Data Integration for the Common Good: Rigorous Inquiry to Improve System Coordination for Individuals with Serious Mental Illness and Justice Involvement

Background: Local governments invest heavily in behavioral health services but frequently lack current, clear data on how people with serious mental illness (many with co-occurring substance use disorders) navigate public systems through funded entities, e.g., jails, shelters, and service providers. Because justice, healthcare, and other public service systems keep separate records, it is challenging to understand how improvements to programs and policies are truly making positive impacts. Cross-system data integration projects are needed yet challenging to achieve due to multilevel government regulations, variations in data systems, and competing organizational priorities.

Methods: The research project used data from 2023-2024 and examined three county-level entities: the jail, the public mental health authority, and the homelessness response system. Prior to the initiation of the study, a data governance committee was established with representatives from each entity and individuals with authority to make decisions regarding the project's legal and technical components. The university acted as the intermediary, managing governance committee meetings and providing independent data analysis across the entities. The project had four distinct phases. First, the data governance committee reviewed relevant state and federal regulations establishing permission to match data across these three systems. Through this process, it was established that we would need three distinct data use agreements (DUA) between each entity and the university. The second stage was executing the DUAs between each entity and the university. Third was the data transfer and matching of the datasets. Due to the nuances of criminal justice data security regulations, the criminal justice data cannot leave the county server. The mental health and homelessness data were transferred to the university using 7-Zip, and the researcher then toggled between a university desktop and a county virtual desktop to perform the matching in SPSS. Due to the limitations of social security numbers within the justice dataset, last names and date of birth were used to create the new identifier to match across the datasets. The final stage was the initial analysis of the combined dataset and discussion within data governance about the initial findings.

Results: The matched dataset comprised n=7010 individuals across the justice and mental health systems. Half of those individuals were also represented in the homelessness response system dataset.

Conclusions: Bringing together justice, behavioral health, and homelessness records can turn fragmented information into useful, actionable insight. The pilot proved the idea works in practice and exposed practical hurdles that can be fixed. However, this project was quite time-consuming and required the lead researcher to draw on many of her macro social work skills to navigate competing priorities and reach an endpoint.

Implications: This project offers a simple, privacy‑minded roadmap for other communities and researchers who want to better understand and support people with SMI. To replicate projects like this elsewhere, we must invest in doctoral social work training that builds robust coalition‑building skills alongside advanced technical data competencies.