How Are Cities Designing These Programs?

Ten design decisions shape every mobile crisis program, and they’re sequential: each one constrains the next.

---

The Decisions Come In Order

There are roughly ten design decisions every city faces when building a mobile crisis program, and they’re sequential, each one constrains the next. The institutional home determines the hiring pipeline. The hiring pipeline shapes team composition. Team composition dictates which calls the program can safely take. Call types drive the dispatch protocol. And the dispatch protocol determines whether anyone actually uses the thing.

Robert Blaine at the National League of Cities, who has tracked program design across dozens of jurisdictions, observed that cities don’t need to pick one model permanently.²¹ Most of the programs with the strongest track records started with a limited scope and expanded as they built evidence. The ones that struggled tended to either overdesign on paper before launching or launch so cautiously that the program never handled enough calls to demonstrate its value.

What follows is how cities have actually addressed each of these decisions, what they chose, what tradeoffs they encountered, and what the documented outcomes show.

Decision 1: Where Does The Program Live?

Where a city houses the program determines its culture, hiring pipeline, funding streams, and how law enforcement and the public perceive it.

Inside Public Health or Behavioral Health

Harris County runs its Holistic Assistance Response Teams through the public health department.¹ Berkeley operates its Specialized Care Unit through Health, Housing & Community Services. San Diego County’s 44-team program sits in the County Behavioral Health division. Austin’s EMCOT is operated by Integral Care, a non-profit organization that is the local mental health authority for Travis County.

This model gives the program clinical autonomy and aligns it with healthcare rather than law enforcement. Clinicians hired through a health department are evaluated by clinical supervisors, follow clinical protocols, and think of themselves as healthcare providers, which is the orientation crisis calls generally require. It also creates a cleaner separation from law enforcement, which matters for communities where police contact carries a history of harm.

The tradeoff is integration with emergency response. Police and fire departments have decades of shared dispatch infrastructure, radio protocols, mutual aid agreements, and shift structures. A program housed in public health has to build or negotiate its way into those systems from scratch. Dispatch integration, a critical determinant of program success — can take significantly longer because you’re bridging two departments that don’t normally share protocols, technology, or culture. Harris County addressed this by embedding staff in the 911 dispatch center, but it took sustained institutional effort.

Inside the Fire Department

Oklahoma City runs its program out of the fire department with four specialized teams.² Tulsa launched its Mobile Crisis Response Team through the fire department and was able to go 24/7 from the start because it leveraged the fire department’s existing round-the-clock operational infrastructure.

Fire departments already have emergency medical expertise, 24/7 dispatch and shift coverage, experience deploying two-person teams to non-criminal calls, and a public reputation that doesn’t carry the same baggage as law enforcement. Firefighters and EMTs already respond to overdoses, medical emergencies, and welfare checks. Adding a behavioral health team to the fire department’s response portfolio feels like a natural extension rather than a radical experiment, which makes it politically easier to launch and institutionally easier to sustain.

The risk is that fire departments have their own institutional culture, one built around command structures, physical response, and hierarchical decision-making that may not align naturally with the de-escalation and client-centered approach that crisis work requires. Programs housed in fire departments that have maintained strong clinical outcomes tend to have independent clinical leadership that prevents the clinical model from being absorbed into the fire department’s operational culture.

Inside the Police Department

St. Petersburg’s CALL program was designed within the police department, in partnership with Gulf Coast Jewish Family & Community Services. The police chief directly shaped the program, chose the call types, and publicly owns the program’s success. The advantage is immediate buy-in from officers and dispatch, when the chief says “route these calls to CALL,” dispatchers listen, because it’s coming from within their chain of command. Olympia’s Crisis Response Unit (CRU), launched in 2019, is a team of unarmed civilian behavioral health specialists housed within the Olympia Police Department that responds independently to mental health, substance use, and homelessness-related calls routed through 911. Despite its home in the police department, CRU operates without officers on scene — responding to over 3,000 calls in 2023 with zero staff injuries caused by clients and only two requests for police backup.¹¹ It’s also worth noting that Eugene’s CAHOOTS program became the national model for mobile crisis response while housed under the police department.

