How the EMD Protocol Guides Emergency Dispatchers

Emergency Medical Dispatch protocol is a structured system that guides 911 dispatchers through scripted questions to figure out what kind of emergency a caller is reporting, how severe it is, and what resources to send. The most widely adopted version, the Medical Priority Dispatch System (MPDS), assigns each call a priority code based on the caller’s answers, so the right level of response heads out the door while the dispatcher stays on the line giving the caller instructions on what to do until help arrives. Far from a passive call-routing exercise, EMD protocol directly shapes patient survival in time-critical emergencies, and the evidence on where it works well and where it falls short is surprisingly granular.

How the System Actually Works

When you call 911 for a medical emergency, the dispatcher does not freelance. They follow a protocol that opens with a chief complaint, essentially the main reason for the call, and then branches into a series of yes-or-no or short-answer questions designed to narrow the situation down. Is the person conscious? Are they breathing normally? Is there severe bleeding? Each answer steers the dispatcher down a decision tree that ends in a determinant code, a shorthand that tells responding crews the likely problem and its severity before they arrive on scene.

The MPDS, developed by the International Academies of Emergency Dispatch, is one of the most common systems in North America and parts of Europe and Australasia. It organizes emergencies into over 30 chief complaint categories, from abdominal pain and allergic reactions to traffic accidents and drownings. Each category has its own protocol card with questions calibrated to that type of emergency. The system also includes pre-arrival instructions, which are step-by-step directions the dispatcher reads to the caller for things like CPR, choking interventions, or controlling bleeding while an ambulance is en route.

Cardiac Arrest and the Survival Connection

The strongest evidence for EMD protocol’s impact comes from cardiac arrest, where every second of delay chips away at the chance of survival. A study of more than 96,000 out-of-hospital cardiac arrest cases found that when EMD protocols were documented as being used, patients were more likely to regain a pulse during EMS care and more likely to survive. Bystander CPR happened in about 64% of cases where EMD was used, compared with 55% where it was not, and bystander use of an automated external defibrillator was also higher in the EMD group.

Those differences translated into meaningful outcome gaps. Patients in the EMD group had roughly 19% survival at the end of the emergency event, versus about 16% in the non-EMD group, even after adjusting for patient and arrest characteristics.

The mechanism behind this is straightforward. When the dispatcher recognizes that a caller is describing someone in cardiac arrest, the protocol immediately pivots to telephone-guided CPR instructions. Getting a bystander to start chest compressions before the ambulance arrives buys the patient time, and protocols are specifically designed to make that happen as fast as possible.

Getting to the First Chest Compression

Speed matters enormously in dispatcher-assisted CPR, and researchers have studied the time from the moment a call comes in to the moment a bystander delivers the first chest compression. One large analysis found a median time of about 176 seconds, just under three minutes, from call receipt to the first compression.² Another study looking at different MPDS versions found a mean of about 240 seconds.³

Those numbers have driven revisions to the scripts dispatchers use. A simplified compression-first script cut the interval from call receipt to first compression by about 25 seconds compared with the conventional script, bringing it down to under 100 seconds in a controlled study.⁴ The idea is to strip away unnecessary questions and get to “push hard and fast on the center of the chest” as quickly as possible. Older protocol versions sometimes spent time confirming details or asking about breathing patterns before launching into CPR instructions. Newer versions front-load the compression instructions.

Stroke Recognition Over the Phone

Cardiac arrest is relatively dramatic to describe over the phone, but stroke is subtler and harder for dispatchers to catch. The symptoms, such as slurred speech, facial drooping, and one-sided weakness, can overlap with a range of other conditions, and callers often do not describe them in textbook terms. This matters because stroke patients benefit enormously from reaching a stroke-capable hospital quickly, and dispatcher recognition of a possible stroke can trigger a faster, more targeted response.

A systematic review found that dispatcher sensitivity for stroke, the ability to correctly identify a stroke when one is actually happening, ranged wildly from about 18% to 83% across studies.⁵ That enormous range reflects differences in protocols, training, and populations studied. One study using the MPDS stroke protocol in San Diego reported a sensitivity of 83%, meaning dispatchers flagged the majority of actual strokes, but the positive predictive value was only 42%, meaning fewer than half the calls they tagged as stroke turned out to be stroke on hospital discharge.⁶ Another study using a different version of the system found a lower sensitivity of 41% but a very high specificity of 96%, meaning dispatchers rarely falsely labeled a non-stroke as a stroke.⁷

The trade-off between sensitivity and specificity is a genuine tension in protocol design. A protocol tuned to catch every possible stroke will also flag many non-strokes, potentially tying up advanced resources for patients who do not need them. One tuned for precision will miss real strokes. Most systems lean toward erring on the side of caution for time-sensitive conditions, accepting some overtriage to avoid missing patients who could benefit from rapid intervention.

