Factors Associated with Imaging Overuse in the Emergency Department: A Systematic Review

Abstract

Background

Emergency departments (ED) are sites of prevalent imaging overuse; however, determinants that drive imaging in this setting are not well-characterized. We systematically reviewed the literature to summarize the determinants of imaging overuse in the ED.

Methods

We searched MEDLINE^® and Embase^® from January 1998 to March 2017. Studies were included if they were written in English, contained original data, pertained to a U.S. population, and identified a determinant associated with overuse of imaging in the ED.

Results

Twenty relevant studies were included. Fourteen evaluated computerized tomography (CT) scanning in patents presenting to a regional ED who were then transferred to a level 1 trauma center; incomplete transfer of data and poor image quality were the most frequently described reasons for repeat scanning. Unnecessary pre-transfer scanning or repeated scanning after transfer, in multiple studies, was highest amongst older patients, those with higher injury severity scores (ISS) and those being transferred further. Six studies explored determinants of overused imaging in the ED in varied conditions, with overuse greater in older patients and those having more comorbid diseases. Defensive imaging reportedly influenced physician behavior. Less integration of services across the health system also predisposed to overuse of imaging.

Conclusions

The literature is heterogeneous with surprisingly few studies of determinants of imaging in minor head injury or of spine imaging. Older patient age and higher ISS were the most consistently associated with ED imaging overuse. This review highlights the need for precise definitions of overuse of imaging in the ED.

Healthcare expenditures in the United States remain exceptionally high, without consistent health benefits.^1,2 The disparity between costs and outcomes of care has received much attention³ and suggests that healthcare services are overused in the US.⁴ Within the emergency department (ED), radiologic imaging is recognized as a healthcare service that is overused. Indeed, use of imaging in the ED has increased disproportionate to the number of visits in recent decades. Kocher and colleagues reported that ED visits increased nationwide by 30 percent from 1996 to 2007 while computed tomography scanning (CT) rose 330 percent.⁵ Data from the National Hospital Ambulatory Medical Care Survey further support this trend; use of CT or magnetic resonance imaging (MRI) tripling and use of ultrasound doubled between 2001 and 2010.⁶

Reduction of unnecessary imaging in the ED has been an area of significant focus recently. In 2013 and 2014, as partners in the Choosing Wisely campaign led by the American Board of Internal Medicine (ABIM) Foundation, the American College of Emergency Physicians (ACEP) released guidelines that cautioned against low-yield imaging for a number of clinical conditions.⁷ In 2015, the Society for Academic Emergency Medicine (SAEM) held a consensus conference focused on a research agenda to optimize diagnostic imaging in the ED.⁸

While there is growing appreciation that overuse of services is prevalent in our healthcare system including in the ED, the determinants of overuse are poorly characterized.⁹ In this systematic review, we focus specifically on summarizing determinants of imaging overuse in the ED. We aimed to systematically synthesize the primary literature describing factors that are positively or negatively associated with overuse of imaging in this setting.

Methods

For this review, overuse was operationally defined in accordance with the definition used by the Agency for Healthcare Research and Quality as “the provision of health care services where the likelihood of harm exceeds the likelihood of benefit.”¹⁰ Because the term overuse was seldom employed by the original study authors, we needed to deduce whether imaging overuse was being described. We determined by consensus that studies describing duplicate imaging, imaging in contradiction to established guidelines, and imaging that was determined by the authors of the included studies to be unnecessary or inappropriate with valid justification (see Table 1) were relevant to this review.

Table 1.

