The story of big data is essentially the same each time it’s told, but the results continue to captivate. In this particular case, the Florida Department of Children and Families (DCF) is using data analysis to identify children and families most at risk, and thus inform how time and money is allocated. When the DCF started this project two years ago, the goal was to see fewer dead children -- and that's what the department says is happening.
A $100,000 contract forged with SAS last fall led to a detailed report that is now shaping how DCF achieves its mission. The SAS report provided an analysis of data taken from 2007 to mid-2013 that came from two case management systems. The findings of the report matched what the department already knew from less-scientific research conducted over the past two years, said Interim DCF Secretary Mike Carroll.
“What the deeper-dive look did was actually helped us to refine what we had already started,” he said. “This type of work and this type of data is going to focus our work in a way we haven’t been able to before, and I hope it leads to better use of resources in that we can prioritize the use of our resources in evidence-based practice where we can get the best outcomes for our most vulnerable kids.”
The SAS report helped DCF identify what the highest-risk children looked like on paper, creating a detailed profile. “We needed to understand a lot more of the common factors in those cases," Carroll said, "and we needed to be able to take that information and refine what we were doing from a case practice standpoint to see if we couldn’t intervene in a more effective way to prevent some of those child deaths."
The report also led to realizations that will inform how the department allocates its resources. For example, the report found that in-home services provided to families reduced the odds of child fatality by 90 percent. Having a statistical backing to what was already presumed to be true, Carroll said, means they will have a good foundation for doing more of that kind of work.
The new profile method of operation was implemented statewide in January, Carroll added, and it’s too soon to know if the impact will be fewer dead children statewide -- but they have a reason to expect success. The department began piloting this project two years ago in Hillsborough County, drawing from their own data analyses, which were less-scientific, but still allowing the department to form a profile of high-risk children, he said. In the past two years, he said, there have been no child deaths in Hillsborough County matching that high-risk profile the department was monitoring.
“Knowing this information has helped us redesign how we go about delivering our services,” he said. “I think the early returns on our investment in this have been very good because I don’t think you can measure a child’s life in dollars. And, to me, the fact that we have seen that we’ve turned that around in Hillsborough County is a huge move forward."
The next step for DCF will be identifying what kinds of interventions or combinations of interventions have the highest likelihood of success, Carroll said, adding that this type of data analysis has value to all the various county and local service providers in government. “We’re only a piece of the puzzle when we deal with these families,” he said. “It’s really my hope that someday we’re all looking at the information that comes out of this. I think it will really help to focus our efforts in local communities.”
When it came to giving SAS the data it needed, he said doing so wasn’t necessarily difficult, but involved different work than the department was used to. The data came from several different systems, and the DCF had to clean it, check it for consistency, and ensure accuracy before handing it over to SAS, he said.
Any state department looking for a similar type of data analysis can do so, said SAS Analytics Expert Albert Blackmon. Those state departments with federally-funded case management systems can use their machine-readable data for analysis. For states with paper-based systems, the department would need to put the data into some kind of digital format, Blackmon said – either way, it’s worth the effort.
“The largest trend you’re seeing -- and this is not just in the child welfare space – is as government agencies recognize the value of the data they have, just like corporations have, they realize they have this incredible asset in data,” Blackmon said. “If they can get to it, arrange it properly and utilize it, they can gain a ton of insight and make their policy and, even more so, provide transparency to the legislature, which is what a lot of these folks are looking for.”
The feedback on this project from Florida state officials and legislators has been great, Blackmon said. “On the DCF side, I’d say that they are starting to position some programming, such as more investigators, pairing investigators, going out to meet with families when they do a home visit,” he said. “On the legislative side, they’re looking for the transparency. They want the information so that it’s easily accessible and the public can access it.”
This first analysis used just two data sources, and the goals are bigger for the next go-around, Blackmon said. “Our report highlights up to 24 [data sources] that could be used,” he said. “I think one of our big goals is taking all those data sources eventually and make this an operational system for DCF, providing that child-centric view, risk-scoring children and helping case-workers to prioritize cases and giving them the tools they need to take care of their kids in the best way possible.”