Defensive medicine and advanced imaging overuse in emergency departments: A focused review of imaging stewardship and overcrowding

Abstract

Emergency departments (EDs) are encountering a significant increase in patient volumes, compounded by greater reliance on advanced radiological imaging modalities. This trend is most evident for computed tomography and magnetic resonance imaging, with expanding use of emergency ultrasonography in selected pathways. These modalities are essential for time-sensitive diagnosis, triage, and risk stratification. Under diagnostic uncertainty, time pressure, crowding, and medico-legal concern, imaging may exceed evidence-based thresholds. Imaging overuse, also described as overutilization, has been associated with radiology bottlenecks, longer ED turnaround times, higher healthcare costs, and worse overcrowding. It may also expose patients to radiation, contrast-related risks, incidental findings, downstream testing, overdiagnosis, and avoidable financial burden. Excessive imaging demand further increases the workload of radiologists and emergency physicians and may contribute to reporting delays, cognitive fatigue, and burnout. This focused narrative review synthesizes the clinical, behavioral, and system-level drivers of overuse of advanced imaging in EDs, and proposes a conceptual reciprocal-pressure framework linking crowding, diagnostic uncertainty, defensive ordering, and radiology bottlenecks. It also discusses mitigation strategies, including clinical decision rules, computerized clinical decision support, Choosing Wisely initiatives, artificial intelligence-supported tools, shared decision-making, audit and feedback, and multidisciplinary imaging stewardship. The evidence supporting these strategies is heterogeneous, and their reported effects are not equivalent. The aim is to support safer, more efficient, and patient-centered diagnostic workflows in emergency care.

Key Words: Defensive medicine; Emergency service, hospital; Diagnostic imaging; Tomography, X-ray computed; Magnetic resonance imaging; Medical overuse; Decision support systems, clinical; Crowding

Core Tip: Advanced imaging overuse in emergency departments is driven by diagnostic uncertainty, time pressure, malpractice concerns, patient expectations, consultant requests, and system-level workflow constraints. This narrative review synthesizes current evidence linking defensive medicine and low-value imaging to radiology bottlenecks, prolonged emergency department throughput, unnecessary costs, incidental findings, and clinician workload. It also proposes a conceptual reciprocal-pressure framework, in which crowding and defensive imaging reinforce each other, that has not been tested experimentally. It also summarizes reported mitigation strategies, including clinical decision rules, computerized clinical decision support, Choosing Wisely initiatives, shared decision-making, audit and feedback, and multidisciplinary imaging stewardship.

INTRODUCTION

Technological developments in radiology have expanded access to advanced imaging in emergency care, particularly computed tomography (CT) and magnetic resonance imaging (MRI). Among these modalities, CT has become the dominant driver of emergency imaging growth because of its rapid acquisition, broad diagnostic scope, and central role in trauma, stroke, pulmonary embolism, and acute abdominal pathways[1-3]. MRI has also gained increasing relevance in selected emergency conditions, supported by improved protocols and wider availability. Ultrasonography (US), in contrast, occupies a dual role: It is an expanding emergency imaging modality, but it also serves as a radiation-free alternative in several stewardship pathways[4,5].

These advances have improved the ability of emergency physicians to detect life-threatening conditions rapidly and to support triage, risk stratification, and treatment planning[6-8]. However, this enhanced diagnostic capability has also created a major dilemma. Although CT and MRI are essential in emergency departments (EDs), their use may exceed evidence-based thresholds under diagnostic uncertainty, time pressure, high patient volume, medico-legal concern, and system-level incentives[9-11].

Emergency imaging has also grown faster than ED attendance in several reported settings[2,3,12]. Utilization growth alone is not evidence of overuse. Case mix, disease prevalence, access, pathway redesign, and scanner availability all shift utilization independently of appropriateness. None of these factors can be separated from overuse without appropriateness, yield, or outcome data[13,14]. Throughout this article we therefore use three terms in distinct senses. Advanced imaging means CT, MRI, and, where it substitutes for cross-sectional imaging in a defined pathway, US. Overuse means imaging performed outside accepted indications. Low-value imaging means an examination unlikely to change management, and we apply this narrower operational concept when appropriateness or yield data exist[15-17].

The central concern is not increased imaging itself, but the proportion of examinations that are unlikely to change clinical management. Such examinations are increasingly described as low-value imaging or imaging overuse, particularly when potential harms, costs, or downstream consequences outweigh the expected diagnostic or therapeutic benefit[15-17]. This pattern is often framed as a physician-level problem, but available evidence suggests a broader system-level issue[9,18]. Physicians face increasing workloads, more complex patient profiles, limited time for observation or follow-up, and continuous concern about missing low-probability but high-consequence diagnoses. In this context, defensive medicine in response to malpractice concerns is one element of a more comprehensive diagnostic risk-management environment that promotes non-evidence-based testing[9,16,17,19].

Therefore, the overuse of advanced medical imaging in EDs should not be viewed solely as a problem of excessive test ordering. Instead, it should be understood as a diagnostic stewardship challenge shaped by clinical uncertainty, professional risk perception, patient expectations, institutional workflows, and resource shortages[20,21]. The focus of this article is more specific than a broad review of emergency imaging overuse. We examine why low-value advanced imaging continues to occur in EDs despite the availability of validated decision rules, and we identify potential intervention points within the imaging-ordering pathway. Previous reviews have described the drivers of imaging overuse in emergency care, while others have evaluated interventions aimed at reducing CT utilization[10,22,23]. However, the connection between these two areas is discussed less often. ED crowding changes the clinical environment in which imaging decisions are made, and increased imaging demand can, in turn, worsen crowding. This review frames that relationship as a reciprocal cycle and discusses where it may be interrupted. It also considers the relative strength of available interventions, rather than presenting all strategies as equivalent.

In this focused review, we aimed to synthesize the complex relationship between defensive medical practices in EDs and the overuse of advanced imaging modalities. In this context, we examine the clinical, behavioral, and system-level factors that trigger unnecessary imaging requests. We investigate the consequences of low-value imaging at the patient, clinician, and system levels, with particular attention to its relationship with ED overcrowding. Finally, we present current strategies developed to reduce this systemic problem. These include: (1) Clinical decision rules; (2) Computerized clinical decision support systems (CDSS); (3) Choosing Wisely initiatives; (4) Shared decision-making; (5) Audit and feedback; and (6) Multidisciplinary imaging stewardship.

Literature identification and review scope

This article was structured as a focused review with a focused synthesis of clinically relevant literature. Relevant publications were identified through targeted searches of PubMed, Web of Science, and Scopus. The searches combined terms for the setting and the modalities: EDs, advanced imaging, CT, MRI, and US. They also combined terms for the problem and the responses to it: Imaging overuse, low-value care, defensive medicine, malpractice, diagnostic uncertainty, overcrowding, clinical decision support, Choosing Wisely, artificial intelligence (AI), shared decision-making, audit and feedback, and imaging stewardship.

Searches were executed on 9 July 2026. The strategy used six concept blocks: Emergency care setting; advanced or diagnostic imaging; overuse and low-value care; drivers, covering defensive medicine and medico-legal factors; overcrowding and throughput; and mitigation or stewardship. Terms within a block were combined with OR, and blocks were combined with AND, giving four search lines. The core line paired emergency care and imaging with overuse. Three thematic lines replaced overuse with drivers, overcrowding, and mitigation in turn. Within-database union yielded 3552 records in PubMed, 2951 in Web of Science, and 4138 in Scopus. An English-language limit, plus a Humans filter in PubMed, brought these down to 2924, 2900, and 4055, respectively. The search did not remove cross-database duplicates, so the pooled figure overstates the number of distinct reports.

