The best available synthesis estimates that 5.7% of emergency department visits involve at least one diagnostic error: approximately one error for every eighteen patients seen. An estimated 2.0% of patients experience preventable harm related to diagnosis and 0.3% experience serious harm. These figures should be interpreted with caution: the overall estimates come from three studies conducted in Canada, Spain and Switzerland, involving a total of 1,758 patients (Newman-Toker et al. 2022). versió catalana
In Catalonia, hospital emergency departments recorded 3,686,790 visits in 2024. The total number of contacts with urgent care services exceeded 6.1 million when urgent primary care centres (CUAPs), the Medical Emergency System (SEM) and psychiatric hospitals are included (Servei Català de la Salut 2025). The review’s estimates cannot be applied directly to this volume of activity because the settings, populations and measurement methods differ. Nevertheless, the scale of emergency care means that even a modest incidence may affect many people and be highly relevant.
What is a diagnostic error?
The National Academies of Sciences, Engineering, and Medicine define diagnostic error as the failure to establish an accurate and timely explanation of the patient’s health problem or to communicate that explanation to the patient and the healthcare team. This definition extends diagnosis beyond assigning a label: it includes when the diagnosis is reached, how it is communicated and whether the information obtained can be acted upon (National Academies of Sciences, Engineering, and Medicine 2015).
In the emergency department, diagnosis is often provisional. A patient may be discharged before the definitive cause of the symptoms is known and still have received safe care: immediately dangerous alternatives have been considered, uncertainty has been explained and follow-up has been arranged. A difference between the emergency diagnosis and the final inpatient diagnosis does not, by itself, prove that an error occurred. Disease evolves, new information emerges during hospitalisation and several teams may be involved (Schols et al. 2026).
This is why it is useful to speak of diagnostic safety and diagnostic excellence. The aim is to reach an explanation that is sufficiently accurate and timely given the information available, communicate it clearly and keep the diagnostic process open when uncertainty remains. Patients and caregivers are part of that process: they provide the history, notice changes and need to know what course to expect and when to seek care again (Mahajan 2025).
Why is diagnosis particularly difficult in emergency care?
We make decisions with incomplete information, often without a prior relationship with the patient and while the clinical picture is still evolving. We do so amid interruptions, constant task switching, time pressure and crowding. Test results arrive at different times, patients move between clinicians and some results become available only after discharge. Every handover, shift change or transfer of care creates an opportunity for information to be lost or responsibility to remain unassigned.
Errors are not evenly distributed
Fractures are the most frequently missed diagnosis in reporting systems and malpractice claims, but they usually cause less severe harm than major vascular, infectious or oncological diseases. These three categories, known as the Big Three in the literature, account for approximately 72% of serious harm associated with diagnostic errors in emergency departments. Fifteen clinical conditions account for nearly 68% of serious harm (Fernholm et al. 2019; Newman-Toker et al. 2022).
The fifteen conditions associated with the greatest share of serious harm
The AHRQ review identifies fifteen conditions that account for 68.1% of serious harm associated with diagnostic errors. Ten belong to the “Big Three”—vascular, infectious and oncological diseases—and five fall outside these groups. The miss rate is the proportion of cases of a disease that were not diagnosed initially; it is not the percentage of claims. Where the review found no condition-specific estimate, the rate is listed as not estimated.
