Predictive Analytics And Child Protective Services
The rapidly changing landscape and continually evolving analytics services and algorithms are helping child protective systems and services in the US, and several other countries identify children at risk of abuse and neglect. While this form of predictive analytics has life-saving potential, it also comes with the risk of removing children from loving, safe households if not used thoughtfully and correctly. Proper training on how to interpret and act on the insights and results these advanced analytics services and solutions produce is essential for the welfare of children and families.
Current studies have indicated that approximately 40 million children below the age of 15 are subject to child abuse and maltreatment every year. Annually, says the WHO, there are over 40,000 reported homicides in children under 15 years of age worldwide—however, since many deaths due to child maltreatment are often wrongfully attributed to accidental means such as drowning or falls, the number of children murdered yearly is actually likely to be much higher. To make matters worse, child welfare services and advocates—especially those in North America—are often overworked and understaffed, meaning that sometimes cases aren’t addressed efficiently or in a timely fashion. For example, a report by the Austin American-Statesman found that in the US state of Texas, more than 30 percent of Child Protective Services investigators leave every year, one out of every six new hires quits within the first six months of their employment, and investigators often are assigned more than 80 cases each (state officials claimed this number was supposed to be much lower, at no more than 20 per person).
Child protection predictive analytics solutions are designed to pick up the slack, in a sense, to more quickly and efficiently identify children who may be at risk
Clearly, there is an issue that needs to be addressed: children are being abused and endangered, and child protection systems are often spread too thin to prevent this on as wide a scale as possible. This is where predictive analytics and advanced analytics services come in. These solutions are not dissimilar to predictive analytics or customer analytics solutions that predict how people will behave towards certain products and businesses; child protection, analytics predicts behavior and foresee possible outcomes, but on a much more serious level. Child protection predictive analytics solutions are designed to pick up the slack, in a sense, to more quickly and efficiently identify children who may be at risk. They analyze large datasets, and calculate probability statistics as well as an overall risk score for individual cases on the basis of things like the child’s age and the age of the parents, whether or not the parents have a criminal history or issues with substance abuse, and if there are any instances of intergenerational abuse within the child’s family. These solutions then flag high-risk cases so that investigators and child welfare services can intervene or review them as soon as possible.
These solutions are not yet perfect. They use data points chosen by humans, and therefore can incorporate societal biases and prejudices into their algorithms; some solutions, for example, use race as an indicator of a child’s risk of being abused. Others simply do not use data points that reflect reality or result in accurate assessments. In Los Angeles County, for example, Child Protective Services recently tested a child protection predictive analytics solution called AURA (short for Approach to Understanding Risk Assessment). In a primary test run, AURA reportedly had a false positive rate of almost 96 percent. Using historical data, it was able to correctly identify 171 high risk children, but wrongfully flagged over 3,800 safe families.
This highlights the importance of critical thinking and human intervention when dealing with analytics in general. Child protective services and investigators need to be able to understand and interpret the results so that children are not wrongfully removed or reviewed. Additionally, they need to make sure that they are using these innovative tools correctly: these predictive solutions are best used as an indicator of which families in a community are most in need of support and assistance rather than which children should be taken from their homes and placed into foster care, and child protective services and systems should not rely on the findings or flags alone to determine whether or not a child will be removed.
When child protective services, systems, advocates, and investigators ensure that they have the capabilities, knowledge, and manpower to properly interpret the results, child protection solutions have the potential to enhance the efficiency of child protection systems and secure the well-being of millions of children around the world. They can save lives and childhoods if we simply apply the proper expertise.