Team-based simulation training for high-stakes operating-room crises
Crisis Resource Management (CRM) does not begin when an emergency starts — it begins minutes or hours earlier, in the ordinary calm of a pre-operative briefing. The habits that determine how a team performs during a crisis are built during routine, low-stakes moments when everyone assumes nothing will go wrong.
Crisis Resource Management traces its lineage directly to aviation. In the 1970s and 1980s, the airline industry recognized that most fatal crashes were not caused by mechanical failure or lack of technical skill, but by breakdowns in communication, leadership, and situational awareness among otherwise competent crews. This insight gave rise to Crew Resource Management (CRM) training, now mandatory for commercial flight crews worldwide.
In the early 1990s, David Gaba and colleagues at Stanford adapted these principles to anesthesiology, coining the term "Anesthesia Crisis Resource Management" (ACRM). The core observation transferred almost unchanged: expert clinicians, like expert pilots, can fail catastrophically not from a lack of individual knowledge, but from a lack of coordinated, team-level cognition under stress. Gaba's group built some of the first high-fidelity simulation programs specifically to train and study this team-level behavior, and the term broadened to "Crisis Resource Management" as it spread beyond anesthesia into surgery, emergency medicine, obstetrics, and critical care.
The founding insight of CRM, borrowed directly from aviation human-factors research: technical expertise does not protect a team from failure. What fails first under stress is usually communication, leadership clarity, and shared awareness — not knowledge.
A "shared mental model" is the state in which every team member holds a compatible understanding of the patient, the plan, the risks, and their own role. It is not automatic — it must be actively constructed, and the briefing is where construction starts.
A well-run briefing typically covers:
• Patient identity, procedure, and site confirmation (echoing surgical safety checklist principles) • Anticipated critical steps, blood loss risk, and estimated duration • Relevant comorbidities and allergies that could shape crisis response (e.g., malignant hyperthermia susceptibility, anticoagulation status, difficult airway history) • Explicit role statement: who is circulating, who is scrubbed, who is managing the airway, who would lead if something goes wrong • Equipment and resource check: is the difficult airway cart present, is blood typed and available, is the emergency manual accessible
This two-to-three-minute investment measurably improves later crisis performance: teams that have already said each other's names and roles out loud recognize deviations from the plan faster and hand off tasks with less friction.
"Know the environment" is often listed as the first formal CRM principle because everything else depends on it. It means knowing, before anything goes wrong:
• Where the emergency equipment lives (difficult airway cart, defibrillator, malignant hyperthermia cart, massive transfusion cooler) • Who else is in the building and how quickly they can be reached (attending backup, blood bank, additional OR staff) • The specific quirks of this OR, this equipment, and this institution's emergency activation pathways • The cognitive aids available and where they are physically located
Teams that "know the environment" convert a crisis from a search problem into an execution problem — they are not discovering where the crash cart is at the same moment they need it.
Every simulated and real crisis has a hidden phase before anyone calls it a crisis: vitals drift, small anomalies accumulate, and the team's attention is still anchored on the routine task at hand. CRM training exists in large part to compress this hidden phase — to help teams notice, name, and react to deterioration before it becomes irreversible.
CRM draws heavily on Mica Endsley's three-level model of situational awareness (SA), originally developed for aviation:
• Level 1 — Perception: noticing the raw data (heart rate climbing, blood in the suction canister, oxygen saturation dipping) • Level 2 — Comprehension: integrating those data points into meaning ("this rate of blood loss is not proportional to what the surgical field shows") • Level 3 — Projection: anticipating where this trajectory leads if unaddressed ("at this rate, we will be hypotensive and coagulopathic within minutes")
Crises are dangerous precisely because early signs are ambiguous and easy to perceive without comprehending. A single dropping number on a monitor rarely triggers action; a team member explicitly connecting three data points out loud usually does. CRM training deliberately rehearses the jump from Level 1 to Level 2 — turning noticed data into spoken concern.
Fixation error — becoming locked onto one diagnosis or task while contradicting evidence accumulates — is one of the most consistently observed cognitive failure modes in simulated OR crises. Structured recognition habits exist specifically to interrupt fixation before it becomes dangerous.
When something feels wrong but is not yet fully understood, CRM teaches a deliberate cognitive pause: stop, take roughly ten seconds to consciously ask "what is actually happening here," and only then commit to the next ten minutes of action. This is not indecision — it is a structured interruption of the reflex to keep doing the last thing that seemed to be working.
In practice this pause looks like:
1. A team member (often but not always the most senior clinician) says a version of "let's stop for a second" 2. The team briefly states the working problem out loud: "we have ongoing brisk bleeding, pressure is falling, differential includes surgical bleeding versus coagulopathy" 3. A provisional leader and provisional plan are named before continuing
The pause typically costs seconds, not minutes, but it converts fragmented individual awareness into a shared, spoken understanding of the problem — the necessary precursor to coordinated action.
