Cities across the United States are developing community responder models to better serve people in crisis. In these models, mental health practitioners and behavioral health clinicians respond to noncriminal emergencies in lieu of or in addition to police units. Denver operates two such programs—the Support Team Assisted Response (STAR) program and the Co-Responder Program—that offer distinct but complementary pathways for responding to behavioral health–related 9-1-1 calls. 

In this report, we provide new evidence about Denver’s programs, including their reach, opportunities to scale services, cost implications, and the first quantification of individual-level criminal justice outcomes for people in the year after they interact with the programs.

Why This Matters

These findings can support Denver’s efforts to further improve its crisis response systems and may be relevant to other communities that are considering starting or expanding community responder models.

What We Found

We looked at program administrative data and found the following:

• STAR and Co-Responder staff responded to over 20,000 calls for service each from June 2020 to September 2024.
• While there were multiple limitations to our approach, we found that in the year following an encounter, both programs were associated with reductions in criminal justice involvement, including the probability of being arrested, compared with a police-only response.
• If demand for STAR services remains consistent, the program could meet full demand by increasing the number of teams working each day to 9 or 10.
• While STAR encounters are more resource-intensive than police-only encounters, STAR encounters also entail more services, including clinical assessment and referrals. STAR costs are partially offset by avoided criminal justice involvement.

How We Did It

We used 9-1-1 call-for-service data, criminal justice system data, Denver program budget documents, and administrative data from WellPower—the organization that staffs the STAR and Co-Responder programs—to characterize program services, estimate program costs, and describe opportunities for scaling.

To evaluate whether STAR and the Co-Responder Program influenced people’s subsequent criminal justice involvement, we used a quasiexperimental research method called propensity score matching, which allowed us to identify a comparison group that was similar to the treatment group of individuals who had encounters with STAR or the Co-Responder Program. Due to data constraints, these methods had multiple limitations, including that some people in the comparison group may have been different from people in the treatment group, which could have influenced our results.