Cardiac Arrest Registry Outcome Benchmarking Dashboard — standardized outcome data, risk-adjusted comparison, and quality-improvement tracking across EMS systems
Meaningful comparison across EMS systems, hospitals, and years is only possible when everyone measures the same thing, the same way. The Utstein template — first agreed in 1991 aboard the Utstein Abbey in Norway and periodically updated since — defines a standardized set of core and supplemental data elements for out-of-hospital cardiac arrest (OHCA) registries, turning fragmented local records into a comparable, poolable dataset.
The template defines precise, reproducible definitions for every step of an OHCA event, removing ambiguity that would otherwise make cross-system comparison meaningless:
Time intervals (all referenced to a single "time zero"): • Collapse time — when the patient was witnessed or estimated to lose consciousness • Call receipt time — when the emergency call reaches dispatch • Dispatch time — when responding units are notified • Vehicle en-route and on-scene time • Time to first compression, first shock, first advanced airway
Event circumstances: • Location (home, public, healthcare facility) • Witnessed status (bystander-witnessed, EMS-witnessed, unwitnessed) • Presumed etiology (cardiac vs. non-cardiac) • Initial monitored rhythm (shockable: VF/pulseless VT; non-shockable: PEA/asystole)
Interventions: • Bystander CPR performed (yes/no), bystander AED use • Dispatch-assisted CPR instructions given • Number of defibrillation attempts, drugs administered
Outcomes: • Return of spontaneous circulation (ROSC) at any point, and sustained ROSC • Survival to hospital admission • Survival to hospital discharge • Neurological status at discharge (Cerebral Performance Category, CPC 1–5)
Without this common scaffolding, one system's "survival rate" might exclude unwitnessed arrests while another includes them — making raw comparisons worse than useless.
A registry that reports a 12% survival rate calculated only from bystander-witnessed shockable arrests is not comparable to one reporting 12% across all-comer arrests including nursing-home unwitnessed events. Utstein-defined denominators are what make benchmarking valid in the first place.
Modern cardiac arrest registries — CARES (Cardiac Arrest Registry to Enhance Survival) in the United States, the Resuscitation Outcomes Consortium (ROC), and EuReCa in Europe — capture Utstein elements through structured electronic case report forms populated from EMS electronic patient care records (ePCR), dispatch center CAD logs, defibrillator data downloads, and hospital discharge abstracts.
Data linkage is the hardest engineering problem: a single patient event must be stitched together across 911/112 dispatch systems, EMS agency records, and hospital EHRs — each with different patient identifiers, timestamps, and data formats. Deterministic and probabilistic matching algorithms link these fragments into one Utstein-compliant case record.
Data quality assurance includes automated range checks (e.g., a response time of 45 minutes flags for review), inter-rater reliability audits on rhythm classification, and quarterly completeness reporting — systems with high missingness on core variables are excluded from benchmarking comparisons until data quality improves.
Four metrics anchor nearly every cardiac arrest benchmarking report: return of spontaneous circulation (ROSC), survival to hospital admission, survival to hospital discharge, and neurologically favorable survival. Each represents a narrower, more clinically meaningful slice of the same funnel — and each attrition point between them tells a different part of the resuscitation story.
Return of spontaneous circulation (ROSC): a palpable pulse or measurable blood pressure sustained for at least 20 seconds, achieved at any point during resuscitation — including transient ROSC that is later lost. This is the earliest and most volatile metric; it reflects immediate response to compressions, defibrillation, and drugs, but does not predict survival on its own.
Survival to hospital admission: the patient has a pulse on arrival at the emergency department and is formally admitted (even if to the ICU for withdrawal of care hours later). This filters out patients who re-arrest and cannot be resuscitated during transport.
Survival to hospital discharge: the patient leaves the hospital alive, regardless of destination (home, rehabilitation facility, skilled nursing). This is the classical "survival rate" quoted in most headlines, but it says nothing about quality of life.
