Evidence for Internal Structure of Established Parenting Measures with a Diverse Population of Fathers Involved in Child Welfare

Background: Existing research demonstrates the importance of engaged fathers in the lives of their children. Federal investments in the United States into community-based parenting programs targeting fathers further demonstrate a broader commitment to equip fathers to impact their families’ lives in positive ways. While tools exist to measure key parenting constructs targeting fathers (e.g., co-parenting, parental efficacy, relationship conflict), their psychometric properties are often tested on samples consisting largely of white, middle-class men. Less is known about the transferability of these constructs and measures to other populations of men engaged in fatherhood programs. It is critical to understand the utility of these tools across populations to ensure sound evaluations of community-based fatherhood programs. This study aimed to contribute in this area by examining the internal factor structure of three parenting measures among a diverse sample of fathers involved in the child welfare system. This presentation will highlight the collaborative process of designing a randomized control trial, the selection of previously constructed parenting measures, and validation analyses of chosen measures with a diverse population using advanced statistical methods.

Methods: In collaboration with the fatherhood program’s implementation team, three established scales measuring co-parenting, parental self-efficacy and satisfaction, and relationship conflict were administered to fatherhood program participants to evaluate the intervention effect over three time points. Preliminary results for this session include 171 participants who completed surveys at the first time point upon program enrollment. Evaluators performed confirmatory factor analyses (CFA) on each measure to examine internal factor structures and assess validity evidence and whether psychometric properties aligned with previous studies that utilized these measures.

Results: Preliminary results from the three models suggested evidence of reliability and validity for each measure’s underlying factor structure. Results also indicated that each construct was comprised of the sub-scale domains that were theoretically established and designed from previous research. The CFAs for each of the three models demonstrated acceptable fits. (CFIs = .83 to .95; TLIs =.81 to .93; RMSEAs = .08 to .10; SRMRs < .08), and all loadings for each measure were statistically significant. Cronbach’s alpha indicated sufficient reliability across each measure and their sub domains (alphas > .84) and were consistent with the original development of the measures.

Conclusions: These analyses provide support that the evaluation team’s collaboration with the community partners was vital to identify previously constructed and validated measures to evaluate community led interventions. By understanding the original goals and purpose of the intervention, the evaluators assisted in identifying previously constructed tools to measure the effects of the intervention while also analyzing the psychometric properties of the measures to ensure that they accurately capture the lived experience of the diverse populations of the program. The evaluators applied complex quantitative strategies to help inform whether the measures were appropriate for the current program intervention and will continue to engage in ongoing conversations with the community partners to adapt or create novel measures if results indicate that the tools were not psychometrically sound to measure their interventions.