Card 10

How Are Cities Designing These Programs?

There are roughly ten design decisions every city faces when building a Community Safety Department, and they are 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 department.

Robert Blaine, a program specialist at the National League of Cities who has tracked program design across dozens of jurisdictions, makes an observation that keeps surfacing in the data: cities do not need to pick one model permanently.¹ Programs that started with a limited scope and expanded as they built evidence tended to perform better than those that either overdesigned on paper before launching or launched so cautiously that the program never handled enough calls to demonstrate its value.

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

Decision 1: Where Does the Department Live?

This is the foundational choice. Where the department sits organizationally determines its culture, its hiring pipeline, its funding streams, its relationship with dispatch, and how law enforcement and the public perceive it. Six documented models exist, each with real-world implementations.

Standalone city department (“fourth branch”). Albuquerque built the first cabinet-level Community Safety Department in the country, co-equal with police and fire.² Durham created a standalone Community Safety Department housing its crisis response divisions.³ Portland formally established Portland’s crisis response department as an equal branch of public safety in June 2025, with staff receiving full first-responder employment benefits.⁴ The advantage is maximum autonomy: no mission confusion, clearest public identity, co-equal budget authority, and the ability to build an institutional culture from scratch rather than inheriting one. The constraint is that everything must be built from scratch: HR classifications, training systems, dispatch integration, facilities, fleet, and political support sufficient to survive the first budget cycle. Albuquerque invested in a 10,800-square-foot headquarters, branded vehicles and uniforms, and a formal training academy.⁵ ⁶ Mayor Tim Keller described the headquarters as a “massive symbol that we are never going to give up on anywhere in our city.”⁵ The infrastructure investment signals permanence, not a pilot.

Public health department. Harris County runs its crisis program through the public health department.⁷ Berkeley operates its Specialized Care Unit through Health, Housing, and Community Services. San Diego County’s 44-team program sits in County Behavioral Health. This model gives the program clinical autonomy, aligns it with healthcare rather than law enforcement, and uses existing Medicaid billing infrastructure. The constraint is that public health departments lack emergency response culture: they must build dispatch integration from scratch, and the shift from office-based care to field-based crisis response requires a fundamentally different operational model. Harris County initially contracted with a nonprofit for field operations, which failed, forcing the county to bring the program fully in-house.⁷

Fire department. At least six programs are now housed in or partnered with fire departments, an emerging national pattern. Eugene’s crisis program was moved from police to fire department oversight after 33 years before collapsing entirely in April 2025.⁸ Philadelphia’s program launched as a fire department and behavioral health partnership. Salt Lake City houses social workers within the fire department. Tulsa, UC Davis, and Oklahoma City have similar arrangements. Fire departments offer existing dispatch integration, 24/7 shift infrastructure, a medical rather than enforcement identity, and high public trust. The constraint is fire union dynamics: some departments resist scope expansion beyond traditional fire and EMS, and adding behavioral health workers to a fire department requires negotiating credential hierarchies.

Police department. St. Petersburg’s crisis program is housed within the police department.⁹ The advantage is immediate dispatch integration and institutional buy-in from officers. The St. Petersburg police chief explicitly chose civilian-only response over co-response, saying: “There are enough calls that are non-violent and non-criminal that we can completely divert these.”⁹ The constraint is risk of mission drift toward enforcement, community trust concerns in neighborhoods where police relationships are strained, and questions about whether clinicians have genuine clinical autonomy when housed within a law enforcement agency.

Nonprofit contractor. Denver’s crisis program uses a hybrid model: one nonprofit employs the clinicians, a public hospital employs the paramedics, and the program sits within the Department of Public Health and Environment.¹⁰ Eugene’s crisis program operated through a nonprofit clinic for its entire 36-year history.⁸ The advantage is workforce flexibility, clinical expertise, Medicaid billing capability, and insulation from government bureaucracy. The constraint is contract vulnerability: Eugene’s program collapsed in April 2025 when its nonprofit operator suffered financial failure,⁸ and Harris County’s contractor failed badly enough to force a full in-house transition.⁷ The documented lesson from both cases: nonprofit contractors can launch programs faster, but they create a dependency that can prove fatal if the contractor fails.

