Citation

Søbjerg LM. Child Abuse Rev. 2022; 31(4): e2749.

Copyright

(Copyright © 2022, John Wiley and Sons)

DOI

10.1002/car.2749

PMID

unavailable

Abstract

Including statistical models in assessment of referrals about children at risk holds promising aspects. This article presents results from an empirical pilot study developing and testing a statistical decision support system in the form of a predictive risk model.The study involved mixed methods and included 208 referrals assessed by 13 social workers in two municipalities in Denmark. The research design involved comparing risk scores provided by the social workers before and after they received a score generated by the model. Social workers' perception of the model was studied through qualitative interviews before, during and after the test. The quantitative results of the study show that presenting a statistical risk score to social workers only marginally changes their initial assessment of a referral. In contrast, the qualitative findings appeared to show that social workers are ready to include statistics in the decision-making process, but that some elements need to be considered. These include the definition of risk being used and the linking of this definition with the available data as well as the needs of the social workers. It is important to remember that a statistical model can support - not define - the assessment. Statistical decision support systems based on predictive risk models might prove rewarding in providing more accurate and homogeneous assessments in child protection. Important issues to consider when developing a model are the definition of risk, the data available to the model and communication of the risk calculated. Risk assessments are based on uncertainties and while statistics might limit the uncertainty, data do not eliminate ambiguity of risk assessment. Social workers should engage in the development of statistical tools, but it is important to acknowledge that decisions should always be made by the professional, not the model.

Language: en

Keywords

child abuse; child maltreatment; child protection; information sharing; multidisciplinary working