Improving our Medicaid members’ care through sharing data

Using data for better care

The systems that serve our most vulnerable populations are typically not integrated, particularly when it comes to data. This lack of integration affects our ability to use valuable data in our work with populations that are most in need. As a result, our systems are often limited in their capacity to provide the best possible care.

The Connecticut Behavioral Health Partnership (CT BHP), administered by Carelon Behavioral Health, launched a new initiative in 2020 in partnership with the Connecticut governor’s Task Force on Housing and Supports for Vulnerable Populations to help improve care for the most vulnerable. The 500 Familiar Faces initiative integrated data between agencies with the goal of improving care for Medicaid recipients experiencing housing instability.

The program integrated data across the six major systems that work with individuals who are unhoused or experiencing housing instability:

  1. Medicaid healthcare services
  2. Housing and homeless management
  3. Child welfare
  4. Criminal justice
  5. Corrections
  6. Behavioral health

The primary goal of the 500 Familiar Faces initiative was to analyze the potential for agencies to use data in cross-agency collaborative care.
 

Working to integrate data

The project developed several new data-sharing agreements while leveraging the existing ones.

When the program began, the first step was to integrate the Medicaid dataset with the Housing Management Information System. One of the program’s primary interests was Medicaid members who were unhoused or experiencing housing instability.

The sample yielded 10,420 individuals and involved at least two agencies.

A six-step matching process matched over 90% of the other agency’s data.

The program identified barriers such as a lack of secure data transfer methods, a lack of data standards across agencies, limited family data, and a lack of standardized time frames for data pulls across agencies.

Datasets

  1. Medicaid: Medicaid healthcare systems data
  2. Homeless Management Information System (HMIS): data on shelter usage and housing status
  3. Department of Children and Families (DCF): child welfare system involvements including home or family removals
  4. Judicial Branch Court Support Services Division (JBCSSD): convictions, violations of probation, legal status
  5. Department of Corrections (DOC): lifetime incarceration data
  6. Department of Mental Health and Addiction Services (DMHAS): behavioral health and substance use service encounters data

 

Sample demographics and agency involvements

Agency involvements

  • In the final sample, between 46% and 59% individuals were involved in each of the four systems.
  • The HMIS dataset included 7,329 individuals and 1,042 families.
  • The sample was predominantly male (60.7%), White (48.9%), non-Hispanic (69.3%), aged 25 to 64 (72.7%), from Central Connecticut (79.8%), and with an average family size of three.
  • Individuals typically interacted with 4.1 of the six agencies, and over 54% interacted with five or six. 
  • Family members showed an average of 3.1 agency involvements, with 81.4% having fewer than four.
  • Families were most likely to be involved with DCF (74%).
  • The analysis indicated higher than anticipated interagency involvements, showing distinct differences across individuals and families.

 

Identifying the highest-need individuals

Agency-specific needs

Each agency had its own definition of high need. For example, DCF defined high need as cases in which a child was removed from their family, and DOC defined the term by the lifetime number of days in which an individual was incarcerated.

The analysis found an overlap between high-need individuals in DOC and in JBCSSD through their experiences with incarceration, parole, and probation.

Ninety percent of DOC and 80% of JBCSSD high-need individuals had five or more agency involvements. The data indicated the impact of individuals’ interactions with the legal system.

Additional overlapping data showed the importance of DMHAS, DOC, and JBCSSD coordinating their efforts.

 

Cross-agency high need

Cross-agency high need is defined as individuals with five or more agency involvements and two or more agency-specific high-need statuses.

In the sample, the highest need tended to be older, male, White, and non-Hispanic.

This group experienced fewer days in shelter, but with more distinct shelter episodes.

Over 95% experienced behavioral health issues, emphasizing the importance of access to behavioral health resources.

This group had more experience with emergency department (ED) and inpatient service admissions. They also experienced more arrests, probation violations, and incarcerations.

 

Cluster and regression analysis

The team used the cluster analysis approach in this sample to identify subgroups of higher-need individuals. The team’s goal was to develop more tailored approaches to care management.

Regression analysis

The team also used multiple regression to determine the odds of an individual being in each cluster for chronic shelter usage. They wanted to learn who was most in need of ongoing housing support.

The cluster was the independent variable, and the number of shelter days was the dependent variable.
 

Findings

  • The sample used for analysis consisted of 5,500 higher-need individuals.
  • The 500 Familiar Faces team selected eight variables for analysis.
  • The team identified six clusters, ranging in size from 580 to 1,234 people.
  • Cluster 3 was the smallest; Cluster 2 was the largest.

Cluster highlights

  • Cluster 1 consisted of 1,171 members with the lowest number of shelter episodes. 
  • Cluster 2 contained older individuals less likely to be involved in the criminal justice or child welfare system.
  • Cluster 3 experienced some of the highest rates of interagency involvement, higher incarceration rates, and child removals.
  • Cluster 4 had disproportionately higher numbers of Black and non-Hispanic individuals, with no behavioral health diagnoses. 
  • Cluster 5 consisted of mostly middle-aged people and was 3.3 times more likely to experience higher shelter use.
  • Cluster 6 had the highest shelter usage for both days and episodes.
     

 

The program’s key takeaways

The 500 Familiar Faces initiative confirmed that the data used by social service agencies is typically not integrated across agencies and is limited by the data they are able to collect. It also demonstrated that cross-agency data integration is achievable even without a universal unique identifier. Data integration across agencies can provide more information about the populations served and improve care management for our most vulnerable populations that utilize services provided by multiple agencies.

This work was supported by CT BHP partner agencies, which include the Department of Social Services, the Department of Children and Families, and the Department of Mental Health and Addiction Services.

This publication does not express the views of these partner agencies or of the State of Connecticut. The views and opinions expressed are those of the authors.