Improving Parent-Child Interactions in Pediatric Health Care: A Two-Site Randomized Controlled Trial

We conducted a single-blind 2-way RCT of SB at two sites: Bellevue Hospital in NYC and University of Pittsburgh Medical Center (UPMC) Children’s Hospital of Pittsburgh. The study was registered at www.clinicaltrials.gov (identifier NCT02459327). Institutional review board approval was obtained from New York University (FY2016-408), NYU Grossman School of Medicine (S14-01764), and University of Pittsburgh (STUDY19040158). Research approval was obtained from NYC Health+Hospitals. Families were not compensated for participation in study interventions but received a modest incentive ($50) for participating in the 6-month assessment.

SB includes (1) VIP as a universal primary prevention strategy,¹⁴  provided for all families randomly assigned to the treatment group at birth, and (2) FCU provided for treatment families with identified psychosocial risks beginning at 6 months.¹⁰ 

VIP was designed as an enhancement to Reach Out and Read,¹⁵  a national program that facilitates literacy promotion during routine well-child care. VIP is a strengths-based, family-centered model that addresses parenting assets and vulnerabilities that mediate the impact of poverty on child development. VIP is posited to enhance positive parenting behaviors and, in turn, promote children’s school readiness. The VIP 0–3 program includes fourteen 25–30 minute sessions delivered in pediatric primary care from birth to 3 years, aligned with well-child visits and delivered by coaches (bachelor’s level) hired for this project. Coaches receive a 3-day training and continued supervision.

Every VIP session follows the same format, including discussion of the child’s development and provision of a developmentally appropriate learning material (book, toy) selected to engage the parent and child in interaction. The coach briefly video records the parent and child interacting with the book or toy and then immediately reviews the video with the parent to identify and reinforce strengths in the interaction and promote self-reflection. The parent is provided with a copy of the video and a personalized pamphlet that includes information about developmental milestones, age-specific suggestions for engaging with their child, and the parent’s goals for interacting with their child at home. For example, at 6 months, the coach would talk to the parent about their child’s development and milestones (eg, making sounds) and tips for interactions (eg, imitating infant sounds), provide the parent with a developmentally appropriate toy (eg, hand puppets), record the parent and infant interacting, review the video together, highlighting strengths, and help the parent plan for interacting with their infant at home.

FCU is an evidence-based home-visiting model that seeks to reduce the development of early disruptive behavior and motivate parents to engage in services that improve parenting practices.¹⁶  Whereas VIP begins at birth in the SB model, families do not begin receiving FCU until the infant is 6 months, following the time period of data reported here. Our description of FCU is therefore brief; details of the program’s evidence are published elsewhere.^10,16 

We used a two-phased enrollment process with consecutive sampling. Mothers and infants were enrolled in the postpartum units of NYC Health+Hospitals and Bellevue from June 2015 to January 2017 and UPMC Magee-Women’s Hospital adjacent to the UPMC Children’s Hospital of Pittsburgh from June 2016 to October 2017. In phase 1, low-income or Medicaid-eligible families were offered enrollment, and informed consent was obtained if they met the following inclusion criteria: (1) the infant is born term, is singleton, has a normal birth weight without significant prenatal or perinatal medical complications, is ineligible for Early Intervention at birth (eligibility is comparable across states),^17,18  and plans to receive pediatric care at the institution; and (2) the parent is the primary caregiver or legal guardian, plans to stay in the birth city for 3 years, speaks English or Spanish as their primary language, has no known significant impairment (eg, intellectual disability, schizophrenia) or medical complication, has no plans to stay in a shelter, infant discharged to mother, and has no previous participation in VIP or FCU. In phase 2, occurring through 6 weeks of age in the outpatient setting, families who presented for a study visit and continued to meet inclusion criteria were randomly assigned to treatment or control groups. Although differences in hospital policies resulted in higher rates of participants declining to complete screening in Pittsburgh, the overall percentage of families signing consent among those assessed for potential eligibility was similar (NYC: 27%; Pittsburgh: 31%; Figs 1 and 2).