Two authors screened records by title and abstract against the review sections, and full texts were retrieved for those judged relevant. Priority went to systematic reviews, meta-analyses, clinical guidelines, policy statements, large observational studies, and implementation studies conducted in or directly applicable to emergency care. Reports were excluded when they addressed elective or outpatient imaging with no emergency pathway, described technical performance with no appropriateness or utilization outcome, or appeared as abstracts only. The authors also hand-searched reference lists of the retrieved articles. They then selected sources purposively for relevance and methodological contribution, rather than by an exhaustive protocol, and the cited literature is the set that survived that selection. The Supplementary material reports the verbatim query strings and the per-line record counts.

This review was not designed as a systematic review or meta-analysis, and no formal PRISMA-based search strategy, formal eligibility algorithm, study selection process, risk-of-bias assessment, or quantitative synthesis was performed. Selection was therefore purposive and open to selection bias, and the counts above describe the retrieval pool rather than a screened and de-duplicated sample.

A proposed reciprocal-pressure framework for emergency imaging overuse

Figure 1 presents the proposed conceptual framework linking diagnostic uncertainty, defensive imaging, and ED overcrowding. The framework includes four sequential stages and two feedback pathways. The first stage is the clinical input: An undifferentiated ED presentation with a low probability of serious disease but a high perceived consequence if that diagnosis is missed. The second stage is the decision environment, which includes time available for reassessment, access to observation, consultant expectations, workload, and the perceived medico-legal risk of a missed diagnosis. The third stage is the imaging-ordering decision. The fourth stage is the operational consequence of that decision, including scanner availability, patient transfer, contrast preparation, image interpretation, consultant review, and continued occupation of ED treatment space until imaging results are available.

Figure 1 Proposed reciprocal-pressure framework linking defensive advanced imaging and emergency department crowding. This figure is a line schematic and not a generated image: Its boxes, arrows, text and geometry were each specified explicitly in a drawing script, and that script was written with the assistance of a general-purpose artificial intelligence assistant. Every word printed in the figure comes from the manuscript text or from Table 2. The four rounded boxes follow a single patient episode: An undifferentiated clinical presentation, the decision environment in which the clinician works, the imaging-ordering decision, and its operational consequence for the department. Two arrow styles are used and no others. A solid navy arrow marks a documented association, that is, a direction of influence reported in the literature cited in the text. A dashed orange arrow marks the proposed feedback limb, which is conceptual only and has not been demonstrated experimentally. Arrow length and position are matters of layout and carry no quantitative meaning. Emergency department crowding is drawn as a proposed amplifying node rather than as a stage of the episode, because the limb that returns from crowding to the decision environment is the part of the loop that has not been tested directly. Taken as a whole, the closed loop is a proposed conceptual framework and not an experimentally established mechanism, and no causal or quantitative claim is made for it. The numbered markers 1 to 7 inside the “Decision environment” and “Imaging-ordering decision” boxes are the seven intervention rows of Table 2, reproduced in the wording of Table 2, and each marker is placed inside the stage that matches the workflow integration point Table 2 assigns to that intervention: 1: Validated clinical decision rules; 2: Computerized clinical decision support at order entry; 3: Audit and feedback; 4: Choosing Wisely and awareness campaigns; 5: Shared decision-making and decision aids; 6: Artificial intelligence-assisted appropriateness and triage tools; 7: Medico-legal reform and legal protection for guideline-concordant care. Table 2 gives the strength of evidence, stakeholders, monitoring metrics and potential unintended consequences for each of these seven interventions.

Two proposed feedback pathways complete the framework. In the first pathway, ED crowding reduces the time available for serial clinical assessment and may lower the threshold for imaging as a substitute for observation or reassessment. In the second pathway, increased imaging demand consumes scanner capacity, radiology workflow, and ED treatment space, which may prolong waiting times and further worsen crowding. The strength of evidence varies across these links. Published data directly support the increase in imaging utilization relative to ED attendance and the association between low-value testing and access block[2,11,13]. By contrast, the complete reciprocal cycle, in which crowding promotes defensive imaging and defensive imaging then contributes to further crowding, should be interpreted as a conceptual synthesis derived from observational and cross-sectional evidence rather than as a causal relationship proven experimentally.

Treating the problem as a loop rather than a list has one practical consequence, and it is the reason we set out the framework. Interventions aimed at a single stage are unlikely to hold. Education addresses the ordering decision, but it leaves the decision environment unchanged. Expanding scanner capacity addresses the operational consequence, yet it may lower the threshold for ordering further. Figure 1 marks the stages at which intervention is plausible, and the Mitigation strategies section grades the interventions available at each.

ADVANCED IMAGING IN EMERGENCY CARE: BENEFIT, NECESSITY, AND THRESHOLD EFFECTS

Advanced diagnostic imaging modalities have substantially evolved emergency medical care[6]. These technologies play a critical and often life-saving role in acute diagnosis, risk stratification, procedural planning, and the rapid triage of time-sensitive conditions[4,24]. For instance, CT is a primary modality for evaluating multiple trauma and enables prompt identification of occult and life-threatening injuries[7,25,26]. CT pulmonary angiography is the principal diagnostic modality for pulmonary embolism in patients in whom imaging is indicated after risk stratification. Current appropriateness criteria emphasize that correlation with clinical status and risk factors improves both diagnostic accuracy and utilization, rather than positioning it as a universal first test[27,28]. Rapid MRI protocols and perfusion CT scans facilitate timely diagnosis and inform treatment decisions in patients with acute stroke[4,29]. Furthermore, US provides immediate, radiation-free assessment and procedural guidance for acute abdominal conditions[5,12]. Therefore, advanced imaging techniques should not be equated with overuse. On the contrary, they are one of the cornerstones of contemporary, high-quality emergency medicine[30,31].

The value of advanced imaging techniques is determined primarily by the clinical context, the patient’s risk profile, the pre-test probability of disease, and the potential for subsequent interventions based on imaging results[9,16]. Since imaging itself cannot treat the patient, its utility is measured solely by its power to alter clinical management or exclude life-threatening conditions[16,21]. Therefore, the ultimate goal is not just to spot a pathology, but to effectively resolve uncertainty and change the pre-test probability of disease[9,32]. A negative imaging result can be extremely valuable if it definitively rules out a life-threatening diagnosis. It may also allow safe patient discharge and help avoid unnecessary interventions or hospitalizations[1,3].

The clinical advantages of advanced imaging are shaped by a diagnostic threshold effect[9,32]. When imaging is used above appropriate diagnostic or risk thresholds, it can offer meaningful clinical benefits that outweigh potential harms[10,32]. By contrast, its marginal value falls rapidly in patients with a very low pre-test probability of disease[32,33]. When imaging is applied to low-risk patients, incidental and clinically unimportant findings are common and may approach the frequency of the target diagnosis. In one emergency series of CT pulmonary angiograms, pulmonary embolism was identified in 13.6% of studies while incidental findings were present in 13.1%, of which 88.3% proved non-significant[34]. Incidental findings were reported in 16.4% of consecutive emergency CT and US examinations in a separate series[35]. As a result, routine imaging for complaints that are unlikely to change established treatment plans converts what could be a necessary diagnostic procedure into a low-value care practice[22,34]. Therefore, to optimize imaging in the ED, clinicians must follow clinical thresholds and limit imaging to circumstances in which the expected diagnostic benefits clearly exceed the risks of radiation exposure, overdiagnosis, and excessive resource use[10,36].