The institutional home question for a CSD specifically. The CSD model, by definition, is the standalone fourth branch. But the component programs within a CSD can draw on any of these institutional models. Albuquerque’s ACS is fully city-operated. Durham’s CSD is fully city-operated. The distinction that makes a CSD a CSD rather than a collection of programs is the unified governance: one director, one budget line, one training system, one dispatch integration, and one political identity as a co-equal branch of public safety. 62% of voters prefer this unified department model over operating the same programs as separate initiatives across different agencies.¹¹

Decision 2: Who Is on the Team?

The team composition determines what the team can do on scene. Clinical capability and workforce constraint pull in opposite directions, and the fundamental tradeoff is capability versus scalability.

Clinician plus peer support specialist. A widely used model.¹² A licensed clinician (typically a master’s-level social worker or counselor) provides clinical assessment, safety planning, and the authority to evaluate for involuntary psychiatric holds. A peer support specialist with lived experience in mental illness, addiction, or homelessness provides relational trust, rapport, and a different kind of expertise that clinical training alone does not produce. The constraint is the clinician: severe national shortage (half the U.S. lives in behavioral health workforce shortage areas), high cost ($60,000 to $90,000 or more), and not all clinicians are suited to or interested in street-level crisis work.

Clinical-medical teams. Some cities, including Denver and Salt Lake City, pair a licensed clinician with an EMT or paramedic, adding medical capability to every response. Because the team includes a licensed clinician, encounters can meet Medicaid billing requirements. Denver reports 96% of encounters are billable through Medicaid under this model (most programs have not reported this metric).¹⁰ The constraint is cost and staffing complexity.

Clinician plus peer plus EMT (three-person team). Durham’s CSD fields three-person teams as the maximum-capability model.³ This team can handle clinical assessment, relational engagement, and medical intervention simultaneously. The tradeoff is cost: three salaries per van. But the scope of practice is the broadest of any documented model, and Durham reports the team has never needed police backup.¹³

Community responder (peer plus case manager, no clinician). Albuquerque’s approach for many of its call types.² Trained civilian responders handle welfare checks, disputes, quality-of-life calls, and homelessness encounters without a licensed clinician on scene. The capability includes rapport, resource connection, basic assessment, and service referral. It cannot include clinical assessment, involuntary commitment evaluation, or medical intervention. The tradeoff is the mirror image of Durham: easier to staff, cheaper per team, broader reach, but narrower scope. Albuquerque must have clear escalation protocols for calls that turn out to need clinical response, and its less-than-1% police backup rate suggests the protocol works for its call mix.¹⁴

The capability-scalability spectrum in practice. Denver and Durham chose maximum capability per team: every van has clinical and medical capability, Durham adds peer support. Albuquerque chose maximum scalability: 140 responders covering the broadest scope of call types at lower credential requirements per team.¹⁴ Both approaches are defensible. The choice depends on call mix (how many calls require clinical assessment versus general support), workforce availability (how many clinicians can a jurisdiction actually recruit), and Medicaid strategy (clinician-staffed teams can bill Medicaid; community responder teams generally cannot for the same encounter types).

Several states have adopted flexible staffing models that stretch scarce licensed professionals across more calls. Colorado allows a peer specialist and a bachelor’s-level clinician to respond together, with a licensed professional available by telehealth. Arizona allows mobile teams staffed with various combinations of behavioral health professionals, peer specialists, and paraprofessionals, with a licensed clinician on call.¹² These models design around the supply constraint rather than requiring a licensed clinician on every team.

Decision 3: Civilian-Only or Co-Response?

This decision has implications for capacity, outcomes, and politics, and the evidence is directional.

Co-response pairs a clinician with a police officer. They respond together. Council members and police chiefs tend to prefer it because an armed officer is present, and it launches faster because it does not require building a separate dispatch and response infrastructure. Denver invested heavily in co-response starting in 2016, building 25 officer-clinician pairs before launching its civilian-only program in 2020.¹⁰

Civilian-only response sends a team without police. The team responds independently. Police are available through fallback protocols if the situation escalates beyond what the civilian team can manage.

