Testing the Latent Factor Structure of Childhood Adversity in a Nationally Representative Sample of Child Welfare-Involved Adolescents

Background and Purpose: Childhood adversity is a complex construct, including a wide range of experiences such as physical abuse, sexual abuse, neglect, parental substance use and/or separation/divorce. Although it is typically measured as a cumulative score, childhood adversity may be better conceptualized as a multi-dimensional construct. One recent model, the Dimensional Model of Adversity and Psychopathology (DMAP), posits two main dimensions, threat (i.e., exposure to violence) and deprivation (i.e., lack of typical cognitive, socioemotional, and/or physical stimuli), that may have partially distinct associations with mental health outcomes. Despite recognition that childhood adversity may include multiple dimensions, few studies have tested competing models to determine the best factor structure, particularly among child welfare-involved youth who often are at disproportionate risk for adverse childhood experiences. Therefore, the purpose of the study was to explore the factor structure of adverse childhood experiences in a sample of adolescents involved with the U.S. child welfare system.

Methods: The study included a nationally representative sample of 1,054 child welfare-involved adolescents between 11-17 years (M = 13.72 years, 55.4% female, 61.8% minoritized racial/ethnic identity) from the National Survey on Child and Adolescent Well-Being II (NSCAW-II). Nine indicators of childhood adversity were assessed via youth, caregiver, and child welfare caseworker report on several validated measures (e.g., Parent-Child Conflict Tactics Scale, Conflict Tactics Scale 2) and NSCAW-specific survey items. Each indicator was scored as exposed (=1) or not exposed (=0). We conducted confirmatory factor analysis (CFA) to test a unidimensional model (model 1) and two-factor threat and deprivation model (model 2). As a supplemental analysis to the CFA, we also conducted exploratory factor analysis (EFA; model 3) without an a priori hypothesis due to the lack of prior research examining the structure of adversity among child welfare-involved adolescents and lack of consensus on the best-fitting measurement model for this population. Missing data were handled using full information maximum likelihood estimation.

Results: CFA models 1 and 2 did not converge or had poor fit and/or low factor loadings that were not improved by model modifications. The model identified via EFA (model 3) indicated that a 3-factor structure fit the data best, X²(12)=6.02, RMSEA<.001, CFI/TLI=1.00, SRMR=.041. Physical abuse and emotional abuse loaded onto the first factor, physical neglect and substance abuse exposure loaded onto factor 2, and emotional neglect and domestic violence exposure loaded onto the third factor. These factor loadings were all >.40. In contrast, sexual abuse, caregiver mental illness, and caregiver separation/divorce cross-loaded onto at least two factors and thus were excluded as they were not well defined by the underlying factors.

Conclusion and Implications: Our findings highlight the importance of continuing to test how to best measure childhood adversity across heterogeneous populations (e.g., community samples, child welfare-involved youth) as different dimensions of adversity may be more (or less) salient for certain populations. Specifically, the results suggest that distinguishing between threat, physical deprivation, and emotional deprivation may be important within child welfare-involved populations. These findings have implications for advancing childhood adversity research and clinical and child welfare practice.