TRENDS IN ED IMAGING UTILIZATION

In recent years, the number of ED visits has climbed markedly. Concurrently, the use of advanced radiological imaging in EDs has risen substantially worldwide[2,3,13]. In several reported settings the growth rate of imaging has exceeded the rate of increase in ED attendance. For instance, Medicare data from the United States show that while ED visits rose by 8.0% between 2006 and 2014, the number of imaging procedures performed during these visits increased by 38.4%[2]. Similarly, a large study in Western Australia found that while ED patient episodes increased by 8% from 2015 to 2022, the number of CT scans performed rose by 90%[13]. Another study at a large tertiary academic medical center in Israel reported a similar divergence: Over an eight-year period, ED visits grew by 23%, whereas the on-call CT workload, measured in relative value units, increased by 52%[37]. These are single-system utilization series. None of them measured appropriateness, and the divergence they describe is therefore compatible with both changing case mix and changing ordering behavior.

Analysis of individual imaging modalities indicates that CT accounts for the largest share of the absolute increase in ED imaging, although the relative growth of MRI exceeded that of CT at one of the two centers reported[13,14]. For example, developments in acute stroke protocols and the adoption of clinical approaches such as “CT angiography (CTA) for all” have driven steep growth in head and neck CTA. In Medicare beneficiaries, ED head and neck CTA rates per 1000 ED visits rose by 14600% and 17781%, respectively, between 2001 and 2015[38], and a single-center series reported a 287% increase in head and neck CTA over eight years[37]. At one health system, CTA was performed in 91% of patients with acute ischemic stroke after a CTA-for-all protocol was adopted, compared with 61% before[39]. There has also been a pronounced rise in the use of MRI and US. Rapid MRI protocols have been developed for conditions such as transient ischemic attacks and acute abdomen in ED observation units. Together with improved accessibility and the absence of radiation, this has been accompanied by sustained growth in emergency MRI. Head MRI utilization per 1000 ED visits in Medicare beneficiaries rose by 1451% between 1994 and 2015[38], and MRI volumes per ED encounter increased by 56.3% over eight years at one level I trauma center[14]. In contrast, requests for traditional plain radiography (X-ray) have remained stable or declined proportionally, as CT and MRI have become more prevalent[14,40]. These trends indicate that ED physicians increasingly rely on advanced cross-sectional imaging modalities rather than classic diagnostic methods[41].

Imaging trends also vary markedly with patient demographics. The increase in CT usage is especially pronounced among older adults and individuals with multiple comorbidities[3,40,42]. In contrast, pediatric emergency imaging has exhibited a more complex, two-phase pattern[43]. CT use in children rose rapidly in the early 2000s. Increased awareness of radiation-induced malignancy risks then led to the “Image Gently” campaign and to pediatric clinical decision rules such as Pediatric Emergency Care Applied Research Network (PECARN), and pediatric CT use declined between 2009 and 2016[44-46]. However, recent data indicate that this downward trend has reversed. In an integrated United States health system, pediatric CT use rose again from 2017 through 2024, with an average annual change of 7.3%, alongside parallel increases in MRI of 6.7% per year and in US of 3.0% per year[46].

Some of this growth is appropriate. Older patients, more complex presentations, and wider access to time-sensitive pathways all raise the number of examinations that are genuinely indicated[3,42]. Volume data alone cannot settle whether the remainder represents overuse. None of the series above reported appropriateness, diagnostic yield, or downstream outcomes[13,14]. The divergence between imaging growth and attendance growth does establish one thing: Utilization is responding to something other than attendance. The sections that follow examine the non-clinical components of that response.

DRIVERS OF IMAGING OVERUSE

The overuse of imaging constitutes a complex, multidimensional, and multifactorial issue that cannot be linked to a single cause[9,23,47]. The disproportionate increase in imaging within EDs is not only the result of individual physicians’ lack of knowledge or irrational ordering habits. Instead, it reflects a systemic problem arising from the high-pressure work environment in which physicians operate[15,20]. The variables contributing to overuse can be categorized into two main groups. The first includes clinician-level factors, such as diagnostic uncertainty, fear of missed high-consequence disease, cognitive workload, and defensive practice. The second includes system-level factors, such as medico-legal pressure, consultant expectations, institutional risk tolerance, crowding, and workflow incentives[9,48,49]. These factors rarely act in isolation[1,19]. Overuse of imaging is rarely explained by a single factor, since these elements typically reinforce one another.

Defensive medicine and diagnostic uncertainty

Emergency medicine physicians routinely encounter patients presenting with undifferentiated symptoms and life-threatening conditions, often with limited clinical information and significant time constraints[50]. In these information-poor and high-risk clinical settings, a primary motivation is the fear of missing a diagnosis, which, although unlikely, could have serious consequences[51]. Emergency physicians themselves rank malpractice concern among the most consistently reported drivers[9,16,52,53]. Yet the distinction matters for what this evidence can support. Most of it comes from clinician surveys and vignette studies, together with cross-sectional practice audits; all measure perceived medico-legal pressure. Studies testing whether legal exposure itself changes imaging behavior are fewer and less consistent: When three states in the United States raised the liability standard for emergency care to gross negligence, CT and MRI use did not decline[54]. Perceived medico-legal pressure and demonstrated medico-legal causation are therefore not interchangeable; we treat only the former as well supported.

Increasing patient volumes in EDs, combined with time constraints and uncertainties about patients’ access to outpatient or specialist follow-up care after discharge, considerably raise physicians' mental demand[11,20]. In these circumstances, emergency medicine physicians often use advanced imaging as a “rule-out” strategy to rapidly address clinical uncertainty and reduce liability risk[15,48,51]. Additionally, the expectations of both patients and their families contribute to this process. Many patients perceive diagnostic tests as indicators of comprehensive care and may request imaging for reassurance or certainty[1,17,24]. Moreover, consulting physicians frequently require advanced imaging as a prerequisite for patient admission to a ward, creating substantial external pressure that further drives imaging overuse[55].

Defensive imaging should therefore not be dismissed simply as irrational behavior. In many cases, it reflects an understandable response to high-risk, time-sensitive decision-making and a clinical culture that places strong emphasis on avoiding missed diagnoses[33,51]. However, when this reflex replaces careful assessment, structured risk stratification, and evidence-based decision rules, it becomes a form of low-value care[56]. In such cases, imaging may expose patients to avoidable radiation, incidental findings, and downstream testing without a proportional clinical benefit[34,35,57].

Medico-legal and organizational culture

Defensive imaging is also shaped by medico-legal and organizational factors that extend beyond individual clinician judgment. In healthcare systems where adverse outcomes may lead to complaints or litigation, physicians may feel vulnerable even when they follow evidence-based recommendations[15,53,54]. The absence of clear legal protection for guideline-concordant decisions may encourage additional testing, not because it is expected to benefit the patient, but because it documents that serious diagnoses were considered and investigated[17,52].

Institutional workflows may further normalize imaging-based decision-making. Consultant physicians may request advanced imaging before accepting patients for admission or further evaluation, shifting part of the diagnostic burden back to the ED[20,51]. Low institutional risk tolerance, limited access to observation pathways, and performance metrics focused on disposition time may also favor imaging over serial clinical assessment[3,20]. When these pressures are not balanced by audit, feedback, and shared diagnostic responsibility, low-value imaging can become embedded in routine practice[18,58].

Emergency physicians may overestimate both the pre-test probability of clinically significant findings and the medico-legal risk associated with not obtaining imaging[18]. This indicates that the main gap may not be knowledge of decision rules, but calibration of diagnostic probability and perceived liability risk. Durable imaging stewardship should therefore include training in diagnostic reasoning under uncertainty, estimation of pre-test probability, and realistic assessment of medico-legal risk as core components rather than optional additions[24,50].

