Date of Degree
Doctor of Philosophy (Ph.D.)
Community Health and Social Sciences
Maternal and Child Health | Public Health | Social Welfare
child maltreatment, child abuse, neglect, child protection, child welfare, child welfare decision-making
Background: Research has shown that adverse childhood experiences are strongly linked with health outcomes over the life course, and that child maltreatment – generally defined as physical abuse, sexual abuse, emotional maltreatment and neglect – can have an immediate, negative impact on child health and development and a longer-term impact on adolescent and adult health, including the leading causes of morbidity and mortality in the U.S. With such significant consequences, the accurate identification of child maltreatment is critical.
Child welfare caseworkers are charged with protecting the safety and fostering the well-being of children who have been identified as maltreated or at risk of maltreatment. Despite the high-stakes nature of child welfare decisions, there is substantial research showing that they have low reliability and are influenced by numerous worker, agency, and community-level factors. There is a conspicuous gap in the literature, however, regarding whether the first question child welfare workers confront – what happened to this child? – is answered either consistently or correctly. The variation in types of substantiated child maltreatment across the U.S. suggests that this question is answered differently state to state and could result in certain types of maltreatment being under- or over-identified by child welfare systems.
This mixed-methods study sought to better understand the factors that contribute to variation in child maltreatment, by type and across states. This was accomplished through two specific aims that were informed by Baumann’s Decision-Making Ecology (DME) framework, which identifies how case (i.e., child/family), decision-maker (i.e., caseworker), organizational and external factors influence child welfare decision-making and are in turn influenced by child welfare outcomes:
- Determining the relative impact of factors at the family level, child welfare system level, and state level on types of substantiated maltreatment across all 50 states through secondary analysis of national child welfare data.
- Exploring factors at the family level, caseworker level, organizational level and community level that may influence the identification of child maltreatment through key informant interviews of child welfare system stakeholders.
Methods: The first aim involved cross-sectional analyses of data from the National Child Abuse and Neglect Data System (NCANDS) from federal fiscal year 2016. The outcome variable was substantiated maltreatment, which had four values: 1) substantiated physical abuse, 2) substantiated sexual abuse, 3) substantiated emotional maltreatment, and 4) substantiated neglect. (Other types of maltreatment, which varied across states, were excluded from the analyses.) The predictor variables included child/family, child welfare system and state-level variables. Univariate, bivariate and multivariate analyses were conducted; multinomial logistic regression (MLR) models were used for the multivariate analyses, which allowed for the simultaneous comparison of the associations between the categorical outcome variable (substantiated maltreatment) and the predictor variables.
The second aim involved key informant interviews (KIIs) with child welfare administrators, child protective caseworkers, family court judges, parent attorneys, and parents with past child welfare system involvement in two jurisdictions in two states (four locations in total). Interview guides covered the following domains: Identifying maltreatment; substantiating maltreatment; parents’ role in identifying maltreatment; the legal system’s role in maltreatment decision-making; and the broader context of child welfare decision-making. Interviews were recorded, transcribed, coded and thematically analyzed; themes were then organized into several domains.
Results: In Aim 1, univariate analysis of child maltreatment types showed a large amount of variation across states. Out of all instances of substantiated maltreatment, the proportion determined to be neglect ranged from 2%-92%, the proportion determined to be physical abuse ranged from 3%-53%, the proportion determined to be sexual abuse ranged from 2%-50%, and the proportion determined to be emotional maltreatment ranged from 0%-41% across all states.
The largest associations from the MLR models involved domestic violence, substance abuse and emotional maltreatment. Specifically, controlling for all other factors, children with a caretaker identified as a domestic violence victim were 5.9 times more likely to be determined to be emotionally abused than neglected, 5.5 times more likely to be determined to be emotionally abused than physically abused, and 19.1 times more likely to be determined to be emotionally abused than sexually abused compared with children who did not have a caretaker identified as a domestic violence victim. Controlling for all other factors, children with a caretaker with a substance abuse problem were 2.55 times more likely to be determined to be emotionally abused than sexually abused compared with children who did not have a caretaker with an alcohol/drug abuse problem.
In Aim 2, analysis identified several themes that corresponded with domains of the DME: family factors such as income and history with the child welfare system; caseworker factors such as managing their dual (social worker-investigator) role; agency factors such as how state maltreatment definitions are operationalized at the local level; and external factors such as how caseworkers anticipate the family court process during the investigation. Additionally, several themes were identified that relate to the relationship between parents and child protective caseworkers, and how that relationship impacts maltreatment identification specifically and child protective investigations generally.