Study Characteristics of Included Studies

Author, year Study design		Data Site		Data Source		Years of Collection			Affected Patient Population				Overuse Event
Duplicated Scans in Transferred Patients with Trauma
Moore, 201325 Mixed - Prospective cohort and survey		Rural, academic medical center in Vermont		Medical record review and interviews of physicians caring for 39 patients			2009 –2010			Transferred patients with trauma		Duplicate CT scan within 24 hours from initial scan
Liepert, 201429 Prospective Cohort		2 academic medical centers in Utah		Prospective data collection			2009– 2010			Patients with trauma transferred to a level 1 trauma center		Duplicate CT scan by receiving hospital
Bible, 201416 Retrospective cohort		Academic medical center in Tennessee		Medical record review			2009 – 2013			Adults with trauma transferred to a level 1 trauma center		Repeat imaging defined as the same imaging modality performed on the same spine region at the receiving hospital
Cook, 201022 Cross sectional study		Academic medical center in North Carolina		Comprehensive patient registry in the hospital’s trauma program			2002 –2007			Children with blunt trauma transferred to level 1 trauma center		Repeated abdominal CT scan
Gupta, 201034 Mixed -retrospective (cross-sectional) and prospective cohort		Rural, academic medical center in New Hampshire		Trauma registry			2006 –2008			Patients with trauma transferred to a level 1 trauma center		Repeated CT imaging at receiving center
Haley, 200931 Prospective cohort		Urban, level 1 tertiary care trauma center in Phoenix, Arizona		Medical records			2006 –2006			Patients with trauma transferred to a level 1 trauma center		Repeated imaging at receiving center
Jones, 201232 Prospective cohort		Level 1 trauma center in Morgantown, West Virginia		Medical record review			2010–2010			Patients with trauma transferred to a level 1 trauma center		Repeated CT imaging at receiving center
Sung, 200926 Prospective cohort and survey		Level 1 trauma academic medical center in Boston, Massachusetts		Medical record review and survey of ordering physicians			2007–2008			Patients transferred to a level 1 trauma center		Repeat CT imaging at receiving center
Farach, 201517 Retrospective cohort		Pediatric trauma center in St. Petersburg, FL		Pediatric trauma registry			2001–2012			All transferred pediatric patients		Repeat CT of the head, chest, and/or abdomen and pelvis within 24 hours of transfer
Emick, 201118 Retrospective cohort		Two academic, level 1 trauma centers in North Carolina		State trauma registry and medical records			2004–2008			Patients with trauma transferred to a level 1 trauma center		Repeat CT imaging of chest and/or abdomen
Young, 201119 Retrospective cohort		Level 1 trauma center in Richmond, Virginia		Trauma registry and medical records			2005–2010			Adult patients with trauma transferred to a level 1 trauma center		Repeat CT scans at receiving center
Liepert, 201127 Prospective cohort and survey		Regional adult level 1 trauma center in Utah serving 4 states		Medical records and survey of ordering physician			2009–2010			Adult patients with trauma transferred to a level 1 trauma center		Repeat CT scans at receiving center
Lee, 200828 Survey		Level 1 trauma center at Lexington, Kentucky		Survey of referring physicians			2005–2006			Patients with trauma transferred to a level 1 trauma center		Expectation of ordering imaging in a patient that will be transferred
Mohan, 201020 Retrospective cohort		Level 1 trauma center in Western Pennsylvania serving 5 states		Trauma registry			2000–2007			Patients with trauma transferred to a level 1 trauma center		Pretransfer CT scan in patients who will be immediately transferred due to trauma severity
Unnecessary, Inappropriate or Defensive Imaging in Non-Transferred Patients in the Emergency Department
Parma, 201421 Cross sectional	Community teaching hospital in Pennsylvania		Trauma registry		2010 –2012			Adult with trauma and Glasgow Coma Score of 15			CT in patients with no visible head injury, loss of consciousness, amnesia, or neurologic complaint
Benarroch-Gampel, 201133 Mixed - retrospective and prospective cohort	Academic medical center in Texas		Medical record review		2005 –2010			Patients with acute gallbladder disease			Abdominal CT in evaluation of acute gallbladder disease
Waxman, 200030 Prospective Cohort	Community hospital in California		Direct observation of patients by research assistants		1997 –1998			Patients with non-traumatic abdominal or chest pain who are non-English speakers and have a language-discordant doctor			Abdominal CT and abdominal radiographs
Chen, 201523 Mixed -Prospective cohort and Survey	One city in Utah		In-person survey of physicians		2012 –2012			Trauma system activation patients presenting at a level 1 trauma center			Self-reported defensive scanning: physicians were asked, in a litigation free environment, which scans would and would not be ordered?
Kanzaria, 201524 Cross sectional - survey only	Medical conferences attended by academic and community-based clinicians		In-person survey of 435 physicians		2013 –2013			Patients of survey respondents (emergency physicians)			Self-report: perception by survey respondents that CT or MRI ordered by themselves and their colleagues was unnecessary
Melnick, 201535 Qualitative	Academic level 1 trauma center in New Haven, Connecticut		Focus groups, cognitive task analysis (direct observation), critical decision method (interviews)		2013–2014			Adult patients presenting to the ED with diagnosis of concussion, head injury or minor traumatic brain injury			Head CT for minor head injury in the ED