This analysis shows that implementing medical-legal reforms or altering malpractice standards alone is insufficient to eliminate defensive medicine[9,54]. Effective reduction of overuse requires legal reforms to be complemented by other measures. These include electronic CDSS, clearly defined institutional roadmaps endorsed by hospitals, feedback mechanisms that monitor physicians’ practices, and a safety culture in which diagnostic risk is shared between the institution and the physician[48,51].

CLINICAL SCENARIOS PRONE TO IMAGING OVERUSE

The overuse of imaging is not uniformly distributed across clinical scenarios[15,21,24,33]. This issue is especially evident in cases where the probability of a serious disease is low, yet the likely consequences are severe, symptoms are nonspecific, and clinicians experience pressure to rapidly exclude rare but critical diagnoses[50]. In these areas of high diagnostic uncertainty, advanced radiological imaging is frequently considered a solution[40]. Although such imaging may appear to provide reassurance, its use without appropriate clinical risk stratification frequently results in low-value care[10]. This pattern is especially prevalent in certain clinical situations encountered in both adult and pediatric EDs.

Table 1 summarizes common adult and pediatric clinical scenarios that are prone to advanced-imaging overuse in the ED. It also outlines the principal drivers of low-value testing, relevant decision rules or stewardship strategies, and recommended first-line imaging approaches. Each rule in the table was derived and validated in a defined population; eligibility limits appear alongside the recommendation. None of the entries is a universal instruction. A rule applied outside the population in which it was validated carries no established safety profile.

Table 1 Clinical scenarios prone to advanced imaging overuse in the emergency department, the decision rules or stewardship strategies that address them, the populations in which those rules apply, and the recommended first-line imaging approach.

Clinical scenario	Population	Principal driver of overuse	Decision rule or stewardship strategy	Eligibility and limitations	Recommended first-line imaging approach	Ref.
Minor head trauma	Adult	Fear of missed intracranial hemorrhage	Canadian CT Head Rule; NEXUS Head CT instrument	Canadian CT Head Rule: Blunt injury with witnessed loss of consciousness, amnesia or disorientation and Glasgow Coma Scale 13-15. Not derived or validated in children	Apply the rule first; non-contrast head CT in rule-positive patients. In validated low-risk patients the rule supports omitting CT	[61,62,64]
Syncope, dizziness or vertigo without focal neurological deficit	Adult	Concern for stroke or intracranial mass despite low yield	American College of Emergency Physicians clinical policy on syncope[63]; Head-Impulse, Nystagmus, Test-of-Skew examination in acute vestibular syndrome[66]	Head-Impulse, Nystagmus, Test-of-Skew applies only to an acute vestibular syndrome, that is, vertigo with nystagmus, nausea or vomiting, head-motion intolerance and unsteady gait, and requires a trained examiner. It is not a screening test for undifferentiated dizziness	Routine cranial CT or magnetic resonance imaging has low yield in patients without focal deficit in the populations studied; risk-stratify before imaging	[31,59,63,66]
Suspected pulmonary embolism	Adult	Nonspecific dyspnea or chest pain; fatal if missed	Wells score; Pulmonary Embolism Rule-out Criteria; age-adjusted D-dimer	Pulmonary Embolism Rule-out Criteria apply only when the clinician has already judged clinical probability to be low; not for moderate or high pre-test probability. Age-adjusted D-dimer, with a cutoff of age multiplied by 10 µg/L, was validated in patients aged 50 years or older with a non-high clinical probability[115]	Risk-stratify first; CT pulmonary angiography in patients with a high clinical probability or a positive D-dimer result, and avoid it in low-risk patients who are negative by the Pulmonary Embolism Rule-out Criteria	[27,28,68,69,115]
Renal colic, recurrent or known stone disease	Adult	Low threshold for CT given high accuracy	Ultrasonography-first pathway; low-dose CT when CT is required	Evidence is from a randomized trial in emergency patients with suspected nephrolithiasis; applies to a typical recurrent presentation. Initial ultrasonography does not exclude the need for subsequent CT	Ultrasonography first; reserve low-dose non-contrast CT for diagnostic uncertainty or suspected complication	[5,70]
Nonspecific, non-traumatic low back pain	Adult	Patient expectation; reassurance imaging	Red-flag criteria; Choosing Wisely	The recommendation applies to non-traumatic low back pain in the absence of red flags or pathologic indicators; the presence of any red flag, such as neurological deficit, history of malignancy or suspected infection, removes the patient from this recommendation	No emergency lumbar imaging in the absence of red flags	[71,72]
Cervical spine trauma	Adult	Reluctance to miss unstable cervical spine injury	NEXUS C-Spine criteria; Canadian C-Spine Rule	Canadian C-Spine Rule applies to alert, stable trauma patients with a Glasgow Coma Scale score of 15 and normal vital signs, and excludes age under 16 years, penetrating trauma, acute paralysis and known vertebral disease. NEXUS requires a reliable examination, which distracting injury or intoxication may preclude	Clinically clear low-risk patients with the rule; cervical CT in rule-positive patients	[47,65,75]
Minor head trauma	Pediatric	Parental anxiety; limited history from a distressed or preverbal child	Pediatric Emergency Care Applied Research Network head injury rule; as low as reasonably achievable principle and Image Gently	Derived and validated in children under 18 years with a Glasgow Coma Scale of 14-15 presenting within 24 hours of blunt head trauma; separate algorithms apply above and below 2 years of age	Observation rather than CT in the very-low-risk group; a systematic review of implementation studies reported a median 27% relative reduction in head CT (relative risk 0.73, interquartile range 0.60 to 0.89), with safety outcomes inconsistently reported across studies	[45,81,82]
Abdominal pain or suspected appendicitis	Pediatric	Radiation risk; diagnostic uncertainty	Ultrasonography-first; rapid magnetic resonance imaging where available	Ultrasonography is operator-dependent and non-visualization of the appendix is common; a non-diagnostic study does not exclude appendicitis. Rapid magnetic resonance imaging availability is limited outside specialist centers	Ultrasonography first; if it is non-diagnostic and suspicion persists, rapid magnetic resonance imaging or contrast-enhanced CT	[5,76,83,113]

NEXUS: National Emergency X-Radiography Utilization Study; CT: Computed tomography.

Common adult presentations

In adult ED settings, specific clinical scenarios have been repeatedly identified in the literature as prone to imaging overuse. In these situations, the presence of a high-risk diagnosis that clinicians are reluctant to miss often leads to an excessive tendency to order diagnostic tests.

Minor head trauma, headache, dizziness, and syncope: The predominant driver of neuroimaging overuse in these nonspecific presentations is concern about missing conditions such as hemorrhage, stroke, or intracranial mass[31,59,60]. In particular, for patients presenting with isolated dizziness without focal neurological deficits, the pooled diagnostic yield of MRI was 5.7% in a meta-analysis of emergency populations, against 13.1% for dizziness or vertigo overall[59]. Comparable yield data for syncope are not available from the sources cited here. Evidence-based decision rules, including the Canadian CT Head Rule (CCHR), the National Emergency X-Radiography Utilization Study (NEXUS) Head CT instrument, and the American College of Emergency Physicians clinical policy on syncope, define low-risk groups in whom imaging can be safely omitted[61-63]. Their effect in routine practice is less consistent: Chart audits of unselected practice report adherence to the CCHR well below the levels reached in validation studies, and a meta-analysis found that non-indicated head CT persists despite the availability of these guidelines[60,64]. Each of these instruments has explicit eligibility criteria. The CCHR was derived in adults with witnessed loss of consciousness, amnesia, or disorientation after blunt head injury and a Glasgow Coma Scale score of 13 to 15, and it was not derived in children[62]. The NEXUS Head CT instrument applies to blunt head injury. It is distinct from the NEXUS C-Spine criteria, which address cervical-spine clearance and not intracranial injury[61,65]. HINTS (Head-Impulse, Nystagmus, Test-of-Skew) is not a screening test for undifferentiated dizziness. Researchers developed it for patients with an acute vestibular syndrome, that is, vertigo with nystagmus, nausea or vomiting, head-motion intolerance and unsteady gait, and it requires an examiner trained in its performance[66].

