What Are the 5 Levels of Triage in Emergency Care?

Emergency departments around the world use a five-level triage system to sort patients by how urgently they need care. The levels range from Level 1, which means you need life-saving intervention right now, to Level 5, which means your condition requires no emergency resources at all. The most widely adopted version in the United States is the Emergency Severity Index (ESI), but several other five-level systems operate internationally, and they do not all use the same logic to assign their levels.

The Five Levels, Briefly

Though the specific names vary slightly between systems, the five tiers follow a consistent structure across major triage scales. Level 1 is reserved for patients who are dying or will die without immediate intervention: cardiac arrest, major trauma with active hemorrhage, respiratory failure. Level 2 covers high-risk situations where a delay could lead to serious harm: chest pain with cardiac features, stroke symptoms, severe allergic reactions, altered mental status, or certain dangerous vital-sign combinations. Levels 3 through 5 cover everything from moderately complex complaints down to minor issues. In the ESI system specifically, the distinction between these lower three levels hinges not on how sick you look, but on how many hospital resources the triage nurse predicts you will need.

A study validating ESI resource predictions found that mean resource use decreased steadily across the levels: Level 1 patients used an average of five distinct resources, Level 2 about four, Level 3 about three, Level 4 roughly one, and Level 5 nearly zero.¹ That monotonic pattern is why the system works: the levels consistently predict how much the emergency department will need to invest in your care.

How ESI Assigns a Level

The ESI uses a branching decision algorithm rather than a symptom checklist. The triage nurse works through a series of decision points. The first question is whether the patient requires an immediate life-saving intervention. If yes, that is Level 1. If not, the nurse asks whether the patient is in a high-risk situation, is confused or lethargic, or is in severe distress. A “yes” to any of those lands the patient at Level 2.

For everyone else, the nurse estimates how many different types of resources the visit will consume. The ESI handbook defines resources as distinct categories: lab tests, imaging, intravenous fluids, nebulizer treatments, specialty consultations, simple procedures (one point each), and complex procedures (two points).² If the nurse expects two or more resource types, the patient is Level 3. If one resource is anticipated, Level 4. If none, Level 5. Vital signs act as a safety net: in the latest version of the ESI, any patient initially sorted into Levels 3 through 5 who has dangerously abnormal vital signs gets bumped up to Level 2.³

This resource-prediction approach is what distinguishes ESI from other five-level systems. Most triage scales ask “how sick is this person?” ESI asks that question at the top, but for the majority of patients who are not critically ill, it shifts to asking “how much will the department need to do for this person?” That pragmatic framing helps the department plan staffing and bed allocation in real time.

Other Five-Level Systems Used Worldwide

The ESI is not the only game in town. Four other major five-level triage instruments are widely used internationally: the Canadian Triage and Acuity Scale (CTAS), the Manchester Triage System (MTS), the Australasian Triage Scale (ATS), and the South African Triage Scale (SATS).⁴ Each one uses five levels, but they get there differently.

The CTAS, for instance, assigns levels based on specific presenting complaints and “second-order modifiers” that adjust severity. It is one of the only systems that explicitly defines a maximum time to physician assessment for each level, a required reassessment interval, and a pain scale integrated into every triage encounter.⁵ The MTS uses flowcharts organized by presenting complaint, and the ATS is the only system that specifies both the time to initial assessment and the maximum duration of a patient examination. The ESI, by contrast, does not define time-to-physician targets for Levels 2 through 5 at all, leaving those operational decisions to each hospital.

These differences matter when studies try to compare performance. A review of triage instruments found that five-level scales are significantly more valid and reliable than three-level scales, and that the CTAS and ESI show the strongest inter-rater reliability, while the ATS and MTS are only moderately reliable.⁶

How Reliable Is the Triage Assignment?

A triage system is only as good as the agreement between the nurses using it. If two nurses looking at the same patient assign different levels, someone is getting the wrong priority. A meta-analysis of 19 ESI studies across six countries found a pooled inter-rater agreement score of about 0.79 on a 0-to-1 scale, which falls in the “substantial” range.⁷ Agreement was higher when the raters were experienced, when the adult version of the scale was used, and when the most current edition was in play.

