Officials say data could help identify at-risk children

Imagine walking into the Oregon State Penitentiary and finding a man convicted of armed robbery. Now imagine rewinding his life to the day he turned 10 years old, or 5 years old, or even the day he was born.

Imagine being able to examine his life and identify what led him to that prison cell.

Maybe it was being pulled in and out of foster care. Maybe it was a father who went to prison, a mother who neglected him or a third-grade teacher who let him move on to fourth grade even though he really couldn't read. Most likely, it was a combination of some of these and more.

No one can turn back the clock on adult lives, but the state of Oregon wants learn from them.

State agencies have begun pooling their data in an effort to predict outcomes for children, from which ones are likely to become criminals to which ones are likely to drop out of high school, and identify ways to help before their lives head irretrievably down those paths.

The Department of Corrections, Oregon Youth Authority, Department of Human Services, Oregon Health Authority and Oregon Department of Education have signed an agreement to share data among them, something they are not typically allowed to do.

They hope to analyze the entire system that works with children and families and develop a list of risk factors and helping factors that could inform earlier interventions for children at risk for bad life outcomes.

It's a cutting-edge idea.

No other state has launched such an ambitious project, OYA spokesman C.J. Drake said, and Oregon has long been known for doing more with public data and research than nearly anyone else.

It also has high level support.

Sen. Jackie Winters, R-Salem, said she has tried to facilitate and advocate for the project and has met little resistance, whether from fellow lawmakers or Gov. John Kitzhaber's office.

"I'm pleased when you can bring agencies together to do more collaboration, and that's what's happening in this case," she said. ""We want to solve these problems early on, rather than being reactionary."

Data collection often raises privacy concerns, but nowhere in this project can the state analyze an individual child.

Rather, the project will operate like the creation of medical diagnostics. For example, doctors have analyzed thousands of heart attack patients and created a list of factors that show a high probability a person will have a heart attack and a list of factors highly likely to prevent one.

A doctor can't say, "You will have a heart attack," but she can say, "Your blood pressure is high and you get very little exercise, so you're at a higher risk for a heart attack. Start walking daily and taking an aspirin to reduce that risk."

State officials want to do the same for children.

They will never be able to say, "This child will drop out of high school," but they want to be able to say, "This child is exhibiting three crucial behaviors that make him likely to drop out of school. Here is a solution that is likely to keep him on track to graduate."

This is an unprecedented use of the state's data and a philosophical change as well. Rather than each agency working independently, this project would allow them to work as a cohesive system.

No one has an estimate yet on how much it would cost, if anything. The Oregon Youth Authority, for example, never needed additional funding to move from data storage to data analysis.

Money may move between programs based on the research findings, but those budget decisions are likely months or years down the line.

The state has collected mountains of data on the people in all of its programs, corrections research analyst Paul Bellatty said. Until now, however, most have used the data for descriptive purposes. (i.e. There are 51,064 registered Democrats in Marion County.)

However, the prison system and particularly the Oregon Youth Authority have started using their data for predictive purposes over the past several years. What is this man's risk of committing another crime? Which program will help this young woman? Which facility will this young man be most successful in?

"(The Juvenile Justice Information System) collects a lot of information, and it's usually very useful," Bellatty said. "The question is: What do you do with it?"

Now social service agencies want to follow that lead and start analyzing data to predict outcomes.

DHS spokesman Gene Evans said his department wants to analyze four areas:

• Risk factors that make a youth in the foster system more likely to end up in the criminal system;

• Which programs or treatment will work best for each family in the system;

• Which institutions or treatments will work best for a child with severe behavioral problems; and

• Which behaviors are most likely to indicate a child is being abused.

For most of history, the answers to those questions have been mostly anecdotal, Evans said. A teacher has a gut feeling that a child is troubled, he said, or a child exhibits such overt behavior as torturing animals.

You can't quantify a gut feeling, he said, but neither do people want to wait to intervene in a child's life until the child is so deeply affected.

However, proving a gut feeling with numbers may not be beyond possibility anymore.

"It's about time," Evans said. "We've reached an age of maturity where there's a new approach to solutions."

The prison system has been using this approach in sentencing for many years.

This kind of statistical analysis asks a computer to look at everyone else who has the same data profile and how they fared under a program or some other experience. It is the same analysis that allows companies to target online advertisements at specific people and allowed the Oakland A's to assemble a winning team as depicted in the book and movie "Moneyball."

The department of corrections uses predictive analytics on every offender who comes through the system by way of a "risk assessment." It takes an offender's information and produces a number that represents the likelihood he will commit another crime.

Craig Prins, executive director of the state's Criminal Justice Commission, said the assessment is used to inform decisions about sentencing, but it is strictly a tool for a judge or district attorney. That number does not decide someone's fate.

"Intuition has driven the justice and juvenile systems," Prins said. "We're trying to move from intuition to analysis."

The youth system uses a similar risk assessment, Bellatty said, but it has gone further. OYA officials now use data to predict which program might be most effective for a particular youth or what kind of custody might serve him best.

It's never a replacement for professional judgment, he said. In fact, many veteran OYA employees have found the data analysis seems to simply back up the intuitive feeling they already had about what a youth needs.

Humans find the same patterns with experience, he said. The computers simply find the patterns faster and can show them in graphs, charts and spreadsheets.

Perhaps no one has more interest in predicting outcomes than educators.

Not every child in a troubled family will wind up in the foster system. Not every child in the foster system will end up in the criminal system. Not every adult in prison had a mother in mental health treatment or the WIC program.

However, nearly every child in every program will pass through Oregon's public schools at some point. Classroom teachers see everyone.

Oregon Education Investment Board analyst Peter Tromba said his agency wants to create a long-term database that can be analyzed the way the youth authority's data can be examined.

Working with other agencies could be crucial, he said, because there is only so much education data can say.

For example, data analysis has shown that third-grade reading skills and ninth-grade attendance both closely correlate with graduating (or not). However, some recent studies have shown a statistical correlation between low birth weight in normal-term babies and that child's likelihood to graduate, Tromba said.

That's information the education department can't collect, but other agencies can.

"Modern predictive analytics is going to tell us things that we never would have considered," he said. "The reason we're building (the database) is so we can really delve deeper into those kinds of questions."

While risk factors are important, so too are "protective" factors, Tromba said. For example, data has shown that a stable, close-knit family can protect against the negative impact of poverty, he said.

Knowing how to help children is just as important as knowing children need help.

"From a school perspective, that's a really great way to look at it," Tromba said.

Prevention is the underlying goal of the project. Corrections Director Colette Peters said she would like her department to keep people from ever needing to go to prison — it's better for society and cheaper for the state.

However, people start becoming who they are at a very young age, and that's the time to catch them, she said. Using data to inform early intervention policies for children and families could be a revolutionary approach, she said.

If it sounds outlandish to focus on small children, Peters said, talk to a teacher.

"There are children in the kindergarten classes now that kindergarten teachers suspect will be in the criminal justice system someday," she said.

Those are the children the state hopes to catch now, while the consequences of their actions are still time outs and not prison sentences.

hhoffman@StatesmanJournal.com or (503) 399-6719

Click here to read the Oregon Youth Authority's plan to analyze the "feeder system" from childhood to prison.