The Overtriage and Undertriage Problem

Triage accuracy is one of the most debated aspects of EMD protocols. Overtriage means sending a high-priority response to a call that turns out to be low-acuity, wasting resources that might be needed elsewhere. Undertriage means assigning a low priority to a call that turns out to be serious, delaying care when it matters most. Both carry real costs.

A study from South Africa found an overtriage rate of about 62% and an undertriage rate of about 15%. Some call types had extreme overtriage, with obstetric complaints overtriaged almost 90% of the time, while respiratory complaints were undertriaged about 31% of the time.⁸ A Swedish comparison of two different protocols found that one system had overtriage of about 38% with undertriage of only 6%, while the other had lower overtriage at 29% but much higher undertriage at 23%.⁹

The Swedish data neatly illustrates the tradeoff: you can have a cautious protocol that sends too many ambulances but misses few true emergencies, or a leaner protocol that conserves resources but lets more genuine emergencies slip through as low-priority. Most systems accept some overtriage as the safer failure mode, since the consequences of undertriage, a patient dying because the ambulance came too slowly, are far worse than the cost of an unnecessary lights-and-sirens response.

Language Barriers and Missed Information

EMD protocols assume the dispatcher and the caller can communicate fluently, which is often not the case in linguistically diverse communities. The consequences of that mismatch are measurable. One study found that language-barrier calls took about 33% longer to get a basic life support unit assigned and 43% longer for advanced life support, largely because connecting to an interpreter service added an average of 49 seconds before the dispatcher could even begin asking protocol questions.¹⁰

For cardiac arrest, where every second counts, those delays compound. Research comparing language-barrier and non-language-barrier cardiac arrest calls found that address acquisition took roughly twice as long, cardiac arrest recognition was delayed by about 18 seconds at the median, and CPR initiation was delayed by over 40 seconds.¹¹ A qualitative analysis of these calls revealed that non-fluent callers often provided crucial information about the patient’s condition, but it came in unexpected phrasing or at unexpected moments in the conversation, and dispatchers frequently did not pick up on it or probe further.¹²

This is not a problem that better scripts alone can fix. Protocol-driven dispatching relies on a predictable conversational structure: the dispatcher asks a question, the caller answers it, and the protocol advances. When a caller’s limited fluency disrupts that structure, the system struggles. Some dispatch centers have experimented with multilingual dispatchers or faster interpreter-connection technology, but the gap remains significant in many jurisdictions.

Caller Emotion and the Challenge of Cooperation

Even without a language barrier, the person on the other end of a 911 call is often panicked, crying, or screaming. That emotional state creates real challenges for protocol-driven dispatching. Dispatchers are trained to use specific verbal techniques, including redirecting the caller’s attention to a task they can do for the patient, which research has found effective at calming callers and sustaining their focus for the rest of the call.¹³

Interestingly, highly emotional callers are not necessarily slower to have cardiac arrest recognized by the dispatcher. One study found that uncooperative callers actually had shorter times to cardiac arrest recognition and to the first chest compression instruction. The problem was downstream: those callers were far less likely to actually receive and follow through on CPR instructions. The rate of dispatcher-assisted CPR delivery dropped from about 86% in the cooperative group to about 54% among uncooperative callers, mainly because callers either refused or were too distraught to continue.¹⁴ The researchers noted that when dispatchers were able to skillfully reassure emotional callers, CPR delivery was still achievable despite the caller’s distress. Directive language and specific, concrete instructions appear to facilitate CPR performance more effectively than open-ended or gentle requests.¹⁵

Quality Assurance and Why Compliance Varies

Having a good protocol on paper does not guarantee dispatchers follow it consistently. Protocol compliance is one of the most studied aspects of EMD performance, and the findings make a strong case for active feedback systems. One early study found that compliance scores jumped from about 76% without performance feedback to about 96% once a structured feedback process was implemented.¹⁶

In a study at a fire department-based dispatch center, a quality improvement intervention that included continuous review of dispatcher performance led to significant improvements in cardiac arrest recognition rates (from about 55% to 69%), normal breathing checks (from about 52% to 77%), and dispatcher-assisted CPR delivery (from 50% to 73%). Perhaps most strikingly, 24-hour patient survival nearly tripled, going from about 8% to 20%.¹⁷ The message is clear: protocols are only as good as the systems that monitor and reinforce how they are used. Dispatch centers without robust quality assurance programs are likely leaving survival gains on the table.