A program housed inside the police department may face pressure to operate under police protocols rather than clinical best practices. Clinicians who report to a police captain may feel less empowered to push back on inappropriate referrals, resist pressure to share client information with investigators, or maintain the clinical independence that effective crisis work requires. And for some communities, a crisis team that operates under the police department’s roof may never be seen as truly separate from law enforcement, regardless of how the team actually behaves on scene.

That said, St. Petersburg’s results are strong. Over 6,500 calls handled with zero serious safety incidents.¹² The police chief explicitly rejected the co-responder model and designed CALL as a civilian-only team from the start. The program works because the institutional champion, the chief — fully committed to the clinical model even though it lives inside his department.

As a Standalone Agency

Some cities have created entirely new departments to house mobile crisis and other non-police public safety functions. This model provides the greatest autonomy and the clearest public identity, when the crisis team reports to its own director, there’s no ambiguity about whose mission it serves. But it requires the most political capital and the most time to build from scratch: new budget lines, new civil service classifications, new leadership positions, new interagency agreements, and new dispatch protocols that don’t exist yet. For example, Albuquerque Community Safety is one of the highest-volume alternative response programs in the country, handling over 100,000 calls since launch.¹³ It has a distinctive model where ACS triage specialists monitor the official CAD system and self-deploy to appropriate calls rather than waiting for police referral. Durham’s Community Safety Department, launched in 2021, is a standalone city department on equal footing with police and fire. Its HEART program, which dispatches unarmed teams of clinicians, peer specialists, and EMTs to behavioral health and quality-of-life 911 calls, has grown from 13 staff in its first year to over 50, responding to more than 32,000 emergency calls with the structural stability that comes from being a full city department.¹⁴

Albuquerque and Durham both created standalone departments with strong mayoral support.

What Actually Drives the Choice

The pattern across documented programs is clear: the institutional home matters far less than the institutional champion. Harris County’s success came through public health because the county judge provided sustained political support. St. Petersburg’s came through the police department because the police chief owned the design from day one. Oklahoma City’s came through fire because the fire department had 24/7 operational capacity. The common factor across programs that succeeded is someone with institutional authority who committed to the program through budget cycles and political resistance. The common factor across programs that stalled is the absence of that champion.

Decision 2: Who’s On The Team?

Team composition varies by city, shaped by both clinical model and labor market reality.

The Clinical Core

Every documented mobile crisis program includes licensed mental health clinicians at the master’s level: social workers, licensed professional counselors, or clinical psychologists. They conduct assessments, de-escalate crises, evaluate suicide risk, make treatment referrals, and determine the appropriate level of care.

The clinical licensing requirement constrains growth. There are roughly 122 million Americans, about a third of the country, living in federally designated mental health professional shortage areas.¹⁵ Twenty-six states report shortages of social workers available for mobile crisis work, with three states reporting what the National Research Institute calls “catastrophic” shortages.¹⁶ Psychiatry positions often remain unfilled for six months or longer.

Peer Support Specialists

Most programs pair clinicians with certified peer specialists, people with lived experience of mental illness, substance use recovery, or homelessness who have completed state-certified training programs. Durham added peer support specialists to its HEART program because, as program leaders described, the specialists “include people in recovery or who have lived unhoused in Durham and understand social service systems that often seem opaque and inaccessible.”²²

Peers are also the fastest-growing part of the crisis workforce for a practical reason: they face fewer credentialing barriers than licensed clinicians. At least 39 states now allow Medicaid reimbursement for peer support services.¹⁹ Several states, Colorado, Georgia, Arizona — have built mobile crisis team staffing models that deliberately pair a peer with a bachelor’s-level or master’s-level clinician, with a licensed clinician available by telehealth for consultation. This extends the reach of scarce licensed professionals without compromising clinical quality.

The risk with peers is burnout and re-traumatization. People with lived experience of crisis are being sent into crisis situations daily. Programs that don’t invest in clinical supervision, regular debriefing, manageable caseloads, and genuine mental health support for their own workers will burn through peer specialists faster than they can train new ones. Harris County’s experience illustrates this directly⁴, when crisis workers were employed through a contractor that couldn’t offer competitive benefits or adequate support, turnover was so high that it undermined program quality and forced the county to transition workers to in-house employment.

