Substantiation of Child Maltreatment Risks Identified By Reporters during the Hotline Assessment

Background and Purpose: Little is known about the specific elements of a child maltreatment report that predict substantiation of the allegation(s). Much of the research on substantiation has focused on information obtained during the investigation and how this information is correlated with case disposition. Far less has addressed information initially conveyed by reporters at the time of the hotline call. 

Methods: Maltreatment reports from the California Child Welfare Service Case Management System (CWS/CMS) were linked to Structured Decision Making (SDM) hotline risk assessments for the years 2010-2014 (N=1,571,079). Report records were aggregated at both the child and the report level to create summary variables corresponding to each hotline assessment. From the SDM hotline risk assessment data, count variables were generated to summarize the total number of risks in each maltreatment category. Summary variables were also generated to represent categories of specific risks reported under each maltreatment category (e.g., sexual contact with an adult caregiver, severe physical abuse, lack of adequate shelter, domestic violence).  Bivariate analyses were used to assess relationships between the specific type and frequency of risks reported during the hotline call and substantiation on specific maltreatment types. Generalized linear models predicted the relative risk of substantiation on each maltreatment type for each risk category identified during the hotline call.

Findings: Findings indicate that specific types of alleged maltreatment are more likely to be substantiated than others. In cases of reported sexual abuse, allegations that abuse took place by a household member or adult caregiver were most likely to be substantiated for sexual abuse (18.4%). For physical abuse, hotline assessments involving severe injury were most likely to predict substantiation of physical abuse, with 27.0% of allegations of severe injury substantiated. In constrast, cruel or excessive punishment was predictive of substantiation in just 7.0% of reported cases.  Hotline assessments involving neglect were substantiated as neglect in 21.7% of reported cases, but some specific types of neglect were more likely to be substantiated than others. Hotline assessments involving prenatal substance use were substantiated for neglect in 62.9% of reported cases, while hotline assessments of inadequate food, shelter, or medical care, were each substantiated in fewer than 20% of reported cases. Reports were most often substantiated for neglect, regardless of the specific type of risk contained in the hotline assessment.  The number of risks contained within one hotline assessment were also predictive of substantiation, with each additional risk making substantiation more likely.

Conclusion and Implications: Given growing interest in predictive risk modeling and the potential to apply automated algorithms to administrative data in real-time, better understanding how combinations of underlying safety concerns recorded using existing risk assessment tools may highlight salient information currently available to screeners of hotline calls, while also documenting additional data that could be used in predictive risk models to improve decisions made under conditions of high uncertainty. These data also provide evidence that despite initial allegations reported as concerns of physical, sexual, or emotional abuse, these reports are more likely to be substantiated as neglect.