Measuring What Matters — A Performance & Expansion Framework for Chicago CARE · IPG
Institute for the Public GoodCivic data · Policy design · Implementation
Performance Management Brief · Behavioral Crisis Response

Measuring What Matters

A benchmarking review and a phased data & dashboard framework for Chicago's Crisis Assistance Response & Engagement (CARE) program — built to prove the coverage gap, track real outcomes and care delivered, count the dollars and hours saved, and open a pathway to community crisis stabilization.
Prepared forCARE Program Leadership — CDPH / OEMC / CFD
Prepared byInstitute for the Public Good (IPG)
CompanionCARE Insight — interactive dashboard prototype
CONTENTS

What's in this brief

01

Executive summary

Chicago expanded CARE to all 77 community areas in May 2026, but availability on a map is not the same as capacity on the street. With teams running Monday–Friday during a daytime window, most behavioral-health 911 calls the city would route to CARE still get a traditional response — because the call was never routed to CARE, because no van was free, or because the call landed outside operating hours.

That gap is not a weakness to hide; it is the single most persuasive number CARE can publish. Every peer program that won expansion did so by measuring the distance between eligible and served. New York's own Comptroller found more than 27,000 B-HEARD-eligible calls went unanswered — roughly half of them simply because teams weren't operating overnight. Albuquerque turned a pilot into a 24/7 cabinet-level department handling 120,000+ calls. In each case, the data told a coverage story first and an outcomes story second.

IPG recommends that CARE build its measurement system around three signature analyses:

  1. The coverage-gap funnel — from all behavioral-health-flagged 911 calls, down through CARE-eligible, routed, in-window, and actually responded, with every drop-off labeled by reason: not routed at triage, not passed by CPD/CFD, no van available, or out of hours. This is the expansion argument, drawn straight from OEMC dispatch data.
  2. The outcome-and-destination tracker — whether the person received care and what kind of care (de-escalation, assessment, medical evaluation, transport to treatment, referrals made), plus — for the subset who needed somewhere to go — where they went, and how often there was nowhere appropriate to take them. That "needed a destination, none available" count is the evidence base for peer-staffed crisis stabilization ("living room") capacity.
  3. The cost tracker — how many police officer-hours and how many dollars are saved when a CARE team responds instead of police officers. This is the budget-defense argument. The hours estimate comes from dispatch data alone, and a base-level dollar figure can be built entirely from city-internal data — no third-party partners required.

The companion CARE Insight prototype shows all three, plus a van data-entry screen designed for a responder entering data on a tablet between calls. The system is staged across four phases so the city can launch on data it already controls (OEMC), layer in van-collected outcomes and care types, load the base-level cost picture from city-internal data, and only then bring on external HIPAA-compliant partners as a refinement — never a dependency.

02

The case we're building

Two audiences read a crisis-response dashboard, and they want different things. Budget and policy decision-makers want to know whether the program works and whether it should grow. The public wants to know it's safe, fair, and real. A good measurement system answers both without drowning either in numbers.

For CARE specifically, the data has three jobs, mirroring the Harvard Government Performance Lab framework:

  • Reactive troubleshooting — spotting in near-real-time where eligible calls leak out of the funnel (never routed, not passed by CPD/CFD, no van, out of area, out of hours).
  • Managing performance — tracking response times, care delivered, on-scene resolution, and connections to services.
  • Making the expansion & sustainability case — quantifying unmet demand and quantified savings to justify more vans, longer hours, and new infrastructure such as crisis stabilization sites.

CARE's three headline numbers

1 — The gap between eligible and served → Legislators · OEMC · CPD · CFD
"X eligible calls got a police or EMS response because they were not routed to CARE."

This headline is deliberately a call-out to the routing system, not just a capacity statistic. It flags two distinct failures that must be counted separately:

  • (i) The call was never routed to CARE at the outset — OEMC triage did not flag or dispatch an eligible call to CARE.
  • (ii) CPD or CFD arrived first and did not pass the call — the responding unit recognized (or should have recognized) a behavioral-health situation within CARE's scope and no handoff occurred.