Paramedics and EMTs

Important when calls involve overdoses, medication issues, or co-occurring medical conditions. Oklahoma City includes paramedics on every team. Denver’s STAR program pairs a WellPower mental health clinician with a Denver Health paramedic or EMT on every van, which means the team can administer Narcan for overdoses, assess vital signs, and handle medical situations that a clinician-only team would need to call EMS for.

Programs without medical capability on the team coordinate with EMS for calls requiring medical intervention. In cities with significant overdose call volume, building medical capacity into the team eliminates a handoff that takes time during an overdose.

Case Managers

Case managers address the underlying conditions after the crisis clinician stabilizes the moment: navigating the behavioral health system, accompanying people to appointments, helping with insurance paperwork, connecting to housing and benefits.

Austin’s EMCOT provides 90 days of transitional crisis services. St. Petersburg’s CALL program has an open-ended commitment, in the words of program leadership, they “never close a client.” Guilford County, North Carolina reduced one group’s non-emergency 911 calls from 344 to 4 in a single month through sustained case management.⁶

Case management is the hardest component to fund. Medicaid reimburses the crisis response more easily than ongoing case management. Caseloads grow as the program matures, requiring continuous hiring. The outcomes from case management, fewer repeat crises, fewer ER visits, families stabilized, take months to materialize, which means programs that launch without follow-up capacity show strong initial call numbers but not cycle-breaking outcomes.

The Workforce Constraint, And What Cities Are Doing About It

Thirty states report workforce shortages as a significant barrier to mobile crisis operations, according to the National Research Institute’s 2023 survey.¹⁶

Cities and states have addressed this through several strategies:

Community college partnerships. Evanston, Illinois partnered with Oakton Community College to create an Emergency Crisis Responder Program, 120 hours of classroom instruction, 30 hours of skills training, and 1,000 hours of supervised field experience. The program was designed before the crisis team launched, which meant Evanston had a local talent pipeline from day one instead of competing with every hospital and clinic in the metro area for the same small pool of licensed clinicians.

State certification programs. Montana invested $500,000 in developing a crisis worker curriculum. New Hampshire’s Division for Behavioral Health created a Crisis Certification program and its community mental health centers are offering sign-on and retention bonuses. Ohio invested in expanding paid internships and scholarships for behavioral health students at two-year and four-year colleges, removing financial barriers to licensing exams, and establishing a Technical Assistance Center to help students navigate available funding, a comprehensive pipeline approach that addresses multiple bottlenecks simultaneously.

Compensation as a retention tool. Harris County brought workers in-house as county employees partly because the contractor model couldn’t offer benefits competitive enough to keep staff.⁴ Portland went further, designating crisis responders as formal first responders⁸ with full employment benefits, explicitly signaling that this is a permanent public safety profession, not grant-funded temp work. South Carolina⁹‘s Department of Mental Health recently implemented a 30% salary increase for mental health professionals in its crisis system.

Flexible staffing models. Several states have adopted team configurations that stretch scarce licensed professionals across more calls. Colorado allows a peer specialist and a bachelor’s-level clinician to respond together in the field, with a licensed professional available by telehealth for real-time consultation. Arizona allows mobile teams to be staffed with various combinations of behavioral health professionals, peer specialists, behavioral health technicians, and paraprofessionals, with a licensed clinician on call if not on scene. Georgia pairs a licensed clinician with a paraprofessional on the team and uses peer specialists more heavily for follow-up contacts and satisfaction surveys.

Decision 3: Co-Responder or Civilian-Only?

Cities choose between sending a clinician with a police officer (co-response) or sending a civilian team without police (civilian-only). The documented evidence favors civilian-only on key outcomes, but co-response remains the more common starting point.

Co-Responder: A Clinician and an Officer Respond Together

A clinician rides with a police officer, or they meet on scene. Co-response is faster to launch because it doesn’t require building separate dispatch and response infrastructure.

Denver invested heavily in co-responder teams starting in 2016, building 25 pairs by the time STAR launched in 2020 and planning to add seven more for round-the-clock coverage. Many cities operate both models simultaneously, co-responders for higher-acuity calls where a weapon or violence history is flagged, and civilian-only teams for lower-acuity calls that pass the safety screening at dispatch.

