California connects datasets to show how mental health services can reduce arrests

Linking disparate datasets and applying customized analytics to them have helped California better map the relationships between mental health services and arrests.

In 2017, the California Mental Health Services Oversight and Accountability Commission partnered with the California Justice Department and received more than 20 years’ worth of criminal justice records that it linked to data from its Full Service Partnership program, which aims to help people experiencing or at risk of institutionalization, homelessness, incarceration or in-patient psychiatric services. As a result, MHSOAC connected data on about 64,000 FSP participants who receive mental health services with 81,000 arrests at three time periods: 12 months before the arrest, during their participation in FSP and one year after exiting FSP.

“We saw a 69% reduction among clients who had three or more arrests prior to participating in the Full Service Partnership program. This is a really, really big deal,” said Dawnté Early, chief of research and evaluation at MHSOAC. “The linked data is helping not only the commission validate that mental health programs can drastically reduce arrests among mental health consumers, but what it’s also showing is that when we invest in the community, going more upstream, that we can reduce contact with the criminal justice system and that’s what we want.”

The research also found that of the 64,000 participants, 70% had no arrests before, during or after participation. For the others, arrests dropped 47% from before to during participation in FSP and 29% from before to after.

The data from FSP is mostly respondent based. Participants answer a series of questions with follow-ups about every three months. Other datasets include key events such as arrests, hospitalizations, hirings, firings, acquiring housing and deaths. From the state Justice Department, the commission gets what Early calls administrative data such as arrests, convictions and whether the defendant was deemed competent to stand trial.

By linking the data, MHSOAC can verify participants’ first-person responses with the DOJ information. That’s because someone who was handcuffed and put in a police car but not actually placed under arrest may say they were arrested because in their mind they were.

“One of the things that we saw when linking these data together that had never touched before was that the positive rates – meaning the rate at which you said, ‘Yes, I was arrested’ – [and] what we actually saw in the system were really different,” Early said. “One of the things that we were able to do was validate what we were seeing in our respondent data to what was actually in the system.”

To combine the datasets and perform the analytics, the commission is using SAS, which helped create the algorithms MHSOAC is using. The idea behind it is called whole-person care, which brings together information from different aspects of someone’s life to have meaningful insight.

Data management, not the analytics, is the hardest part, said Jennifer Robinson, SAS’ director of local government. That’s because entities use different data formats, making governance and usage and sharing policies crucial to have in place before any analytics can happen.

“One of the aspects that’s very difficult in whole-person sharing is ensuring that the person in one system is the same person in another system, so if you’re matching records, making sure you’re matching the records that relate to the right person,” Robinson said. “To accomplish that, SAS analytics uses entity resolution. That concept is scoring the information to be able to give confidence ratings to how accurate we think the match is.”

Data is anonymized before it’s analyzed and arranged into reports for customers such as MHSOAC, she added.

The commission creates data dashboards to share their results and engage the community on what they are doing. They also produce policy reports and MHSOAC adopts recommendations based on what the research shows. Additionally, they write briefs and provide infographics, Early said.

“There are different audiences for this information and so for us, if the information is not being used or isn’t understandable, then we’re not doing our job,” she said.

The commission recently began integrating data with other California state agencies that intersect with mental health consumers. For instance, this year it started using quarterly wage data from the Employment Development Department. Because the respondent data captures information on whether and FSP participant was employed, lost or changed jobs, MHSOAC can validate and quantify what being in FSP means for employment and livable wages, Early said.

Last month, MHSOAC got death records from 2000 to 2019 and birth records from 2000 to 2018 to study suicide and maternal mental health, and at the end of June it will get data from the California Education Department from 2000 on.

“What we have are at the base are mental health consumers and then we go out linking to schools, going back so that you can follow an individual from school all the way to adulthood to see what works, what was a part of that success and how do we invest in what goes to those successes,” Early said. “Where could we have intervened earlier to see a successful outcome?”

Ultimately, the commission will be able to link that school data to birth, employment, criminal justice data and death records, providing a full view of someone’s life span to see what’s most impactful.

“We begin to have an accurate picture of what impact are the mental health services…having on reducing some of these negative impacts that are associated with being a mental health consumer,” Early said.