Vascular conditions
| Condition | Quantified risk | Associated factors and presentations | Vulnerable population or situation |
|---|---|---|---|
| Stroke | 17% initially missed; 13.5% of serious harm | Dizziness or vertigo, headache, confusion, gait disturbance, nausea or vomiting, absence of a focal deficit, and posterior circulation involvement. The miss rate is 4% when focal weakness is present and may reach 40% when dizziness or vertigo predominates. | Younger people, women, and Black or Hispanic patients in several studies; mild presentations, low triage priority and arrival without emergency medical transport. |
| Acute myocardial infarction | 1.5% missed; 26% delayed among admitted patients; 8.3% of serious harm | Absence of chest pain, dyspnoea, syncope or a fall, nausea or vomiting, fatigue, weakness, confusion and low triage priority. | Women, older people and Black patients were at higher risk in some studies, although findings were not consistent. Particular vigilance is required when symptoms are not centred on the chest. |
| Aortic aneurysm or dissection | 36% missed; 6.1% of serious harm | Absent or mild pain, abdominal or back pain, syncope, dyspnoea, fever and a presentation resembling an acute coronary syndrome, including ECG changes or a positive troponin result. | Patients without classic aortic pain or a suggestive history. Some studies report longer delays in women; demographic evidence is inconsistent. |
| Venous thromboembolism | 20% missed; 5.1% of serious harm | Cough, haemoptysis, syncope, pulmonary infiltrates, absence of dyspnoea in some series, and failure to assess clinical probability or order a D-dimer when appropriate. | Patients with atypical presentations or underestimated pre-test probability. No demographic group has been identified as consistently vulnerable. |
| Arterial thromboembolism | 15% delayed ≥24 h; 38% delayed ≥6 h; 2.8% of serious harm | Non-specific abdominal pain, mesenteric ischaemia not suspected before CT, subtle vascular signs, and delays in imaging or vascular assessment. | Patients with pain out of proportion to examination findings or subtle signs of ischaemia. Evidence about demographic subgroups is limited. |
Infectious conditions
| Condition | Quantified risk | Associated factors and presentations | Vulnerable population or situation |
|---|---|---|---|
| Meningitis or encephalitis | 22% missed; 4.7% of serious harm | Headache or altered mental status with absent or unclear meningeal signs, a non-specific early phase and lack of paediatric consultation when needed. | Infants and children with subtle signs, and older people with afebrile presentations or confusion. In one paediatric study, cases in children aged 91 days to 5 years were missed more often than in those aged 30 to 90 days. |
| Sepsis | 10% missed; 4.7% of serious harm | Weakness or fatigue, confusion, absence of fever, only minor abnormalities in initial vital signs and low triage priority. | Older people and children with non-specific manifestations. In paediatrics, lower patient volume and lack of specialist consultation have been associated with greater risk. |
| Spinal or intracranial abscess | 56% missed*; 2.6% of serious harm | Back pain or headache without fever or an initial neurological deficit, progressive symptoms, repeat visits and low suspicion because the condition is rare. | Patients who re-present, are at risk of bacteraemia or immunosuppression, or have early neurological deficits. The available figure combines emergency and outpatient care. |
| Pneumonia | 14% missed; 2.1% of serious harm | Dyspnoea, weakness or confusion without fever or typical respiratory symptoms, equivocal initial imaging, and coexisting heart failure or other alternative diagnoses. | Older, frail or multimorbid patients with non-specific manifestations. |
Oncological conditions
| Condition | Quantified risk | Associated factors and presentations | Vulnerable population or situation |
|---|---|---|---|
| Lung cancer | Rate not estimated; 3.9% of serious harm | An incidental nodule or lesion not identified on imaging, a report not reviewed or a result not communicated, and lack of a closed-loop follow-up process. | Patients with incidental findings on radiography or CT and no named person or team responsible for confirming follow-up. |
Other high-risk conditions outside the “Big Three”
| Condition | Quantified risk | Associated factors and presentations | Vulnerable population or situation |
|---|---|---|---|
| Spinal cord compression or injury | Rate not estimated; 5.4% of serious harm | Spinal pain, mild or evolving neurological deficit, trauma with competing injuries, and delayed MRI or specialist assessment. | Patients with multiple trauma, a difficult examination, a history of cancer or infection risk, and early neurological symptoms. |
| Traumatic brain injury or traumatic intracranial haemorrhage | Rate not estimated; 3.6% of serious harm | Multiple trauma, injuries competing for attention, altered consciousness or intoxication, and difficulty deciding on or interpreting imaging. | Patients with multiple trauma or an unreliable examination and history; the risk of secondary injury rises with trauma severity. |
| Cardiac arrhythmia | Rate not estimated; no condition-specific studies found; 2.2% of serious harm | Intermittent symptoms, resolved syncope or palpitations, a single normal ECG, and lack of adequate monitoring or follow-up. | Patients with transient or recurrent episodes, especially when assessed between episodes. Evidence about subgroups is insufficient. |
| Gastrointestinal perforation or rupture | Rate not estimated; 1.6% of serious harm | Initially non-specific abdominal pain, absent or subtle peritonism, and delayed reassessment or imaging. | Older, frail or immunocompromised patients, who may have attenuated inflammatory signs; condition-specific evidence is limited. |
| Intestinal obstruction, with or without hernia | Rate not estimated; 1.4% of serious harm | Early or intermittent symptoms, absence of vomiting, inconclusive early examination or imaging, and failure to integrate a history of abdominal surgery or hernia into decision-making. | Older people and patients with previous abdominal surgery or hernia; specific evidence on vulnerability is limited. |
* The 56% estimate for spinal abscess comes from a study that combined emergency and outpatient care and should therefore not be interpreted as an emergency-department-only rate. Percentages of “serious harm” indicate the share attributable to each condition among analysed claims involving death or permanent disability, not an individual patient’s probability of experiencing an error. Source: Newman-Toker et al. 2022, tables 3, 7 and 9.
