having increased social risk if they had 1 or more of the following: homelessness, being a victim of violence, involvement with child protection, financial difficulties, food insecurity, smoking or alcohol use during pregnancy, or previous mental illness, including depression.

At 3 and 4.5 years, we assessed children’s social-emotional development on the basis of parent reports using 4 subscales from the Parent Rating Scales of the Behavior Assessment System for Children, Second Edition (BASC-2)²⁹ (Social Skills, Attention Problems, Hyperactivity, and Aggression) and an Externalizing Problems composite (Hyperactivity and Aggression). The BASC-2 has been normed in English and Spanish. T-scores (mean = 50; SD = 10) were calculated. Recommended cut points were used to dichotomize scores as “Clinically At-Risk” (T-score ≥60, 1 SD above the mean) and “Clinically Significant” (T-score ≥70, 2 SD above the mean) for Attention Problems, Hyperactivity, Aggression, and Externalizing; Social Skills were categorized as Clinically At-Risk for a T-score ≤40 (1 SD below the mean).

The analytic sample for aim 1 (sustained VIP 0–3 impacts) consisted of 275 families randomly assigned postpartum to VIP 0-3 or control 0 to 3 years who had the 4.5-year assessment. The analytic sample for aim 2 (early VIP 3–5 impacts) and aim 3 (additive and/or synergistic impacts) consisted of 252 families randomly assigned a second time at age 3 years who had the 4.5-year assessment. Twenty-three families who were randomly assigned postpartum were not randomly assigned a second time because of either missing the 3-year assessment (n = 22) or not consenting for second random assignment (n = 1) but nonetheless received the 4.5-year assessment because the original enrollment consent included follow-up beyond age 3 years. These 22 families were within the random assignment protocol for aim 1 and were included in the aim 1 analytic sample but were outside the random assignment protocol for aims 2 and 3 and therefore were not included in the aim 2 and aim 3 analytic samples.

Power analyses, previously described for postpartum enrollment, were also performed for the current sample and revealed that 275 dyads would provide 80% power to detect 0.35 SD differences with 2-tailed α = .05. Statistical analyses were performed on the basis of intent to treat for each aim. χ² and t tests and/or analysis of variance tests were used to compare families contributing data at 4.5 years with those who did not and also to compare across groups for sociodemographics and behavior scores at age 3 years after phase VIP 0-3 completion and before phase VIP 3-5 initiation. For aim 1 (sustained VIP 0–3 impacts), we compared VIP 0-3 and control 0 to 3 years children for mean behavior scores and proportions meeting clinical thresholds using multiple linear and logistic regressions, respectively, that adjusted for VIP 3-5 random assignment; comparisons of cells with 0% present were based on Fisher’s exact test. For aim 2 (early VIP 3–5 impacts), we similarly compared VIP 3-5 and control 3 to 5 years children with adjustment for VIP 0-3 random assignment. For aims 1 and 2, Cohen’s d (mean difference and population SD), relative risk reduction (RRR), absolute risk reduction (ARR), number needed to treat (NNT; calculated as 1/ARR), and 95% confidence intervals (CIs) were calculated. Subgroup analyses were performed for families at increased psychosocial risk, given greater impacts on 3-year behavior.²³ For aim 3 (additive and synergistic impacts), we performed cross-sectional analyses as follows: (1) compared mean 4.5-year behavior scores by using linear regressions in which the predictor variable was the number of VIP doses (0–2), and (2) estimated effect sizes by using multiple linear regressions dummy coded separately for receipt of VIP “double dose” (combined VIP 0–3 and VIP 3–5) or “single dose” (either VIP 0–3 or VIP 3–5) and compared with receipt of no VIP (control 0–3 years and control 3–5 years) as the reference group. Finally, we analyzed trajectories from 3 to 4.5 years using multilevel models (MLMs) with Stata/SE 14 (StataCorp, College Station, TX). MLM analyses allowed us to determine sustained impacts of VIP 0-3 (linear coefficient for VIP 0–3) and in separate models, impacts of VIP 3-5 (VIP 3–5 × age interaction), while accounting for repeated measures and random slopes; additional models allowed us to test for potential synergy (ie, additional increase in impact due to inclusion of both components) through inclusion of a double interaction term (VIP 0–3 × VIP 3–5 × age). Multiple comparisons were addressed through the Benjamini and Hochberg³⁰ procedure by using a false discovery rate of 10% (ie, 1 in 10 statistically significant tests expected to be false positive)³¹; application of this approach yielded equivalent findings.