Independent research supports the civilian-only model. A Wayne State University study across five Michigan localities found that civilian-only teams showed higher diversion rates from criminal legal and medical systems, while co-response produced arrest outcomes similar to police-only response.¹⁵ A Community Mental Health Journal review reached the same conclusion: civilian-only crisis response was the only model with significant arrest reductions.¹⁶

CSD programs are civilian-only by definition. Albuquerque has completed 100,000-plus calls with less than 1% requiring police backup.¹⁴ Durham operates both under one umbrella: civilian-only field teams for non-violent calls and an officer-clinician unit for calls involving weapons or violence threats.³ In practice, civilian-only and co-response function as complementary segments serving different call types. The civilian-only model handles the vast majority. The co-response model handles the narrow set of calls where a weapon or violence threat has been reported but a behavioral health component is also present.

Decision 4: What Calls Does the Department Handle?

Every CSD maintains a hard safety boundary: weapons, credible violence threats, active physical altercation, and crime in progress go to police. This boundary is enforced at dispatch. No documented CSD sends unarmed responders to calls with reported weapons.

Within that boundary, the design question is how broad to go. Three documented approaches exist along a specialist-to-generalist spectrum.

Specialist (narrow scope). Denver’s crisis program handles seven specific behavioral health call types: intoxicated persons, suicidal series, welfare checks, indecent exposure, trespassing or unwanted persons, assist calls, and syringe disposal.¹⁰ This narrow scope means every call is well-matched to the team’s clinical capability. It also means the department handles a smaller volume.

Generalist (broad scope). Albuquerque ACS handles the broadest range: “down-and-outs” (intoxicated or incapacitated individuals), abandoned vehicles, non-injury traffic accidents, needle pickups, welfare checks, mental health crises, homelessness encounters, substance use calls, parking disputes, trespassing, and neighbor disputes.² This broad scope captures more call volume and more fully relieves police workload, but requires community responders to handle situations that may be outside their primary training.

Hybrid (segmented by acuity). Durham’s CSD operates four segments, each handling different call types at different acuity levels.³ Community Response (civilian-only) handles non-violent behavioral health calls. Co-Response (officer plus clinician) handles calls with weapons or violence threats. Crisis Call Diversion (clinician in the 911 center) handles calls that can be resolved by phone. Care Navigation (case managers) handles follow-up for repeat callers. This segmented approach matches team capability to call acuity, but requires more complex dispatch routing.

The documented expansion pattern is consistent across jurisdictions: start narrow, build a safety record, then expand. Denver launched with seven call types and maintained that scope. Oklahoma City and San Diego County both started narrow and expanded their eligible call types as the safety record accumulated and police leadership gained confidence.¹⁷ The cities that tried to launch with a broad scope before building a track record faced more political resistance.

Decision 5: How Do Calls Reach the Department?

Dispatch integration determines whether a CSD is used. A department that dispatchers do not route calls to handles no calls.

Three principal dispatch models have been documented, and the evidence distinguishes them clearly.

911 call-center-led dispatch. The dispatcher decides whether a call goes to police or the CSD, using a decision tree, specific call codes, or screening protocols. Denver, St. Petersburg, and Albuquerque use this model. The advantage is efficiency: one decision, one dispatch, no handoff. The constraint is dispatcher hesitancy. NYU researchers documented Denver dispatchers who “forget” the alternative response option, worry about personal liability, and default to sending police.¹⁸ Denver’s crisis program reaches approximately 50% of eligible calls, with dispatch hesitancy identified as a primary cause.¹⁰

Embedded clinician dispatch. Clinicians or behavioral health professionals are physically placed inside the 911 center. When a call-taker recognizes a potential crisis call, the caller is transferred to the embedded clinician, who can resolve the situation by phone or dispatch a field team. Austin places 24 of its 71 crisis employees directly in the dispatch center.¹⁹ Philadelphia embeds behavioral health navigators who consult with call-takers. Houston’s Crisis Call Diversion program, one of the earliest embedded models, has processed tens of thousands of calls. The advantage is clinical quality: the person making the dispatch decision has clinical training, and roughly 80% of calls that reach an embedded clinician can be resolved by phone without sending anyone to the field.¹⁹ The constraint is cost and the warm handoff itself, which adds time.

