Analyzing Systemic Racial Disparities With Statistical Learning Models and HMIS Data

Different populations experience homelessness—and risk factors of homelessness—differently. Understanding those differences and responding to clients’ vulnerabilities effectively requires nuance and precision. But how can CoCs identify subtle differences in large datasets? Recent advances in open-source software and computing languages, statistical learning, and data visualization tools have made the simultaneous analysis of millions of data points efficient and interpretable. This session will walk participants through a case study of how Pierce County, Washington, is integrating machine learning tools and methods into its HMIS analysis pipeline to inform a racial equity analysis of its Coordinated Entry system. After identifying a disparity in different subpopulations’ housing prioritization scores, the County used a statistical learning model to identify which Coordinated Entry assessment responses were predictive of a given race—and in turn, help guide future research and policymaking accordingly.