Stroke heads this list. The review estimates an average miss rate of 17%, but the risk varies markedly by presentation: it is approximately 4% when focal weakness is present and rises to 40% when dizziness or vertigo predominates. A normal head CT does not exclude a posterior circulation ischaemic stroke and may create false reassurance if it is not integrated with the history and clinical examination.
Acute myocardial infarction has a much lower miss rate, around 1.5%, but its frequency means that the absolute number of affected patients remains relevant. Furthermore, identifying the disease during admission does not rule out a delay in treatment caused by an initially non-specific diagnostic impression. Other high-risk vascular conditions include acute aortic syndrome and venous thromboembolism.
Among the infections responsible for the greatest share of serious harm are sepsis, meningitis, encephalitis, pneumonia, and spinal or intracranial abscesses. Spinal cord compression or injury, which ranks fourth in the AHRQ list, is a distinct neurological emergency.
In the oncological group, lung cancer is the condition ranked highest among diseases associated with serious harm from diagnostic error.
The same pattern recurs: risk increases when a serious disease presents with mild, transient, non-specific or unexpected symptoms. Stroke in a young person, afebrile sepsis in an older person, acute coronary syndrome without chest pain, or a spinal abscess beginning as back pain are situations in which the real presentation differs from the classic picture taught in textbooks.
Biases explain part of the problem
In malpractice claims analysed in the AHRQ review, 89% of cases included some failure in clinical assessment, reasoning or decision-making. This figure comes from cases that had already caused harm and led to a claim; it does not mean that clinicians are responsible for 89% of diagnostic errors. Errors usually have several causes and arise from interactions among the patient, clinicians, the team, the organisation and technology (Kachalia et al. 2007; Newman-Toker et al. 2022).
Some cognitive biases are particularly recognisable in emergency care:
- Premature closure: accepting the first plausible explanation before checking relevant alternatives (an inadequate differential diagnosis).
- Anchoring: maintaining the initial hypothesis despite new information that should alter it.
- Confirmation bias: giving greater weight to information that supports our impression and discounting evidence that contradicts it.
- Availability bias: overestimating a diagnosis because we have seen it recently or remember a particularly striking case.
- Framing: allowing a prior label, such as “frequent attender” or “anxiety”, to shape the interpretation of symptoms.
- Overconfidence: subjective certainty exceeding the actual accuracy of the diagnosis.
Knowing these biases helps us describe what may have happened. Training about bias alone, however, has shown limited benefit unless it is accompanied by changes in processes and working conditions. Interventions must operate where and when decisions are made.
What can we do during emergency care?
Take a diagnostic pause before discharge
Not every discharge needs to become an endless checklist. In selected patients—for example, those who may have one of the fifteen conditions listed above—a brief pause can help us review the diagnostic hypothesis before closing the case:
- What does not fit the working diagnosis?
- Which serious diagnosis can I not afford to miss in this presentation?
- What evidence makes that serious alternative diagnosis sufficiently unlikely?
- Is there any discordant finding that I have not yet explained?
Create a safety net at discharge
A safe discharge should explain the working diagnosis, the degree of uncertainty that remains, the expected course and the changes that should prompt the patient to seek care again. Warning signs must be specific and understandable, and patients must know where and when to seek help. Saying “return to the emergency department if you get worse” provides little guidance if the patient does not know what “worse” means in their particular case (Schiff et al. 2018).
Listening to family members or caregivers is especially important when they describe changes in behaviour, speech, gait or level of consciousness that may not be apparent during a brief visit.
Close the loop on pending results
Every pending culture, laboratory result or imaging report needs an identified responsible person, an alert mechanism and confirmation that the result has been reviewed and communicated. If we assume that “someone will see it”, follow-up is left to chance.
Facilitate a second opinion
A second clinical assessment or second reading of a selected test, such as an ECG or radiograph, may help identify discordant findings. In specific settings, a validated artificial intelligence system can provide additional support, with professional oversight and a clear pathway for managing alerts. Its usefulness depends on real-world performance, the target population and integration into clinical workflow (Wright et al. 2019).