Pediatric Emergencies and Age-Related Blind Spots

Children present unique challenges for EMD protocols designed primarily around adult emergencies. One practical problem is medication dosing. Pediatric drug doses are weight-based, and paramedics arriving on scene often have to guess or estimate a child’s weight before they can give medications. Research has found that dispatchers can actually obtain reasonably accurate weights from callers for children up to about seven years old, giving crews extra preparation time en route.¹⁸

Recognizing cardiac arrest in children, however, is harder than in adults, and it gets paradoxically harder with older children. Dispatchers recognized cardiac arrest most often in infants under one year old (about 62% of the time) and least often in teenagers aged 14 to 18 (about 47%). Older children and teens had significantly lower odds of having their cardiac arrest recognized over the phone.¹⁹ One likely explanation is that callers and dispatchers both have stronger mental models of what an infant in distress looks like, while a teenager who collapses may be described in ways that overlap with fainting, seizures, or intoxication.

General pediatric dispatch protocols also tend to overtriage, sending high-priority responses for calls that turn out to be lower-acuity, though they perform reasonably well at identifying truly low-risk patients who could be managed without emergent transport.²⁰ Refining pediatric protocols remains an active area of work.

Opioid Overdose and Naloxone Instructions

As the opioid crisis has deepened, EMD protocols have expanded to include instructions for bystander naloxone administration. Newer versions of the MPDS include prompts for the dispatcher to ask callers whether naloxone is available and to guide them through administering it to a suspected overdose victim.²¹

But the integration has been uneven. A study of heroin overdose calls in Victoria, Australia, found that nearly half were dispatched under a protocol category, such as cardiac arrest or respiratory distress, that would not trigger naloxone instructions, even during a period when take-home naloxone programs were already running in the community. The authors urged EMS agencies to adopt updated dispatch system versions that allow naloxone prompts regardless of which protocol pathway the call initially falls into.²² The core issue is that an overdose victim who is not breathing may get classified as “unconscious” or “cardiac arrest” before the dispatcher ever reaches a question about possible drug use, and older protocol versions did not circle back to naloxone for those classifications.

Comparing Dispatch Models Across Countries

The MPDS is not the only dispatch system in use globally. Criteria Based Dispatch (CBD), used in parts of Scandinavia and elsewhere, takes a different approach, relying on trained dispatchers to exercise more clinical judgment rather than following a rigid question tree. Research comparing the two systems on cardiac arrest calls has produced nuanced results.

One comparison found that the MPDS system dispatched ambulances faster, with a median dispatch time of about 15 seconds compared to 33 seconds for CBD, but the CBD system got callers to actually perform chest compressions at a significantly higher rate. In one study, 70% of callers in the CBD system performed compressions versus 32% in the MPDS system, even though both systems offered CPR instructions to callers at roughly equal rates.²³ A follow-up study found that cardiac arrest recognition rates were similar between the two systems (about 82% for MPDS versus 77% for CBD), and the time to actual chest compression delivery remained slightly longer in the MPDS system.²⁴

The takeaway is not that one system is universally better. MPDS achieves faster ambulance dispatch and provides a standardized framework that works well even with less-experienced dispatchers. CBD may produce better bystander CPR compliance because its more conversational style allows dispatchers to adapt their language to individual callers. The best system for a given region depends on local workforce training, call volume, and resource availability.

Artificial Intelligence as a Dispatcher Aid

Machine learning tools are beginning to appear in dispatch centers, particularly for cardiac arrest recognition. A randomized trial tested an AI system that listened to 911 calls in real time and flagged possible cardiac arrests for the dispatcher. Dispatchers using the AI tool recognized 93% of confirmed cardiac arrests, compared with 91% for dispatchers using standard protocols alone, a difference that was not statistically significant. However, the AI alerts by themselves had higher sensitivity than dispatchers without the tool (85% versus 78%), suggesting the technology picks up on patterns that humans miss, though at the cost of more false alarms.²⁵

A separate line of research has explored using AI to detect the emotional state of callers from the audio itself, with the idea that recognizing extreme distress early could help dispatchers adjust their communication strategy. One model achieved a positive predictive value of about 85% using only the first ten seconds of a call recording.²⁶ These tools are still early-stage, but they point toward a future where dispatchers work alongside algorithms rather than relying entirely on scripted protocols.

The Mental Health Cost for Dispatchers

Dispatchers process emergencies all day, every shift, often hearing people at the worst moments of their lives. The toll is real and underappreciated. A cross-sectional study of emergency call-takers and dispatchers found that about 11% screened positive for PTSD, roughly 16% screened positive for depression, and about 7% for anxiety disorders. Nearly 3% met criteria for severe secondary traumatic stress.²⁷ Unlike paramedics and firefighters, dispatchers often do not get resolution on their calls. They guide a caller through CPR and then move to the next call without ever learning whether the patient survived. That lack of closure, combined with high call volumes and the emotional intensity of the work, creates a chronic stress environment that dispatch centers are only beginning to address with formal support programs.