Suspected pulmonary embolism: Pulmonary embolism, although potentially fatal if undiagnosed, often presents with nonspecific symptoms such as shortness of breath or chest pain[28]. This diagnostic uncertainty has accompanied a substantial increase in pulmonary CTA utilization[28]. However, performing CTA in low-risk patients exposes them to risks from radiation and contrast agents and has been associated with the detection of clinically unimportant emboli. After CT pulmonary angiography came into use in the United States, the reported incidence of pulmonary embolism rose by 81% while mortality changed minimally, a pattern interpreted as overdiagnosis, and complications presumed to arise from anticoagulation increased[67]. Several clinical scoring systems have been developed to address this overuse, including the Wells score, the Pulmonary Embolism Rule-out Criteria (PERC), and age-adjusted D-dimer strategies[68,69]. PERC is intended only for patients in whom the physician has already judged the clinical probability to be low, and it should not be applied to patients with a moderate or high pre-test probability[69].

Renal colic, back pain, and acute abdomen: The threshold for utilizing CT in cases of abdominal pain has progressively lowered due to its high diagnostic accuracy[4,5]. Although CT is necessary in acute abdomen and suspected appendicitis, routine non-contrast CT for recurrent renal colic results in unnecessary radiation exposure. A multicenter randomized trial enrolled ED patients with suspected nephrolithiasis. It found no difference between initial US and initial CT in high-risk missed diagnoses with complications, serious adverse events, pain scores, return visits, or hospitalizations; cumulative radiation exposure was lower in the US groups[70]. A US-first pathway is therefore a reasonable initial strategy for patients with known stone disease and a typical recurrent presentation. CT should be reserved for diagnostic uncertainty or suspected complication[5,70]. Nonspecific, non-traumatic low back pain also creates challenges. For patients without “red flag” symptoms, such as neurological deficits, cancer history, or suspected infection, emergency lumbar imaging is considered low-value and is discouraged by the Choosing Wisely initiative[71,72]. Additionally, unnecessary abdominal and pelvic CT or MRI scans frequently detect incidental findings, or “incidentalomas”, which can increase patient anxiety and initiate a cascade of further diagnostic testing[26,73,74].

Cervical spine trauma: CT imaging is often employed to exclude cervical spine injury after trauma[47]. However, ordering cervical CT scans for low-risk patients who can be safely evaluated using the NEXUS C-Spine or Canadian C-spine criteria exemplifies unnecessary imaging overuse[47,65,75]. Both instruments exclude specific groups. The Canadian C-Spine Rule applies to alert, stable trauma patients with a Glasgow Coma Scale score of 15 and normal vital signs; it excludes patients under 16 years of age, penetrating trauma, acute paralysis and known vertebral disease[75]. NEXUS is different. Its criteria require a reliable clinical examination, and distracting injury or intoxication may preclude this[65].

Pediatric emergency imaging: A distinct stewardship priority

Advanced imaging requires a different risk-benefit assessment in children than in adults. Children are more sensitive to the carcinogenic effects of ionizing radiation, and their longer remaining lifespan increases the time window for radiation-associated malignancies to develop[76,77]. Large cohort studies have linked childhood CT exposure with a higher subsequent risk of leukemia and brain tumors, although the absolute risk for an individual child remains low[78,79]. For this reason, pediatric emergency imaging represents a particularly important area for diagnostic stewardship.

Despite these concerns, advanced imaging may still be overused in pediatric EDs[80]. Common examples include head CT after minor head trauma, cervical spine imaging in low-risk trauma, whole-body CT in selected low-risk injury mechanisms, and CT for abdominal pain, suspected appendicitis, or renal colic. The drivers extend beyond fear of missed diagnosis. Parental anxiety, the limited history a distressed or preverbal child can offer, and the difficulty of arranging timely follow-up after discharge all raise the perceived value of an immediate definitive test[51].

Reducing avoidable CT in children is a priority for emergency care[81]. Deciding when to use neuroimaging for minor head injuries is particularly difficult, and clinical decision rules provide structure for that decision. The PECARN rule identifies children at very low risk of clinically important traumatic brain injury with high sensitivity in the populations in which it was derived and validated[81,82]. A systematic review examined interventions designed to reduce imaging in children with minor traumatic head injury. Strategies including the PECARN rule decreased head CT orders by a median of 27%, with a relative risk of 0.73 and an interquartile range of 0.60 to 0.89[81]. That figure is a median across non-randomized before-and-after studies rather than a pooled meta-analytic estimate, the effect varied between sites, and the review noted that safety outcomes were inconsistently reported.

Pediatric abdominal pain and suspected appendicitis provide another important example of imaging stewardship. US is the recommended initial modality in this setting because it avoids ionizing radiation, and it is the first-line approach in pediatric appropriateness guidance[76,83]. If US does not give a clear result, rapid MRI protocols, where available, are an alternative to CT[5]. When radiation cannot be avoided, such as in severe trauma, the as low as reasonably achievable principle and the Image Gently recommendations apply[45]. Low-value imaging is not confined to advanced cross-sectional imaging. Ordering chest X-rays for uncomplicated asthma or bronchiolitis, or abdominal X-rays for constipation, are practices that quality-improvement programs have targeted for reduction[84,85].

Pediatric emergency care therefore requires its own threshold rather than an adjusted adult one. The radiation risk per examination is higher, and the diagnostic yield of many tests is lower. Yet the consequences of a missed injury remain no less serious[76]. Parents should be included in these decisions by clearly explaining what is and is not known about radiation risk. Involving families in the decision has been described as a component of pediatric imaging stewardship, although the evidence that it reduces imaging volume is limited[10,51,76].

Scenarios in which withholding or deferring imaging is preferred

Stewardship is usually framed as choosing between imaging modalities. In a smaller group of situations, though, the better decision is to defer imaging altogether or to accept a less definitive test. The examination itself carries a risk that the diagnostic gain does not offset. These situations deserve separate mention: They are where the competing risk runs counter to the rest of this review.

Pregnancy is a clear example of this balance. When US or MRI can answer the clinical question, these modalities are generally preferred over CT. However, the fetal radiation dose from a single clinically indicated CT examination should be considered alongside the potential harm of delayed or missed maternal diagnosis. The same principle of weighing an avoidable dose against the harm of a delayed diagnosis underlies current pediatric imaging guidance[76]. Renal impairment represents a similar situation. Consensus guidance from the American College of Radiology and the National Kidney Foundation indicates that the risk of contrast-induced acute kidney injury with modern intravenous iodinated contrast media has often been overestimated for many patients. The same guidance emphasizes that withholding an indicated contrast-enhanced examination solely because of renal function may also cause harm[86]. In both scenarios, the appropriate approach is not automatic avoidance of CT or contrast, but a documented, patient-specific modality decision.

A third group is the patient for whom imaging is unlikely to change management, irrespective of the result. This includes repeat cross-sectional imaging in recurrent presentations with an established diagnosis and unchanged symptoms. It also includes imaging ordered after the disposition decision has effectively been made. In each of these cases, the appropriate question is not which modality to use, but whether the result will alter what happens next[16,21].