Dual-mechanism dispatch. Durham combines both: an automated CAD system (One Solution/Central Square) flags all eligible calls based on call type codes around the clock, independent of any individual dispatcher’s judgment, AND embedded clinicians in the 911 center screen calls in real time.³ Two independent systems catch calls. This dual approach produces the highest documented call capture rate. The automated system removes the dispatcher discretion problem. The embedded clinician adds clinical judgment. The consistent finding across jurisdictions is that investing in dispatch infrastructure (systems and embedded personnel) produces higher call capture than investing in dispatcher training alone.

For a CSD specifically, dispatch integration is the architectural question that separates a department from a collection of programs. A CSD with its own dedicated dispatch code, CAD flagging, and ideally an embedded clinician is integrated into the emergency response system. A CSD that depends on individual dispatchers remembering to send it calls is operationally fragile regardless of how good the field teams are.

Decision 6: Hours and Geography

Almost no CSD launches citywide 24/7. The documented pattern is to start in high-demand zones during high-demand hours and expand as the workforce pipeline allows.

Operating hours. Albuquerque achieved 24/7 coverage in September 2023.¹⁴ Durham expanded to 12 hours daily.²⁰ Denver’s crisis program operates 6 a.m. to 10 p.m.¹⁰ Seattle’s department runs 11 a.m. to 11 p.m. The overnight gap remains common across most CSDs. Three smaller programs (Indianapolis, Placer County, and UC Davis) have also achieved 24/7 coverage, but the majority of large-city programs have not.

The consequence of the overnight gap was documented in Albuquerque before its 24/7 expansion.²¹ During hours when the CSD was closed, police shot and killed 27-year-old Keshawn Thomas on a call type ACS handles routinely (Q09 covers this in detail). Albuquerque’s decision to expand to round-the-clock operations was driven in part by incidents like this. The gap persists in cities that have not reached 24/7 because clinicians willing to work overnight shifts in crisis field response are a small subset of an already constrained pipeline.

Geographic expansion. Most programs launch in a defined zone and expand. San Diego County grew from a single team to 44 teams covering the full county, including 700-plus schools, tribal communities, and Marine Corps Base Camp Pendleton.²² This expansion took years and was driven by demonstrated demand and accumulated safety data. The jurisdictions that expanded fastest were those where police leadership publicly requested more coverage in additional zones. San Diego County’s experience shows that the call types expand beyond the original scope: the team handles “mothers experiencing postpartum depression,” “parents whose children have physically assaulted them,” and “combative loved ones with dementia.”²² The biggest complaint from police agencies in the county is not that the teams are ineffective but that there are not enough of them.

Zone selection. Albuquerque deliberately located its headquarters in an area with high concentration of addiction and behavioral health issues.⁵ Station 14, a vacant fire station being converted into an ACS substation, follows the same logic.

The workforce pipeline determines the expansion timeline. Any city planning a CSD launch that has not started workforce development before it begins hiring is already behind. The cities that avoided staffing crises during expansion built their pipeline in advance. Albuquerque’s training academy produces cohorts of new responders, creating a renewable pipeline rather than one-time recruitment.⁶ Portland’s resolution directing first-responder designation with full employment benefits is a retention strategy as much as a political statement: competitive compensation reduces the turnover that forces constant recruitment.⁴ South Carolina’s Department of Mental Health recently implemented a 30% salary increase for mental health professionals in its crisis system, an acknowledgment that the pipeline problem is partly a compensation problem. The Evanston/Oakton community college partnership represents an emerging model: academic institutions producing credentialed workers specifically for alternative response roles.²⁹

Decision 7: City Employees or Contractors?

This decision has immediate consequences for quality control, retention, political identity, and vulnerability.

City employees (in-house). Durham made department staff direct city employees with full benefits and enrollment in the North Carolina State Retirement System.²⁰ Portland passed a resolution directing first-responder designation with full employment benefits, explicitly signaling permanence.⁴ Albuquerque’s responders are city employees working within a cabinet-level department.² The advantage is accountability (staff cannot refuse calls within the department’s scope), retention (government benefits and retirement attract and keep workers), and institutional identity (the department is a real government entity, not a contract arrangement). Director Smith noted that because Durham’s staffers work “for the city, rather than a private contractor, they can’t refuse calls that fall within the program’s criteria.”²⁰

Nonprofit contractor. Denver’s crisis program employs clinicians through a nonprofit contractor.¹⁰ Eugene’s crisis program was operated entirely by a nonprofit clinic.⁸ The advantage is speed (nonprofits can hire faster than government HR systems), clinical expertise (nonprofits may already have a behavioral health workforce), and Medicaid billing infrastructure (many nonprofits are established Medicaid providers). The disadvantage is contract vulnerability: Eugene’s program collapsed when its nonprofit operator failed financially.⁸ Harris County’s contractor failed badly enough to force a full transition to in-house operations.⁷ After that transition, service linkages increased 228%.