Co-response has structural limitations. The officer is still tied up on the call, so the workload relief that police departments value doesn’t materialize. A 2025 Wayne State University study found that co-responder programs produced arrest outcomes similar to law enforcement-only responses, while civilian-only teams achieved significantly better diversion and stronger connections to services.¹⁷

Civilian-Only: Unarmed Teams Respond Independently

The team deploys without police. Officers are available as backup if the situation escalates, but the team makes the initial contact alone. This is the model used by STAR in Denver, HEART in Durham, CALL in St. Petersburg, HART in Harris County, and the majority of the newer programs launching across the country.

St. Petersburg’s police chief explicitly rejected the co-responder model when designing CALL. His reasoning was direct: there are enough non-violent, non-criminal calls that can be “completely diverted” from police without any officer involvement. He didn’t want to launch a program that still required an officer on every call, that would defeat the purpose.

Every civilian-only program uses the same bright-line protocol: any indication of weapons, active violence, threats, or a crime in progress gets routed to police. The safety record across documented civilian-only programs:

• 6,500+ calls, zero incidents, St. Petersburg

• 16,000+ calls, zero injuries, Minneapolis

• 13,000+ calls, 98% no police, San Diego County

Denver has never called police for backup during a STAR intervention. And as the Marshall Project reported in 2024 after surveying experts across the field, there have been no known major injuries to any community responder on the job in any documented program in the United States.¹⁸

The Tradeoff

Multiple cities started with co-responder models and evolved toward civilian-only as the safety record built trust. Programs that remain co-responder permanently have not fully solved the workload problem and, if the Wayne State data holds across more jurisdictions, have not achieved the arrest diversion outcomes that civilian-only teams demonstrate.¹⁷

Decision 4: What Calls Does The Team Handle?

The Bright Line

Across every documented program: any situation involving a weapon, credible threats of violence, active physical altercation, or a crime in progress goes to police.²³

The Standard Call Types

The core call types that nearly every program handles from launch include welfare checks, suicidal ideation without a weapon, panic attacks, psychotic episodes, substance use and intoxication including overdoses (when medical capability is on the team), dementia-related confusion and wandering, family behavioral crises without violence, homelessness-related calls, and low-level disturbances like loitering or trespassing where no crime is in progress.

The Gray Zone

Every program confronts a set of call types where the right response isn’t obvious. Domestic calls where there’s no weapon but a volatile history. Scenes where the dispatcher can’t confirm safety conditions. Callers with prior encounters that involved escalation. Situations where the person is described as “agitated”, a word that covers everything from someone pacing and talking loudly to someone about to throw a punch.

Different cities draw this line differently, and it’s a genuine design decision with real consequences. The wider the net, the more impact the program has. The narrower the net, the easier the launch and the lower the perceived political risk. Some programs have limited themselves to a small percentage of eligible calls by excluding third-party callers, conflict resolution calls without a clear behavioral health nexus, and any call where the person is described as “aggressive”, which, as the Center for American Progress documented, can exclude calls where the person is simply shouting or upset but not dangerous.²⁴

The highest-volume programs expanded beyond initial narrow definitions. Denver STAR’s call types now include intoxication, suicidal ideation, welfare checks, indecent exposure, trespassing, syringe disposal, and requests for resources like shelter, food, vouchers, and transportation.²⁵ San Diego County’s teams respond on college campuses, tribal communities, grade schools, and a military base.²⁶

The Expansion Pattern

Starting narrow and widening over time can create the political space to expand, once dispatchers see the team handle thousands of calls without incident, the resistance to adding new call types drops. Denver started as a pilot in a single district with limited hours and a handful of call codes. Now it operates seven days a week across the city. The data from the first six months is what cities have used to make the case for expanding call types.

Decision 5: How Do Calls Reach The Team?

Dispatch integration is the operational bottleneck that determines whether programs get used. A well-designed program with a well-trained team accomplishes nothing if 911 dispatchers don’t route calls to it. Evidence from multiple cities shows that dispatcher hesitancy is the most persistent and underestimated operational barrier.

