A Multilevel Analysis of Recurrence in Alternative Response

Background and Purpose: Alternative response (AR) is an approach to child protective services (CPS) where caseworkers are required to provide a family-centered, strengths-based approach as opposed to making a final determination of abuse/neglect. One area that continues to vex child welfare administrators and researchers is recurrence; or the return of a family into the child welfare system, particularly among AR families as AR is typically used for lower risk families. Moreover, the main purpose of AR is to provide appropriate supports and resources in which to stabilize the family. Using known predictors of recurrence (e.g., previous CPS experience, race/ethnicity, number of children in the home, etc.), a multilevel analysis was performed to determine whether the adoption of an AR helped to reduce the number of recurring investigations in one mid-Atlantic state.

Method: A cross nested multilevel logistic analysis was performed using administrative data for 23,871 families. Level 1 used child characteristics (e.g., age, race/ethnicity, gender) and level 2 used caregiver characteristics (e.g., same as child plus type of response, allegation type, number of children in home, number of previous investigations, and Medicaid receipt); families were nested by county. Differences were examined between families who received a traditional response (TR) versus an AR. These same characteristics were used to predict which families would receive a subsequent investigation, and among those, what predicted a substantiated recurrence. Families were followed for 18-months post investigation.

Results: Bivariate analysis determined differences among TR and AR families for child and caregiver race, maltreatment allegation, Medicaid receipt, and recurrence. In the predictive models, some of these differences disappeared. The number of children, child gender, and Medicaid receipt predicted a subsequent investigation. Child age, Medicaid receipt, previous investigative finding/response, and county predicted a substantiated recurrence; however, those in receipt of Medicaid were less likely to have a substantiated recurrence, whereas in the first model they were more likely to endure a subsequent investigation. Of interest was the predictive nature of a previous investigative finding in that families who previously received an unsubstantiated investigation were more likely to have substantiated recurrence as compared to families who received an AR response.

 Conclusions and Implications: The results suggest that additional research is needed to fully understand how specific case factors influence family outcomes as some findings were in line with previous research (e.g., more children in a home predicts recurrences) whereas others are still in limbo (e.g., whether AR adequately keeps children safe – this study did not find AR to predict recurrences). This study is unique in that it analyzed an entire state’s population; however, due to the idiosyncratic implementation of AR within each county, there is caution in generalizing beyond this state. Future research should consider going back a step and determining factors that predict the type of response a family will receive at intake and then follow the family through the life of the case. Further, operationalizing organizational factors, though difficult could illuminate an additional understanding of the efficacy of AR and its relationship to recurrence.