Of 675 families enrolled postpartum, 450 were randomly assigned to VIP 0-3 and control 0 to 3 years, and 275 (61.1%) completed the 4.5-year assessment and were included in aim 1 analyses (Fig 1). Two hundred and ninety-six of the 450 families (65.8%) were randomly assigned a second time at 3 years, of whom 252 families (85.1% of second random assignment) completed the 4.5-year assessment and were included in aims 2 and 3 analyses.

Participants in both the aim 1 and aim 2 and aim 3 samples were comparable across study conditions for all baseline characteristics (Table 1), although VIP 0-3 families had somewhat lower literacy (P < .10), and VIP 3-5 families had lower maternal age (P < .05). The 275 families included in 1 or more analytic samples were similar to nonparticipants for maternal age, marital status, social risk, and firstborn child; however, in the analytic samples, there were more who had lower education (P < .001), literacy (P < .05), and SES (P < .01) and were more likely Hispanic and/or Latino (P < .001), immigrants (P < .001), and Spanish-speaking (P < .001), with somewhat more female children (P < .10). Among these families, a median of 10 out of 15 VIP 0 -3 visits were attended, and a median of 4 out of 6 possible VIP 3-5 visits before 4.5 years were attended.

VIP 0-3 was associated with sustained reductions in Attention Problems (d = −0.25), Hyperactivity (d = −0.31), and Externalizing (d = −0.24) at age 4.5 years, 1.5 years after completion of the program (Table 2), with 59.6% RRR in meeting the Behavior Assessment System for Children (BASC) Externalizing Clinically At-Risk criterion for the overall sample and an 80.2% RRR for those with increased psychosocial risk (Table 3). In addition, VIP 0-3 was associated with a reduction in meeting the BASC Hyperactivity Clinically Significant criterion for the overall sample (2.8% [VIP 0–3] versus 9.1% [control 0–3 years]; RRR: 69.2%; 95% CI: 7.0 to 89.8; P = .03) and for those at increased psychosocial risk (0% [VIP 0–3] versus 15.6% [control 0–3 years]; RRR: 100%; 95% CI not defined because of 0% proportion; P = .006 by Fisher’s exact test). The ARR for Clinically Significant Hyperactivity was 6.3% for the overall sample and 15.8% for those at increased psychosocial risk, respectively, corresponding to NNTs of ∼16 (95% CI: 8 to 156) and 6 (95% CI: 3 to 26), respectively. No significant reductions were found for any of the other scales for the clinical criteria. No differences were found for Social Skills.

VIP 3-5 was associated with early reductions (Table 4) for Aggression (d = −0.22) and Externalizing Problems (d = −0.26) and marginally for Hyperactivity (d = −0.26). There was a 60.7% RRR in meeting the BASC Externalizing Clinically At-Risk criterion for the overall sample and a 72.3% reduction (not significant) for those with increased psychosocial risk (Table 5). No differences were found for Social Skills or Attention Problems.

Each dose (ie, receipt of either VIP 0–3 or VIP 3–5) was associated with reductions in Attention Problems, Hyperactivity, Aggression, and Externalizing Problems, with d’s ranging from −0.18 to −0.32 per dose received (Table 6). Receipt of double dose (ie, both VIP 0–3 and VIP 3–5) was associated with reductions in each of these domains, with d = −0.63 for Hyperactivity and d = −0.54 for Externalizing Problems. MLMs of trajectories for Attention Problems, Hyperactivity, Aggression, and Externalizing (Table 7, Fig 2) revealed significant linear coefficients for VIP 0-3 across the age range but no significant age × VIP 0-3 interactions (latter not shown), indicating that differences in VIP 0-3 at 3 years were sustained through 4.5 years. In contrast, trajectory models revealed significant age × VIP 3-5 interactions for Hyperactivity, Aggression, and Externalizing but no significant linear coefficients for VIP 3-5, indicating increasing impacts for VIP 3-5 after second random assignment. Furthermore, a nonsignificant trend was seen in which VIP 0-3 and VIP 3-5 each potentiated impacts of the other on Attention Problems, as indicated by the age × VIP 0-3 × VIP 3-5 interaction (Fig 2B; P = .06).

We found that promotion of positive parenting activities such as reading aloud, play, and talking resulted in enhanced social-emotional development for children in low-income families. VIP 0-3 (infant through toddler) resulted in sustained impacts on Externalizing Behavior and Attention Problems 1.5 years after program completion. VIP 3-5 (preschool period) resulted in additional, independent impacts, with some evidence of synergy for Attention Problems in which impacts were seen only for children participating in both components.

