AISP Q&A with Christopher Kelleher of Pay for Prevention at OHSU’s Center for Evidence-based Policy

Christopher Kelleher, from OHSU’s Center for Evidence-based Policy, discusses Pay for Prevention Program

Q1. How did the Pay for Prevention program start?

CK Response: In 2009, Congressman Earl Blumenauer and former Governor John Kitzhaber initiated an effort, subsequently titled Pay for Prevention, to answer two questions: (A) How can Oregon harness high-quality evidence and data be to improve outcomes for the state’s most vulnerable children and (B) What financing methods can be employed to direct outcomes-based funding to maltreatment-prevention programs that show real effectiveness? To begin answering those questions, the Oregon legislature appropriated funds in 2014 to the Center for Evidence-based Policy, part of Oregon Health & Science University.

Q2. What factors led the program to link children’s data with their parent’s data?

CK Response: Our focus was on preventing the maltreatment of young children (ages 0-6), and our review of the English-language literature confirmed the hypothesis that maltreatment risk could be tied almost entirely to parental and family indicators. We committed ourselves to developing a statistical model, based on data about children and their parents, that could predict the risk of maltreatment and foster care placement.

Q3. What methodology did you use for linking children’s data with their parents’ data?

CK Response:  We’re fortunate that the Integrated Client Services (ICS) division of the Department of Human Services (DHS) has developed excellent methods for linking data and individuals across datasets. We structured our Data Use Agreements so that each participating agency sent data to ICS. They then performed the necessary linking, assigned unique Study IDs, and provided us with a de-identified dataset.

I have to note that this work has all been done in partnership with ECONorthwest. They took the lead in cleaning the data, developing our analysis plan, and preparing the statistical models. It’s been a great partnership.

Q4.  What state administrative data will the Pay for Prevention program utilize?

CK Response:  A reliable statistical model focused on abuse prevention needs information on children who experienced maltreatment and children who didn’t. We used Oregon birth records as the foundation. The Department of Vital Statistics provided ICS with information on all births between 2000 and 2010 (children and parents). ICS then linked that data with records from Child Welfare and from most of the programs that DHS administers (e.g. self-sufficiency, substance-abuse treatment, substance-use rehabilitation), as well as with some data on criminal sentencing. The fact that the data coming to us is de-identified and that we’re operating under a strong IRB protocol has been crucial in building support and maintaining it.

Q5.  What are some of the outcomes the program has found by linking the data?

CK Response:  Linking the data has allowed us to develop and validate a statistical model that uses information available in the birth record to predict the likelihood of a child being maltreated by age two and placed in foster care by age six. The model looks at children born in 2002 or 2003. It includes two years of information about the parents prior to birth and follows children for six years after birth.

We’ve also developed an economic model that breaks down different types of cost associated with predictive factors and child-welfare outcomes. Neither model would be solid without great support from the participating agencies. We’ve relied on their input to make sure that we’re interpreting data properly and that, in the case of the economic model, we understand the financial realities of the affected systems.

Because the dataset contains geocoded information down to the census-block level, we’ve been able to create user-adjustable maps that reveal how child maltreatment and foster-care-placement rates vary by location. This “hotspot” data has already proved useful in decision-making. Long term, we think it can be valuable in aligning efforts and directing resources to the areas of greatest need.

As a step toward making the data accessible, we’ve developed a website ( that contains a number of interactive maps and infographics.

Q6.  Do you have plans to integrate additional state data into the Pay for Prevention program?

CK Response:  We’ve just finished amending our original agreements to update the data, so we’ll soon have information from 2000 to 2015. Just as important, we’re bringing in new data (once again, all children and their parents) from the Oregon Youth Authority and the Department of Education. The new data will significantly enhance our understanding of risk factors as well as outcomes. We’ll also be adding linked information from Head Start, special education, and the state’s primary home-visiting program.

Q7.  What data would be most helpful to add to your dataset and why?

CK Response:  We’ve begun investigating the introduction of healthcare utilization data, including state Medicaid data. Adding health information would increase our understanding of child-welfare outcomes, open up new avenues for collaboration, and significantly expand the potential range of actionable intelligence.

This site is registered on as a development site.