Evidence-Based Implementation of a Child Welfare Informatics (CWI) System Loc H. Nguyen, Dr.P.H., M.S.W. Nguyen Page 1
Introduction Child Welfare Informatics (CWI) is the integration of child welfare practice with information technology and computer science to proactively analyze child welfare information in order to develop and implement best practices (Nguyen, 2007). The CWI process can be thought of in five stages, with each stage evolving from the previous stage. In each of the stages, there are five possible areas that may influence overall child welfare outcomes. These five factors include societal factors (e.g., economy, employment, health of population, etc.), family/child factors (e.g., education, health, mental health of child, etc.), non-child protective services factors (CPS) (e.g., non-protective social services, mental health services, education, etc.), worker education and training factors (e.g., social work vs. non social work degree, new hire worker training, in-service training, etc.), and human resources factors (e.g., hiring process, performance evaluations, worker retention, etc.). Within each of these factors, there can be metrics that influence prior to the time of child welfare services (i.e., pre-cws) or metrics that lead to current child welfare services (CWS). Currently, the state of CWI can be described as being in Stage 1 (Figure 1). There is a great degree of disconnect between pre-cws and CWS as well as a great degree of disconnect between the factors, though some research exists in each factor. For example, Sedlak and Broadhurst (1996) noted that children from low income families (less than $15,000) are 22 times more likely to be maltreated, 15 times more likely to be sexually abused, and 56 times more likely Nguyen Page 2
to be educationally neglected than children from moderate income families (greater than $30,000). However, there is no research on how pre-cws societal factors are related to current societal factors affecting CWS (e.g., how do low income families fare while in the system). The same disconnect between pre- CWS metrics and current CWS metrics is true with family/child factors, non-child welfare services, worker education and training factors, and human resource factors. Further, there is very little research or applications that link any of the current knowledge between the five factors. Thus, if a child is taken into temporary custody, most likely the placement of that child will be because of the next available space on a worker s caseload, rather than what is best for that child. The assignment of the case management of that child is because of the next available vacancy on a caseload rather than identifying what workers would have the experience and background to provide services to that child. In essence, the child welfare outcomes are currently driven only by family/child factors during CWS. Most importantly, there is very little research that utilizes data from Los Angeles County, and the research that is available are from jurisdictions that look very different from Los Angeles County. In the second stage of the CWI process (Figure 2), we being to see some links between the Pre-CWS indicators and the current CWS indicators, but we do not quite see the links across the factors. There is very little research among some of the factors and no research on other factors. One area that has been looked at is disproportionality or the over-representation of children from certain backgrounds. In 2010, only 8% of the Los Angeles County population were Nguyen Page 3
African-Americans. However, African-American children make up about 19% of the reported allegations for child abuse and neglect (i.e., pre-cws) in Los Angeles County, and nearly 34% of children in foster care (i.e., current CWS) and under the supervision of DCFS (Needell et. al., 2007). In the third stage of the CWI process (Figure 3), we begin to better utilize information by linking not only the Pre-CWS indicators and metrics to the current CWS indicators and metrics, but also link information between the five factors. The child welfare outcomes are still reactive, but at least there is the beginning of the understanding of how these five linked factors lead to outcomes. In fact, different partnerships within the County would begin to comprehensively examine current and historical information using accepted research and practice methodologies to link some of the information. We would evaluate if there is any link between societal factors (e.g., economy) and family/child factors (e.g., the over-representation of certain ethnicities in foster care). We can also look at whether education and training factors (e.g., worker training on specific social work practices) are linked to human resources factors (e.g., worker performance through performance evaluations). In fact, we already have some research in this area. For example, there is research to suggest that newly-hired Children s Social Workers (CSWs) who have graduate social work degrees tend to score higher on social work knowledge tests before and after trainings than those with non-social work degrees (Inter-University Consortium, 2010). For the fourth stage of the CWI process (Figure 4), we begin to better merge the pre-cws and CWS metrics and the five factors. We leverage Nguyen Page 4
