Michigan’s Office of Child Support (OCS) plays a crucial role in helping to ensure children receive the financial support they need even when their parents are not together.
The OCS wanted to address a key challenge: that as many as one in five child support cases kept getting “stuck” in pre-obligation, the first step in obtaining an obligation to pay child support.
As a result, too many children were waiting to receive support, and the OCS was at risk of not meeting federal child support guidelines for the percentage of open cases for which obligation is established.
Reflecting Michigan’s strategic goal of using analytics to improve child support services and outcomes for families, the OCS decided to test predictive analytics as a tool for understanding—and reversing—that trend.
Michigan’s OCS collaborated with Accenture to pilot use of analytics to drive deeper understanding of how and why some child support cases stall in pre-obligation. The team brought together subject-matter experts with deep understanding of child support; data scientists with knowledge of artificial intelligence and machine learning techniques; and OCS’s longstanding program experts.
Data scientists selected 50 tables from the OCS’s enterprise data warehouse to create the Michigan Case Analytic Record (MiCAR), which consolidates information from 40+ disparate data sets at a case level. The result is a 360-degree view of the pre-obligation process for each case. By applying machine learning models to four years’ worth of case data, the team began to uncover the factors that correlate with higher risk of delay, and to predict which cases would become delayed. Based on those factors, they built models for two key points of delay.
In 16 weeks, the pilot showed the potential of predictive analytics and led to the formation of Michigan’s Analytics Hub.
It identified regional, demographic, racial and socio-economic differences in the prevalence of delays (e.g., Medicaid cases and cases with younger children and younger custodial parents were more likely to experience delays). These and other insights are enabling proactive mitigation of more than half of Michigan’s delayed child support cases.
The OCS continues to explore how it will apply the initial set of findings, including integrating dynamic risk scores into its case management system. The insights also revealed opportunities to redesign how the OCS interacts with various participants in the child support ecosystem.
Above all, the pilot is building momentum for the OCS’s transformation to become a data-driven organization and to collaborate more effectively with other Health and Human Services organizations.