Back to top

Applying Predictive Analytics to Child Maltreatment Response and Prevention

Monday, October 27, 2014

The below article by Daniel Heimpel was published in the The Chronicle of Social Change on October 27, 2014: 

How can big data help improve child maltreatment response and prevention?

At 18, Miranda Sheffield got the shock of her life when a pregnancy test came up positive.

Sheffield’s first reaction was “devastation, disappointment, depression, tragedy: any and everything that could be negative.”

She was a track star and a senior at Pomona High School, set in the vast urban tracts surrounding Los Angeles. Unlike anyone else in her family, Sheffield was on her way to college.

Even if she didn’t fully realize it as a pregnant teen in 2004, the prospects for her baby girl were less than optimal. She was single, on public health insurance and had been in foster care for almost half of her quickly fading childhood.

In 2013, Emily Putnam-Hornstein and a team of researchers from the University of Southern California and the University of California-Berkeley released a groundbreaking study that tracked what happens to the children of young mothers who had experienced the child protection system in Los Angeles.

By age five, 40 percent of children with mothers who had been suspected victims of abuse were reported themselves. For sons and daughters of moms who had been confirmed victims, 18 percent would wind up as confirmed victims themselves.

“She had this face of disappointment,” Sheffield said. But Ms. Lisa wiped it clear and simply asked: “How are we gonna move forward?”
 
Sheffield had the baby.
 
Applying Risk Models after Abuse is Suspected
 
Los Angeles County’s blue ribbon commission issued its final report in April, where it recommended that the county “immediately implement the process used by Eckerd in Hillsborough County, Florida and in other industries to achieve remarkable safety results.”
 
Richard Martinez, the superintendent of Pomona Unified School District, was a member of the commission and also serves on the blue ribbon transition team, which has worked to move the county along in implementing the commission’s recommendations.
 
Martinez said the transition team hasn’t made any headway on seeing a predictive analytics tool like Eckerd’s is implemented.
 
“That is something we truly need to implement from my perspective,” Martinez said. “To predict potential risk based on family history or risk factors, I think that would be critical for this community.”
 
The Eckerd model, called Rapid Safety Feedback, was developed by looking at 1,500 cases of child maltreatment in Hillsborough County Florida, where Eckerd is the private provider operating child protection services. Eight of those cases were fatalities.
 
Eckerd found that fatality prevention had to focus on children under the age of three who are left in their home. Every child who had died in the previous two years fell into that category.
 
Eckerd also created a nine-point risk assessment checklist that child protection workers must fill out. If the issues aren’t resolved rapidly, the case is moved up from the manager to the director level so that an adequate response is forthcoming.
 
While the blue ribbon commission was showing off the Eckerd model, DCFS was inching ever closer to launching a risk prediction tool of its own. As far back as 2008, the department started developing a program called Approach to Understanding Risk assessment, or AURA, according to Francesca LeRúe, who heads the 50-person Risk Management Division.
 
AURA analyzed risk factors for cases that were deemed critical incidents, near fatalities or fatalities over a three-year period starting in 2011. The results from 2011 and 2012 yielded a risk score, which was then tested against the 2013 data.
 
The risk scores were helpful in indicating which children would end up with the most serious cases of abuse, according to Le Rúe.
 
LeRúe, who spent a good part of her 26 years at DCFS investigating child abuse, sees AURA as an invaluable tool for child protection workers. But even so there are limitations.
 
“People ask me, why don’t you just take that child away from that family?” she said.  “Risk of maltreatment doesn’t mean maltreatment has happened. You detain a child if there is an immediate risk and can’t develop a safety plan. You can’t detain, but more services can be directed – absolutely.”
 
Armand Montiel, spokesman for Los Angeles County’s Department of Children and Family Services (DCFS), agrees that using data to determine risk is critical, and also thinks that predictive analytics’ real potential lies before a call of abuse is ever made.
 
“Instead of predictive analytics, lets do preventive analytics,” Montiel said in an interview. “DCFS is not the proper agency for social engineering. The school system, mental health, public health, probation, law enforcement, they all see these families before we do.”
 
Offering Services Before Abuse is Reported
 
Margaret Henry is a juvenile dependency judge in LA County, and is a member of the blue ribbon transition team. Judge Henry remembers a presentation given by Putnam-Hornstein before a group of child protection leaders last year.
 
“There was dispute about labeling people,” Henry said. “Some people thought it was labeling and discriminating.”
 
Indeed, Putnam-Hornstein is grappling with this very same question.
 
“It is just a tool that will hopefully allow for the better classification of risk, but we need to think carefully about the downside if we misclassify an individual,” she said. “I think there is actually very little downside to using predictive analytics in the context of prevention. When we don’t have enough slots to serve everyone – we are simply talking about trying to identify the children and families who most need available voluntary services.”
 
Judge Henry also pointed to the upsides of better targeting voluntary services.
 
“You do not want DCFS detaining every child just because of the predictive analysis,” she said. “But if it were through public health, if young women in situations [of heightened risk] were identified by public health and offered additional services, that is the way to do it.”
 
For Sheffield, no such services were forthcoming.
 
Now 28, she sees having her daughter as the single best thing in her life. The situation sure didn’t seem that way at first.
 
She had her baby in 2004, eight years before California would extend foster care to age 21. But even then, Los Angeles County’s juvenile courts often allowed for children to stay in past their 18th birthday on the county’s dime.
 
Sheffield recalls her social worker and supervisor coming over to Ms. Lisa’s house with a stack of papers for her to sign shortly after her 19th birthday, and within months of her baby being born.
 
Sheffield said she wasn’t ready, that she needed help, at which point she says the county workers became forceful.
 
“If you won’t close your case,” she remembers them saying, “We will say there is some form of neglect taking place.”
 
“They were saying if I do not close my case they could have some ability to show that I have no income, that I have nothing to provide my daughter,” she said. “They were bullying me.”
 
Scared, she signed the papers. Now a youth representative working for the Children’s Law Center of California at the children’s court in Monterey Park every day, she can see ways she could have advocated for herself more. She said she didn’t even call her county-appointed lawyer, because, as is often the case, he had only recently been assigned to her case and she barely knew him.
 
At that time, there was no transitional housing for young mothers leaving foster care that Sheffield could find. Her only option was to go to a shelter for homeless mothers. Luckily, Ms. Lisa let her stay on. Sheffield’s church also pitched in.
 
But for all intents and purposes, the system had left her, and her baby, to face the risks we now know await teen mothers with CPS histories.
 
In the 10 years since she had her child, there have been no reports of abuse or neglect. Sheffield would go on to get a bachelor’s degree, intern for Representative Nancy Pelosi (D-Calif.) and now serves as a peer coordinator for youth who are eligible for extended foster care benefits.
 
By all accounts, Sheffield is a success story. But that her child, now ten and happily situated in a fifth grade classroom, didn’t endure the abuse and neglect that Sheffield faced is the larger victory.
 
This happened in absence of a comprehensive system.
 
What could be done for the thousands of other Miranda Sheffields if such a system was in place?
 
Daniel Heimpel is the founder of Fostering Media Connections and the publisher of The Chronicle of Social Change.