research findings on societal factors and family/child factors pre-cws to understand that there are certain characteristics that may lead a child to be in the system longer than need be (preventable long-term stays), or a child that could be at high risk for re-abuse or re-neglect (preventable re-abuse/re-neglect). On the worker education/training side, we leverage research informed data to find out which workers would likely make poor child welfare workers (preventable future poor child welfare workers). For the HR process, we identify characteristics of those who are likely to leave the agency (preventable worker turnover). Then, we begin to create ways to mitigate adverse child welfare outcomes through a proactive process of the merge. Indeed, if we know certain children may have adverse outcomes because of known domestic violence issues, we assign the case management of those children to workers who have had the education and training who have the best chance to address the needs of those children because the workers have had the education and training, especially in domestic violence. Research outside of County gives us some preliminary clues for Stage 3. For example, research from the National Council on Crime and Delinquency (2006) indicates that California counties with high annual worker turnover (greater than 25%) have twice the number of overly long investigations (>60 days) and more than 250% times the number of substantiated re-abuse allegations within three months than those California counties with low annual worker turnover (less than 10%). Interestingly enough, information from the new-hire CSWs from the first year of the IUC Academies in FY1991-FY1992 Nguyen Page 5
indicate that they have had less than a 5% annual turnover, and that about 45% of them were still employed with the County after more than 18 years. Stage 5 of the CWI process involves the ability to leverage the information from Stage 4 to influence the outcomes before the next child comes into the system. Indeed, because of the evidence-based CWI, we will be able to prevent children from even entering the child protective services system. For those that enter the child protective services system, we will have a fully integrated CWI process (Stage 4) which maximizes the reduction of their potential adverse outcomes. For example, we will be able to identify characteristics of groups of children in the community who have a high likelihood of being abused or neglected, and DCFS would work with other County Departments and community-based agencies to employ targeted prevention programs to prevent the abuse or neglect. From these same communities, we will have identified the characteristics of those individuals who would become excellent social workers, and DCFS would work with other County Departments and community-based agencies to employ targeted programs to recruit these individuals as early as high school. From high school, the programs would create incentivized opportunities with clear tracts for these individuals to ultimately become DCFS employees. For the children who do end up coming into contact with DCFS, we will have developed a model to predict how well these children will fare in the system, including the ability to identify children who are likely to stay in the system longer, who are likely to be re-abused while in the system, etc. We will also develop a model that will identify the workers with the background, Nguyen Page 6
education, and expertise to be able to be in the best position to mitigate the potential adverse outcomes, and be assigned that child s case. There are estimates of the incredible economic impact of child abuse and neglect. In fact, the economic estimates of child abuse and neglect in the U.S. in approximately $108.0 billion in 2010 dollars, of which only about 25% is due to direct child welfare service costs (Wang & Holton, 2008). For example, in Los Angeles County, the local public child welfare agency spent $1.8 billion in direct child welfare service costs in one fiscal year (Chief Executive Office, 2010). The total economic impact of child abuse and neglect in Los Angeles County is, conservatively, at least $7.0 billion. This impact will be lessened with a better integrated evidenced-based CWI system. Future In the next few years, we should be able to leverage the large amount of information that is available in the any state s child welfare data base to develop a community profile of what child abuse and neglect may look like in the community. From this profile, we should be able to develop prevention programs that begin to address and ameliorate child abuse and neglect in the community. For those children that still come to the attention of a child welfare, multivariate modeling will assist in determining the potential outcomes of such children. From the worker perspective, and using the latest in technology (e.g., tablets used in the field), we would be able to determine the background, training, education, and expertise of a worker. Ultimately, we would be able to assign a child with specific needs to a worker with the expertise to address those needs. Nguyen Page 7