That overall figure masks some trouble spots. Level 2, the “emergent” category, is consistently the hardest level for nurses to agree on. A Japanese simulation study found that overall ESI agreement between expert and triage nurses was excellent, but agreement specifically on Level 2 assignments dropped to only moderate.⁸ Level 2 sits in a gray zone between “this person could die” and “this person needs a lot of resources but is stable,” and that ambiguity creates disagreement. Meanwhile, a study tracking early ESI adoption in Jamaica found overall agreement was low initially, though it improved over time with continued training.⁹

Nurse experience is a major factor. An evidence review found that triage accuracy rates range from about 59% to 82%, with experience in triage consistently linked to higher accuracy.¹⁰ Spending more clinical time doing triage and taking longer on each individual assessment both improved classification accuracy in another study.¹¹ The implication is straightforward: a rushed assessment by a less experienced nurse is more likely to sort you into the wrong level.

What Happens When Triage Gets It Wrong

Being assigned too low a level, known as undertriage, is the more dangerous error. Undertriaged high-acuity patients in one large study experienced an average eight-minute delay in care compared with patients correctly identified as high acuity.¹² Eight minutes may not sound like much, but in conditions like stroke or sepsis, every minute counts. The same study found that undertriaged patients tended to have more chronic health conditions and higher rates of recent healthcare use, suggesting their complexity made it harder to classify them quickly at the front door.

In pediatric emergency departments, the consequences of undertriage are particularly stark. Undertriaged children had longer waits for IV placement, nebulized treatments, and hospital admission. They were also nearly four times as likely to return to the emergency department needing admission or emergency intervention compared with properly triaged children.¹³

Overtriage, assigning someone a higher level than warranted, is less immediately dangerous to the individual patient but carries its own costs. It pulls staff and beds away from genuinely sicker patients and contributes to emergency department crowding. A study of children presenting with behavioral health conditions found that over half of visits were overtriaged, while about 8.5% were undertriaged.¹⁴ The undertriage rate was higher among younger children, boys, Hispanic patients, and non-Hispanic Black patients compared to non-Hispanic white patients, pointing to potential biases baked into how triage decisions are made for certain groups.

Racial Disparities in Triage

The pattern of unequal triage extends beyond pediatric behavioral health. A study comparing triage scores between African American and white patients presenting with the same chief complaints found that African American patients received significantly lower acuity ratings on average. That translated into longer wait times to reach a treatment area, with much larger gaps in certain complaint categories.¹⁵ Because triage is partly subjective, especially in systems that rely on a nurse’s assessment of distress or pain severity, implicit bias can influence which level a patient receives. The evidence is clear enough that researchers and professional organizations increasingly flag triage equity as a priority for system redesign.

Vital Signs as a Safety Check

One of the built-in safeguards in modern triage systems is the use of vital signs to catch patients who might otherwise be undertriaged. In the latest ESI revision, every patient assigned to Levels 3 through 5 gets a vital-sign check, and if their numbers fall into a predefined danger zone, they are automatically bumped up to Level 2. This prevents a patient who seems stable on the surface but has, say, a dangerously low blood pressure or a racing heart from sitting in the waiting room.

The specific thresholds can be a problem for older adults, whose baseline vitals often differ from younger populations. Research on this question found that elderly patients with systolic blood pressure below 80, a heart rate below 40 or above 140, body temperature below 35°C, or severely depressed consciousness had critical event rates above 20% and should be moved up to Level 1.¹⁶ Using age-adjusted thresholds captured more genuinely sick elderly patients and improved the proportion of critical cases identified at the highest triage levels. This is an area of active refinement across triage systems, because one-size-fits-all vital sign cutoffs miss people at the extremes of age.

Psychiatric Emergencies and the Triage Challenge

Mental health presentations are notoriously difficult to triage using systems designed around physical complaints and resource use. A patient in acute psychosis, for example, may not need lab tests or imaging, but could be at serious risk. The ESI’s resource-prediction approach can undercount the severity of these visits, since many psychiatric emergencies are resource-light in the traditional sense yet demand urgent attention.

The CTAS handles psychiatric presentations more explicitly. For a complaint like “depression, suicidal, or deliberate self-harm,” the system uses specific modifiers: an active suicide attempt or clear plan results in Level 2, suicidal thoughts without a plan in Level 3, and depression with no suicidal ideation in Level 4.¹⁷ But even the CTAS has been criticized for overtriaging psychiatric patients. When CTAS scores were compared with the Australian Emergency Mental Health Scale, the CTAS rated about half of psychiatric patients as “urgent” and nearly a third as “emergent,” while the Australian system classified three-quarters of the same patients as presenting no danger to self or others.¹⁸ That gap suggests general-purpose triage tools may not map cleanly onto the spectrum of psychiatric urgency. Some hospitals have adopted specialty psychiatric triage instruments that assign risk levels based on danger to self or others, behavioral agitation, and psychotic symptoms rather than physical vital signs and resource estimates.¹⁹

