Evaluating Approaches to Linking Evictions Records:: Assessing the Feasibility of Research with Integrated Data.

Abstract

Objectives: This study investigates different approaches to integrating evictions data with Medicaid and homeless shelter utilization records at the individual level for the state of Delaware. We especially focus on evaluating the feasibility of creating an integrated dataset focused on children and adolescents through different approaches to matching.

Methods: We attempt to link existing statewide records on evictions, Medicaid, and shelter from 2017-2019. We first compare direct match and probabilistic match approaches to linking evictions and Medicaid records, and then incorporate shelter records. Finally, we consider a limited set of characteristics relevant to potential future public health research among children who experienced eviction, had a shelter stay, and were enrolled in Medicaid.

Results: Direct matching resulted in a lower match (14%) rate than probabilistic matching (22%) of eviction records to Medicaid data. Homeless shelter records had a high match rate to Medicaid records, even when using a direct match (75%). A sizeable subset of children (n=216) were linked across the three data sources, though this was from a small percentage of cases in the evictions data. Among this subset of children, most (71%) were enrolled in Medicaid in all three years considered by this study and Black children were greatly overrepresented (75%).

Conclusions: Integrating evictions records with other health and human service data involves a number of challenges. Probabilistic matching yielded a considerably higher number of matches after manual review, resulting in a possible study sample of children who have experienced eviction, a homeless shelter stay, and were enrolled in Medicaid. Strategies to increase the match rate for eviction records through using records from other, more universal services may be necessary for investigations that require more comprehensive coverage of the population.