"Anticipate and plan" is a core CRM principle distinct from reaction. Experienced teams do not wait for a crisis to fully declare itself before preparing for it:
• During a case with known high blood-loss risk, blood may be pre-typed and cross-matched before any bleeding occurs • In a patient with malignant hyperthermia susceptibility, dantrolene location and reconstitution procedure are reviewed before induction, not after the first temperature spike • Verbalizing a "what if" — "if bleeding accelerates, our first three moves are X, Y, Z" — pre-loads the team's response so that recognition and action can happen almost simultaneously
Anticipation converts a crisis response from an improvisation into a rehearsal — the team is executing a plan it already discussed rather than inventing one under pressure.
Human working memory reliably fails under acute stress — this is not a character flaw, it is a well-documented feature of the stress response. CRM addresses this directly by offloading memory onto external cognitive aids and by converting ambiguous group responsibility into named, individual task ownership.
One of the most reliable predictors of poor crisis performance in simulation is an unclear or unclaimed leadership role — not the absence of a competent person, but the absence of an explicit statement of who is directing the response.
CRM teaches leaders to declare themselves out loud: "I'll lead this, please give me updates every two minutes." This single sentence does several things at once: it tells the rest of the team where to route information, it frees other skilled clinicians to become effective followers rather than competing decision-makers, and it creates a single point of situational integration.
Good followership is an equally trained skill: closed-loop acknowledgment of orders, proactively surfacing new information to the leader rather than acting unilaterally on it, and speaking up assertively (using structured techniques such as graded assertiveness or "CUS" — I'm Concerned, I'm Uncomfortable, this is a Safety issue) when something seems wrong.
Leadership in CRM is understood as a role, not a rank. A circulating nurse or junior resident can and should hold provisional leadership of a specific task (e.g., "I've got the code cart") even while a senior surgeon leads the overall response — CRM explicitly separates the leadership of the whole crisis from ownership of individual tasks.
The Stanford Emergency Manual (and its many institutional derivatives) is the best-known example of a modern perioperative cognitive aid: a structured, tabbed booklet of one-page algorithms for crises such as malignant hyperthermia, anaphylaxis, massive hemorrhage, local anesthetic systemic toxicity, and cardiac arrest. Similar structured cognitive aids underpin ACLS algorithms in cardiac arrest management.
The design logic behind these aids is deliberately simple and consistent:
• One crisis per page — no scanning through unrelated content under time pressure • Action-oriented checklist format rather than prose explanation • A designated "reader" role — someone other than the person actively performing hands-on interventions reads the checklist aloud, step by step • Boxes or checkmarks so the team can visually track what has and has not been done
Critically, cognitive aids are used by calling them out loud, not by silently glancing at them. "Someone grab the MH cart and the emergency manual, MH tab" turns a private thought into a team-wide, verifiable action.
A crisis generates far more simultaneous tasks than any one person can perform: securing the airway, establishing additional IV access, drawing labs, preparing medications, documenting events, communicating with family, and coordinating with blood bank or additional staff, often all within the same two minutes.
CRM explicitly teaches task distribution by name, not by broadcast:
• Ineffective: "someone get more IV access" (diffusion of responsibility — everyone assumes someone else will do it) • Effective: "Maria, please get a second large-bore IV" (named ownership, verifiable completion)
This single behavioral shift — from broadcasting a need to assigning a name — is one of the most reproducible, trainable improvements in simulated crisis performance, because it eliminates the ambiguity that causes both duplicated effort and, more dangerously, unclaimed critical tasks.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Malignant Hyperthermia | Rising ETCO2, masseter rigidity, tachycardia, hyperthermia | Stop triggers, hyperventilate 100% O2, dantrolene 2.5 mg/kg IV repeat, active cooling | Single-page algorithm prevents dosing/sequence errors under stress |
| Massive Hemorrhage | Rapid blood loss, hemodynamic instability | Activate massive transfusion protocol, 1:1:1 product ratio, permissive hypotension, surgical control | Coordinates blood bank, nursing, and surgical teams via one shared checklist |
| Anaphylaxis | Hypotension, bronchospasm, rash, elevated airway pressure | Stop suspected trigger, epinephrine titrated to severity, fluids, antihistamine/steroid adjuncts | Distinguishes anaphylaxis from anesthetic causes of hypotension quickly |
| Local Anesthetic Systemic Toxicity | Seizure, arrhythmia, cardiovascular collapse after LA administration | Stop injection, airway support, lipid emulsion therapy bolus + infusion, modified ACLS | Prompts lipid emulsion — an easily forgotten, non-intuitive intervention |
A well-recognized crisis with a clear leader can still fail if the team does not reach outside itself for help early enough. CRM treats "call for help early" as a distinct, trainable skill — not an admission of inadequacy, but a proactive resource-management decision made before the situation demands it.
Clinical culture has historically rewarded self-sufficiency, which can create a dangerous incentive to delay calling for help until a situation is undeniably dire. CRM deliberately inverts this: calling for help while a problem is still manageable is framed as a sign of good judgment, not weakness.
Early mobilization has compounding benefits: additional hands arrive while there is still time to brief them properly, backup clinicians arrive before exhaustion sets in for the primary team, and blood products or specialized equipment can be in transit well before they become critically needed rather than being requested at the point of desperation.
