Does It Work?

The evidence base for mental health dispatch integration is consistent in direction and limited in verification. No randomized controlled trial or independent evaluation with comparison groups has specifically studied the embedded dispatch clinician model. The outcome figures that programs report — Austin’s 90%+ diversion rate, Dakota County’s 83% resolution without police, Sacramento’s 90%+ phone resolution — are program-reported, not independently verified. That distinction matters when these numbers are cited in budget hearings and a skeptic checks the source.

What follows maps what the evidence shows, at what tier, and where the gaps are.

What the Evidence Shows: Program-Reported Outcomes

Call diversion rates. Every documented dispatch integration program reports high rates of calls resolved without police response. These rates are all program-reported.

Austin EMCOT: More than 90% of calls handled by embedded dispatch clinicians diverted from police response in FY2025, according to program data reported to KVUE.
Dakota County: 83% of mental health calls handled without law enforcement field response, reported by operations director Brent Anderson.
Sacramento County: 988 counselors integrated with 911 dispatch resolve over 90% of transferred calls by phone, according to CalMatters April 2025 reporting.
Houston CCD: Diverted nearly 7,500 calls from law enforcement between 2016–2021, saving more than 11,000 police hours and generating $1.67 million in annual savings by 2020, according to Houston CIT program data.
Nebraska (statewide 988 integration): More than 96% of calls de-escalated by crisis counselors without activating a Mobile Crisis Response Team, according to KETV reporting.

These figures come from different program designs and are not directly comparable. A diversion rate in one city may count police diversion from the call center; another may count the rate of calls resolved without any field response; another may count the rate at which callers reached the clinical pathway rather than the police pathway. The Wayne State University 2025 study documented the methodological variation as a field-wide pattern: programs measure different things under similar labels, making cross-program comparison unreliable without knowing each program’s exact measurement definition.

Cost figures. Three documented program-reported cost calculations:

Austin: $12 million annual savings projected, based on the cost differential between a diverted call and the full police-response cost chain. Program-generated, not independently verified. The comparison baseline is the full police-response cost chain (officer time, potential ER, potential booking).
Houston CCD: $1.67 million in annual savings by 2020, calculated against the difference between a diverted call and the 104 minutes officers spent on average with a CIT call, at a program budget of approximately $460,000 annually.
Guilford County ART: $400,000 in savings and 170 public safety staff hours recovered in one program period, per Spectrum Local News reporting.

All three use different comparison baselines and are not additive. No independent analysis has verified any of these calculations.

Phone resolution — the core dispatch claim. The most significant metric for the dispatch layer specifically is the rate at which calls are resolved by phone with no field dispatch at all. Austin reports approximately 80% of calls handled by its dispatch clinicians resolved entirely by phone without dispatching anyone. This is the metric that most directly demonstrates what dispatch integration does that mobile crisis response does not — most calls never leave the building.

The phone resolution rate is also the most policy-relevant metric because it quantifies the call volume that would otherwise have resulted in some kind of field response. If 80% of behavioral health calls can be resolved by a clinician on the phone, then the remaining 20% is the call population that actually needs field response — and that population is smaller, better screened, and more appropriate for mobile crisis or police response than the full undifferentiated call set that 911 currently dispatches to police.

Minneapolis comparison. Minneapolis’s Behavioral Crisis Response team — which operates both a dispatch function and a field response — reached approximately 30,000 calls by April 2025, nearly double the historical annual volume of mental health-related police calls, according to the National League of Cities. This figure documents call volume growth as programs expand and dispatchers develop confidence routing behavioral health calls to clinical staff. It does not disaggregate the dispatch-layer calls from the field-response calls — the two functions are combined in BCR’s reporting.

What the diversion rates don’t tell us. Diversion rates measure routing outcomes: the call was directed to a clinician rather than police. They don’t measure clinical outcomes: whether the person’s crisis resolved, whether they received appropriate ongoing care, whether they called again. The routing outcome and the clinical outcome are related but not identical. A call that routes to a clinician and is “resolved by phone” may have been genuinely resolved — crisis de-escalated, resources connected, stability restored. Or it may have been handled in a way that closed the call without changing the underlying situation. Most programs do not collect data that would distinguish between these outcomes at scale.

The Evidence Tier Limitation

No independent research institution has conducted a controlled evaluation of embedded dispatch clinicians with comparison groups, pre/post measurement, and controls for confounding factors. The closest published independent research:

The Wayne State University 2025 study of 911 dispatch coding in Michigan documented that 911 professionals coded just over half of mental health-component calls correctly at dispatch — establishing that the routing problem dispatch integration addresses is real and measurable. The study’s focus is coding accuracy at dispatch, not program outcomes. It establishes the problem; it does not measure what embedded clinicians do to solve it.