Why Dispatchers Hesitate

The problem is real, documented, and rational. Dispatchers have spent their entire careers sending police to crisis calls. They are trained to think about worst-case scenarios. They know about the one time twenty years ago that an officer arrived at a noise complaint and someone started shooting. As Daut’e Martin of the Law Enforcement Action Partnership puts it, the fear is rarely about an entire category of calls, it’s about one incident that sticks in institutional memory.²⁷

NYU’s Policing Project, which has studied dispatch integration in Denver, San Francisco, Tucson, Chicago, and Minneapolis, documented “significant hesitation” from dispatchers.²⁸ Some add police backup to every diverted call. Some forget to use the diversion option. Some refuse to route calls on liability grounds.

“If STAR can’t protect themselves, I’m not sending them by themselves.”

Denver 911 dispatcher²⁸

Researchers from NYU studying Denver and San Francisco found that existing dispatch protocols work reasonably well under ideal conditions, when callers provide clear information, when the call fits neatly into a recognized category, when the crisis team is available. But real calls are messy. Callers are vague, frightened, intoxicated, or reporting what a neighbor told them secondhand. The protocols most cities use rely on rigid yes/no prompts that leave little room for the kind of nuanced judgment that dispatchers actually have to make. Some of the most experienced dispatchers developed creative workarounds — reframing questions, stretching definitions, or entering responses after they’d already decided who to send — to route calls to the crisis team despite protocol constraints rather than because of them.

Three Dispatch Models

The Center for American Progress, the Law Enforcement Action Partnership, and NYU’s Policing Project produced the most comprehensive analysis of dispatch models in a 2023 report.²⁴ They identify three principal models:

911 Call Center-Led Dispatch. The 911 call-taker decides whether a call goes to police or the crisis team, using a decision tree, specific call codes, or clinical screening protocols. The dispatcher then routes the call through the same CAD system used for police. This is the model used in Denver, Olympia, Chicago, Cincinnati, Dayton, St. Petersburg, and San Francisco. The advantage is efficiency: no call transfer, no warm handoff, no second phone conversation. The call-taker makes one decision and the team deploys. The disadvantage is that it puts the clinical judgment squarely on call-takers who aren’t clinically trained, and the cultural change required is the hardest part of the implementation.

Embedded Professional-Led Dispatch. Clinicians or behavioral health navigators are physically placed inside the 911 call center. When a call-taker recognizes a potential crisis call, they do a “warm handoff”, staying on the line while introducing the caller to the embedded professional, who can resolve the situation over the phone or dispatch a mobile team. This is the model used in Austin, Louisville, Houston, and St. Louis. Austin puts 24 of its 71 crisis employees directly in the dispatch center.²⁰ Philadelphia’s embedded behavioral health navigators serve as consultants to call-takers, provide education, and host monthly sessions to build rapport and trust. Houston’s Crisis Call Diversion program, launched in 2015, was one of the earliest embedded models and has processed tens of thousands of calls. The advantage is clinical quality — the person making the dispatch decision has clinical training. The 80% of calls that can be resolved over the phone by a clinician never need a field response at all. The disadvantage is cost and the warm handoff itself, which adds time and can lose callers who don’t want to be transferred.

External Hotline-Led Dispatch. 911 call-takers transfer eligible calls to an external line, a behavioral health crisis hotline, a 311 social services line, or a direct program line — which handles the call and dispatches the team. This is the model used in Atlanta, Toronto, Baltimore, San Diego, and Dayton. The advantage is that callers who don’t want to interact with 911 at all can call the crisis line directly. In communities where people refuse to call 911 because of prior negative experiences with police, this creates an entirely separate entry point. The disadvantage is operational complexity — bridging two separate call centers with different technology systems, different protocols, and different institutional cultures creates opportunities for calls to fall through the cracks.

What Actually Works to Overcome Dispatcher Hesitancy

The research and field experience point to several concrete practices, and cities that implemented multiple strategies simultaneously saw faster buy-in than cities that relied on training alone:

Clear, published protocols. Denver 911 created an easy-to-understand flowchart showing exactly which call types get routed to STAR and exactly which circumstances trigger a police response instead. Call-takers are more likely to use the new system when policy, not personal judgment — dictates which calls go where. The NYU Policing Project emphasizes that call-takers need to feel that the institution has their back: if they follow the protocol and something goes wrong, the protocol is what gets reviewed, not their individual decision.

