IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

New Tech Improves Outcomes for Children in Allegheny County

The implementation of a natural language processing platform has helped the Allegheny County Department of Human Services better interpret the data on the children it serves through natural language processing.

A young person's face, in black and white, sits on top of a block of binary code.
Allegheny County, Pa., is working to improve the health and safety of residents with a new natural language processing (NLP) platform in one of its most critical agencies.

The platform, provided by Augintel, has changed the way employees of the Office of Children, Youth and Families (CYF) within the county’s Department of Human Services (DHS) access and interpret data about the people it serves by pulling data from text-rich notes.

“So much of this rich information is in notes,” said Katy Collins, chief analytics officer for the Allegheny DHS. “And so, are there ways that we can kind of combine this information and help our workers get more insights into what’s happening with the families — with the ultimate goal of helping to serve them better?”

The pilot program officially kicked off on Jan. 19 for select users, and the full rollout began May 3.

As Augintel CEO Marty Elisco explained, the platform uses natural language processing to pull information that caseworkers might need from narrative data. A caseworker may have hundreds of pages of case notes about a child, and within those pages, important information is buried that could give insights into risk factors and social determinants of health.

As Elisco explained, the platform has helped save caseworkers around five hours per week that would be spent digging for data that is buried in case notes.

The platform allows users to search specific keywords as well as conceptually related keywords. For example, if a caseworker searches “death,” the system would also find mentions of “funeral.”

It also provides caseworkers a dashboard view of case information and flags notes with either a risk factor or a protective factor.

Regarding social determinants of health, sleep safety is an important factor of a child’s well-being, and a caseworker might need to know related information like whether there is a crib in the house. This is where the machine learning piece comes in, Elisco said, as the application can help detect information in the text that signifies an issue with safety management and can alert county staff to any potential risks.

Collins provided another example of how getting information related to early warning signs for opioid use can help caseworkers understand, and in some ways, get information earlier in the process to mitigate risk.

“So, it’s almost an early warning system that helps them address those problems. Now that’s one piece,” said Elisco. “The other piece is just helping caseworkers find the information that they’re looking for in text.”

If a caseworker needs to identify family members of a child to address an issue, rather than having to look back at years’ worth of data to find mentions of caregivers, the system gives users a breakdown of the people’s names that are mentioned in the text, how often they are referenced and when.

This is also helpful with case transfers, Collins said, as it gives caseworkers insight as to what has happened within a case prior to them reading multiple years of case notes.

As Collins explained, the agency is doing work around quality improvement initiatives as well. Caseworkers were the employees primarily impacted by the pilot, but with the agency rollout, supervisors and others — like frontline workers — will also see the impact, as will quality improvement and management staff.

“Anything that’s being spot-checked for in case notes today as part of any oversight processes, we can automate,” Elisco stated.

Ultimately, there is an opportunity to use this tool within other DHS programs, like housing and family support. To do so, Elisco said the tool would extract data from multiple systems, like the Homeless Management Information System and the Family Support system. Accessing the narrative data housed within these systems would create a unified or “whole-person” view of the family and other factors that may impact a child’s well-being, rather than a caseworker having to access different systems to piece together a family’s story.

Editor's note: This story was adjusted to remove references to the other agencies using the platform.
Julia Edinger is a staff writer for Government Technology. She has a bachelor's degree in English from the University of Toledo and has since worked in publishing and media. She's currently located in Southern California.