Contents lists available at ScienceDirect Children and Youth Services Review journal homepage: www.elsevier.com/locate/childyouth Discussion Risk assessment and decision making in child protective services: Predictive risk modeling in context Stephanie Cuccaro-Alamin a,b , Regan Foust a , Rhema Vaithianathan c , Emily Putnam-Hornstein a,b,⁎ a Children's Data Network, Suzanne Dworak-Peck School of Social Work, University of Southern California, USA b California Child Welfare Indicators Project, School of Social Welfare, University of California at Berkeley, USA c Centre for Social Data Analytics, Auckland University of Technology, New Zealand ARTICLE INFO Keywords: Predictive risk modeling Risk assessment Child welfare Actuarial models ABSTRACT In an era in which child protective service agencies face increased demands on their time and in an environment of stable or shrinking resources, great interest exists in improving risk assessment and decision support. In this article, we review the literature and provide a context for predictive risk modeling in the current risk assessment paradigm in child protective services. We describe how predictive analytics or predictive risk modeling using linked administrative data may provide a useful complement to current approaches. We argue that leveraging technology and using existing data to improve initial triage and assessment decisions will enable caseworkers to focus on what they do best: engaging families and providing needed services. 1. Introduction In 2014, the U.S. child protective services (CPS) system received 3.6 million allegations of child abuse and neglect, involving an estimated 6.6 million children. Of these, approximately 3.2 million children ex- perienced an investigation or received an alternative response and an estimated 702,000 were found to have been victims of abuse or neglect. From there, 21% (n = 147,462) entered foster care (U.S. Department of Health and Human Services, 2016). Thus, every day through a series of decisions often made by multiple individuals, children and families are referred to CPS and then triaged. Yet a comprehensive understanding of how most effectively to screen and then serve children and their fa- milies is still emerging. Correctly ascertaining levels of acute and chronic maltreatment risk among the millions of children referred to CPS agencies each year is no easy task, nor is matching and tailoring services to meet the needs of these children and families. The risk factors for child maltreatment have been well documented for decades. Multiple individual, family, and community risks are often present for these vulnerable children, including poverty (Gil, 1971; Jones & McCurdy, 1992; Pelton, 1989, 1994; Sedlak & Broadhurst, 1996; Wolock & Horowitz, 1979) and its many correlates, such as female-headed families (Brown, Cohen, Johnson, & Salzinger, 1998; Gelles, 1989, 1992; Gillham et al., 1998; Sedlak & Broadhurst, 1996), low parental education (Brown et al., 1998; Kotch et al., 1995; Zuravin & DiBlasio, 1996; Zuravin & Grief, 1989), unemployment (Gelles, 1989; Gillham et al., 1998; Kotch et al., 1995), welfare receipt (Brown et al., 1998; Jones & McCurdy, 1992; Needell, Cuccaro-Alamin, Brookhart, & Lee, 1999; Paxson & Waldfogel, 2002), and impoverished neighborhoods (Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2007; Coulton, Korbin, Su, & Chow, 1995; Drake & Pandey, 1996). Characteristics observable and universally collected at the time of birth also have been documented as related to risk of CPS referral, including early maternal age, late or absent prenatal care, low birth weight, birth ab- normalities, and positive toxicology (Hussey, Chang, & Kotch, 2006; Putnam-Hornstein & Needell, 2011; Stith et al., 2009). Higher rates of CPS reporting also have been found among Black and Native American children relative to their White and Hispanic counterparts (Ards, Myers, Malkis, Sugrue, & Zhou, 2003; Drake, Lee, & Jonson-Reid, 2009; Font, Berger, & Slack, 2012; Putnam-Hornstein & Needell, 2011). Although child maltreatment is found disproportionately among non-White and teen-parent families, considerable evidence suggests that socioeconomic status also may confound these relationships because minorities and adolescent parents are disproportionately likely to be single and poor (Bolton, Laner, & Kane, 1980; Garfinkel & McLanahan, 1986; Gil, 1971; Kinard & Klerman, 1980; Saunders, Nelson, & Landsman, 1993). Despite the wealth of literature regarding risk factors for child maltreatment, the accurate identification of referred children for whom the threat of maltreatment is most immediate and consequential has proven difficult. High rates of subsequent maltreatment referrals among children with initially unfounded allegations (Drake, 1996; Fluke, Shusterman, Hollinshead, & Yuan, 2005; Jonson-Reid, Drake, Chung, & Way, 2003) and increased risk of child maltreatment deaths http://dx.doi.org/10.1016/j.childyouth.2017.06.027 Received 12 January 2017; Received in revised form 8 June 2017; Accepted 8 June 2017 ⁎ Corresponding author at: Suzanne Dworak-Peck School of Social Work, University of Southern California, 1150 S. Olive St., Ste. 1400, Los Angeles, CA 90015, USA. E-mail address: ehornste@usc.edu (E. Putnam-Hornstein). Children and Youth Services Review 79 (2017) 291–298 Available online 15 June 2017 0190-7409/ © 2017 Elsevier Ltd. All rights reserved. MARK