Alongside these sit the two capacity gaps: no van was available (buys vans) and the call came in out of hours (buys hours). Each of the four reasons points at a different owner and a different fix — which is exactly what makes the funnel actionable rather than just descriptive.

Stretch analysis — flagged, not required

Ideally the funnel would also show calls that could have been passed to CARE based on the final disposition codes CPD and CFD mark after the response (e.g., a call closed as a mental-health disturbance that arrived as a generic disturbance). This retrospective re-coding analysis is powerful — it converts "the call wasn't eligible" into "the call was eligible and we can prove it" — but it depends on access to CPD/CFD post-response coding and adds real matching complexity. IPG recommends treating it as a Phase 2+ enhancement rather than a launch requirement.

2 — Care delivered, and the destination problem → CDPH · Providers · Stabilization advocates
"Of the people who needed somewhere to go, X had no appropriate destination available."

This tracker has two layers. First, for every response: did the person receive care, and what kind — de-escalation, mental-health assessment, medical evaluation, harm-reduction supplies, safety planning, transport to treatment, referrals. That is the program's core value story.

Second — and scoped deliberately — the destination measure counts only individuals who needed a destination: someone who could not safely stay in place and for whom an ER or jail was not the right level of care. Calls resolved calmly on scene do not belong in this denominator; including them would dilute the number into meaninglessness. The scoped count is the demand signal for peer-staffed living-room crisis stabilization — a model Illinois already funds (IDHS Living Room Program) and Chicago already runs (Thresholds' Forever Hope on the North Side).

3 — Hours and dollars saved → Budget office · City Council
"X police officer-hours — worth $Y — were redirected by sending CARE teams instead of police officers."

Every CARE primary response to a call that would previously have drawn a police response frees officer time (typically two officers for the duration of the call, plus any transport and paperwork). Multiplying diverted calls by the average police time-on-call for comparable call codes yields an hours-saved estimate from dispatch data alone — available in Phase 1. Applying a loaded hourly cost, plus the city's own cost for avoided CFD/EMS transports, produces a base-level dollar figure entirely from city-internal data — loaded in Phase 3, with no third-party partners required. External HIPAA-compliant partners (hospital ER costs, involuntary-commitment costs) arrive in Phase 4 purely as a refinement on top. This is the number that defends the budget line — the lesson of CAHOOTS, below.

03

City-by-city profiles

Four peer programs chosen because they are well-documented and publish data — plus one cautionary tale. For each: what they track, how it's housed, what the dashboard looks like, and the one lesson Chicago should carry forward.

Albuquerque Community Safety (ACS)

Albuquerque, NM · Standalone cabinet department

The most institutionally ambitious model: a third branch of public safety alongside police and fire, with its own academy, its own dispatch integration, and specialized teams for behavioral health, co-response, and street outreach. ACS publishes detailed quarterly reports built from CAD (computer-aided dispatch) data: call volume by source (911 vs 311 vs self-initiated), response times, outcome frequencies (e.g., transport to a provider, "no person found"), and report counts by council district with maps.

120k+
calls to date
85%+
diverted from police/fire
<1%
calls needing police assist
24/7
citywide coverage

Lesson for Chicago: ACS's quarterly metric stack — volume, response time, outcome frequency, geography — is a clean template, and its 85%-diverted figure is the model for CARE's hours-and-dollars-saved headline. Note their honest reporting of "no person found" (~25% of calls); Chicago's equivalent is "gone on arrival," and publishing it builds credibility rather than undermining it.

NYC B-HEARD

New York, NY · FDNY EMS + Health+Hospitals

Chicago's closest cousin: a health-led partnership pairing EMTs with H+H mental-health professionals, dispatched only through 911, deliberately not requestable by name, with weapons/violence calls excluded. B-HEARD publishes quarterly data and patient-experience surveys (99% felt treated with respect; 43% served in community vs. hospital). But its defining public document is the 2025 Comptroller audit, which quantified the coverage gap in brutal detail.