The cities that achieved the most stable operations (Albuquerque, Durham, Portland post-transition) are those where staff are direct government employees. 68% of voters say CSDs should be staffed by city employees.¹¹ The compromise model is to use a contractor for launch and plan a transition to in-house operations within a defined timeline.

Decision 8: Follow-Up and Case Management

The emergency response is the visible part of a CSD. The follow-up and case management component is where the repeat-crisis cycle actually breaks.

Integrated follow-up within the department. Durham’s case management division segment is a dedicated case management function within the department.³ Care navigators receive referrals from field teams and provide sustained engagement with people who cycle through crisis repeatedly. This integrated model means the department owns the full lifecycle from initial response through sustained follow-up. Guilford County, North Carolina documented the impact: one individual with 344 repeat 911 calls was reduced to 4 after sustained follow-up.²³

Contracted follow-up. Denver’s crisis program contracts follow-up to a community nonprofit ($2.3 million contract).¹⁰ The team responds to the crisis; the contractor handles what comes after. The advantage is that the contractor may have specialized community connections. The constraint is the handoff: the person must agree to engage with a new organization, and the contractor’s performance depends on a contract that can be terminated or underfunded.

No formal follow-up. Some programs handle the acute encounter and rely on referral to existing community resources. This model is the easiest to staff and fund, but it does not address the underlying pattern that produces repeat crises. Without follow-up, the same individuals cycle through emergency response indefinitely. St. Petersburg’s approach, which “never closes a client,” represents the opposite end of this spectrum: sustained engagement until the person’s situation stabilizes.⁹ Chicago’s FACT program provides 30-day follow-up after the initial crisis encounter. Ohio’s statewide youth crisis program provides six-week follow-up. The variation in follow-up intensity is large, and the comparative data isolating follow-up as the causal factor in reducing repeat crises is limited. But the pattern from programs that track this metric is consistent: longer and more intensive follow-up is associated with fewer repeat 911 calls.

The response philosophy shift. Traditional police response prioritizes quickly clearing calls to remain available for the next emergency. CSD responders can spend extended time addressing underlying issues. CSD responders routinely spend 30 minutes or more on a single encounter, and over an hour on suicide-related calls. This extended engagement time is not a sign of inefficiency; it is the mechanism through which the department identifies needs, builds rapport, and initiates the service connections that reduce future calls. Director Smith described what the department looks for in staff: “people who will draw near to people at that moment of crisis, that sense of physical proximity, of not othering or judging or being afraid.”²⁰ Durham deliberately calls the people they serve “neighbors” rather than “clients” or “consumers.”

The follow-up question connects directly to cost-effectiveness. The Durham fiscal evaluation found that the program “pays for itself through fiscal externalities,” with $902 in net savings per call.²⁴ Those savings come primarily from reduced downstream costs: fewer ER visits, fewer repeat police dispatches, fewer jail bookings. Programs without follow-up are less likely to generate these downstream savings because they address the episode without addressing the pattern.

Decision 9: Safety Protocols and Fallback

The safety record across documented CSDs is consistent: zero or near-zero serious injuries to responders across hundreds of thousands of encounters. The question is what protocols produce this record.

The hard boundary at dispatch. Every documented CSD enforces the same bright line: no weapons, no credible violence threats, no active physical altercation, no crime in progress. This boundary is enforced at the dispatch level, before the team is sent. The screening happens through call type codes, dispatcher assessment, and (in dual-mechanism systems) clinician review.