CONSEQUENCES OF LOW-VALUE IMAGING

Low-value imaging refers to radiological examinations that are unlikely to change clinical management, treatment decisions, or patient outcomes[87]. Although often perceived as harmless when results are negative, such imaging can still produce meaningful harm[15]. Its consequences extend beyond the individual patient and may affect clinicians, radiology workflow, hospital resources, and ED flow[10,17]. It is therefore important to assess the harms associated with low-value imaging across three interconnected areas: Consequences at the patient level, consequences for clinicians and the health system, and the way such imaging relates to ED overcrowding.

Patient-level consequences

The most direct patient-level risks of unnecessary imaging arise from ionizing radiation and contrast administration[1]. Although the radiation risk associated with a single CT scan in adults is generally low, cumulative exposure at the population level remains an important public health concern[88]. A recent risk-projection analysis estimated that CT examinations performed in the United States in 2023 could be associated with approximately 103000 future cancers, an estimate derived from dose modeling and risk-transfer assumptions rather than from observed cancer outcomes[88]. Iodinated contrast agents introduce additional risks, including allergic reactions and extravasation. The risk of contrast-induced acute kidney injury from modern intravenous contrast media is now considered lower than was previously assumed, and consensus guidance recommends against withholding an indicated examination on that basis alone[5,86].

Incidental findings represent another important consequence of unnecessary imaging[35,89,90]. Advanced imaging may reveal abnormalities unrelated to the patient's presenting complaint, many of which are benign or clinically insignificant[91,92]. Once detected, such findings often generate further assessment, which may create anxiety for patients and families and may trigger a diagnostic cascade involving follow-up imaging, specialist referral, invasive testing, or biopsy[74]. In some cases, this process leads to overdiagnosis and overtreatment without meaningful clinical benefit[17,74].

Clinician and system-level consequences

The overuse of advanced imaging imposes substantial strain on clinicians, radiology departments, and the wider healthcare system[93]. Emergency and radiology departments frequently operate at or near maximum capacity. Imaging requests for patients with low clinical probability consume limited scanner availability, radiologist time, and operational resources. This may delay access for patients with urgent, time-sensitive conditions. In presentations such as acute stroke or major trauma, delays in imaging may postpone definitive treatment; the extent to which low-value imaging contributes to such delays has not been quantified directly[94].

High imaging volume also increases cognitive workload for radiologists and emergency physicians. Persistent demand for image interpretation, particularly during emergency and night shifts, may contribute to fatigue, burnout, and diagnostic vulnerability[17,20]. Work overload has been associated with a higher risk of diagnostic errors in radiology[49,95]. Excessive testing may also be accompanied by reporting delays, communication failures, and delayed recognition of clinically important findings[95]. These associations come from cross-sectional and observational work, and the direction of effect between workload and error has not been established experimentally. Therefore, low-value imaging may impair system performance not only by consuming resources but also by increasing the complexity and workload of emergency diagnostic pathways.

Relationship between imaging overuse and ED overcrowding

Imaging overuse and ED overcrowding are associated, and the association is plausibly bidirectional[96]. As EDs become overloaded, clinicians have less time for detailed evaluation. In such situations, physicians may choose imaging as a faster route to making decisions and enabling discharge[51]. This may in turn add to congestion, although the net effect has not been measured directly.

The time a patient spends in the ED for imaging involves multiple bottlenecks. These include test approval, patient transfer to the radiology unit, scanner availability, renal function testing when contrast is required, establishment of intravenous access, contrast preparation, image acquisition, radiologist interpretation, and consultant physician review[10,30]. Patients often remain in the ED and occupy a treatment space until these steps are completed. When imaging is unlikely to change management, this waiting time represents avoidable resource use and may contribute to lost bed-hours[11]. Reduced bed availability then limits capacity for newly arriving patients and may further worsen crowding[11].

Low-value imaging is therefore best described as both a plausible consequence and a plausible contributor to ED overcrowding, rather than as an established cause of it. Crowding may lower the threshold for imaging by limiting time for clinical reassessment; unnecessary imaging, in turn, may prolong ED length of stay, consume radiology capacity, and delay patient flow[51,96]. Overcrowding is multifactorial. Inpatient boarding, staffing levels, inpatient bed capacity, consultant availability, and patient-flow processes all contribute substantially, and imaging is one element among these rather than the dominant one[11,96]. A single-site cross-sectional study found that low-value diagnostic testing was associated with access block. This finding is consistent with the proposed loop, but it does not establish its direction[11]. Breaking this cycle requires imaging stewardship strategies that are integrated into ED workflows and that preserve rapid access for patients with time-sensitive diagnostic needs.

MITIGATION STRATEGIES

Reducing imaging overuse requires more than asking clinicians to order fewer tests[71,72]. It represents a comprehensive system redesign challenge. The main goal is not to promote underuse of imaging, but to ensure its most appropriate and clinically valuable application[97]. High-value, time-sensitive imaging services should remain rapid and accessible in emergency situations[58,98]. In contrast, low-value tests should be restructured and reduced through evidence-based interventions that correspond with ED workflows[73,97]. Effective strategies include clinical decision rules, Choosing Wisely initiatives, computerized clinical decision support, AI-supported tools, shared decision-making, audit and feedback, and multidisciplinary imaging stewardship. These strategies do not rest on evidence of equal strength, and Table 2 grades them accordingly.

Table 2 Emergency department imaging stewardship interventions graded by the strength of the supporting evidence, with the stakeholders involved, the point of workflow integration, suggested monitoring metrics, and potential unintended consequences.

Intervention	Strength of evidence in emergency populations	Principal stakeholders	Workflow integration point	Suggested monitoring metrics	Potential unintended consequences	Ref.
Validated clinical decision rules	Strongest. Multiple prospective derivation and validation cohorts with defined eligibility criteria; implementation effect varies by site	Emergency physicians; nursing triage; radiology	At clinical assessment, before the order is placed	Rule documentation rate; rule-concordant imaging; missed clinically important injury; unscheduled return within 72 hours	Application outside the validated population; rule fatigue; documentation performed after the imaging decision	[61,62,65,68,69,75,82]
Computerized clinical decision support at order entry	Moderate to strong. Observational and implementation studies report improved appropriateness; effect depends on the guideline library and local configuration	Emergency physicians; radiology; informatics	Embedded in computerized order entry	Appropriateness score distribution; alert override rate; duplicate examinations avoided; time from order to acquisition	Alert fatigue; workflow interruption; systematic overriding; gaming of indication fields	[99,102,105]
Audit and feedback	Moderate but variable. Effects across sites are inconsistent and depend on timeliness and credibility of the data	Department leadership; quality improvement teams; individual clinicians	Periodic, outside the clinical encounter	Individual and peer ordering rates; diagnostic yield; guideline adherence; case mix adjustment	Perceived as punitive; distrust of unadjusted data; attention shifts to the measured test only	[22,48,58]
Choosing Wisely and awareness campaigns	Limited when used alone; greater when embedded in pathways and order sets	Professional societies; department leadership; patients	Local pathway and patient communication material	Uptake of targeted recommendations; rate of the specific low-value examinations named	Little effect without structural reinforcement; symbolic adoption without workflow change	[58,72,97,98]
Shared decision-making and decision aids	Limited and indirect in the emergency setting; most evidence is from non-emergency contexts or small studies	Emergency physicians; patients and families	During the clinical consultation	Documented discussion; patient-reported understanding and satisfaction; imaging rate in eligible presentations	Time cost in a crowded department; unequal uptake across language and health-literacy groups	[9,24,51]
Artificial intelligence-assisted appropriateness and triage tools	Emerging. Evidence of effect on imaging appropriateness is limited; where external validation has been performed, most algorithms lose performance on the external dataset	Radiology; informatics; emergency physicians; governance committee	Order entry, record extraction, or radiology worklist, depending on the tool	Local validation performance; drift monitoring; override rate; effect on report turnaround; equity of performance across subgroups	Automation bias; performance loss under dataset shift; opacity of the recommendation; unclear liability	[101,103,104,106,108,109]
Medico-legal reform and legal protection for guideline-concordant care	Weakest. Evaluated in few jurisdictions; the principal natural experiment found no accompanying reduction in advanced imaging	Legislators; hospital risk management; professional bodies	Outside the clinical pathway; institutional and legislative	Imaging rates before and after reform; malpractice claim frequency; clinician-reported medico-legal concern	Reform alone may not change behavior if institutional and consultant expectations are unchanged	[49,53,54]