At study enrollment (postpartum), baseline sociodemographic characteristics were assessed by parent survey. At child age 6 months, primary and secondary outcomes related to parenting including cognitive stimulation and parent-child interaction were assessed through survey and observation. All research assistants were blind to the randomization group and trained by project investigators and coordinators, receiving periodic follow-up to maintain fidelity.

StimQ₂ is a structured interview measure of caregiver cognitive stimulation. It was developed and validated in English and Spanish for low-income populations.^19,20  Cronbach’s α in the current sample was 0.76. Three subscales of the StimQ₂ Infant were administered: (1) Parent Verbal Responsivity (PVR), measuring verbal interactions across 2 subdimensions (Everyday Routines, Play and Pretend); (2) Parental Involvement in Developmental Advance (PIDA), measuring teaching activities; and (3) Reading Activities (READ), with 3 subdimensions (Quantity, Quality, Diversity of Concepts). To ensure accuracy and limit social desirability bias, administration includes prompts for examples and follow-up questions. We calculated total, subscale, and subdimension scores.

Parental warmth was examined as a secondary outcome by using the Supporting and Enjoying subscale of the Parenting Your Baby questionnaire. This scale has shown high construct, convergent, and predictive validity.²¹  Cronbach’s α in this sample ranged from 0.67 to 0.73.

Observational measures of global parent-child interaction quality were assessed by using video recording and review of 10 minutes of free play. Mothers played with their infant by using toys typically available in homes but not provided in VIP. Coding of the videos was performed by using the Parent-Child Interaction Rating Scales – Infant Adaptation (PCIRS-IA), which assesses the quality of the parent’s interactions with their child (C.S. Tamis-LeMonda, PhD, P. Ahuja, PhD, B. Hannibal, PhD, J. Shannon, PhD, M. Spellmann, PhD, unpublished observations).^22,23  Coders blind to participants’ treatment condition rated interactions on a 1 to 7 scale. An additional 15% of the interactions were scored by a second coder. Weighted κ ranged from 0.7 to 0.85, indicating satisfactory agreement. Five domains of the PCIRS-IA relevant to child development were coded: parental sensitivity (awareness of child’s needs, interests, and capabilities), parental intrusiveness (imposition of own agenda), parental support for cognitive development (intention to support learning), parental support for language quantity (verbal stimulation), and parental support for language quality (richness of language). We then developed a composite measure of “cognitive stimulation” on the basis of exploratory and confirmatory factor analysis demonstrating strong loading of 3 subscales (parental support for cognitive development, language quantity, and language quality). Factor loadings were 0.79, 0.91, and 0.89, respectively (root mean square error of approximation = 0.00). The resulting Cronbach’s α for this factor was 0.91.

A total of 400 families in the full sample were needed to provide at least 80% power for a minimally detectable effect size of 0.3 after accounting 25% attrition.²⁴  Analyses were performed with Stata 14 (Stata Corp, College Station, TX)²⁵  and on the basis of intent to treat (ie, by group assignment regardless of level of intervention participation). We conducted baseline equivalence tests across the full sample and within each site using t tests and χ². Participation in VIP was calculated as the mean number of visits through 6 months and compared across groups by using a t test. Because random assignment occurred within two sites, comparisons of outcomes for the full sample used multiple regression analyses that included fixed effects for the site. For each of these analyses, we calculated mean difference and 95% confidence interval (CI) and Cohen’s d as a measure of effect size in SD units. We also performed secondary analyses within each site separately; however, because the study was not powered to detect effect sizes within each site, within-site analyses are exploratory.

In total, 403 families were randomly assigned: 200 in NYC and 203 in Pittsburgh (Figs 1 and 2). There were many between-site differences, with NYC primarily Latinx and Pittsburgh primarily Black/African American (Table 1). Mothers in NYC had higher rates of marriage and cohabitation and were less likely to be high school graduates. No significant differences emerged between the treatment and control groups across baseline characteristics (Table 2). There were also no significant differences for baseline characteristics among families who were or were not assessed in the full sample (omnibus, P = .48) or in either site (omnibus NYC, P = .26; omnibus Pittsburgh, P = .21).