On-scene escalation protocols. When a situation changes after the team arrives, documented protocols include immediate disengagement and relocation to a safe distance, radio request for police backup, and clear criteria for when to disengage versus when to attempt continued de-escalation. Durham’s CSD staff reported that 99% felt safe on calls.¹³ Denver’s crisis program has not reported any police backup requests for safety through 2023 (the latest independently documented period).¹⁰

The fallback to police. Every CSD maintains the ability to request police if a situation escalates. Albuquerque’s less-than-1% backup rate over 100,000-plus calls shows the fallback is rarely needed.¹⁴

What the safety record does and does not tell us. The field is young, the scale of operations has not yet been tested over decades at the volume that police and fire departments handle, and dispatch screening means CSD teams are systematically routed away from the calls most likely to produce violence. The safety record reflects both protocol design and call selection.

Decision 10: Metrics and Accountability

The metrics a CSD reports determine whether it survives its next budget cycle. The documented pattern from cities that sustained political support is consistent: report metrics that answer the questions each stakeholder asks.

For the budget office: cost per response, downstream cost avoidance (ER diversions, jail diversions, reduced repeat 911 calls), Medicaid revenue generated. Durham’s NBER evaluation found $902 net savings per call.²⁴

For police leadership: officer hours freed, calls diverted from police dispatch, backup request rate. Durham documented 10,000-plus officer hours freed.²⁵ Oklahoma City achieved a 57% reduction in mental health dispatches.¹⁷ Albuquerque Police Commander Luke Languit credited ACS with enabling crime reduction by freeing officers to focus on felony cases.²⁶

For the swing-vote council member: quarterly response volume by district, safety incidents (or lack thereof), constituent feedback, and comparison to pilot-phase projections. The swing vote wants to see that the program is performing within the bounds of what was promised.

For the public: resolution rates without police involvement, response times, safety record, and stories. Numbers provide the structural case; individual encounters provide the human dimension. The department’s language choices matter here: Durham calls the people it serves “neighbors,” and Portland clients describe feeling “treated with compassion and dignity.”

The metric trap. The number of calls handled is a common metric but a misleading one in isolation. A CSD that handles 100,000 calls but provides no follow-up and generates no downstream cost reduction may be less effective than a CSD that handles 20,000 calls with integrated care navigation that breaks the repeat-crisis cycle. Volume measures activity. Outcome measures impact. Denver’s crisis program survived a mayoral transition because its Stanford evaluation (34% crime reduction, independently verified) provided an outcome metric that stood up to political scrutiny.²⁷ Albuquerque’s 100,000-call volume provides the activity metric.¹⁴

Cross-Program Patterns

Across documented CSDs, several patterns recur:

What every program shares. A hard safety boundary enforced at dispatch (no weapons, no active violence). Two-person team minimum. Some form of follow-up or care navigation. Police leadership support on the record. Survival through at least one leadership transition.

Dispatch infrastructure over dispatcher training. Durham’s dual-mechanism dispatch produces the highest documented capture rate.³ Denver’s training-only model produces documented hesitancy and approximately 50% capture.¹⁸ The dispatch design determines whether a CSD handles its intended call volume.

The capability-scalability tradeoff determines everything downstream. Denver and Durham chose maximum capability per team: every van carries clinical and medical capability. Albuquerque chose maximum scalability: 140 responders covering the broadest scope at lower credential requirements. There is no single correct answer. The choice depends on the jurisdiction’s call mix, workforce pipeline, Medicaid strategy, and how much capacity it needs to demonstrate in its first two years.

The overnight gap remains the norm. Albuquerque achieved 24/7 in September 2023, but most large-city CSDs still operate less than around the clock.¹⁴ The workforce constraint is the binding limit. Albuquerque and three smaller programs have reached 24/7, suggesting it is achievable but resource-intensive.

Compensation is a design decision, not an afterthought. Programs that pay crisis workers competitively retain them. Programs that do not lose them to hospitals, private practice, and other settings that pay better and demand less. Denver’s crisis program clinicians earn $65,000 to $78,200, with managers at $80,000 to $100,000 and a language differential of $1 to $2.50 per hour.¹⁰ Harris County brought workers in-house as county employees partly because the contractor model could not offer benefits competitive enough to retain staff.⁷ Portland went further: passing a resolution directing first-responder designation with full employment benefits, explicitly signaling that this is a permanent public safety profession, not grant-funded temporary work.⁴ Voters agree: 57% say CSD positions require at least three years of relevant work experience, 54% want specific college or post-graduate coursework, and 73% say approximately 1,000 hours of training is sufficient, comparable to 800-900 hours for police.¹¹