18
teams (9 per shift)
27k+
eligible calls unanswered
14.2k
missed only b/c overnight
43%
served in community

Lesson for Chicago — the central one: The audit's most damaging finding was also its most useful: ~14,200 eligible calls got a traditional response purely because no team was operating at that hour. That is the exact argument for extending CARE's window. The audit also flagged that the city didn't track why calls fell out of scope — a gap CARE should design into its system from day one, with routing and pass-through reasons counted explicitly.

Portland Street Response (PSR)

Portland, OR · Public Safety service area

A medic + clinician + community-health-worker model with a dedicated aftercare unit that logs thousands of follow-up visits — useful precedent for CARE's 1/7/30-day follow-up. PSR's public dashboard (updated weekly) is a model of plain-language transparency, and its evaluations quantified reduced police workload on welfare checks and "unwanted person" calls — the time-savings evidence in action.

~52
staff
15k+
calls/yr
5,617
aftercare follow-ups/yr
19%
↓ police welfare-check workload

Lesson for Chicago: A weekly-updated public dashboard sustains political support, and PSR's evaluations show how to express impact in police-workload terms — the same framing CARE's cost tracker should use.

Durham HEART

Durham, NC · Community Safety Department

The program the Harvard tool profiles, and the best dashboard in the set. Durham's live dashboard (Power BI) does exactly what Chicago needs: it shows both the number of HEART responses and the number of calls that were eligible for a HEART response side by side — roughly 13,000 responses against ~56,000 eligible calls. That single comparison is the coverage-gap argument, made visible to the public.

56k
eligible calls
13k
actual responses
5.7m
avg min to scene
29m
avg min on scene

Lesson for Chicago — the dashboard blueprint: Durham's tabs (Over Time, By Program, By Location, By Day & Hour, Eligible Calls) and its "eligible vs. responded" framing should be CARE Insight's spine. The By Day & Hour view is what makes the case for which hours to add.

CAHOOTS — the sustainability cautionary tale

Eugene/Springfield, OR · White Bird Clinic (nonprofit)

The 35-year-old model everyone copied. In April 2025 it abruptly ended in Eugene. The cause was not effectiveness — it was funding architecture. The city contract covered only ~40% of the cost; the rest leaned on federal grants that tightened, and street-response work is hard to bill to Medicaid. A nonprofit-and-contract model with thin, volatile funding collapsed despite decades of proven impact.

~40%
of cost covered by city contract
35 yrs
of service, ended abruptly
~900
calls missed in first 3 weeks

Lesson for Chicago: Measurement protects funding. CARE's advantage is that it's embedded in city government (CDPH/CFD/OEMC) rather than a single nonprofit contract — but it should still publish cost-per-response and hours/dollars saved so the program can defend its line item, and explore Medicaid reimbursement pathways deliberately, as Oregon's experience shows that's where these models live or die.

04

What they track & how

Stripped of branding, the programs converge on a common metric backbone — and on a common set of data sources and tools.

The shared metric backbone

Table 1 — Metrics common across the benchmarked programs, by core question
Core questionMetrics in common useTypical source
Triage & dispatch
Are we reaching the right calls?
Call volume; call origin (911/988/311/self); call type (initial & final code); response time (to-dispatch, travel, total); time on scene; response type (primary / co-response / secondary); geographyCAD / dispatch system
Coverage gap
What are we missing?
Eligible calls not served + reason (not routed, not passed by police/fire, no team available, out of hours, out of area); eligible-vs-responded ratioCAD + eligibility logic
System impact & cost
What do we relieve?
Police calls diverted; police hours freed; EMS/ER transports avoided; use of force / arrests (near-zero is the story); cost avoidanceCAD + police/EMS/hospital data
Staff & safetyResponder felt-safe rate; staff retention/turnover; training coverageResponder survey; HR system
EquityService-recipient demographics; staff demographics; complaints; repeated calls (same person/address)Van-collected; HR; dispatch
Care & connectionCall outcome (resolved on scene / transported / backup / no contact); care delivered on scene, by type; referral types; transport destination; follow-upsVan case management
Community impactAwareness of the program; willingness to call emergency servicesResident survey