Families had high rates of participation in VIP. The mean number of VIP visits completed was 3.49 (SD = 0.97) of four possible visits. The mean number of visits completed did not differ between NYC (mean = 3.40; SD = 1.13) and Pittsburgh (mean = 3.58; SD = 0.77), P = .17.

Analyses of outcomes related to cognitive stimulation and parent-child interactions for the full sample (Table 3) and for separate exploratory analyses of NYC and Pittsburgh (Table 4) are shown.

In the full sample, treatment families scored significantly higher than controls on most dimensions and subdimensions of the StimQ₂. This included StimQ₂ total (Cohen’s d = 0.28; P = .01), StimQ₂ READ (d = 0.23; P = .02), including Quality (d = 0.26; P = .01) and Diversity of Concepts (d = 0.23; P = .02), StimQ₂ PVR Play and Pretend (d = 0.21; P = .04), and StimQ₂ PIDA (d = 0.25; P = .01). In exploratory analyses, NYC treatment families scored significantly higher than controls on StimQ₂ READ quality (d = 0.32; P = .04) and Diversity of Concepts (d = 0.27; P = .049) and StimQ₂ PIDA (d = 0.34; P = .02). Pittsburgh treatment families scored significantly higher than control group families on StimQ₂ total (d = 0.27; P = .046). Although scores for treatment families were higher than scores for controls in Pittsburgh for StimQ₂ PVR (d = 0.23; P = .08), differences were not statistically significant. There were no differences in mother-reported Parenting Your Baby: Supporting and Enjoying for the full sample or at either site.

In the full sample, treatment families scored significantly higher than controls on several dimensions of the observations. These included parental support for cognitive development (d = 0.36; P < .001), language quantity (d = 0.40; P < .001), and language quality (d = 0.37; P < .001), as well as the cognitive stimulation factor (d = 0.40; P < .001). In exploratory analyses, NYC treatment families scored significantly higher than controls on parental support for cognitive development (d = 0.47; P = .01), language quantity (d = 0.70; P < .001), language quality (d = 0.52; P < .001), and the cognitive stimulation factor (d = 0.60; P < .001). In Pittsburgh, treatment families had higher scores for parental support of cognitive development (d = 0.26; P = .07); however, this was not statistically significant. In Pittsburgh, no differences emerged for language quantity, quality, or the cognitive stimulation factor. There were no differences in sensitivity or intrusiveness either in the full sample or at either site.

In this study, we examined impacts of VIP delivered in primary care on cognitive stimulation and parent-child interactions for the first phase of SB. In the full sample, including two sites with racially and ethnically diverse participants, VIP had positive impacts on both survey and observed measures of cognitive stimulation and parent-child interactions at child age 6 months. More specifically, participation in VIP impacted multiple domains of parent-reported cognitive stimulation, including reading, verbal responsivity, and teaching behaviors, and observed measures of parent-child interaction, including parental support for cognitive development, language quantity, and language quality.

Findings replicate previous research revealing VIP impacts on cognitive stimulation and parent-child interactions and extend them by revealing these impacts across a more comprehensive set of survey and observational measures and with families from diverse racial and ethnic groups. These impacts are notable given substantial research demonstrating that such behaviors mediate the relationship between poverty and school readiness,^26,27  with implications for long-term educational trajectories. This replication of findings is especially important in light of challenges in replication in the behavioral sciences more broadly.²⁸ 

Although we did not replicate impacts on parenting warmth using the parent-reported Parenting Your Baby, this null finding may reflect the limited variability within the current sample (93% scored within 1 point of ceiling). Lack of impacts on observed sensitivity and intrusiveness may suggest specificity of VIP in impacting positive aspects of cognitive stimulation or could be a consequence of the assessment being performed in early infancy.²⁹  It is also possible that additional exposure to VIP or using the full complement of strategies tailored to risk heterogeneity provided in the full SB model may be necessary to affect these behaviors.

