What Are the Risks?

Mental health dispatch integration has documented failure modes, structural fragilities, and honest gaps in what the current evidence can tell decision-makers. Four failures are documented with named programs and identified causes.

Documented Failures and Near-Collapses

Austin EMCOT budget cut (2024–2025). Austin’s EMCOT program, designated a “national role model” by the U.S. Department of Health and Human Services, faced a 33% budget cut after voters rejected Proposition Q in November 2024. The city’s contribution dropped from $3 million to $2 million annually for FY2026. The program maintained 24/7 operations but required staffing adjustments. A well-functioning program with national recognition can still lose a ballot measure when its funding depends on a single city appropriation subject to voter approval.

Houston CCD overnight gap. Houston’s Crisis Call Diversion program launched in 2015 and has operated 6 AM to 10 PM since its inception. From 10 PM to 6 AM, behavioral health calls route to police — the default the program was built to replace. Houston has not published documentation explaining why the overnight gap has persisted for a decade despite the program’s documented savings and success rate. The gap illustrates a common pattern: programs that launch at partial coverage find organic expansion to 24/7 difficult without a specific budget event driving it.

Baltimore ARPA funding cliff. Baltimore’s behavioral health 911 diversion program was funded through ARPA through 2026. The program added embedded behavioral health clinicians in the 911 center starting in 2024 and expanded eligible call types. What happens when ARPA funds expire has not been publicly resolved as of early 2026 — documenting the pattern of programs launching on time-limited federal grants without a documented transition plan to sustainable funding.

The Baltimore pattern is not unique: many dispatch integration programs that launched or significantly expanded between 2021 and 2024 used ARPA funding that expires in 2026. Programs without documented transition plans to general fund appropriation, state contracts, or telecom fee revenue face a structural fragility at the exact moment they are publishing their early outcomes. The programs that survive ARPA expiration tend to be those that treated ARPA as startup capital rather than operating budget — and built a recurring revenue base during the ARPA period.

National coverage hours gap. A 2024 National Research Institute survey found approximately 40% of programs claiming 24/7 crisis response coverage actually staff all shifts. The gap between claimed availability and actual staffing is a national pattern, not a single program failure. Programs that claim 24/7 availability in their public materials but do not staff overnight shifts are providing callers a false expectation about what will happen when they call at 2 AM.

Columbus Right Response Unit: low capture under training-only design. Columbus launched the Right Response Unit in 2021 with specially trained 911 dispatchers who could divert calls to social workers. In its first year, the program diverted approximately 10% of eligible mental health calls. The LEAP/CAP/NYU report identifies this as the structural limitation of dispatcher-training-only models: the routing decision stays with a non-clinical dispatcher who defaults to police under uncertainty.

The Dispatcher Override Problem

The LEAP/CAP/NYU report and the Wayne State University 2025 study both identify dispatcher hesitancy as a persistent implementation gap — dispatchers who have the option to route to clinical staff but continue to send police out of habit, risk-aversion, or distrust of the new system.

Baltimore’s consent decree documentation identified a specific version: call-takers diverted calls to the 988 pathway only when a caller explicitly stated a mental health diagnosis. The behavioral health component was present but unstated in the calls that were missed. The program responded with training refreshes and job-aid redesign — but the gap demonstrates that even programs with active clinical staff in the system face a capture problem at the dispatcher recognition point.

The Wayne State 2025 study found 911 professionals coded just over half of calls that law enforcement later classified as mental health-related — direct evidence that dispatcher recognition gaps exist even in jurisdictions with some mental health routing capacity.

The dispatcher override problem is most acute in Model 1A programs (dispatcher-training-only) and Model 1B programs (warm transfer to 988), where the dispatcher still makes the initial behavioral health identification. In Model 1C programs (embedded clinician), the clinician can monitor the call stream and flag calls the dispatcher might have missed — but the warm handoff from dispatcher to clinician still requires the dispatcher to recognize the call type first. The LEAP/CAP/NYU report identified this as the residual capture gap even in embedded programs: calls the dispatcher routes to police before the clinician has visibility.

Mitigation approaches documented in practice: Austin’s continuous dispatcher training through “coach reviewing game film” feedback sessions; Durham’s automated CAD flagging that catches behavioral health calls without depending on dispatcher recognition; and Dayton’s ride-along program where dispatchers accompanied clinical response teams, building confidence in what the program could handle.

Safety Risk: Ambiguous Calls

Every documented program has built a category of excluded calls — situations where weapons, active threats, or unknown safety factors require police response. The documented programs report strong self-reported safety records: Durham’s HEART responders reported feeling safe 99% of the time in nearly 25,000 calls; Minneapolis reported no serious responder injuries in 16,000+ calls. Both figures are program-reported.