Effect sizes per dose of VIP (up to 0.32) were comparable to many home visiting programs, whereas those for double dose (up to 0.63) were even greater.^32,33 Notably, these impacts were found for a primary care–based intervention costing ∼1/5 to 1/25 of that of home visiting; however, additional study across multiple sites and diverse populations will be needed to determine if impacts are maintained. Large reductions in risk together with relatively low NNTs (NNT = 15 and 6 for Clinically Significant Hyperactivity for the overall sample and highest risk families, respectively) suggest potential for substantial benefit to cost ratio, given high costs related to education and health care associated with clinical range disruptive behaviors^34,35 and low cost for VIP 0-3.

With these analyses, we extend previous research and can inform policies seeking prevention of poverty-related disparities. In demonstrating that primary care programs promoting reading aloud and play can impact social-emotional development at school entry, we build on our previous work in which we documented 3-year impacts at VIP 0-3's conclusion and on evidence linking these activities to enhanced parent and child relationships.³⁶ With our current findings, we provide not only direct support for the VIP as the program under study but also implicit support for primary care programs more broadly seeking to promote these activities (eg, ROR). In demonstrating additive impacts of infant through toddler and preschool age intervention, our findings converge with a factorial RCT of the Play and Learning Strategies home visiting program, which found additive impacts resulting from extension of an infant through toddler parenting intervention through the late toddler to early preschool period.¹² In our current findings, we support the use of ongoing intervention from birth to school entry in primary care. In demonstrating impacts in a low-income sample, we build on studies of ROR and Healthy Steps in which impacts in similar populations were demonstrated.^14,37,–42 With our current findings, we support a role for pediatric health care for location of universal programs seeking primary prevention before emergence of child, parent and child, and/or family challenges, in addition to secondary and tertiary prevention programs providing screening and referral for services (eg, Healthy Steps Montefiore model,⁴² Help Me Grow,¹⁰ Assuring Better Child Development⁴³).

This study had a number of strengths, including the following: (1) a factorial RCT, which is a strong design for estimating independent and additive effects; (2) high follow-up of families randomly assigned a second time at 3 years (85.2%); and (3) consistency across cross-sectional and trajectory analyses. There were also important limitations. First, attrition resulted in analytic samples comprising greater proportions of Hispanic and/or Latino immigrants with lower SES, and it is possible that less-engaged families could have experienced differences in program impact. Second, assessments were conducted by parent report, which can differ from report by other observers such as teachers and be susceptible to response bias; however, the primary focus of the VIP on reading aloud and play may have reduced socially desirable responses. Third, clinical-level Hyperactivity and/or Externalizing Behaviors on the BASC do not necessarily indicate a diagnosis of attention-deficit/hyperactivity disorder, and clinical assessments were not available for participating families.

In this study, we provide strong support for the use of pediatric primary care to promote positive parenting activities such as reading aloud and play and the potential for such programs to promote social-emotional development as reflected through reductions in disruptive behaviors. With the effect sizes, it is suggested that such programs can result in clinically important differences on long-term educational outcomes, given the central role of behavior for child learning.² There is high potential for scalability of the VIP together with relatively low-cost support integration and alignment with existing initiatives through home visiting and community-based intervention (eg, Administration for Children and Families’ Maternal, Infant, and Early Child Home Visiting,⁴⁴ Bridging the Word Gap,⁴⁵ Providence Talks,⁴⁶ City’s First Readers,⁴⁷ and Thirty Million Words⁴⁸) and dissemination within well-child care.

ARR

absolute risk reduction

BASC

Behavior Assessment System for Children

BASC-2

Behavior Assessment System for Children, Second Edition

BB

Building Blocks

BHC

Bellevue Hospital Center

CI

confidence interval

MLM

multilevel model

NNT

number needed to treat

RCT

randomized controlled trial

ROR

Reach Out and Read

RRR

relative risk reduction

SES

socioeconomic status

VIP

Video Interaction Project

VIP 0-3

Video Interaction Project phase birth to 3 years

VIP 3-5

Video Interaction Project phase 3 to 5 years

We thank many colleagues for their guidance and support, including J. Lawrence Aber, PhD, Clancy Blair, PhD, David Dickinson, EdD, Arthur Fierman, MD, Virginia Flynn, MS, Gilbert Foley, EdD, Perri Klass, MD, Lesley Morrow, PhD, Erin O’Connor, EdD, Cybele Raver, PhD, Catherine Tamis-Lemonda, PhD, Wendy Tineo, PhD, and Linda van Schaick, MSEd. We also thank those who contributed to this project, including Angelica Alonso, Jenny Arevalo, Nina Burtchen, Diego Catalan Molina, Aida Custode, Yuliya Gurevich, Jennifer Ledesma, Maya Matalon, Andrea Paloian, Caroline Raak, Kristina Vlahovicova, and Lisa White. We also thank the parents and children who participated.