How and where they track it

  • Department housing varies, but dispatch is always the anchor. Whether health-led (NYC, Chicago) or a standalone department (ABQ), the call-side truth lives in the CAD/dispatch system. Chicago's anchor is OEMC.
  • The van/case-management system carries the outcome side. What happened on scene, what care was delivered, what referral was made, follow-up — these are entered by responders. This is the data most vulnerable to being incomplete, which is why the input experience matters (see §08).
  • Dashboards are built on mainstream BI tools. Durham uses Power BI; Portland uses a weekly-refreshed public dashboard; ABQ publishes PDF quarterly reports off CAD extracts. None of this requires custom engineering — it requires clean data pipes and data-sharing agreements.
  • Surveys fill the gaps that admin data can't. Patient experience (B-HEARD), responder safety, and community awareness all come from lightweight surveys layered on top of the operational data.
IPG recommendation

IPG recommends against relying on PDF quarterly reports as the primary reporting vehicle. The timeframe is far too infrequent for performance management — a quarter-old number can't drive a staffing or routing correction — and the production process is manual, consuming analyst time on formatting instead of analysis. The recommended pattern is a regularly updated dashboard fed directly through data piping from OEMC and the van system, with quarterly PDFs (if required at all) generated as snapshots from the dashboard, never assembled by hand. The survey instruments — both the responder-facing entry in the van and any patient-experience survey — can be built HIPAA-compliant and secure from the start: consented collection, encrypted transmission and storage, role-based access, and aggregate-only public display. Compliance is a design requirement, not a reason to fall back to manual processes.

05

Four lessons for Chicago

  1. Publish the gap, not just the wins. Durham shows eligible-vs-responded; NYC's auditor forced the number into the open. CARE should own it first — including the routing and pass-through failures, named as such. Measuring unmet demand is how every peer justified more vans and longer hours.
  2. Hours are the highest-leverage expansion. B-HEARD lost ~14,200 calls purely to nighttime non-coverage. CARE's M–F daytime window means a large, countable share of eligible calls can never reach a team. The By-Day-&-Hour view turns that into a staffing plan.
  3. Track the destination problem to build the destination. Counting the people who needed somewhere to go but had no appropriate sub-acute option is the evidence for peer-staffed living-room stabilization — which Illinois funds (IDHS Living Room Program) and Chicago already has (Thresholds' Forever Hope).
  4. Measurement is funding insurance. CAHOOTS proves a beloved, effective program can still collapse on funding architecture. Publishing hours and dollars saved — and pursuing a deliberate Medicaid-reimbursement strategy — is how CARE defends its line item through every budget cycle.
06

Recommended metric set for CARE

Mapped to the two data sources CARE actually has — OEMC dispatch and in-van interaction data — and prioritized so the program can start with what it controls. The three signature analyses anchor the set.

Signature analysis 1 — the coverage funnel (illustrative monthly figures)

911 calls flagged behavioral-health
3,200
100%
Screened CARE-eligible
1,850
58%
Eligible & within operating window
640
20%
Routed to CARE & van available → responded
388
12%
Gap A — not routed / not passed by CPD·CFD
84
3%
Gap B — routed, but no van free
168
5%
Gap C — eligible but out of hours
1,210
38%

The headline this produces: "1,462 eligible calls got a police or EMS response because they were not routed to CARE — 84 because triage never routed them or CPD/CFD didn't pass the call, 168 because every van was busy, and 1,210 because they happened outside operating hours." Gap A is a protocol-and-training fix for OEMC, CPD, and CFD. Gap B buys vans. Gap C buys hours.

Signature analysis 2 — care delivered & the destination tracker

Two linked measures from van data: care delivered on scene, by type (every response), and the destination outcome for the subset who needed one — placed in an appropriate setting vs. no appropriate destination available. The denominator discipline matters: only people who could not safely remain in place and did not belong in an ER or jail count toward the destination measure.