Neurologically favorable survival: survival to discharge AND a Cerebral Performance Category (CPC) of 1 (good cerebral performance) or 2 (moderate disability, independent for daily activities). This is increasingly regarded as the single most important outcome — it captures what patients and families actually care about: not just being alive, but returning to a meaningful life.
A system can have a high raw survival-to-discharge rate while a large share of survivors have severe neurological disability (CPC 3–4). Benchmarking on favorable survival alone avoids rewarding systems that merely convert cardiac deaths into permanent vegetative states.
The gap between each funnel stage is itself diagnostic:
• Large drop between ROSC and admission survival → suggests problems with post-ROSC care during transport (airway management, hemodynamic support, re-arrest prevention) • Large drop between admission and discharge survival → points to in-hospital care quality: targeted temperature management, early coronary angiography for suspected cardiac cause, ICU-level neuroprotective care • Large drop between discharge survival and favorable neurological survival → signals problems earlier in the chain: time to first compression and time to defibrillation are the strongest predictors of neurologically intact survival, because cerebral ischemia accumulates every minute circulation is absent
Benchmarking dashboards typically display all four metrics simultaneously as a funnel or waterfall, so a system with strong ROSC rates but weak favorable-survival rates immediately reveals where along the chain intervention is needed.
Two EMS systems can have identical clinical protocols and identical crew skill, yet report very different raw survival rates simply because their patient populations differ. A system serving a dense urban core with many public, witnessed arrests and rapid bystander CPR will always outperform a rural system with longer response times and more unwitnessed home arrests — unless outcomes are adjusted for these differences before comparison.
Risk-adjustment models for cardiac arrest outcomes typically control for a well-established set of covariates that strongly predict survival independent of system quality:
• Witnessed status: bystander-witnessed arrests have roughly 2–3× the survival of unwitnessed arrests, because CPR and defibrillation can start immediately rather than after an unknown "downtime" • Initial rhythm: shockable rhythms (ventricular fibrillation, pulseless ventricular tachycardia) carry substantially better prognosis than non-shockable rhythms (asystole, pulseless electrical activity) — a system serving an older, sicker population with more PEA/asystole arrests will show lower raw survival for reasons unrelated to care quality • Location: public-location arrests benefit from bystander presence, AED availability, and shorter EMS response; home arrests, especially in single-occupant households, are more often unwitnessed • Response time distribution: rural and suburban systems with longer transport distances face inherently longer response intervals than dense urban systems • Age and comorbidity burden: older patients and those with more comorbidities have lower baseline survival potential regardless of resuscitation quality
Ignoring these factors and simply ranking systems by raw survival rate produces a league table that mostly reflects population geography and demographics, not clinical performance.
A rural EMS system with 9-minute average response times and mostly unwitnessed home arrests could deliver textbook-perfect resuscitation care and still show a lower raw survival rate than an urban system with 4-minute response times and abundant bystander CPR in public spaces. Risk adjustment is what allows a fair comparison of the care itself.
Risk-adjusted benchmarking typically uses multivariable logistic regression (or similar modeling) to predict the expected outcome for each system given its specific case mix, then compares the observed outcome to that expectation:
1. Fit a model across the full registry population: outcome ~ witnessed status + initial rhythm + location + age + response time + comorbidity index 2. For each contributing system, compute the expected outcome rate given that system's actual distribution of these covariates 3. Compute an observed-to-expected (O/E) ratio: O/E > 1.0 means the system outperforms what its case mix would predict; O/E < 1.0 means it underperforms 4. Standardize this into a risk-adjusted rate comparable across systems, analogous to standardized mortality ratios used in other quality domains
This approach — long used in cardiac surgery and trauma outcome benchmarking — lets a rural system with a difficult case mix demonstrate excellent risk-adjusted performance even with modest raw numbers, and prevents an urban system with a favorable case mix from appearing better than its actual care quality warrants.
A single risk-adjusted outcome number, viewed in isolation, is of limited use. The real value of a benchmarking dashboard comes from placing that number against a reference band — regional or national peer performance — and watching how it moves quarter over quarter, which reveals whether a system is improving, holding steady, or slipping.
