by Marty Elisco
Natural language processing (NLP) refers to the ability of computers to ingest and “understand” free text data in the same way humans do. This data can be in the form of spoken words, hand-written or typed text, and other unstructured data. In the context of child welfare, NLP therefore can read, understand, and analyze the unstructured data in case notes, giving staff access to that wealth of data and providing analysis of it. In addition to giving caseworkers and supervisors access to this greater knowledge, NLP also can help them to more readily identify risks, strengths, relationships, early warning signs, and social determinants of health (SDOH) that affect children and families.
The Value of NLP: Pennsylvania and Hawaii
In Allegheny County, Pennsylvania, the Department of Human Services (DHS) serves the more than 1.4 million children and families who live in the Pittsburgh area. DHS has long recognized the value of the unstructured data in case notes. “We have incredibly rich administrative data,” says Katy Collins, DHS Chief Analytics Officer. “But our caseworkers were continuing to find that so much of that rich information was buried within case notes and unstructured data.” Ultimately, DHS decided to turn to NLP to address the problem. Thanks to their NLP platform, Augintel, county case teams now have a better understanding of the families they serve and are quickly able to identify risks and strengths as well as the SDOH at play. One important example of this is a history of drug use; it could take several weeks or months for a new caseworker or service provider to uncover this information because all the detail and deeper descriptions of substance use live within the unstructured data. The NLP platform ensures that staff newly assigned to the case can immediately surface that information. Allegheny County also is seeing productivity results with field teams. Caseworkers estimate that using Augintel is saving them five hours per week that were previously spent combing through case notes for information and that can now be spent focusing on more impactful tasks.
The Hawaii Department of Human Services (DHS) manages more than 2,500 child welfare cases each year. This work is supported by two information systems designed for managing cases, identifying and tracking delivery of services, and satisfying federal reporting requirements. Duality of systems, staffing shortages, and aging technology made it exceedingly difficult for caseworkers to find critical information within their case notes. Hawaii DHS leadership recognized this challenge and made a commitment to making the move to the Comprehensive Child Welfare Information System (CCWIS), but they also knew that CCWIS system selection and implementation is a multi-year process and that their problems were immediate. Hawaii DHS leadership identified NLP as technology that could address their current challenges as well as add significant value to CCWIS when implemented in the future. Hawaii DHS now accesses case notes from both of their case management systems using Augintel NLP to digest, summarize, and present critical information that is often lost in the case notes. Previously, caseworkers had to click through notes one-by-one to try to find that information. One DHS caseworker has found this helps her be more responsive to families: “A service provider called me and asked if the family had received services with them in the past. Previously, I would have to wait to return to the office to get into the file. But I was able to access Augintel on the spot and tell them right away.” DHS leadership also is seeing improvements: Caseworkers have become more motivated to add in strong notes because they know that it will be easy to go back and find them—and that their manager or colleague will be able to quickly find what they need, as well.
Successful client outcomes depend on the observations and insights of the caseworkers, clinicians, and specialists who deliver services and manage their clients’ cases. Those insights are captured in case notes—unstructured data—which can number in the hundreds of pages per case. This makes consuming case knowledge across an entire case team incredibly time-consuming and difficult. NLP solves for this problem by methodically extracting information from narrative data, then analyzing and presenting that data in easy-to-understand ways. In parallel, the data becomes searchable in the same way that internet browsers help us search the internet. With this information, teams have easy access to the data they need to streamline case management, make better informed decisions, and achieve better client outcomes.
Marty Elisco is the CEO of Augintel, a SaaS (Software as a Service) company that improves the ability of health and human services professionals to deliver care. To find out more about Augintel and its child welfare-focused projects, see https://www.augintel.us/child-welfare.