For example, as soon as a worker completes a training course, they would be able to take a competency test about the course on a computer tablet. That information would be immediately uploaded and analyzed in concert with current child welfare databases. Predictive analytics would identify which workers have the potential to address a child s needs. Information would be sent via email to a worker s supervisor to make the final call. The goal is to develop a real-time application of an evidence-based datadriven decision-making model at a public child welfare agency that is fully supported by predictive analytics. One of the first steps is to develop a proof-ofconcept (POC) which would examine the potential of whether this modeling could be done with one metric in a pilot office. One generation from the POC would determine the potential of modeling one metric to the entire agency. Two generations from the POC would determine the potential of modeling multiple metrics to the entire agency. Three generations from the POC would see the most exciting potential of a CWI system. A CWI system will identify incredibly high-risk children (i.e., children at substantial risk for a serious incidence or even death). A child welfare agency would be able to assemble high-risk multi-disciplinary, multi-agency response teams. An emergency response (ER) public child welfare worker sent into the field with tablet and blue-tooth capabilities linked to the support team at their regional office. Office team members can constantly update ER worker with information. For example, mental health team member identifies mental health issues and provides information to ER worker in order to help guide the workers Nguyen Page 8
questions. Furthermore, the ER worker will provide new information to team members when interviewing people. For example, the ER worker finds out mother was a divorcee two years ago with non-child protective services contact. A social services team member identifies more children from previous marriage that were not on the original referral, and in turn, guides the ER worker to ask specific questions about those children. Another example is that ER worker finds out parent moved to the current state here from another state three years ago. A team member contacts child protective services from another state and finds significant domestic violence and medical issues that lead to child abuse in that other state. Team members, including a public health team member, then guides the ER worker to ask further specific questions. Once information is added to the child welfare database, the CWI system begins to identify future workers and services to best address the child s needs. Nguyen Page 9
Author Information At the time of the Symposium, Loc H. Nguyen, Dr.P.H., M.S.W., was the director of the Inter-University Consortium (IUC). The IUC is a partnership between the social work programs in six local universities (California State University, Dominguez Hills; California State University, Long Beach; California State University, Los Angeles; California State University, Northridge; University of California, Los Angeles; and University of Southern California) and the public agencies in Los Angeles County. The biggest partnership is with the Los Angeles County Department of Children and Family Services (DCFS). For nearly 21 years, the IUC has provided training to over 125,000 staff representing more than 2.4 million staff-training hours on a myriad of issues, including, but not limited to general child protective services, safety and risk, child development, mental health, and disproportionality. Nguyen Page 10
References Chief Executive Office. (2010). The County of Los Angeles Annual Report 2009-2010. CEO: Los Angeles, CA. Available at http://file.lacounty.gov/lac/cms1_146766.pdf. Nguyen, L.H. (2007). Child welfare informatics: A new definition for an established practice. Social Work, 52:361-363. Inter-University Consortium. (2010). Inter-University Consortium Department of Children and Family Services Training Project Year XVIII annual report (2008-2009). National Council on Crime and Delinquency. (2006). Relationship between staff turnover, child welfare system functioning, and recurrent child abuse. Cornerstone for Kids: Houston, TX. Needell, B., Webster, D., Armijo, M., Lee, S., Cuccaro-Alamin, S., Shaw, T., Dawson, W., Piccus, W., Magruder, J., Exel, M., Smith, J., Dunn, A., Frerer, K., Putnam Hornstein, E., Ataie, Y., Atkinson, L., & Lee, S. H. (2007). Child welfare services (CWS/CMS) reports. Retrieved from http://cssr.berkeley.edu/cwscmsreports/. Sedlak, A.J., and Broadhurst, D.D. (1996). Third national incidence study of child abuse and neglect (NIS-3).U.S. Department of Health and Human Services [HHS], National Center on Child Abuse and Neglect, Washington, DC. Wang, C. and Holton, H. (2008). Total estimated cost of child abuse and neglect. Prevent Child Abuse America: Chicago, IL. Available at http://www.preventchildabuse.org/about_us/media_releases/pcaa_pew_ec onomic_impact_study_final.pdf. Nguyen Page 11
Figure 1 The first stage of a Child Welfare Informatics (CWI) system. Nguyen Page 12
Figure 2 The second stage of a Child Welfare Informatics (CWI) system. Nguyen Page 13
Figure 3 The third stage of a Child Welfare Informatics (CWI) system. Nguyen Page 14
Figure 4 The fourth stage of a Child Welfare Informatics (CWI) system. Nguyen Page 15
Figure 5 The fifth stage of a Child Welfare Informatics (CWI) system. Nguyen Page 16