Signature analysis 3 — the cost tracker

From Phase 1: CARE primary responses × avg. police time-on-call for comparable codes × officers-per-call = officer-hours freed, convertible to dollars with a loaded hourly rate. From Phase 3: the base-level dollar figure — officer time plus the city's own cost for avoided CFD/EMS transports — built entirely from city-internal data, with no third-party partners needed. From Phase 4: external HIPAA-compliant partners (hospital ER costs, involuntary commitments) refine the total upward.

Phase-tagged metric list

Table 2 — Recommended CARE metrics, source, and the phase they come online
MetricPrimary sourcePhase
Call volume; origin; type (initial & final code)OEMC dispatch1
Response time (phone → dispatch → travel → on-scene); time on sceneOEMC + van timestamps1
Response type (primary / co-response / self-dispatch); geography by community areaOEMC dispatch1
Coverage funnel: eligible vs. served, by reason — incl. not-routed and not-passed countsOEMC + eligibility logic1
Eligible-but-out-of-hours calls (the hours argument)OEMC dispatch1
Police officer-hours freed (est.) — the time-saved headlineOEMC dispatch (comparable-code baselines)1
Police calls diverted; arrests & use of force on CARE calls (expect ~0)OEMC + van outcome field1→2
Call outcome (resolved / transported / backup / no contact / hold)Van case mgmt2
Care delivered on scene, by type (de-escalation, assessment, medical eval, supplies, safety plan, transport to care)Van case mgmt2
Referral types (mental health, SUD, housing, benefits, primary care, stabilization)Van case mgmt2
Transport destination (home, clinic, shelter, living room, ED)Van case mgmt2
Needed a destination → placed vs. "no appropriate destination available" (living-room case)Van case mgmt2
Service-recipient demographics; repeated need (same person/address)Van + dispatch2
Responder felt-safe rate; staff retention; complaintsResponder survey; HR2
Retrospective "could-have-been-passed" analysis from CPD/CFD post-response codes (stretch)CPD/CFD disposition data2+
Follow-up completion at 1 / 7 / 30 daysVan case mgmt3
Base-level cost avoidance in dollars (officer time + avoided CFD/EMS transports, at city-internal cost)OEMC + CFD internal cost data3
Community awareness; willingness to call emergency servicesResident survey3
Refined cost avoidance (hospital ER costs, involuntary commitments) via external partnersHospital data share4
07

Data sources & the agreements they require

The biggest implementation risk is not analytics — it's getting data to flow between agencies. The Harvard tool is blunt about this: establish data-sharing processes before launch, via MOUs and data agreements.

Table 3 — Data CARE needs, who holds it, and the agreement to put in place
DataHeld byAgreement needed
911 call records, codes, dispatch & CAD timestamps, eligibility flags, routing decisionsOEMCInter-agency data agreement (city-internal); define a behavioral-health eligibility flag and a routed-to-CARE flag in CAD
On-scene outcome, care delivered, referrals, destination, demographics, follow-upCARE / CDPH (van case mgmt)Internal CDPH data governance + HIPAA-compliant handling built in from the start
Diversion confirmation, arrests/use-of-force on CARE calls, post-response disposition codes (for the could-have-been-passed analysis)CPD / CFDMOU for matched-call data exchange
City-internal transport & response cost baselines (for base-level dollar figure)CFD / OEMC / OBMCity-internal cost-data request — no external agreement needed
ER admission & involuntary-commitment costs (cost refinement only)Hospitals, external providersHIPAA-compliant data-use agreements; honest-broker / matched-record process — Phase 4
Practical sequencing

Phase 1 needs only the OEMC agreement, which is city-internal and therefore fastest — and it powers two of the three signature analyses (the coverage funnel and the hours-saved estimate). The CPD/CFD MOU follows in Phase 2. Phase 3's base-level dollar figure uses only city-internal cost data, so all three signature analyses are fully operational — including costs — before any external party signs anything. The external HIPAA-compliant agreements (hospitals) are isolated in Phase 4 as a refinement, so they can take as long as they take without ever blocking the program's core numbers. Treat the agreements as a parallel workstream owned by a named person, not an afterthought.

