A Recipe for Found Data

September 21, 2015 | Andrea Bogie, Researcher

bogieandrea

Have you ever gone to the grocery store and thrown a bunch of random items in your cart without thinking about what you were going to make? When you got home, were your cupboards full but you could not figure out what to make for dinner? Maybe you had what you needed to make a really great breakfast or some delicious snacks, but what you really needed to make was dinner for four.  

Have you ever gone to the grocery store and thrown a bunch of random items in your cart without thinking about what you were going to make? When you got home, were your cupboards full but you could not figure out what to make for dinner? Maybe you had what you needed to make a really great breakfast or some delicious snacks, but what you really needed to make was dinner for four.  

There is a more efficient and logical way to shop to make sure you have the ingredients for a meal that is exactly what you need—decide what you want to make and then go to the store to buy the ingredients. 

Just like there are different ways to approach grocery shopping, there are different ways to approach social service analytics. You can take all available data and see what types of questions you can answer, or you can ask a question and then use the data available to answer that question.  

Just like with shopping, if you start with the pieces, you may end up with answers to questions that are not relevant to practice or aren’t feasible to implement due to limited resources. For example, you may end up answering questions about which children are more likely to experience education difficulties when you need to answer questions about which children are more likely to experience being removed from their homes. It makes more sense to start with a question and see if you have the ingredients (i.e., data) to answer it. In 2010, Los Angeles County did just that. 

The LA County Department of Children and Family Services (DCFS) was already using Structured Decision Making® (SDM) assessments to help workers make decisions at different points during a child welfare case. One of the assessments they were using was an actuarial risk assessment to classify families by their likelihood of becoming involved with the department again in the future so that DCFS could direct resources to families at the highest risk.   

DCFS was also working with youth who were already involved in both the child welfare and juvenile justice systems as part of the county’s crossover program. While the crossover effort was working well for youth who were already in both systems, DCFS leaders were interested in taking a step toward preventing youth from becoming involved with the juvenile justice system. DCFS approached NCCD with a question: is it possible, using data already collected by workers, to create an assessment similar to the maltreatment risk assessment that can help us identify youth in the child welfare system who are at the highest risk of becoming involved with the juvenile justice system? DCFS staff were interested using the assessment to target services aimed at meeting the needs of those youth that put the youth at the greatest risk of delinquency. 

To begin to answer this question, NCCD researched prior studies that identified risk factors for delinquency. With those factors in mind, we then queried California’s Child Welfare Services Case Management System (CWS/CMS) and NCCD’s SDM® database to gather all available information about a cohort of children receiving ongoing child welfare services in Los Angeles County. The information included characteristics about the children and their families, for example, child age, prior child welfare history (allegations, investigations, cases open for services, child removals), substance use, education, domestic violence, and involvement with mental health services, among others. With some additional outcome data from the LA County Probation Department, NCCD was able to examine whether there were relationships between each characteristic and/or combinations of those characteristics and subsequent arrests and/or adjudications.  

After analyzing the available data, NCCD was able to answer the question that DCFS had asked: YES, it is possible to identify child welfare-involved youth who are at the highest risk of subsequent delinquency using data already collected in the field. The analysis resulted in development of a screening assessment to classify youth by their risk of subsequent delinquency.  

By starting with a question, NCCD was able to focus data analyses on providing an answer that was directly related to a problem that DCFS wanted to solve. So rather than trying to scrape a reasonable meal together from whatever you happen to have in the cupboard, be purposeful with the data “ingredients” you buy and make the analytics meal you really want to eat!