Clinical decision rules and Choosing Wisely initiatives

Evidence-based clinical decision rules and Choosing Wisely initiatives act as essential tools to guide clinicians in reducing low-value imaging[72]. For minor head trauma in adults, the CCHR and the NEXUS Head CT instrument are applicable in appropriately selected patients, while the NEXUS C-Spine criteria are used for cervical spine trauma[61,62,65,75]. In suspected pulmonary embolism, structured strategies such as the Wells score, PERC, and age-adjusted D-dimer testing can reduce unnecessary CT pulmonary angiography in appropriately selected patients[68,69]. The PECARN rules, developed for pediatric head trauma, have demonstrated the ability to safely reduce head CT orders in the populations in which they were validated[82]. Similarly, a US-first approach for suspected renal colic in patients with known stone disease, and the application of red-flag criteria in non-specific back pain, can avoid examinations that are unlikely to change management[70-72].

However, there are significant limitations in the application of the decision rules. These decision rules should not be viewed as simple instructions to “scan” or “not scan”. Their value depends on the clinical purpose of the rule, the selected threshold, the patient’s pre-test probability, disease prevalence, and the consequences of false-positive and false-negative results[99,100]. Rule-out strategies should prioritize sensitivity and patient safety, whereas rule-in strategies should emphasize specificity and resource efficiency. For example, risk stratification tools such as the Wells score use different thresholds to guide different clinical actions rather than relying on a single universal cutoff[68]. A rule also performs differently outside its derivation population, and adherence in routine practice is frequently lower than in the validation studies[60,64]. Therefore, effective imaging stewardship requires not only adoption of decision rules, but also careful attention to how thresholds are selected, validated, reported, and implemented in local ED populations[81,99,100].

Choosing Wisely recommendations complement clinical decision rules by promoting awareness of low-value imaging among clinicians, patients, and institutions[10,58]. Their impact is likely to be greatest when they are embedded into local pathways, order-entry systems, audit and feedback programs, and patient communication tools. Used in isolation, awareness campaigns may have limited effect[98]. Used as part of a broader stewardship strategy, they can help shift institutional culture toward more appropriate imaging.

AI and digital decision support

AI and computerized CDSS are increasingly being evaluated as tools for imaging stewardship in emergency care[101,102]. These tools are often discussed as a single category, but they occupy four distinct roles, and the validation each role requires differs. Grouping them obscures which applications have demonstrated an effect on imaging appropriateness and which remain proposals.

The first role is appropriateness support at the point of order entry. Rule-based or model-based CDSS present guideline-concordant recommendations to the clinician while the order is being placed, and they flag duplicate or guideline-discordant examinations. This role carries the most direct evidence in emergency practice: A retrospective observational study of radiology clinical decision support in a quaternary center reported improved imaging appropriateness after implementation of the system[99]. The validation such systems require is local. The recommendation depends on the guideline library and order set currently in use, as well as on the case mix of the department[102].

The second role involves extracting clinical variables from free-text records. Language models can retrieve the elements a decision rule needs, such as mechanism of injury, anticoagulation status, or prior imaging, from notes that were never structured for that purpose. Researchers have evaluated an open-source model for information extraction from radiology reports[103]. The relevant validation is extraction accuracy against a reference standard. The characteristic failure is silent omission: A missing variable changes the rule output without any indication to the clinician.

The third role is prioritization of the radiology worklist. Algorithms triage acquired studies so that examinations likely to contain a time-critical finding are reported first; this may shorten time to diagnosis for those patients[8,101]. This role does not change whether an examination is ordered, so it cannot reduce overuse. Its benefit lies in throughput. Its risk is that studies sorted to the back of the queue wait longer.

The fourth role is image interpretation. Researchers have evaluated generative and discriminative models for interpretation tasks in emergency settings, including chest radiography[104]. Interpretation carries the highest validation burden of the four, because an error reaches the patient through the report itself, and prospective evidence in emergency populations remains the most limited for this role.

However, the effectiveness of these tools depends on how they are designed, validated, and integrated into clinical practice. Poorly implemented CDSS may interrupt workflow, increase alert fatigue, or be ignored by clinicians[105]. AI-based recommendations may also introduce automation bias, in which physicians accept algorithmic output without sufficient clinical scrutiny[102,106]. This is particularly important in emergency medicine, where decisions are often made under time pressure and with incomplete information.

Several additional concerns remain unresolved. Algorithms trained on limited or non-representative datasets may perform less reliably across different populations and clinical settings[107]. Strong internal performance is not evidence of reliable performance elsewhere. A model developed to detect pneumonia on chest radiographs generalized inconsistently across hospital systems: Performance at one external site matched internal performance, while performance on another site's data was inferior. The model could also identify the hospital of origin, indicating that it had partly learned site-specific acquisition characteristics rather than disease features[108]. A systematic review restricted to deep learning algorithms for radiologic diagnosis that had undergone external validation found that 81%, or 70 of 86 algorithms, performed worse on the external dataset than internally, and that 24%, or 21 of 86, showed a substantial decrease[109]. Changes in population, scanner, protocol, or documentation practice after deployment produce the same effect. Clinicians recognize this dataset shift as a known failure mode for clinical algorithms in routine use[106]. Any imaging-stewardship algorithm therefore requires local validation before deployment, along with monitoring after it.

Limited explainability may reduce clinician trust or make it difficult to understand why a specific imaging recommendation was generated. Medico-legal uncertainty may also influence adoption, especially if physicians feel pressured to follow AI recommendations to reduce liability risk, even when the recommendation does not fully match the clinical context[8,110].

Therefore, AI and CDSS should be viewed as adjuncts to, rather than replacements for, clinical judgment. Their role in imaging stewardship should be to support structured decision-making, reduce unwarranted variation, and improve workflow efficiency. They should be implemented with local validation, transparent governance, clinician feedback, and ongoing monitoring for unintended consequences[102,104].

Shared decision-making, audit, and feedback

Shared decision-making can help reduce low-value imaging when the probability of serious disease is low, the expected benefit of imaging is uncertain, and patient expectations influence clinical decisions[9,24]. In these situations, patients may view imaging as reassurance or as evidence of a complete evaluation. Clear communication about diagnostic uncertainty, expected benefits, potential harms, radiation exposure, incidental findings, and follow-up options can support more appropriate decisions[24]. Decision aids may be particularly useful when they are brief, clinically relevant, and adapted to the time-sensitive environment of the ED.

Audit and feedback address imaging behavior at the clinician and institutional levels. Reports that compare individual ordering patterns, guideline adherence, imaging yield, and peer or departmental benchmarks may encourage more appropriate test ordering[48,58]. However, feedback is most effective when it is timely, non-punitive, actionable, and linked to education or local quality improvement initiatives[58,99]. Simply reporting imaging volumes is unlikely to change behavior if clinicians do not trust the data or do not see how it applies to their clinical context[102].