CSD responders carry more than people expect. Department field teams typically carry naloxone, basic medical supplies, material support (clothing, hygiene products, blankets), and extensive knowledge of local service networks. Some departments authorize responders to enter private residences. Many transport people to services directly. Albuquerque Division Manager Joshua Reeves described talking down a man threatening to jump from a bridge: “Officers now are learning to defer to us as sort of the people who might de-escalate.”²⁸ The toolkit reflects the department philosophy: meet people where they are, address immediate needs, and connect them to what comes next.

The cities that started narrow and expanded outperformed the cities that tried to launch at full scale. Denver, Oklahoma City, San Diego County, and Durham all began with limited scope and expanded as their safety record and political support accumulated. The expansion was driven by demand (police leadership requesting more coverage) rather than by design ambition.

The Bottom Line

Ten design decisions, made in sequence, determine what a Community Safety Department looks like in practice. No two cities have made the same choices at every decision point. The documented pattern across established CSDs: build for the jurisdiction’s specific context (workforce, politics, call mix, budget), prioritize dispatch infrastructure,³ start narrow and expand on evidence, and plan for the workforce pipeline before hiring begins.⁶

Source Appendix

1. Robert Blaine, National League of Cities, technical assistance and guidance on alternative response program design. NLC publications on municipal crisis response infrastructure.

2. City of Albuquerque, “What Is Community Safety?” ACS services and department page. https://www.cabq.gov/acs/services

3. CSG Justice Center, “Durham, NC — Expanding First Response Program Highlights,” updated December 2024. Four-division structure, dual-mechanism dispatch, three-person teams. https://csgjusticecenter.org/publications/expanding-first-response/program-highlights/durham-nc/

4. Portland City Council Resolution No. 37709, June 25, 2025. First-responder designation, full benefits. https://www.portland.gov/council/documents/resolution/adopted/37709

5. KRQE News 13, “Albuquerque Community Safety Dept. headquarters is now open,” June 2024. 10,800 sq ft, Keller quote. https://www.krqe.com/news/albuquerque-metro/albuquerque-community-safety-department-headquarters-is-now-open/

6. CNM, “CNM and City of Albuquerque Launch Annual ACS Academy to Train Alternative First Responders,” October 2025. https://www.cnm.edu/news/cnm-and-city-of-albuquerque-launch-annual-acs-academy-to-train-alternative-first-responders

7. FOX 26 Houston, Harris County HART, May 2024. Contractor collapse, 228% service linkage increase. https://www.fox26houston.com/news/harris-county-mental-health-program-faces-uncertain-future-amid-financial-concerns

8. NPR / KFF Health News, Aaron Bolton, February 5, 2026. Eugene shutdown, Montana defunding. https://www.npr.org/2026/02/05/nx-s1-5693908/police-mental-health-calls-988-911-mobile-crisis-teams

9. Results for America, “Non-police emergency response: St. Petersburg, FL.” Zero incidents, civilian-only choice. https://catalog.results4america.org/case-studies/non-police-emergency-response-st-petersburg-fl

10. Urban Institute, Denver STAR interim evaluation, 2024. Team composition, salary ranges, dispatch capture, contractor model. https://www.urban.org/research/publication/evaluating-alternative-crisis-response-denvers-support-team-assisted-response

11. Safer Cities, national survey of 2,400 registered voters. 62% unified model, 68% city employees, workforce expectations.

12. Goldman, Looper, and Odes, “National Mobile Crisis Team Survey Report,” NRI, November 2024. Team composition national patterns, state staffing flexibility. https://988crisissystemshelp.samhsa.gov/sites/default/files/2024-11/National%20Mobile%20Crisis%20Survey%20ReportvFINAL.pdf

13. Tradeoffs / The Marshall Project, “How Durham Got Police Onboard with Unarmed Crisis Response,” May 2, 2025. 99% felt safe, zero backup for safety threats. https://tradeoffs.org/2025/05/02/how-durham-north-carolina-got-police-onboard-with-unarmed-crisis-response/

