An Effective Model for Using Administrative Data to Inform Social Services Policy, Practice, and Management

Purpose: Administrative data serves as one of the best sources to assess the experiences of children and families involved with social services. However, using it to improve outcomes is challenging. It requires an information system that can provide the necessary information in a suitable form to the appropriate people in a timely manner. With advances in KDD (Knowledge Discovery and Databases) technology providing consistent information to all related parties in diverse form is possible. KDD can achieve consistency and diversity simultaneously, a key factor for using administrative data in social services. To effectively address the many local problems, diversity is a must while accountability requires consistency of measurements. In this presentation, we will describe the KDD model designed to streamline the use of administrative data in social services. Properly implemented, the system is easy to maintain and use to generate timely information which can provide a comprehensive view of the social services as well as provide a rich source of data for more fundamental research.

Methods: This is a case study of how KDD technology was applied in North Carolina to effectively use administrative data for policy, management, and practice for social services. In 1997, the Jordan Institute for Families at UNC-CH, in collaboration with the NC-DHHS built a KDD information system to support self-evaluation efforts at the local level for Workfirst. Over the 13 years, the project grew to include data for child welfare, Food and Nutritional Services, and Medicaid. The project adapted and modified the KDD model developed and used for business intelligence for social services. The model will demonstrate the essential building blocks that any research group using administrative data should have.

Results : The complete KDD model for administrative data includes the full cycle of the KDD process from data acquisition to consumption of the information via a public website. In addition, longitudinal files in the system are used for in depth research as well as on-going and ad hoc analysis for policy. The model has three stages (1) data repository, (2) information generation, and (3) information delivery. In the first stage, raw data tables are acquired from state systems to build comprehensive longitudinal files depicting the experiences of children and families over time. In the second stage, the longitudinal files are analyzed for various reports, research, and monitoring of performance measures. Finally, each of these diverse forms of information are presented to the public using the appropriate forum. In particular, a comprehensive set of county level performance measures can be shared with the counties using dynamic web technology.

Conclusions and Implications: We have learned that trust, real support through policies and funds, access to good technical expertise in both social work and IT, and training are key factors to successfully implementing the KDD model in a state. Access to the required expertise can be obtained through strong partnerships between government agencies and interdisciplinary teams at universities. The partnerships can lead to effective use of administrative data while providing a priceless opportunity for in depth research.