Examining Equitable Decision-Making in Child Welfare: Child Vulnerability and Family Protective Factors As Predictors of Risk Assessment

Background/Purpose: Equitable service to families reported for maltreatment depends on consistent and reliable child welfare decision-making. However, research has shown that individual biases, cognitive shortcuts, and organizational factors can lead to inconsistent and inaccurate decision-making. Many child welfare agencies use the Structured Decision Making^® (SDM) model (NCCD, 2015) to standardize decision-making and to improve accuracy. Yet, even when utilizing SDM, there are still decision points that require worker discretion. One such decision is the choice whether or not to conduct an in-depth risk assessment (RA) for referred families who have no safety threats identified on their safety assessment (SA). Because SDM recommends but does not require a RA for these families, workers’ discretion determines which families receive further assessment. If workers inconsistently administer RAs to these families, some proportion of future maltreatment risk may go unassessed.

To better understand this step of SDM decision-making, we examined why some families with no safety threats receive RAs while others do not. Specifically, we assessed whether the probability of receiving a RA depends on the number and nature of child vulnerabilities or family protective factors indicated on the SA.

Methods: We examined data from 2,001 initial SAs and 990 associated RAs administered in 2011 in one urban county in a large western state. Using multiple logistic regression, we analyzed five child vulnerabilities and 10 family protective factors as possible predictors of the odds of receiving a RA. We controlled for allegation type and disposition in all models.

Results: Among the 1,496 SAs where the worker indicated no safety threats and determined that the household was safe, 37.8% were subject to follow-up RA and 62.2% were not. The odds of receiving a RA were associated neither with total number of child vulnerabilities, nor any individual child vulnerability. Protective factors were associated with the odds of receiving a RA. For every additional protective factor, the estimated odds of receiving a RA increased by 11% (OR=1.11, p<0.001). Furthermore, three specific protective factors were significantly associated with the estimated odds of receiving a RA: (1) a child’s capacity to participate in safety interventions (OR=3.69, p<0.001), (2) a caretaker’s willingness to recognize problems (OR=1.86, p=0.014), and (3) a caretaker’s commitment to meeting the needs of the child (OR=0.60, p=0.043).

Conclusions/Implications: We found that child vulnerabilities were non-significant predictors of RA, and that protective factors increased, not decreased, odds of RA. These unexpected results raise the question of whether factors other than those indicated on the initial SA may be influencing the decision to administer RA. If this is not the case, our results suggest that some discretionary use of SAs may be mistargeted. There may be a need to better train workers on discretionary SDM decision-making, including documenting additional factors that influence the decision to administer RA. Standardizing child welfare decision-making will prevent mistargeting of resources, under-identification of maltreatment risk, and further system penetration of families with no safety threats. Furthermore, it will ultimately contribute to more reliable and equitable assessment and intervention for families reported for maltreatment.