08

Designing for low-information, low-technology users

Two very different humans touch this system: a responder entering data from a van on a tablet, and a viewer (a council member, a resident, a program manager) who needs the story in ten seconds. Both are low-tech by assumption. Design accordingly.

For the van responder (data entry)

  • Tap, don't type. Outcomes, care types, referral types, and destinations are chip/button selectors and dropdowns, not free text. Free text is where data quality dies.
  • One screen, big targets, finishes in under a minute. A responder logs between calls, sometimes one-handed. Large buttons, a single clear "Save response" action with an unmistakable confirmation.
  • Pre-fill from dispatch. Call time, location, and initial code come from OEMC automatically; the responder only adds what happened.
  • Make the hard metrics easy. "Needed a destination" and "no appropriate destination available" are two prominent taps — because if they're hard to log, they won't be, and the living-room case never gets built.
  • Plain language, no codes to memorize. "Resolved on scene," not a numeric disposition code.

For the dashboard viewer

  • One sentence before any chart. Each view opens with a plain-language takeaway ("Last month CARE reached 1 in 5 eligible calls").
  • "What is this?" on every metric. Inline definitions so no one needs a glossary.
  • Lead with the gap, then the care story, then the savings. Order encodes the argument.
  • High contrast, large type, keyboard- and screen-reader-friendly. A public dashboard is an accessibility obligation, not a nicety.

The companion CARE Insight prototype implements both: a tap-first "Log a response" screen that feeds live tiles, and a viewer dashboard that opens every tab with a sentence. A phase switcher shows exactly which tiles light up as each data source comes online.

09

Phased rollout & dashboard designs

Start on data CARE controls today. Add the care-and-destination story once van entry is solid. Load the costs from city-internal data — no third parties needed. Bring on external HIPAA-compliant partners last, as refinement rather than dependency. Each phase ships a working dashboard.

Phase 1 · Prove the gap, estimate the savings

Launch on OEMC data · Quarter 1

Goal: Stand up the coverage-gap funnel — with routing and pass-through reasons — plus the officer-hours-saved estimate, using only OEMC dispatch data, so two of the three signature analyses exist from day one.

Metrics live

Call volume, origin, type; response time; on-scene duration; response type; geography; eligible-vs-served funnel with reason codes (not routed / not passed / no van / out of hours); police officer-hours freed (est.).

Data & agreements

OEMC inter-agency data agreement (city-internal). Define a behavioral-health eligibility flag and a routed-to-CARE flag in CAD. No external/HIPAA agreements required.

Dashboard views

Overview KPIs · Coverage Gap funnel · Calls (volume/type/geography) · By Day & Hour heatmap · Cost & Time Saved (hours estimate).

Tooling

A regularly refreshed dashboard fed by direct data piping from OEMC — per the IPG recommendation, no manually produced PDF reporting. Weekly refresh minimum.

Phase 2 · Tell the care story

Add van data entry · Quarter 2

Goal: Capture what happened on scene through a low-tech van entry tool — the care delivered and its type, and the scoped destination tracker that justifies crisis-stabilization investment.

Metrics added

Call outcome; care delivered by type; referral types; transport destination; needed-a-destination → placed vs. none available; demographics; repeated need; responder felt-safe rate; complaints; confirmed police diversion. Stretch: retrospective could-have-been-passed analysis from CPD/CFD disposition codes.

Data & agreements

CDPH van case-management system with HIPAA-compliant collection built in; CPD/CFD MOU for matched-call diversion, use-of-force, and disposition-code data; responder & recipient micro-surveys.

Dashboard views

Outcomes & Care Delivered · Destination tracker (needed / placed / none available) · Equity (demographics, repeat callers) · Responder safety.

Design focus

Tap-first van entry: chips, dropdowns, pre-filled dispatch fields, prominent needed-destination + no-destination taps, <60-second flow.

