B School for Public Policy

Professor Dennis Culhane

Summary: 
Advancing Evidence-Based Social Policies through Intergovernmental Data Sharing Partnerships

Seminar by Professor Dennis Culhane

There is increasingly broad recognition that policymaking can be done more effectively when decisions regarding support for public programs are made strategically, based on the rigorous analysis of evidence. In several key areas of social policy, including housing and education, such evidence-based policymaking at the federal level needs to rely on data collected and evaluated at the state and local levels. This seminar provided staffers with a better understanding about how the state and local evidence base is gathered and how that base can inform their own work.

In response to the challenge of addressing complex social problems with limited resources, the Office of Management and Budget (OMB) and Congress have called for evidence-based initiatives to facilitate program evaluation and policy research. In 2016, Congress established the Commission on Evidence Based Policy Making to make recommendations on how to accomplish this. While Congress considers mechanisms to link data from federal agencies on a national level, there is much that the federal government can learn from the use of integrated data systems (IDS) at the state and local levels.

Common Policy Challenges

States are facing common policy challenges that are multi-faceted in nature and require an integrated approach for using data to identify solutions.

Superutilizers in healthcare: Superutilizers are often people with complex social disadvantages. For instance, 65% of Medicaid expenditures can be attributed to just 5% of Medicaid users, and one-fifth of those 5% are also homeless.

Education achievement gaps: Research shows that the achievement gap is impacted by factors beyond schools, and reflects health, developmental, family, and community influences. Narrowing the achievement gap necessitates addressing this broader spectrum of issues.

The opiate crisis: Currently, there are nine states doing linked administrative data projects, bringing together pharmacy data, EMT (emergency medical transport) data, and hospital records, to try to understand the dynamics of opioid abuse.

Prisoner Reentry: About 400,000 people are coming out of state prison every year, many of whom have been there for 15-25 years. They’re mostly in their 50s, and they have a lot of health and social issues, in addition to housing and treatment needs.

Child Abuse and Neglect: Research shows that adults who have the longest running challenges with poverty, behavioral and mental health issues, and substance abuse are people who have experienced significant child abuse and neglect. They often struggle for decades, and cost society hundreds of thousands of dollars per person. This strain can be mitigated through better informed care.

What Is IDS?

IDS brings together the data administrators from the different agencies to consider policy challenges. Rather than focusing on a specific program within an agency using only the data for that program, IDS focuses more comprehensively on the person or issue at hand.

These systems are designed to address a full breadth of administrative data collected over the course of a lifetime. The data are vast and include birth records, public school and state testing, higher education, workforce training, wages, health care histories, and death certificates.

Data governance process and sharing protocols

The Actionable Intelligence for Social Policy (AISP) team at the University of Pennsylvania has developed best practices for proper implementation of the necessary governance structure needed to facilitate data linkage by exploring IDS governance, proper legal agreements, data security, and data standards.

Governance is the foundation of IDS use. In order to build one of these systems, all of the participating stakeholders—government leadership, service providers, researchers, and the public—must be represented in a memorandum of understanding (MOU). Every agency has the ability to protect its own data, and veto use of its data on any project if it doesn’t comply with its standards of data usage.

The MOU must address the legal protections governing the use of administrative data. Data collected by government agencies are protected by several federal laws including the Health Insurance Portability and Accountability Act (HIPAA), the Family Educational Rights and Privacy Act (FERPA), the Privacy Act of 1974, and 42 CFR Part 2. Within each of these there is an exemption for analysis, audit, and evaluation including a specific research exemption.

While the MOU authorizes the use of data, a Data Use Agreement (DUA) ensures end-user compliance. Increased access to data and increased data security may sound antithetical, but effective research platforms currently exist and are evolving. Canada, Australia, New Zealand, and many countries in Europe each have derived integrated data systems with similar features.

Policy Innovations and Integrated DATA

Homeless vs. Housed

A New York City initiative[1] in 2000 placed 5,000 chronically homeless people with severe mental illness into subsidized housing. The records of 5,000 people who didn’t get into the housing program were used as a comparison group. Using data from a variety of sources including the NY Office of Mental Health, Department of Homeless Services, and the Human Resource Administration, AISP was able to compare the services these folks were using before and after they got into the subsidized housing, and quantify the cost of the program.

The data showed that the average chronically homeless person was using $40,500 in public services annually, including time in emergency rooms, hospitals, shelters, etc. Multiplied by over 10,000 people over a ten year period, the cost of public services used accumulated to billions of dollars. Placement into housing reduced the use of services to about $24,000 per person per year, a decline of $16,200. The cost of the housing was $17,200 per unit per year, so the net cost to government was about $980 per person per year for that program. In short, 95% of the cost of housing was accounted for by reductions in use of supportive services.

This study has now been replicated over 50 times in different states and countries. In the US, the Bush Administration’s chronic homelessness initiative and the Obama Administration’s veteran homelessness initiative were based on the results of this work.

LA Youth Exiter Study: Outcomes by Domain

This study looked at three groups of young people “emancipated” from the state system at the age of 18: those in foster care, in the juvenile justice system, and a crossover group who started in the former but wound up in the latter.[2] Using the integrated data system from LA County, AISP assessed the extent to which these groups as adults (out to age 25) used public welfare services, how much these services cost, and what level of education and employment they achieved.

The biggest finding was a heavy user phenomenon: 25% of the kids accounted for 75% of the public dollars spent over the 8-year period studied. A high percentage of severe mental illness among the juvenile justice group, including prodromality for schizophrenia, also became clear, revealing opportunities for intervention. This study has been replicated in Washington State, New York City, and Ohio, and is the baseline from which California is measuring the impact of extending foster care to age 21.

Advantages of this data from a research perspective:

  • Population based: Compared to data derived from samples, IDS entails significant coverage and better generalizability, and data can be linked at the individual level. In fact, the U.S. Census Bureau has its own administrative unit that simulates the results of the census and has been able to identify 96% of the population through administrative records.
  • Low to no cost: data collection is built into agencies’ operating costs.
  • Longitudinal: administrative data allows for the study of people over 10-30 years.
  • Policy Relevant: The datasets track how government agencies spend their money.

Homeless vs. Housed, pre/post, propensity score - matched groups

Homeless vs. Housed, pre/post, propensity score - matched groups

LA Youth Exiter Study: Outcomes by Domain

LA Youth Exiter Study: Outcomes by Domain

IDS Costs

From a research perspective, IDS are incredibly cost efficient. For a researcher to follow 500 people for ten years using traditional primary data collection approaches could cost $5 million or more. Projects that track thousands of individuals over 10-20 years using administrative data can cost around $200,000.

Learn more by visiting https://www.aisp.upenn.edu.

[1] Culhane, D. P., Metraux, S., & Hadley, T. (2002). Public service reductions associated with placement of homeless persons with severe mental illness in supportive housing. Housing Policy Debate, 13(1), 107-163.

[2] Culhane, D. P., Bryne, T., Metreaux, S., Moreno, M., Toros, H., & Stephens, M. (2011). Young Adult Outcomes Of Youth Exiting Dependent Or Delinquent Care In Los Angeles County. Retrieved from https://hilton-production.s3.amazonaws.com/documents/97/attachments/Hilton_Foundation_Report_Final.pdf?1440966405