Shared decision-making and audit and feedback address different points in the pathway. They are plausibly complementary: The first acts on the individual consultation, the second on aggregate ordering patterns. We are not aware of comparative evidence showing that this particular combination outperforms either component delivered alone. We therefore present them as potentially complementary elements of a broader stewardship program, rather than as a superior package[22,58].

Multidisciplinary imaging stewardship

Effective imaging stewardship in EDs depends on a multidisciplinary, collaborative approach. Shared responsibility among emergency medicine specialists, radiologists, consultant physicians, hospital administrators, IT specialists, quality improvement teams, and risk managers is essential[10,17].

System redesign must prioritize structured test order indications, integration of decision support systems into the electronic health record workflow, and standardized imaging pathways[99,102,111]. Formalizing radiology consultations in areas of diagnostic uncertainty supports integration between radiology and clinical practice[30,102]. Core elements include accelerated imaging procedures for time-critical cases, regular audit and feedback, patient-oriented communication tools, and a safety culture that emphasizes shared accountability for diagnostic risk[58,99]. The main objective of imaging stewardship is to reduce low-value imaging while maintaining timely access to fundamental, high-value tests[10,45,87].

Table 2 sets out the interventions discussed above against the strength of the evidence supporting them, the stakeholders and workflow point each one acts on, the metrics by which it can be monitored, and its plausible unintended effects. The gradation is deliberate. Validated decision rules and order-entry decision support rest on the strongest evidence in emergency populations, while audit and feedback and Choosing Wisely programs have supporting evidence that varies more across sites. Shared decision-making and AI-assisted tools remain promising, yet they are supported mainly by small or indirect studies, and medico-legal reform has been evaluated least of all. Ordering a stewardship program by that gradation, rather than adopting all components simultaneously, also makes it possible to attribute any observed change to a specific component.

LIMITATIONS

This article is a focused narrative review, and its limitations follow from that design. Literature was identified through structured searches but selected purposively rather than by a protocol-driven algorithm. There was no duplicate independent screening, no formal eligibility criteria applied to every record, no risk-of-bias assessment, and no quantitative synthesis. Selection bias cannot be excluded. A different team applying the same searches would not necessarily arrive at the same set of cited sources.

The evidence base itself constrains what can be concluded. Most studies underpinning the relationship between imaging use, crowding, clinician workload, and diagnostic error are observational, cross-sectional, or single-center. Several of the utilization series report volume without appropriateness or outcome data. Associations described in this review should therefore not be read as causal effects. The reciprocal-pressure framework in Figure 1 is an interpretive structure, not a validated model.

Context varies in ways that limit transferability. Scanner access, observation capacity, radiologist availability, out-of-hours staffing, reimbursement structures, malpractice systems, and disease prevalence differ substantially between health systems. An intervention effective in one setting may not be reproducible in another. Most of the cited implementation evidence comes from high-income systems with well-resourced radiology departments.

Finally, the competing risk deserves explicit statement. Reducing imaging carries the potential for missed or delayed diagnosis, and a delayed diagnosis can itself prolong length of stay and worsen outcomes, exactly the opposite of the intended effect. Restricting low-value examinations without simultaneously monitoring diagnostic safety is not a neutral act. Any stewardship program should therefore report missed and delayed diagnoses together with unscheduled return visits, and it should track adverse events carefully alongside imaging volume. The program should treat an unexplained fall in volume as a signal requiring review rather than as success.

FUTURE DIRECTIONS

Although initiatives to reduce imaging overuse are expanding, various key research priorities remain unaddressed in the literature. Future investigations should move beyond quantifying changes in imaging volume[22]. Prospective studies ought to focus on multidimensional clinical and systemic outcomes, including diagnostic safety, rates of missed or delayed diagnoses, ED length of stay, patient experience, cost-effectiveness, physician workload, and the environmental viability of radiology[10,99,112].

Prospective studies should systematically evaluate the impact of narrowing imaging thresholds on patient safety. Implementing a rigorous safety monitoring framework is necessary to determine whether restricting low-value tests results in missed major diagnoses or an increase in adverse clinical events[10,113].

Equity is another major priority. Existing evidence suggests that diagnostic imaging use may vary according to race and ethnicity, insurance status, and the pediatric capability of the treating hospital[80,114]. Stewardship strategies must therefore be designed carefully to avoid reducing access for patients who already face barriers to care. An appropriateness intervention that lowers imaging overall can still widen an existing gap. This happens if it is applied most consistently where clinical documentation is most complete. Stewardship programs should therefore report imaging rates and diagnostic outcomes by patient group, rather than in aggregate alone[80,114]. Future research should examine whether decision support tools, audit programs, and imaging pathways affect disadvantaged groups differently. Pediatric imaging also requires dedicated research, because children differ from adults in radiation sensitivity, disease prevalence, communication barriers, and acceptable diagnostic thresholds[51,76].

From a technological perspective, combining AI-powered tools and real-time decision support systems into electronic health records offers notable promise[101,102,104]. However, it is essential to prospectively evaluate whether these algorithms contribute to overconfidence or alert fatigue, even as they aim to reduce clinicians' cognitive workload[105,107]. Prospective evaluation should also cover external validity, since an algorithm that performs well at its development site may not transfer to other sites[108,109].

Finally, achieving sustained behavioral change may require addressing physicians' medico-legal concerns. Legal protection for clinicians who follow evidence-based decision rules has been proposed as one mechanism. It is an option that warrants formal evaluation, not a conclusion this review can support. The available evidence is limited and jurisdiction-specific. In the one substantial natural experiment, three US states raised the liability standard for emergency care, and researchers found no accompanying reduction in advanced imaging use[54]. Future legal and institutional research should examine the real effects of such medico-legal reforms on physicians' defensive test-ordering behaviors[49,53].

CONCLUSION

Advanced imaging is crucial in modern emergency care. It provides rapid diagnosis, risk stratification, and treatment planning for potentially life-threatening conditions. However, its value depends on appropriate clinical thresholds, pre-test probability, the expected impact on management, and available resources. When used outside evidence-based indications, imaging ceases to be a high-value diagnostic tool and becomes a low-value form of care. In this case, it exposes patients to unnecessary radiation, contrast-related risks, incidental findings, overdiagnosis, sequential testing, and increased costs. Imaging overuse is also associated with increasing radiologist workload, reporting delays and cognitive fatigue, consumption of scanner capacity, and worse ED congestion. The reciprocal-pressure framework proposed in this review is a conceptual synthesis of these observations rather than an experimentally established mechanism.

Reducing imaging overuse requires a balanced stewardship approach. The goal should not be to curtail imaging indiscriminately; instead, it should maintain rapid access for patients who are most likely to benefit while safely reducing examinations that offer limited clinical value. Imaging choices are often driven by defensive medicine, diagnostic uncertainty, medico-legal pressure, patient expectations, requests from consultants, and established institutional processes. What follows from the evidence reviewed here is not that every listed intervention should be adopted routinely. Effects differ across institutions, and several interventions rest on indirect or observational evidence, some of which carries unintended consequences. We recommend instead that EDs implement imaging stewardship programs that are locally validated and continuously monitored, selecting components according to the strength of the evidence behind them and the constraints of the local setting. Such programs should define explicit safety outcomes at the outset: Missed and delayed diagnoses, unscheduled return visits, ED length of stay, patient experience, and equity of access. Programs should also report these outcomes alongside imaging volume. Future work should focus on putting ideas into practice while keeping diagnostics safe, advancing equity, centering what patients value, proving cost-effectiveness, and building medico-legal structures that support imaging choices grounded in evidence.