Predictive Risk Modelling to Prevent Child Maltreatment and Other Adverse Outcomes for Service Users: Inside the ‘Black Box’ of Machine Learning

Recent developments in digital technology have facilitated the recording and retrieval of administrative data from multiple sources about children and their families. Combined with new ways to mine such data using algorithms which can ‘learn’, it has been claimed that it is possible to develop tools that can predict which individual children within a population are most likely to be maltreated. The proposed benefit is that interventions can then be targeted to the most vulnerable children and their families to prevent maltreatment from occurring. As expertise in predictive modelling increases, the approach may also be applied in other areas of social work to predict and prevent adverse outcomes for vulnerable service users. In this article, a glimpse inside the ‘black box’ of predictive tools is provided to demonstrate how their development for use in social work may not be straightforward, given the nature of the data recorded about service users and service activity. The development of predictive risk modelling (PRM) in New Zealand is focused on as an example as it may be the first such tool to be applied as part of ongoing reforms to child protection services.

[1]  Dick Schoech,et al.  Expert Systems: Artificial Intelligence for Professional Decisions , 1985 .

[2]  Rhema Vaithianathan,et al.  Vulnerable Children: Can administrative data be used to identify children at risk of adverse outcomes? , 2012 .

[3]  T. Stanley.  Making Decisions: Social work processes and the construction of risk(s) in child protection work , 2005 .

[4]  Daryl J. Higgins,et al.  The limitations of using statutory child protection data for research into child maltreatment , 2004 .

[5]  D. English,et al.  Risk assessment as a practice method in child protective services. , 1994, Child welfare.

[6]  Emily Keddell.  Theorising the signs of safety approach to child protection social work: positioning, codes and power , 2014 .

[7]  R. Barn,et al.  A Contested Identity: An Exploration of the Competing Social and Political Discourse Concerning the Identification and Positioning of Young People of Inter-Racial Parentage , 2005 .

[8]  Ethem Alpaydin.  Introduction to machine learning, 2rd ed , 2014 .

[9]  Nigel R. Parton.  Risk, Advanced Liberalism and Child Welfare: The Need to Rediscover Uncertainty and Ambiguity , 1998 .

[10]  Integrated health and human services information systems to enhance population-based and person-centered service. , 2013, American journal of preventive medicine.

[11]  P. Bourdieu,et al.  Language and Symbolic Power , 1991 .

[12]  E. Munro.  The DAnGers OF inFOrMATiOn shArinG , 2007 .

[13]  Richard Pugh.  Effective Language in Health and Social Work , 1995 .

[14]  V. LeBlanc,et al.  Stress responses and decision making in child protection workers faced with high conflict situations. , 2012, Child abuse & neglect.

[15]  Nan Jiang,et al.  Children in the public benefit system at risk of maltreatment: identification via predictive modeling. , 2013, American journal of preventive medicine.

[16]  D. Fergusson,et al.  Randomized Trial of the Early Start Program of Home Visitation: Parent and Family Outcomes , 2006, Pediatrics.

[17]  Gary King,et al.  Substantiation as a Multitier Process: The Results of a NIS-3 Analysis , 2003, Child maltreatment.

[18]  T. Cross,et al.  Caseworker Judgments and Substantiation , 2009, Child maltreatment.

[19]  M. Kubát.  An Introduction to Machine Learning , 2017, Springer International Publishing.

[20]  Philip Gillingham,et al.  The use of assessment tools in child protection: an ethnomethodological study , 2009 .

[21]  Shoshana Pollack,et al.  Labelling Clients ‘Risky’: Social Work and the Neo-liberal Welfare State , 2010 .

[22]  Desmond K. Runyan,et al.  The dimensions of maltreatment: introduction. , 2005, Child abuse & neglect.

[23]  Bob Lonne,et al.  A comparison of systems and outcomes for safeguarding children in Australia and Norway , 2012 .

[24]  P. Gillingham,et al.  Child Protection Practitioners and Decision-Making Tools: Observations and Reflections from the Front Line , 2010 .

[25]  B. Maclaurin,et al.  Differentiating Between Substantiated, Suspected, and Unsubstantiated Maltreatment in Canada , 2009, Child maltreatment.

