In part of our Digital Communities Special Report, we look at how Allegheny County, Pa.'s Department of Human Services put data in a central location to solve a host of social problems.
Editor's note: The Digital Communities Special Report, which appears twice a year in Government Technology magazine, offers in-depth coverage for local government leaders and technology professionals. View links to the entire report here.
In 1999, in Allegheny County, Pa., the Department of Human Services DHS) did something few local governments would have considered. It built a data warehouse as a central repository for the county’s health and human services data, part of an overarching strategy to deliver integrated support to its most vulnerable citizens. While most data warehouses were oriented toward research or one-shot projects at the time, Allegheny County took a different tack and decided to use it for decision support for a range of social problems: behavioral health, child welfare, homelessness, aging and other disabilities.
The warehouse uses software from IBM and Oracle and has steadily built up the number of data sets it contains. As the data has become richer, the ability to conduct more comprehensive analysis, particularly around gaps in client coverage, has steadily improved. For example, in 2005, DHS developed an algorithm that helped officials spot clients who had multiple needs. The algorithm boosted the number of children in child welfare who received mental health support from 26 percent to 44 percent.
In 2009, DHS signed an agreement with the Pittsburgh Public School District to share data and improve both the education and well-being outcomes of children under its care. Since then, there have been data-sharing agreements established with another 20 school districts.
Last year, DHS began using its data warehouse to conduct the kind of predictive modeling that Goldsmith called the hardest for local governments to establish. The Allegheny Family Screening Tool helps with decision-making when a worker screens a call regarding a child who is at risk of possible maltreatment. It predicts possible outcomes with a fair degree of accuracy by creating a score for intervention based on the integration and analysis of hundreds of data elements. The higher the score, the more likely the possibility of future maltreatment, giving decision-makers the kind of information they need to intervene and investigate. Studies have shown the tool to be ethically appropriate because it is more accurate than the alternatives currently being used.
Another tool used by DHS is a data dashboard. DHS has been able to help inform health providers whether someone who has died from a drug overdose had recently been receiving mental health or substance abuse treatment. It may sound like a simple correlation to make, but in government that kind of social service data sharing is not the norm, according to Erin Dalton, deputy director for DHS. “It seems straightforward, but there has been historically and funding-wise, real walls between mental health and substance services,” she said.
In the realm of social issues, data sharing can be the most rewarding process, but also the most vexing. “When it comes to hurdles, there are technical aspects to data mining and integration that are fairly complex,” said Goldsmith, “but data sharing is the most important and it’s the one issue that can slow everybody down.”
According to Dalton, Allegheny County has had to pay for strong, external legal guidance to manage data-sharing issues. But she believes too many government agencies misinterpret laws so they can hold on to data rather than share it. “I think it’s become pretty obvious that there isn’t a legal issue in most cases, but really a policy decision that needs to be made,” she said. “Those kinds of decisions should not be made by relatively low-level bureaucrats.”
Most government officials who are trying to fashion data-driven solutions for vexing social problems would agree with that. What is required are strong data governance practices. A paper from the Ash Center at the Harvard Kennedy School, Lessons from Leading CDOs, states that strong data governance fosters interoperability and reuse of data across the enterprise. Data governance also addresses cross-agency sharing of who collects what, and for what purpose, to avoid duplication of data collection in different or competing formats.
An example of the kind of policy decision-making Dalton has in mind can be found at the state level. Indiana set out to improve data sharing in a secure and legal way while removing technical barriers. The starting point was an executive order signed by former Gov. Mike Pence, creating the Management and Performance Hub, where appropriate state data could be held securely and used to drive better government performance through analytics.
The problem the state wanted to address was a troubling health issue: Why were some infants dying in their first year of life while others weren’t? For years, Indiana had been pursuing a policy that wasn’t producing results. It was based on the prevailing belief that infant mortality was caused by pregnant moms who either smoked cigarettes, did drugs or drank too much alcohol.
With Gov. Pence’s executive order, policymakers were freed up to share data between multiple agencies. Using analytics and data science, the state was able to uncover the real culprit in its high rate of infant mortality: lack of access to prenatal health care.
Taking down barriers to data sharing at the local level can be just as beneficial in the long term. Dalton’s advice is to avoid having to go through a memorandum of understanding agreement every time data is to be shared. While she makes it clear that Allegheny always works within the law, she believes that signed agreements aren’t necessary, just good governance and rules for data use. “Once you get into agreement mode, you might as well add three more years to the project,” she warned.