The Stanford Gardner Center’s 2024 evaluation of a California co-response program — clinicians paired with police officers responding to scenes — found a 16% reduction in involuntary psychiatric detentions and a 17% reduction in mental health-coded calls in communities with the program. The study used comparison communities and two years of data. This evaluates field co-response, not dispatch-embedded clinicians. It is the closest independently evaluated adjacent evidence, and it establishes that clinical expertise in the response chain changes outcomes. Whether the same theory of change holds at the phone level — without a field visit, without direct observation of the person in crisis — is a different question the Stanford study does not answer.

A 2025 Massachusetts analysis of 10,904 behavioral health crisis incident records found that co-response clinicians achieved the highest emergency department diversion rates across responder types. Again, this is a field co-response finding. It documents the value of clinical expertise in the response chain but says nothing specific about the phone-based dispatch layer.

The distinction matters practically: a city council member who cites these adjacent studies as evidence for the dispatch integration model specifically is making a logical leap the evidence doesn’t support. The adjacent evidence supports the theory of change. It does not verify the specific implementation.

What the Repeat-Caller Evidence Shows

Guilford County’s Adult Resource Team documented a specific outcome for high-frequency callers: a group making 344 non-emergency 911 calls in a 30-day period dropped to four calls in the subsequent period after being connected to appropriate ongoing services, according to WFMY News2 and Spectrum Local News reporting. This is a single program’s reported outcome from a single population segment. It cannot be generalized to the broader call population. But it illustrates the mechanism programs describe: when dispatch-level clinical triage connects a high-frequency caller to appropriate services, the person’s relationship with the emergency system changes. The test is whether the effect holds across programs and over time — which has not been studied.

The Measurement Gap Across the Field

The National Research Institute’s 2024 survey of crisis response programs found approximately 40% of programs claiming 24/7 availability actually staff all shifts. The gap between claimed availability and actual staffing is the most concrete evidence that the field’s self-reported performance data overstates real-world program reach.

This measurement gap matters for interpreting diversion rates. A program that runs 12 hours per day and claims 90% diversion during those hours is not a 90% diversion program — it is a 90% diversion program during business hours. The 10 PM to 6 AM window runs at 0% clinical diversion regardless of the daytime figure. Published outcome numbers for programs with coverage-hours gaps reflect a subset of calls, not the full call population.

The field has begun building more rigorous measurement infrastructure. Baltimore’s consent decree requires a public quarterly dashboard showing diversion volumes, call types, and outcomes, reviewed by the Harvard Government Performance Lab. Durham’s Community Safety Department publishes monthly performance reports. Albuquerque ACS publishes quarterly reports including call volume, response time, and backup rates. These programs are building the data foundations for independent evaluation — but the independent evaluation itself has not been commissioned or published.

What We Don’t Know

Whether outcomes hold across jurisdiction types. The best-documented programs are in mid-to-large cities — Austin, Houston, Sacramento, Dakota County — with established behavioral health infrastructure and in most cases a designated local mental health authority that provides embedded clinical staff. Whether dispatch integration produces comparable diversion rates in rural jurisdictions, counties without a local mental health authority, or communities with weaker downstream service infrastructure is not documented. The Minnesota mandate produced 85/87-county compliance, but the outcome data for lower-capacity county implementations is not publicly available.

Whether phone resolution translates to outcomes for callers. Diversion rates measure what happens to the routing decision. They don’t measure what happened to the person: whether their crisis resolved, whether they accessed services, whether they called again. Austin’s same-day follow-up call creates a continuity touchpoint. But field-wide individual outcome data — the evidence that would show whether people who go through clinical dispatch routing fare better than people who don’t — is largely absent.

Whether coverage gaps distort reported figures. Programs operating 12 hours per day capture a different call population — concentrated in daytime hours when callers may have more support, services are open, and acuity may differ — than programs operating 24/7. Comparing diversion rates across programs without controlling for coverage hours could reflect call timing differences as much as model quality.

Bottom Line

Program-reported diversion rates from Austin, Houston, Dakota County, Sacramento, and Nebraska are consistent: the large majority of behavioral health calls routed to clinical staff are resolved by phone without field dispatch. No independent evaluation of the embedded dispatch clinician model specifically has been published. The independently evaluated adjacent evidence — Wayne State on coding accuracy, Stanford Gardner Center on co-response field outcomes — establishes that clinical expertise changes routing and response outcomes without directly verifying the phone-only dispatch layer. The evidence supports investment in the model; it does not yet provide the independent verification that a skeptic demanding Tier 1 evidence would require.