Analysis of the KIIs also captured variation in how different types of child maltreatment are understood, identified and managed by child protective staff and agencies. Distinctions were made between types of maltreatment that can be ‘seen’ and types that require further assessment and/or evidence to substantiate. KII participants also identified discrepancies between how certain types of maltreatment are defined in state law compared with how they are operationalized by agencies and/or understood by the general public. Emotional maltreatment and domestic violence were identified as events that raise particular challenges for child protective staff.
Discussion: Decision-making in the child welfare system is a complex process that, as previous research has demonstrated and this study confirms, is influenced by multiple factors. While this study supports the overall premise behind the DME framework, it also revealed that as it stands the DME framework does not fully capture the dynamic way different levels of the ecology intersect with each other (specifically through the relationship between parents and caseworkers) and how such interplay can influence the decision-making process on its own.
Another finding from this study relates to the association between family poverty and child maltreatment, particularly neglect. While the MLR analyses found that children from poor families are more likely to have been identified as experiencing neglect than other types of maltreatment compared with children from higher income families, the KII findings raise questions about the factors that drive these differences and whether the differences overstate the variation between poor and higher-income children’s experiences. Additionally, the KIIs and additional review of states’ statues revealed that many states’ definitions of neglect are extremely broad and encompass actions (such as extreme corporal punishment and sexual touching) that many would consider to be other types of maltreatment.
The KIIs revealed that domestic violence may be the family risk factor that is managed with the most variation across child welfare agencies, and at times among staff within the same agency, with different individuals categorizing exposure to domestic violence as neglect, physical abuse and/or emotional maltreatment. While the largest associations found in the MLR analyses involved emotional maltreatment, EM is by far the least identified and substantiated type of maltreatment in all but a handful of states, and in the KIIs emotional maltreatment was described as being at times challenging to identify and almost always difficult to prove. This is concerning given the research demonstrating the high prevalence of emotional maltreatment across the population, links between EM and a range of physical and mental health problems across the life course, and a lack of focus on EM in evidence-based parenting programs, likely because child welfare data do not indicate a need for such a focus.
The KIIs identified two agency-level factors that appear to influence the maltreatment decision-making process and resulting maltreatment data. The first is the presence of Alternative Response (AR) protocols, which by removing families with lower levels of risk from the investigation process affect maltreatment data used by researchers, policy makers and administrators. The second agency-level factor that was identified was the maltreatment screening process. Data suggest that hotline staffs’ decision-making may be influenced by the same kinds of agency-level factors that influence child protective caseworkers’ decision-making, and that those factors can result in variation in the categorization of maltreatment across jurisdictions that are operating under the same law. Data also suggest that the categorization done by hotline staff plays a critical role in driving the remainder of the maltreatment-related decision-making.
Based on this study’s findings, recommendation for child welfare administrators include ensuring that accurate and complete data are entered into administrative systems; clarifying what actions constitute child maltreatment for both families and the broader community; identifying and addressing decision-making and parent engagement processes that are coercive or lack transparency; examining variation in decision-making across staff, including hotline staff; and tracking and reporting maltreatment determinations across subpopulations to ensure consistency and equity in decision-making. Recommendations for policy makers include improving data mapping and data entry practices; identifying and addressing vagueness, variation and misalignment of maltreatment definitions; and addressing specific concerns related to neglect, emotional maltreatment and domestic violence. Recommendations for researchers include accounting for the variation in child maltreatment definitions in child welfare research; further exploring how neglect, emotional maltreatment and domestic violence are managed by child welfare agencies; and further exploring the role of hotline staff in the identification and categorization of child maltreatment.
Conclusion: While past child welfare studies have largely assumed that the first decision that is made by child welfare staff is correct, this dissertation research demonstrates that child maltreatment decision-making is a complex, emotional and often subjective process that is influenced by multiple factors at the family, caseworker, agency and community levels, both separately and in combination. As a result, available child welfare data do not accurately or completely capture maltreatment experienced by children that become known to the system.
The reliability of data matters both on the individual level, where it affects the type and effectiveness of support provided to both children and parents, and on the aggregate level, where it drives policy, funding, and future research and program development. Child welfare systems should be more curious about what story their data tell – about their state’s legal definitions, about their own practices, and about the families they serve – and be more transparent with their data and practices. Maltreatment-related definitions and data should be shared with and explained to the public, so that families, community members and political leaders better understand children’s and parents’ experiences and can partner with child welfare agencies to ensure that interventions are applied fairly, appropriately and effectively.
Tullberg, Erika, "What happened to this child? Identifying factors that influence the identification and categorization of child maltreatment in the United States" (2019). CUNY Academic Works.