Abstract: Improving the Usability and Clinical Application of the Parent Anger Inventory: Implications for Research and Practice (Society for Social Work and Research 20th Annual Conference - Grand Challenges for Social Work: Setting a Research Agenda for the Future)

526P Improving the Usability and Clinical Application of the Parent Anger Inventory: Implications for Research and Practice

Schedule:
Sunday, January 17, 2016
Ballroom Level-Grand Ballroom South Salon (Renaissance Washington, DC Downtown Hotel)
* noted as presenting author
Allison E. Dunnigan, MSW, Doctoral Candidate, Washington University in Saint Louis, St.Louis, MO
Megan Feely, MSW, Doctoral Student, Washington University in Saint Louis, St. Louis, MO
Patricia L. Kohl, PhD, Associate Professor, Washington University in Saint Louis, St. Louis, MO
Wendy Auslander, PhD, Barbara A Bailey Professor of Social Work, Washington University in Saint Louis, St. Louis, MO
Background and Purpose: Parenting-related anger has a negative impact on parent-child interactions and is associated with an increased risk of child maltreatment.  A key mechanism to preventing maltreatment is reducing anger among parents, and as such, evidence-based parenting interventions target anger as a critical intermediate outcome for evaluating treatment effectiveness. The Parent Anger Inventory (PAI) is   commonly used to measure parenting-related anger in parenting interventions.  The PAI assesses which of the 50 listed child behaviors parents “have found to be a problem in the last month,” and then assesses how angry the parent gets in response to the behavior on a five-point scale from “not at all” to “extremely”.  However, the 50-item measure is long and tiring for respondents, and there are no clearly defined subscales.  Identifying latent factors that form subscales could improve the usability of the measure and enhance clinicians’ ability to target interventions to a parent’s needs.

The purpose of this study was to identify meaningful and interpretable subscales for the PAI.   Theoretically informed domains for subscales were identified using the framework for child conduct and personality problems originally developed by Peterson (1961).  The items were grouped into the following domains:  Developmental problems, Problems at school/public, Disobedience, Interpersonal interactions and Defiance.  These subscales were tested in a confirmatory factor analysis (CFA) and compared to a CFA suggested by an exploratory factor analysis (EFA) model.  

Methods:  Baseline data from an intervention study of parents investigated by child welfare services were used for the current study (n=142).  Data were collected using a structured in-person interview.  The sample was 66% African American or biracial, 56% of the children were male and 92% of the parents were female.  Items that were rarely endorsed on the PAI were removed leaving 42 items.   Models were assessed for fit using the available fit indices.

Results:  The CFA of the PAI as a single-factor had poor fit (RMSEA=.069, CFI=.85, TLI=.84).  The theoretically derived model had a better fit to the data (RMSEA=.061, CFI=.884, TLI=.873) with standardized factor loadings ranging from 0.46-.86.  The EFA informed confirmatory model did not fit the data as well as the theoretically informed model (RMSEA=.097, CFI=.902, TLI=.890).

Conclusion:  The five subscales of the theoretically informed model is the most promising solution for a meaningful and usable assessment of parenting anger. In contrast, the subscales suggested by the EFA did not correspond to child behavior theory.  The theoretically informed subscales can be used by practitioners to identify specific categories of problems.  Increasing the precision of the measure would allow practitioners to target an individual parent’s most problematic issues. Moreover, researchers who develop and evaluate the effectiveness of parenting interventions may be better able to refine their interventions, as well as to make more precise conclusions about the strength of their evidence. Further research is needed to test the validity and test-retest reliability of the identified subscales.