14. City of Albuquerque, “Albuquerque Community Safety Department Marks Four Years,” September 2025. 120,000+ calls, <1% backup, 140 staff, 24/7. https://www.cabq.gov/acs/news/albuquerque-community-safety-department-marks-four-years-of-impact-and-innovation

15. Leonard Swanson et al., Wayne State Michigan study, Psychiatric Research and Clinical Practice, May 2025. https://pmc.ncbi.nlm.nih.gov/articles/PMC12418740/

16. Matthew Bakko et al., Community Mental Health Journal, January 2025. https://doi.org/10.1007/s10597-024-01447-4

17. Oklahoma City expansion: KOSU, February 6, 2026. https://www.kosu.org/health/2026-02-06/oklahoma-city-plans-to-expand-mental-health-crisis-response-with-state-opioid-settlement-money

18. Policing Project at NYU School of Law, “Transforming Denver’s First Response Model,” 2023. Dispatcher hesitancy. https://www.policingproject.org/transforming-denvers-first-response-model-report

19. Austin embedded clinician model: KVUE (Austin), reporting on EMCOT expansion. 24 of 71 crisis employees work in the 911 dispatch center; 80% phone resolution rate. https://www.kvue.com/article/news/politics/austin-mayor-and-council/austin-city-council-expand-integral-cares-mental-health-crisis-outreach-team/269-e85ce07b-74b0-4fd7-b420-f6fc6d92e855. See also KUT Austin coverage of EMCOT operations.

20. Jeff Billman, “A New Model for Public Safety in Durham,” The Assembly NC, June 26, 2024. Smith quotes, city employee model, 12-hour coverage. https://www.theassemblync.com/politics/criminal-justice/durhams-new-model-for-public-safety/

21. Murat Oztaskin, “Sending Help Instead of the Police in Albuquerque,” The New Yorker, February 4, 2023. Keshawn Thomas, overnight gap. https://www.newyorker.com/news/dispatch/sending-help-instead-of-the-police-in-albuquerque

22. San Diego County News Center, Cassie N. Saunders, “Mobile Crisis Response Program Celebrates Four Years,” January 22, 2025. 2 pilot teams to 44 teams, K-12 school coverage. https://www.countynewscenter.com/mobile-crisis-response-program-celebrates-four-years/. Times of San Diego, January 26, 2025: 700+ schools. https://timesofsandiego.com/health/2025/01/26/mobile-crisis-response-teams-mark-4-years-of-tackling-mental-health-challenges-across-san-diego/. San Diego Union-Tribune, Tammy Murga, April 20, 2024: tribal communities (Viejas Reservation), Camp Pendleton, university campuses, unexpected call types. https://www.sandiegouniontribune.com/2024/04/20/thousands-of-people-seeking-help-did-not-get-a-police-response-thats-a-good-thing/

23. NACo / WFMY, Guilford County. 344→4 calls. https://www.naco.org/news/north-carolina-county-decreases-911-non-emergency-calls-connecting-residents-services-they

24. Bocar A. Ba et al., NBER Working Paper No. 34344, October 2025. $902 net savings. https://www.nber.org/system/files/working_papers/w34344/w34344.pdf

25. City of Durham / ICMA 2025 Award. 32,000+ calls, 10,000+ hours. https://icma.org/page/2025-community-health-safety-award-city-durham

26. Commander Luke Languit quote. https://www.cabq.gov/acs/news/albuquerque-community-safety-marks-100-000-calls-for-service-milestone

27. Thomas S. Dee and Jaymes Pyne, Science Advances, June 2022. 34% crime reduction. https://pmc.ncbi.nlm.nih.gov/articles/PMC9176742/

28. Joshua Reeves, ACS Division Manager. KOB4, Griffin Rushton reporting on ACS milestones. Reporter bio: https://www.kob.com/news_team/griffin-rushton/. See also KRQE coverage of ACS operations.

29. Oakton College, “City of Evanston Partners with Oakton for C.A.R.E. Responders Training,” May 28, 2024. 28-hour Emergency Crisis Responder certification, nine-employee inaugural cohort. https://www.oakton.edu/about/news/2024/05/city-of-evanston-partners-with-oakton-for-care-responders-training.php. City of Evanston: https://www.cityofevanston.org/Home/Components/News/News/6297/17

Sources