Embedded clinical support. Putting clinicians in the dispatch center, even without a full embedded-professional model — gives call-takers someone to turn to in real time when a call is ambiguous. It removes the burden of making clinical judgments from people who aren’t clinically trained.

Ride-alongs for dispatchers. The NYU Policing Project recommends bringing dispatchers on ride-alongs with crisis teams so they can see firsthand how calls are handled. When a dispatcher watches a two-person unarmed team calmly de-escalate a situation that the dispatcher would have sent three patrol cars to, the hesitancy drops.

Regular data feedback. Chicago’s CARE program sends quarterly updates to dispatchers with data on diverted calls and examples of people connected to services. When dispatchers see that hundreds or thousands of calls were handled safely, the abstract fear of “what if something goes wrong” gets counterweighted by concrete evidence that things are going right.

Self-dispatch capability. Some cities give crisis teams the ability to view pending calls in the CAD system and volunteer for ones they believe are appropriate. Dayton’s MRU responders can view calls pending for police and assign the calls to themselves. In Olympia, the Crisis Response Unit team can radio in to “poach” calls from the police channel. Denver recently gave STAR teams in-vehicle CAD terminals²⁵ so they can view call details directly rather than relying solely on radio communication.

Adding mental health as a formal 911 category. Austin’s dispatchers now ask callers: “Do you need fire, EMS, police, or mental health?” That structural change means mental health calls are identified from first contact, not discovered partway through a call that was initially coded as something else.

Building 988 as a parallel entry point. King County enables warm transfers between 911 and 988³⁴, where mental health counselors can dispatch mobile crisis teams directly. Cincinnati enables 911 call-takers to divert appropriate calls to a 988 Lifeline provider at Talbert House for de-escalation, even while a traditional response is on the way.

The Timeline to Expect

Dispatch buy-in is measured in months to years, not days to weeks. Andrew Dameron, the Director of Emergency Communications and 911 in Denver, has observed that dispatchers may feel nervous about deviating from traditional protocols.²⁹ The culture shift requires legal indemnity guarantees, clear and repeatedly reinforced flowcharts, sustained training, embedded clinical support, and consistent data showing that diverted calls resolved safely.

44%

The dispatch gap. Denver’s STAR program handles only 44% of eligible calls despite having 8 vans operating.²⁵

Decision 6: Hours and Geography

Almost no program launches at 24/7 citywide coverage.

Where Most Programs Start

Daytime or extended hours in the highest-need geographic zone. Denver STAR started as a pilot in a single police district with two teams operating Monday through Friday, 10am to 6pm. The program now operates seven days a week, 6am to 10pm, with 8 vans covering the city. Durham’s HEART began covering a small portion of the city before expanding citywide. The pilot model is standard because it lets the program build its safety record, train dispatchers, and work out operational problems before scaling.

The 24/7 Aspiration

Mental health crises don’t follow business hours. Suicide attempts peak in evening and early morning hours.

Programs that have achieved or are pushing toward 24/7 coverage include Tulsa (which launched at 24/7 by leveraging fire department infrastructure), King County (27 units operating around the clock), San Diego County (44 teams providing round-the-clock coverage), and Portland (city council passed a resolution with expanded budget to reach 24/7). The 2024 National Research Institute survey of 1,800 mobile crisis teams found that while 70% of programs report 24/7 availability, only 40% actually meet the minimum staffing threshold for true 24/7 coverage.³⁰

The Scale Gap Is Real

Denver’s 44% eligible-call capture rate tells the story.²⁵ Austin needs roughly three times its current staffing to reach all eligible calls.²⁰

Decision 7: In-House or Contracted?

Recent program closures have turned the in-house vs. contractor debate from theoretical to documented.

Contracting Is Faster

No new city job classifications, no civil service navigation, no HR infrastructure to build. St. Petersburg partners with Gulf Coast Jewish Family & Community Services. Austin’s EMCOT is operated by Integral Care. San Diego County contracts with multiple behavioral health providers across its 44-team system. For a city that needs to launch in months rather than years, contracting removes a major bureaucratic bottleneck.