Comparable VIP participation across sites supports feasibility for parents from diverse geographic locations and racial and ethnic backgrounds. Although not sufficiently powered (and thus not always resulting in findings that meet standards of statistical significance), exploratory analyses within sites suggest that the VIP intervention resulted in clinically meaningful effect sizes at both sites for both survey and observational measures. For the survey measures, the effect sizes for StimQ₂ overall and the StimQ₂ READ subscale were similar across sites. Whereas the PVR subscale had a higher effect size in Pittsburgh, the PIDA subscale had a higher effect size in NYC. For the observational measures, although impacts were found in both sites, effect sizes were greater overall in NYC. To put this in perspective, effect sizes of 0.2 to 0.3 (found in both sites for survey measures and in Pittsburgh for several observational measures) are similar to previous reports and are at levels comparable to more intensive programs (eg, home visiting)^30,31  that are considered to be clinically significant.³²  Effect sizes of 0.5 to 0.7 (found for observational measures in NYC) are higher than many studies of other preventive interventions.^30–33  Future studies are needed to interpret potential differences across subdomains and subdimensions between the two sites and determine if the patterns obtained from survey and observation measurements persist longitudinally. Data from later assessment points within the study will help to address these questions.

This study had many strengths, including a multimethod assessment across a geographically and racially and ethnically diverse sample. There were also some limitations. First, the study was not powered to show differences across the two sites and populations, resulting in exploratory subgroup analyses. Second, because race and ethnicity fully confounded site, the design did not allow consideration of either characteristic separately in interpreting findings. Third, this study took place during a period of specific stressors for immigrant and other racial and ethnic minority families, including heightened racism and discrimination.³⁴  Experience of stress in these communities may have impacted enrollment and participation in assessments and could have implications for generalizability. Finally, in the current study, we excluded several high-risk populations including newborns qualifying for Early Interventions and preterm or low birth weight infants. Although these exclusions limit generalizability of findings to higher-risk groups, pilot adaptation of SB is currently being implemented for these populations, children in foster care, and those with prenatal opioid exposure. Future research will be necessary to examine the efficacy of SB for parenting and child outcomes in these groups.

SB is one of the first tiered models linking and integrating evidence-based interventions across pediatric primary care and home visiting to prevent disparities in early development and school readiness. This study revealed that VIP delivered as the first phase of SB had impacts on parenting behaviors across two geographically distant sites with mothers from racially and ethnically diverse backgrounds. Although exploratory analyses suggest some potential site differences in impact, clinically meaningful effect sizes were found at both sites, suggesting utility in both communities and generalizability of the program to sites led by individuals outside the original VIP project. Future study at subsequent assessments will further determine impacts and feasibility of the integrated, comprehensive SB model.

We thank many individuals who contributed to this project, including Angelica Alonso, Jenny Arevalo, Adriana Chung, Aida Custode, Fernanda Fernandez, Ashley Griffin, Yuliya Gurevich, Julianna Gutierrez, Jessica Harding, Spenser Kinsey, Jennifer Ledesma, Maya Matalon, Elise Mauren, Maritza Morales-Gracia, Luciane Piccolo, Christopher Rodrigues, Anne Seery, Tracie Stuft, and Mackenzie Whipps. We also thank all of the children and parents who participated in this research and the many research interns who assisted with this project.

CI

confidence interval

FCU

Family Check-Up

PCIRS-IA

Parent-Child Interaction Rating Scales – Infant Adaptation

PIDA

Parental Involvement in Developmental Advance

PVR

Parent Verbal Responsivity

RCT

randomized controlled trial

READ

Reading Activities

SB

Smart Beginnings

UPMC

University of Pittsburgh Medical Center

VIP

Video Interaction Project