Phase 3 · Load the costs — city data only

City-internal cost data · Quarter 3

Goal: Convert hours saved into a defensible base-level dollar figure using only city-internal data — officer time at loaded rates plus avoided CFD/EMS transports at the city's own cost — plus close the follow-up loop. No third-party partners are needed for any of this, so the cost headline never waits on an external signature.

Metrics added

Base-level cost avoidance in dollars (officer time + avoided CFD/EMS transports at city-internal cost); follow-up completion (1/7/30 day); community awareness & willingness-to-call.

Data & agreements

City-internal cost-data requests only (CFD transport cost, OBM loaded officer rates). Resident survey instrument. No HIPAA or external agreements required.

Dashboard views

Cost & Time Saved upgraded to full base-level dollars · Follow-up completion · Living-Room Pathway (demand identified vs. placed) · Public transparency view.

Sustainability

Hours-and-dollars-saved reporting to defend the budget line every cycle — the CAHOOTS lesson, now operational with all three signature analyses live.

Phase 4 · External HIPAA-compliant partners — refinement, not dependency

External data shares · Quarter 4 onward

Goal: Bring on external partners to refine what already works. Hospital ER admission costs and involuntary-commitment costs sharpen the Phase 3 dollar figure upward; provider-side data deepens the destination and follow-up picture. Because costs were loaded first from city data, these agreements can take as long as they take without ever blocking a headline number.

Metrics added

Refined cost avoidance (hospital ER costs, involuntary commitments); receiving-provider confirmations for warm hand-offs to living rooms and treatment settings.

Data & agreements

HIPAA-compliant data-use agreements with hospitals and external providers; honest-broker / matched-record processes; client-consent workflow; Medicaid-reimbursement groundwork.

Dashboard views

Cost & Time Saved refined tier (external-verified dollars) · Living-Room Pathway with receiving-side confirmation.

Sustainability

Deliberate Medicaid-reimbursement strategy — the funding architecture that Oregon's experience shows these programs live or die on.

The living-room pathway, made concrete

Phase 2's scoped "needed a destination, none available" count is the demand signal. When it's large and persistent, it justifies dedicated peer-staffed crisis stabilization capacity that CARE vans can route to directly — diverting people from EDs and jail. Illinois already funds this through the IDHS Living Room Program (20+ statewide, peer-led, voluntary, accepting referrals from police/fire/EMS), and Chicago has a working example in Thresholds' peer-led Forever Hope living room. The metric chain is: needed a destination → could not place → repeat → demand quantified → capacity funded → CARE routes there → ED/jail diversion measured.

10

Sources

  • Harvard Kennedy School Government Performance Lab — Essential Metrics for Alternative Emergency Response Programs (2024).
  • City of Chicago — CARE program pages, dashboard, and citywide-expansion announcement (May 2026); chicago.gov/CARE.
  • University of Chicago Health Lab / Urban Labs — CARE implementation evaluation.
  • City of Albuquerque — ACS quarterly reports FY25 Q2–Q4; ACS four-year anniversary release (2025).
  • NYC Mayor's Office of Community Mental Health — B-HEARD data & FAQ; NYC Comptroller — B-HEARD audit (May 2025); NYC H+H operational-model announcement (Nov 2025).
  • City of Portland — PSR data dashboard & safety dashboards; Portland State University HRAC evaluations; local reporting (2022–2026).
  • Durham Community Safety Department — HEART public dashboard; figures via the Harvard GPL tool.
  • OPB / KLCC / KVAL / Daily Emerald — CAHOOTS / White Bird Clinic closure reporting (2025–2026).
  • IDHS Living Room Program; Turning Point Behavioral Health (Skokie/Evanston); Thresholds Forever Hope (Chicago); Heyland, Emery & Shattell (2013), The Living Room.

Figures in the funnel and dashboard prototype are illustrative, constructed to be plausible for Chicago's scale; they are not actual CARE counts. Replace with live OEMC and van data on implementation.

Institute for the Public Good
Measuring What Matters · CARE performance & expansion framework · Companion: CARE Insight prototype