The safety risk documented in practice is not programs being sent to clearly dangerous situations. It is misassessment of ambiguous situations at the dispatch level: a caller who does not mention a weapon that is present, a welfare check on someone with no violence history who is acutely destabilized. Albuquerque’s ACS required police backup on less than 1% of FY2025 calls — but that 1% represents real situations where dispatch-level clinical assessment was insufficient.

Workforce Constraint

Programs that have built embedded clinical capacity face a shared workforce constraint: licensed mental health clinicians are scarce, and the dispatch environment requires a specific professional profile.

The federal HRSA designates more than half of all U.S. counties as behavioral health workforce shortage areas. Programs staffing 911 centers with licensed clinicians compete with hospitals, outpatient clinics, and private practices for workers from an already constrained pool.

The dispatch environment adds a second filter: not all clinicians suited to mental health work are suited to phone-only, high-velocity, non-continuous-care work. Austin’s EMCOT has invested in training and support for dispatch clinicians, including the “coach reviewing game film” feedback model — recognizing that the dispatch environment is professionally distinct from other clinical settings. Kedra Priest, Practice Administrator of Crisis Services at Integral Care, has noted that clinicians must be comfortable handling crises at the phone level without the full information a field visit would provide.

Evidence Gaps

No independent evaluation of embedded dispatch clinicians specifically. The outcome data for dispatch integration is primarily program-reported. No independent evaluation with comparison groups, pre/post design, and controls for confounding factors has been published on the embedded dispatch clinician model. The Stanford Gardner Center’s 2024 study evaluated field co-response in California. The Wayne State 2025 study evaluated dispatch coding accuracy, not program outcomes. Neither directly evaluates what embedded dispatch clinicians produce.

Outcome data for callers is largely absent. Programs track call diversion rates. Few track what happened to the person: whether the crisis resolved, whether they connected to ongoing services, whether they experienced subsequent emergency contacts. The field can demonstrate that routing changed; it cannot yet demonstrate that outcomes for people in crisis improved.

Coverage gaps make cross-program comparison unreliable. Programs operating 8–12 hours per day capture a different call population than programs operating 24/7. Comparing diversion rates across programs without controlling for coverage hours produces misleading comparisons.

Scope Limitations

The program only reaches callers who dial 911. People who fear police contact and don’t call 911 are not reached by dispatch integration regardless of how good the clinical routing is. The 911 call stream is the entry point; callers who avoid 911 are outside the program’s reach entirely.

Follow-up capacity affects whether phone resolution persists. When a dispatch clinician resolves a call by phone, the person received a clinical response at the moment of crisis. If nothing changes in their ongoing situation, the next crisis produces the next call. Philadelphia’s development of six weeks of follow-up care after initial crisis contact is an attempt to address this structural limit, but follow-up capacity depends on the downstream behavioral health service infrastructure that dispatch programs do not control and that varies dramatically across jurisdictions. A dispatch program in a jurisdiction with inadequate outpatient capacity will cycle the same individuals through repeat contacts without the underlying situation changing.

The coverage hours gap leaves the overnight window unchanged. Austin, Durham’s CCD, and Albuquerque have achieved 24/7 coverage. Most documented programs have not. Houston’s overnight gap has persisted for more than a decade. Durham’s CCD had no overnight coverage through mid-2025. A program operating part-time improves outcomes during its operating hours and leaves the overnight window — often the most resource-scarce hours of the day — exactly as it was before the program existed.

The eligible call volume is larger than any current program captures. RTI International’s analysis of Durham 911 call data found approximately 15% of calls were behavioral health or quality-of-life calls, while only 1–2% were CAD-coded as mental health at dispatch. The gap between 15% and 1–2% represents eligible calls that existing programs miss even before accounting for coverage-hour limits. The full behavioral health need in any 911 system is larger than the subset currently reaching clinical routing.

Bottom Line

The documented risks in mental health dispatch integration are: funding fragility (Austin Prop Q cut, Baltimore ARPA cliff), dispatcher capture gaps (Columbus 10% capture, Wayne State miscoding study, Baltimore fidelity gap), workforce shortages (HRSA shortage designations), and evidence gaps (no independent evaluation of embedded dispatch clinicians specifically). Coverage hours limitations are documented in Houston and Columbus with identified causes. The funding structure that explains most program fragility — encounter-based crisis funding against capacity-based police and fire funding — is documented in the Angela Kimball / Inseparable analysis.