A benchmark is not a single target number but a distribution. Registries typically construct benchmark bands using:
• Percentile bands from a large peer cohort of similarly structured EMS systems (matched on population density, call volume, or region) • Statistical process control limits (funnel plots) that account for the fact that smaller-volume systems naturally show more month-to-month variance around the mean — a system with only 40 arrests per year will swing more than one with 4,000, purely from sampling noise • National reference rates published by registries such as CARES, against which any individual system, region, or state can be compared
Displaying a system's risk-adjusted rate as a point estimate with a confidence interval against this band avoids over-reacting to random quarter-to-quarter noise, especially for lower-volume systems.
Trend analysis plots the risk-adjusted outcome metric over consecutive quarters (typically a rolling 8–12 quarter window) and classifies the trajectory:
• Improving: statistically significant upward slope, often following a specific intervention (e.g., dispatch-assisted CPR protocol rollout, new defibrillator placement program) • Stable: outcome oscillates within the benchmark band with no significant slope — this is the expected pattern for a mature, well-functioning system • Declining: downward slope that crosses below the benchmark band, which should trigger a root-cause review before the trend becomes entrenched
Because single quarters are noisy, most dashboards apply a moving average or a formal trend test (e.g., CUSUM — cumulative sum control chart) to detect a real shift in performance level rather than random fluctuation.
CUSUM charts are widely borrowed from industrial quality control specifically because they detect small, sustained shifts in performance faster than simple quarter-over-quarter comparison — critical for catching a slow erosion in bystander CPR rates before it shows up as a statistically significant drop in survival.
When a system falls below benchmark on risk-adjusted outcomes, the dashboard's job shifts from measurement to diagnosis. Because the chain of survival is sequential — each link depends on the one before it — systematically working back through bystander recognition, dispatch-assisted CPR, response time, and in-hospital care isolates exactly where the weakest link sits, so improvement resources go where they will do the most good.
The chain of survival — early recognition and call for help, early high-quality CPR, early defibrillation, basic and advanced EMS care, and post-arrest care — provides a natural root-cause checklist:
1. Bystander CPR rate: is the community trained? CPR education programs in schools and workplaces, public awareness campaigns, and CPR-friendly 911/112 dispatch culture all move this number. A system below benchmark here should invest in community training and hands-only CPR public messaging.
2. Dispatch-assisted CPR (T-CPR): does the dispatch center provide real-time phone-guided CPR instructions to callers before EMS arrives? This single protocol change is one of the most cost-effective interventions available, often lifting bystander CPR rates by 2–3× within a year of implementation.
3. Response time performance: are first-responder and ambulance response times within target (e.g., 90th percentile under 8 minutes)? This may require resource redeployment, first-responder co-response (fire/police with AEDs), or public-access defibrillator placement in high-incidence locations.
4. In-hospital care quality: does the receiving hospital deliver targeted temperature management, early cardiac catheterization for suspected coronary occlusion, and structured neuroprognostication before withdrawal-of-care decisions? Routing patients to designated cardiac arrest receiving centers, when available, measurably improves favorable survival independent of prehospital care.
A 2020 analysis of systems that improved dispatch-assisted CPR protocols showed bystander CPR rates rising from the 30s into the 60s (%) within 18 months, with a corresponding measurable increase in neurologically favorable survival — demonstrating that a single, well-targeted link in the chain can move the entire outcome funnel.
Mature registry programs close the loop between measurement and action using a structured quality-improvement cycle:
• Quarterly benchmarking review identifies systems or regions below the benchmark band • Root-cause analysis, guided by chain-of-survival metrics, identifies the specific weak link (e.g., low bystander CPR in a specific zip code cluster, response-time drift in a specific station's coverage area) • A targeted intervention is designed and piloted (CPR training drive, T-CPR protocol update, station relocation, receiving-hospital pathway change) • The same risk-adjusted metrics are tracked post-intervention on the same quarterly cadence to confirm the intervention actually moved the needle, not just anecdotally but in the risk-adjusted trend line • Successful interventions are documented and shared across the peer network, turning individual system learning into system-wide improvement — the ultimate purpose of registry benchmarking.