Assessing the feasibility of using actuarial risk assessment tool to identify risks in child protection cases.

Abstract

The problem: Risk and uncertainty are integral to child protection decisions because of the need to protect children from intrafamilial maltreatment, which is more common than abuse by strangers and occurs in the private sphere of the home. The problem of assessing intrafamilial risks to children following abuse and neglect referrals has received little attention in the UK compared with North America. Improved assessment of such risks would inform child protection plans, enabling risk reduction and improved safeguarding of children. There are currently no UK recommended valid and reliable risk assessment tools available for child protection teams, who rely on unaided professional judgement where information may be incomplete, expertise is variable and the process open to bias. Currently, UK child protection risk assessments are based on professional opinions about the range and weighting of factors associated with families where children have been abused or neglected. Study purpose: Selecting and weighting the most significant predictive factors of risk to children by using actuarial statistical methods is more likely to lead to yield accurate risk ratings. Whilst not perfect predictors, such actuarial tools categorise cases into low or high risk groups better than unaided professional judgements or consensus based assessments. Actuarial risk assessments have not previously been tested for UK child protection work, so this study selected and applied the best available North American tool. Method: The Michigan Family Risk Assessment for Abuse or Neglect (FRAAN) was selected on the basis of its published evidence base and applied to a cohort of UK Serious Case Review reports from Thirteen Counties (SCRs) to identify and rate pre-existing risks before the abusive event. Further analysis of risk factors singly and in combination was extended to compare the Thirteen Counties data with two large recently published SCR studies and one study of UK child homicide perpetrators. Findings: The FRAAN assessment scores correctly identified most cases as High or Intensive Risk. FRAAN performed adequately in identifying these very high risk cases (Sensitivity = 88%) but specificity could not be assessed using these exclusively high risk cases. Most frequent neglect risk factors were inadequate physical care of children, unsupported primary carers and their inability to prioritise the child’s needs over their own. Most frequent abuse risk factors were families not cooperating with a need for parenting improvement, prior abuse incidents, a youngest child aged <6 years and domestic violence. Comparing those cases where children died versus those where they survived, risk factor frequencies for Deceased and Survivor groups were strongly positively correlated, showing no statistical differences between the direction of scores for the two groups. Parental substance misuse was more common in the Deceased group, whilst households where a child had a disability or delinquency problems were both more common in the Survivor group. FRAAN risk scores could not effectively distinguish between fatal and non fatal outcomes. Comparisons between this Thirteen Counties study dataset and a UK study of child homicides and two major recent reviews of SCRs demonstrated no statistical differences between the ages and sexes of the children, the causes of death and parental characteristics, except for in the cases of men with convictions for violence, which were possibly under recorded in SCRs. Key risk factors in all the studies were parental mental illness, chaotic neglectful families and substance misuse. These factors appear to distinguish high risk families where there are further risks of fatal child abuse. Conclusions: The use of formal risk assessment tools is likely to help child protection teams identify cases where extra support, or substitute care, is required to protect children at high risk of further intrafamilial maltreatment. The use of such tools in low and medium risk cases was not tested in this study due to lack of access to a mix of cases. Policy, practice and further research directions are recommended to extend the testing of the FRAAN risk assessment tool.