Clinical decision support systems
In high-risk situations, electronic health records can incorporate validated decision rules or risk-stratification tools, such as the Canadian C-Spine Rule, the Wells score for pulmonary embolism, the HEART score or NEWS2. These tools are useful when applied to the population and purpose for which they were validated and when presented at the right point in the workflow without creating alert overload.
Learn what happened afterwards
We often do not know the final diagnosis of a patient we treated. Reviewing return visits to the emergency department within 72 hours of discharge, or discrepancies between the diagnosis at hospital admission and the diagnosis at hospital discharge, can turn clinical activity into learning. These signals are not automatically equivalent to an error and must be reviewed in clinical context. A recent systematic review identified eleven studies of diagnostic feedback in emergency departments. The results suggest a possible benefit, but most of the evidence is of low quality and the most useful format, frequency and timing remain unknown (Schols et al. 2026).
Reduce the conditions that foster error
Interruptions should be reduced during clinical assessment and, especially, before making and communicating a discharge decision. We must also address crowding, standardise handovers (for example, with IDEAS or similar tools), ensure supervision when needed, and facilitate observation or follow-up when uncertainty remains too high. Diagnostic safety is both a clinical and an organisational responsibility.
Conclusions for emergency care practice
- Diagnostic error is common and clinically relevant, although overall estimates still carry considerable uncertainty.
- Serious harm is concentrated in a small number of diseases, particularly vascular and infectious conditions.
- Mild, atypical or non-specific presentations of serious disease pose the greatest threat to diagnostic safety.
- Before discharge, clinicians should review discordant findings and the serious diagnosis that could still explain the presentation.
- Communicating uncertainty, specifying warning signs and assigning responsibility for pending results are part of the diagnostic process.
- Feedback and second opinions should be integrated into routine work and assessed using meaningful indicators.
- Improvement requires action on clinical reasoning, communication and system conditions.
References
- Newman-Toker DE, Peterson SM, Badihian S, et al. Diagnostic Errors in the Emergency Department: A Systematic Review. Comparative Effectiveness Review No. 258. Rockville (MD): Agency for Healthcare Research and Quality; 2022. Errata and addendum, 2023. doi:10.23970/AHRQEPCCER258.
- National Academies of Sciences, Engineering, and Medicine. Improving Diagnosis in Health Care. Washington (DC): National Academies Press; 2015. doi:10.17226/21794.
- Mahajan P. From diagnostic errors to diagnostic excellence in emergency care: Time to flip the script. Acad Emerg Med. 2025;32(3):366-368. doi:10.1111/acem.15033.
- Wright B, Faulkner N, Bragge P, Graber M. What interventions could reduce diagnostic error in emergency departments? A review of evidence, practice and consumer perspectives. Diagnosis (Berl). 2019;6(4):325-334. doi:10.1515/dx-2018-0104.
- Schols LA, Zwaan L, Woltman AM, Scheepers RA, Haagsma JA, Rood PPM. The effect of feedback on the diagnostic process of physicians at the emergency department: a systematic review. Eur J Emerg Med. 2026;33(4):228-236. doi:10.1097/MEJ.0000000000001337.
- Fernholm R, Pukk Härenstam K, Wachtler C, Robinson KA, Carlsson AC, Holzmann MJ. Diagnostic errors reported in primary healthcare and emergency departments: a retrospective and descriptive cohort study of 4830 reported cases of preventable harm in Sweden. Eur J Gen Pract. 2019;25(3):128-135. doi:10.1080/13814788.2019.1625886.
- Kachalia A, Gandhi TK, Puopolo AL, et al. Missed and delayed diagnoses in the emergency department: a study of closed malpractice claims from 4 liability insurers. Ann Emerg Med. 2007;49(2):196-205. doi:10.1016/j.annemergmed.2006.06.035.
- Schiff GD, Martin SA, Eidelman DH, et al. Ten principles for more conservative, care-full diagnosis. Ann Intern Med. 2018;169(9):643-645. doi:10.7326/M18-1468.
- Servei Català de la Salut. Memòria 2024. Barcelona: CatSalut; 2025. Available at: https://scientiasalut.gencat.cat/handle/11351/13733.
If you prefer, you can listen to episode 58 of EMpod.cat or use any music or podcast platform to hear a more detailed discussion of this topic and other stories related to emergency medicine.
Xavier Basurto xbasurto@urgem.cat