[26]  Philip Gillingham,et al.  Decision-making tools and the development of expertise in child protection practitioners: are we ‘just breeding workers who are good at ticking boxes’? , 2011 .

[27]  K. Manion,et al.  eQuiVOcaTiNG OVer The care aND PrOTecTiON cONTiNuum: aN eXPLOra TiON OF FamiLies NOT meeTiNG The ThreshOLD FOr sT aTuTOry iNTer VeNTiON , 2008 .

[28]  Kevin P. Murphy,et al.  Machine learning - a probabilistic perspective , 2012, Adaptive computation and machine learning series.

[29]  Pablo J. Boczkowski,et al.  The Relevance of Algorithms , 2013 .

[30]  Kristen Johnson,et al.  Actuarial risk assessment in child protective services: Construction methodology and performance criteria , 2013 .

[31]  J. Hippisley-Cox,et al.  Performance of the QRISK cardiovascular risk prediction algorithm in an independent UK sample of patients from general practice: a validation study , 2007, Heart.

[32]  E. Munro.  The Munro review of child protection: final report, a child-centred system , 2011 .

[33]  P. Gillingham.  Practitioner perspectives on the family risk evaluation tool : Ab aide to decision making or 'just another form to fill in'? , 2009 .

[34]  Aron Shlonsky,et al.  The next step: Integrating actuarial risk assessment and clinical judgment into an evidence-based practice framework in CPS case management , 2005 .

[35]  Jason F. Jent,et al.  Multidisciplinary Child Protection Decision Making About Physical Abuse: Determining Substantiation Thresholds and Biases. , 2011, Children and youth services review.

[36]  Dick Schoech.  Interoperability and the Future of Human Services , 2010 .

[37]  K. Heffernan.  Social Work, New Public Management and the Language of ‘Service User’ , 2005 .

[38]  Marit Skivenes,et al.  Risk assessment and domestic violence – how do child welfare workers in three countries assess and substantiate the risk level of a 5‐year‐old girl? , 2015 .

[39]  Emily Keddell.  The ethics of predictive risk modelling in the Aotearoa/New Zealand child welfare context: Child abuse prevention or neo-liberal tool? , 2015 .

[40]  Colleen Vojak.  Choosing Language: Social Service Framing and Social Justice , 2009 .

[41]  Peter Scourfield.  A Critical Reflection on the Involvement of ‘Experts by Experience’ in Inspections , 2010 .

[42]  P. Gillingham,et al.  Child protection, risk assessment and blame ideology , 2008, Children Australia.

[43]  R. C. Herrenkohl.  The definition of child maltreatment: from case study to construct. , 2005, Child abuse & neglect.

[44]  P. Bourdieu,et al.  Language and Symbolic Power , 1991 .

[45]  H. DCruz.  Constructing meanings and identities in child protection practice , 2004 .

[46]  A. Lau,et al.  Defining maltreatment according to substantiation: distinction without a difference? , 2005, Child abuse & neglect.

[47]  Ira M. Schwartz,et al.  Computational intelligence techniques for risk assessment and decision support , 2004 .

[48]  Emily Keddell.  Reasoning Processes in Child Protection Decision Making: Negotiating Moral Minefields and Risky Relationships , 2011 .

[49]  John Fluke,et al.  Allegory of the Cave , 2009, Child maltreatment.

[50]  Patricia L. Kohl,et al.  Time to Leave Substantiation Behind , 2009, Child maltreatment.

[51]  P. Gillingham.  Electronic Information Systems in Human Service Organisations: The What, Who, Why and How of Information , 2015 .

[52]  Charles Gene Lyle,et al.  Looks can be deceiving: Using a risk assessment instrument to evaluate the out-comes of child protection services , 2000 .

[53]  F. Stanley,et al.  Child abuse and neglect — is it time for a public health approach? , 2008, Australian and New Zealand journal of public health.

[54]  D. Wennberg,et al.  Case finding for patients at risk of readmission to hospital: development of algorithm to identify high risk patients , 2006, BMJ : British Medical Journal.