In-House Is More Durable

Harris County learned this the hard way. The county initially contracted with external providers for crisis response staffing, but contractor turnover was so high that it was undermining program quality. Workers employed by contractors couldn’t get benefits competitive enough to stay, they’d build skills on the crisis team and then leave for a hospital or clinic job that offered health insurance, retirement, and predictability. Harris County transitioned to county employment, with full benefits, and retention improved.

Durham’s HEART staff are city employees. Portland designated crisis responders as formal first responders with full employment benefits.⁸

The Polling Backs the Permanence Argument

In a national survey of 2,503 registered voters conducted in 2024, 78% of Harris County residents agreed¹⁰ that “all first responders responsible for public safety should be public employees working for a county agency.” Seventy-four percent agreed that “public agencies are harder to eliminate when political administrations or budget situations change.” That’s the durability case in public opinion terms: residents understand intuitively that a contracted program can be quietly defunded when political winds shift, while a program staffed by city employees with civil service protections is institutionally harder to eliminate.

The Cautionary Case: CAHOOTS

The most visible example of what happens when a program depends on contract funding without adequate public investment is CAHOOTS itself. For 36 years, CAHOOTS was the nation’s most recognized mobile crisis program, the model that inspired hundreds of others. It operated through a contract between the City of Eugene and White Bird Clinic⁵, a nonprofit.

In April 2025, the program shut down entirely in Eugene. White Bird Clinic’s Interim Executive Director Amée Markwardt told reporters that the city’s contract covered only about 40% of CAHOOTS’ operating costs. The rest depended on federal grants that evaporated when the new administration took office. The city and clinic severed their contract. Approximately 20 workers were laid off. Justin Madeira, the program coordinator, said bluntly that the program hadn’t foreseen the cost pressures from inflation and cost of living increases. A grassroots “Friends of CAHOOTS” coalition organized to bring the program back, and by late 2025 the City of Eugene issued an RFP for a new crisis response program, but the damage was done. The nation’s oldest and most celebrated mobile crisis program collapsed because its public funding covered less than half the operating cost.

In Montana, two mobile crisis programs in Great Falls and Billings also recently shut down, leaving just six units statewide. Montana’s Medicaid system only reimburses mobile crisis teams for time spent actively responding to calls in the field, not for documenting encounters, waiting between calls, training, or any of the infrastructure time that makes the service possible. Angela Kimball of Inseparable, a mental health policy advocacy organization, frames the structural problem: cities pay for fire and police capacity to be at the ready regardless of whether they’re actively on a call — but crisis programs are often funded only for active response time, which makes it impossible to maintain a stable workforce.

The Pattern

Many cities start with contractors for speed and transition to in-house as the program matures. The transition itself is a significant administrative project, new job classifications, benefits packages, sometimes union negotiations. But cities that have done it report improved retention and accountability. The pattern from CAHOOTS, Great Falls, and Billings is consistent: programs built on expiring funding without a transition plan proved structurally fragile, regardless of how well they performed.

Decision 8: Follow-Up and Case Management

Follow-up capacity varies widely across programs.

The Minimum

Follow-up within 24 to 48 hours to check that the person is connected to services and stable.

The Standard

Active case management for days or weeks, navigating the behavioral health system, accompanying people to appointments, helping with insurance paperwork, connecting to housing, coordinating with family members. This requires dedicated case management staff, not just crisis response clinicians pulling double duty between field calls. The clinician who de-escalated the crisis at 2am should not also be the person making the Medicaid eligibility call at 9am the next morning — that’s a different skill set, a different workflow, and a recipe for burnout when programs try to collapse them into one role.

The High End

Austin provides 90 days of transitional crisis services after the initial response.³¹ St. Petersburg’s CALL program maintains an open-ended commitment; program leadership describes it as “never closing a client.”¹²

The Number That Tells the Story

344 → 4

Guilford County, North Carolina. 344 non-emergency 911 calls from one group in a single month. After social workers provided sustained case management following the initial crisis contacts, that number dropped to 4. A 99% reduction in repeat calls from the highest-frequency users of the 911 system.

The Funding Tension

Follow-up is the hardest component to fund because the payoff is delayed and distributed. Medicaid reimburses crisis response more readily than ongoing case management. The outcomes from case management — fewer ER visits, fewer repeat 911 calls, fewer arrests, families that stay intact — materialize over months, not days, which means the budget justification always lags behind the budget ask. And case manager caseloads grow as the program matures, requiring continuous hiring just to maintain the same level of service per client.