A simple heuristic taught in many CRM courses: if you find yourself thinking "we might need help," that thought itself is the trigger to call for help — not a signal to wait and see.
In massive transfusion scenarios specifically, the interval between recognizing a hemorrhage and activating a formal massive transfusion protocol is one of the most consistently measured performance metrics in CRM simulation research — shorter activation time is strongly and repeatedly associated with better outcomes.
Standard conversational communication assumes a shared, low-stakes context that simply does not exist during a crisis — background noise, competing priorities, and cognitive load all cause spoken instructions to be missed, misheard, or silently ignored. CRM formalizes a three-part loop to prevent this:
1. Sender issues a clear, directed instruction to a named individual ("Tom, please push one gram of tranexamic acid now") 2. Receiver repeats back the instruction to confirm understanding ("Pushing one gram of TXA now") 3. Sender confirms the read-back is correct, or corrects it immediately
This loop takes only a few seconds longer than a one-way instruction, but it converts communication from probabilistic ("I said it, so presumably it happened") to verified ("I confirmed it was heard correctly and is happening"). In simulation-based training, closed-loop compliance is one of the most improvable — and most frequently graded — CRM behaviors.
Resource mobilization is not a single phone call but a continuously updated assessment of what the team will need next:
• Personnel: additional physicians, nurses, or technicians appropriate to the specific crisis (e.g., a second anesthesiologist for airway backup, a hematologist for coagulopathy management) • Blood products and pharmacy: activating massive transfusion protocols, expediting factor concentrates or reversal agents • Equipment: point-of-care coagulation testing, additional monitoring, specialized airway or vascular access equipment • Space and logistics: whether the case needs to move to a different room, whether ICU or interventional radiology needs early notification
Effective teams reassess this list at intervals throughout the crisis rather than only at the initial call for help — resource needs evolve as the clinical picture evolves, and CRM trains teams to keep asking "what do we need next" rather than treating the first mobilization as sufficient.
The crisis is not over when the patient is stable — from a systems-safety perspective, it is only half finished. The structured debrief converts a single stressful event into durable individual and organizational learning, and is often the single most consistently under-performed step in real clinical practice.
Two widely taught frameworks structure how debriefs are conducted so that they generate honest reflection rather than defensiveness or a superficial pass/fail judgment:
• Debriefing with Good Judgment (Rudolph, Simon, Raemer, and colleagues): built on the "advocacy-inquiry" technique. Instead of only stating an observation ("I noticed the epinephrine was delayed"), the facilitator pairs it with a genuine question about the reasoning behind it ("I noticed the epinephrine was delayed — I'm curious what you were thinking at that point"). This assumes participants had a sensible internal logic ("good judgment") even when the outcome was suboptimal, and surfaces the actual mental model that drove the decision rather than assigning blame after the fact.
• PEARLS (Promoting Excellence and Reflective Learning in Simulation): a blended framework combining self-assessment, focused facilitation, and directive teaching as needed, moving through phases — reactions, description, analysis, and application/summary — giving facilitators a flexible structure that adapts to how much guidance a given team needs.
Both frameworks share a foundational assumption: debriefing is a facilitated conversation aimed at understanding reasoning, not a scorecard aimed at assigning fault.
The "basic assumption" underlying modern debriefing practice: we believe everyone participating in this event was intelligent, well-trained, and cared about doing well. This single stated assumption measurably changes how honestly participants engage with a debrief.
A "latent safety threat" (LST) is a system-level hazard — a missing piece of equipment, a confusing drug label, an unclear escalation pathway — that exists quietly in normal operations and only becomes visible when a crisis stresses the system. Unlike an individual's knowledge gap, an LST will reproduce the same failure for any team that encounters it.
Examples surfaced in real debriefs and in-situ simulations include: a malignant hyperthermia cart stored in a locked room without after-hours access, dantrolene reconstitution instructions that are unclear under time pressure, a difficult airway cart missing a specific size of equipment, or an emergency phone number that connects to a disconnected line.
Because LSTs are structural rather than personal, addressing them (relocating equipment, revising a protocol, fixing a label) produces safety improvement that benefits every future team — this is why in-situ simulation programs treat LST identification as one of their primary outputs, sometimes even more valuable than the training benefit to the individual participants.
A complete CRM learning loop closes at three levels simultaneously:
• Individual: each team member reflects on their own decision-making, communication, and technical performance during the event • Team: the group reviews shared mental model breakdowns, leadership clarity, task distribution, and communication loop failures specific to that event • System: identified latent safety threats and process gaps are routed to the appropriate committee or department for structural correction — equipment relocation, protocol revision, or additional training design
An after-action report typically documents: a timeline of key events and interventions, what worked well, specific opportunities for improvement, any latent safety threats identified, and concrete action items with named owners and follow-up dates. Without this final documentation and follow-up step, even an excellent debrief conversation risks generating insight that is never converted into lasting change — the debrief is only as valuable as the loop it closes.