Dimensional Analyses of Childhood Adversity: Contributions and Challenges

• In Event: Conceptualizing & Analyzing Childhood Adversity

Abstract

Introduction: Research on adverse childhood experiences (ACEs) has helped to elucidate the harmful effects of early life stressors. Nevertheless, several critiques of ACE measurement standards have been levied, including the use of rudimentary count scores to operationalize cumulative risk. Some scholars contend that dimensional models of adversity will advance our capacity to unveil causal mechanisms that link ACEs and their consequences. Drawing on a neuroscientific perspective, McLaughlin et al. (2014) distinguished experiences of threat and deprivation, while Ellis et al. (2012) inferred from evolutionary theory that threatening and harsh events have different developmental implications than unpredictable events. Recently, Ellis and colleagues (2022) synthesized the models into a three-dimensional framework: (1) threat/harshness; (2) deprivation/harshness; (3) unpredictability.
Evidence has been marshaled to support all of these models, and methodologies have varied widely in this rapidly evolving area. Further studies are needed to reconcile discrepancies and demonstrate the reproducibility and utility of dimensional approaches—especially in marginalized populations that are often exposed to many interrelated adversities. Using data from a sample of 1,641 low-income women, this study explores whether two different dimensional approaches fit an expanded ACE assessment and improve the prediction of adult outcomes relative to an ACE count score.

Study Population and Methods: Data originated from the Families and Children Thriving Study, a longitudinal investigation of families that received home visiting services. ACEs were assessed using the Childhood Experiences Survey, which yielded 10 conventional indicators of child maltreatment and household dysfunction along with seven other adversities: financial problems, food insecurity, homelessness, parental absence, parent/sibling death, peer victimization, and violent crime victimization. Using confirmatory factor analysis (CFA), 11 ACEs were categorized as aligning with the McLaughlin two-factor model while all 17 items were used to fit the Ellis-McLaughlin three-factor model. A second-order CFA also was performed for the three-factor model, producing a latent ACE variable that was regressed on six outcomes: global health, sleep disturbance, depression, PTSD, substance misuse, and smoking. The resulting standardized solutions and R2 values were compared to coefficients derived from regression analyses of ACE count scores.

Results: A CFA of 11 ACEs confirmed that a threat-deprivation model fit the data well (CFI=.954, RMSEA=.067, SRMR=.066), while the threat-deprivation-unpredictability model was a good fit for the expanded set of 17 ACEs (CFI=.950, RMSEA=.053, SRMR=.063).
However, between-factor correlations were high for the two-factor model (.775) and three-factor model (range=.654-.784); structural equation modeling showed the outcomes could not be regressed on independent factors due to collinearity. Moreover, correlations between all factor scores and outcomes were comparable in magnitude. A second-order CFA produced a well-fitting, parsimonious solution by modeling three subdomains of ACEs under one global latent construct. Regression analyses indicated that, compared to ACE count scores, the latent ACE variable was associated with larger standardized coefficients and R2 values.

Conclusion: ACEs can be configured to fit many models. While dimensional models are intriguing conceptually, high intercorrelations between ACEs present methodological challenges to fulfilling their promise. Further research on dimensional models is needed, with an eye toward consilience and practical significance.

Authors

• Joshua Mersky, University of Wisconsin - Milwaukee
  Presenting Author

• ChienTi Plummer Lee, University of Wisconsin - Milwaukee

• Colleen Janczewski, University of Wisconsin - Milwaukee