Programs that launch without follow-up capacity show strong initial call numbers but not cycle-breaking outcomes.

Decision 9: Safety Protocols

Every documented program includes protocols for when a call exceeds what an unarmed team can safely handle.

The Universal Standard

These elements appear in every documented program:

Pre-arrival scene assessment. The dispatcher screens for weapons, violence history, active threats, and any prior encounters with law enforcement at the address. The crisis team receives this information before they arrive. Washington State’s mobile crisis response program guide specifies that information regarding dangerousness must be made available to responding staff without unduly delaying the response.

Two-person minimum. Every team deploys with at least two staff members. Washington State’s program guide makes this a regulatory requirement.³²

Direct communication channel with police. Crisis teams carry radios on the police channel or have a dedicated communication line to dispatch for requesting backup.

Clear threshold for disengaging. Any sign of a weapon, any credible threat of violence, any physical aggression toward the team triggers an immediate departure. The team does not negotiate, does not try to de-escalate a situation that has become physically dangerous, does not wait for a supervisor’s approval. The team’s judgment on scene safety is final.

Authority to decline any call. Teams can refuse to respond to a call before arrival if the dispatch information raises concerns, or leave a scene at any time after arrival.

The Safety Record

The numbers across documented programs are consistent :

6,500+

calls, zero incidents — St. Petersburg

16,000+

calls, zero injuries — Minneapolis

13,000+

calls, 98% no police — San Diego County

Denver has never called police for backup during a STAR intervention.³³ As the Marshall Project reported in 2024, there have been no known major injuries to any community responder on the job in any documented program in the United States.¹⁸

What Programs Have Built in Advance

Cities with the longest-running programs documented pre-built crisis communications plans, established media relationships before any incident occurred, and maintained data systems capable of producing context immediately.²⁸

Decision 10: Metrics That Tell the Story

Cities with durable programs built data infrastructure before launch.

Track These From Day One

Documented programs report tracking these metrics from launch:

Calls handled. Total call volume.

Calls resolved without police. San Diego County reports 98%; Chicago FACT reports 94%.³⁴

Officer hours saved. Durham publishes 10,000+ officer hours freed.³⁵

Repeat calls reduced. Guilford County’s 344-to-4 figure is the most documented example.⁶

ER diversions. SAMHSA documented 79% reduction in hospitalization costs.³⁶

Arrests avoided. Wayne State found mobile crisis was the only model to produce statistically significant arrest reductions.¹⁷

Responder safety incidents. Minneapolis: zero serious injuries across 16,000+ calls.³⁷

Connection-to-services rate. Harris County reported a 228% increase in service linkages after transitioning to in-house staffing.³⁸

How Cities Use the Data

Oklahoma City reports year-over-year police call volume changes — a 57% decline⁷ in calls to which police previously responded — which is a system-level metric that shows impact beyond the program itself. They accomplished this by integrating 911 and 988 to seamlessly hand off most mental health calls from 911 to 988 dispatchers who can resolve the call or send mobile crisis units. Minneapolis tracks responder safety across all calls (zero serious injuries) — the number that directly answers the safety concern. Harris County polls public perception before and after exposure to program data, showing that support rises from 78% to 88% when residents learn about results — demonstrating that transparency about performance is an asset, not a risk.

The Tension: Numbers and Stories

Cities with the most durable programs combined hard metrics with individual case narratives in their reporting to councils and the public.²⁸

The Bottom Line

The pattern across the cities with the strongest programs is consistent: an institutional champion mattered more than the institutional home. Heavy investment in dispatch integration separated programs that got used from programs that didn’t. Starting narrow on call types and hours — with a documented plan to expand — produced better outcomes than ambitious launches that couldn’t sustain their scope. Follow-up capacity built from day one generated the data that funded the program’s second year. And the programs built on expiring grant funding without a transition to stable public revenue — CAHOOTS being the most visible example after 36 years — proved structurally fragile regardless of their performance record.

The gaps in design knowledge — timelines for how long each phase takes, head-to-head comparisons between different program architectures, long-term workforce pipeline outcomes — are covered in What Are The Risks?