Triage for Children

Kids are not small adults, and triage systems have had to adapt accordingly. Pediatric vital sign norms differ by age, children cannot always describe their symptoms, and certain pediatric conditions (like high fevers in infants or breathing difficulties in toddlers) carry risks that are easy to underestimate. Pediatric-adapted versions exist for the major triage scales. A review of the evidence found that the pediatric CTAS, ESI version 4, and the MTS all have validated pediatric tracks with acceptable reliability.²⁰ In addition to these adapted triage scales, some departments use supplemental early warning scores, like the Pediatric Early Warning Score, to flag children whose condition is deteriorating even if their initial triage level seemed stable.

The inter-rater reliability meta-analysis noted that agreement scores for pediatric triage versions tend to be lower than for adult versions.⁷ Part of the difficulty is that the resource-prediction step in the ESI is harder to perform for children, whose presentations are less predictable and whose resource needs can change rapidly.

How Triage Levels Predict What Happens to You

Beyond deciding wait order, triage levels carry real predictive weight for patient outcomes. A large retrospective study of prehospital triage in Taiwan found that Level 1 patients had dramatically elevated risks: the odds of death were about 84 times higher than baseline, odds of ICU admission roughly 13 times higher, and odds of a critical event about 25 times higher. Level 2 patients also had significantly increased risks across all outcomes, with odds of death about four times higher than lower-acuity patients.²¹ These numbers come from one system in one country, but the general pattern, that higher triage levels predict worse outcomes, has been validated across multiple systems and settings.²²

This is part of what researchers mean when they talk about a triage system’s “construct validity”: if Level 1 and Level 2 patients consistently end up sicker, admitted more often, and dying at higher rates, then the system is measuring something real about urgency. It does not mean the system is perfect at the individual level, but the population-level sorting works.

Why Your Wait Time Feels Random (and What Actually Helps)

One of the most common frustrations for emergency department patients is watching someone who arrived later get taken back first. Understanding the five levels makes this less mysterious: a Level 2 chest pain patient who walks in after you will always jump ahead of your Level 4 ankle sprain, regardless of arrival order. But even within the same level, wait times can vary depending on bed availability, staffing, and the current mix of patients already being treated.

Research on patient satisfaction consistently shows that communication matters as much as actual wait duration. A study that tested giving patients estimated wait times at triage found that patients who received any information about expected delays were significantly more satisfied, even when their actual waits were long. Patients who received no communication about delays were up to five times more likely to rate their visit poorly compared with those who were kept informed.²³ The takeaway for departments is simple: telling someone “your wait will be about two hours” does more for satisfaction than shaving fifteen minutes off that wait in silence.

Mass Casualty Triage Is a Different System Entirely

If you have seen color-coded tags in disaster movies, those represent mass casualty incident (MCI) triage, which is a fundamentally different approach from emergency department triage. Hospital-based triage aims to identify and prioritize the sickest patients for comprehensive care. MCI triage, by contrast, is designed to do the greatest good for the greatest number, which sometimes means passing over the most severely injured in favor of those more likely to survive with immediate treatment.²⁴ In MCI protocols, a patient who would be a clear Level 1 in a normal emergency department might be tagged “expectant” (unlikely to survive) and receive comfort care only, because the resources needed to save them could save three other people instead. The ethical logic flips when demand overwhelms capacity.

Artificial Intelligence as a Second Opinion

Machine learning models are increasingly being tested as triage decision support tools. Scoping and systematic reviews have found that these models generally outperform conventional triage scores at predicting outcomes like hospital admission, ICU transfer, and length of stay.²⁵²⁶ One model using a gradient-boosting algorithm achieved an area under the curve above 0.91 for predicting critical outcomes, outperforming the standard CTAS triage system on the same patient data.²⁷

None of these models have replaced human triage nurses, and their current role is advisory. The appeal is that an algorithm can incorporate dozens of variables simultaneously, including vital signs, lab values, past medical history from the electronic health record, and time-of-day patterns, and flag patients whose risk profile does not match their assigned triage level. In theory, this could catch some of the undertriage cases that lead to delayed care. In practice, the field is still working through questions about how these tools interact with clinical workflow, whether they reduce disparities or reinforce them, and how to handle the inevitable cases where the algorithm disagrees with the nurse’s gut feeling.