CT=computed tomography imaging, MRI=magnetic resonance imaging

Data Sources and Searches

We began with a scoping review by searching MEDLINE^® and Embase^® from January 1998 through July 2016. Our initial search broadly included terms reflecting use and overuse of healthcare services, including procedures and diagnostic tests. We searched using the medical subject heading terms and keywords related to the overuse of healthcare services: “medical overuse” OR “health services misuse” OR health services overutilization OR “unnecessary procedures” OR medically unnecessary procedures OR Diagnostic Tests, Routine/utilization OR Defensive Medicine OR Practice Patterns OR Health Services Abuse OR Health Services Overuse OR medical overutilization OR inappropriate utilization. We followed with a targeted search through March 1, 2017 with specific terms for articles addressing overuse of imaging in the ED: “diagnostic imaging” OR radiography OR tomography OR scanning OR scans OR scan OR imaging OR “magnetic resonance” OR “diagnostic testing” OR “diagnostic evaluation.” We hand searched the reference lists of each included article as well as related systematic reviews for additional articles. Searches were limited to human studies in the English language. Our protocol followed PRISMA guidelines and was registered in Prospero (#CRD42015029482) as part of a broader review examining overuse of several types of healthcare services.

Study Selection

Two reviewers independently screened titles, abstracts, and full-text for inclusion. Differences between reviewers were resolved through consensus adjudication. Studies were included if they were written in English, contained original data, pertained to a U.S. population, and identified a determinant associated with overuse of imaging in this setting. Studies describing only the prevalence of overuse, and not its determinants, were excluded. We further restricted the study to data collected after 1996, given the substantial changes in the U.S. healthcare system in the past two decades regarding availability of imaging technologies. We had no restrictions regarding study design.

Data Extraction, Quality, and Applicability Assessment

Using standardized forms, reviewers extracted information on the general study characteristics, study participant characteristics, the methods of data collection, the overuse event under investigation, the determinants evaluated by the investigators and the determinants identified as being significantly associated with the overuse event. The determinants were classified as being related to the patient, the clinician, or the environment including the region and health system. A second reviewer confirmed the accuracy of the abstracted data. We used the criteria for determining statistical significance as had been defined by each article. For studies that analyzed significance with both bivariate and multivariate methods, we extracted only the significant results from the multivariate analyses.

Two reviewers independently assessed the risk of bias in included articles using previously validated instruments. The Critical Appraisal Checklist (from the Center for Evidence Based Management) was used for cohort studies and surveys.¹¹ The single qualitative study was assessed using the Checklist for Qualitative Research from the Joanna Briggs Institute.¹²

Data Synthesis and Analysis

We created detailed evidence tables. We synthesized the results by the type of imaging and then by the determinants, organized as patient-level, clinician-level, and organizational or environmental. We created summary tables of these results. The results were not amenable to quantitative pooling given the heterogeneity in design across studies.

Role of the Funding Source

The funders had no role in this project.

RESULTS

We identified 10,859 titles meeting our inclusion criteria. Of these, 484 articles proceeded to full-text review. (Appendix Figure) We identified 20 studies meeting our inclusion criteria that examined determinants of overuse of imaging in EDs.

Characteristics of included studies

The included studies were five retrospective cohort studies,^13–17 two cross-sectional studies,^18,19 six surveys,^20–25 four prospective cohort studies,^26–29 two studies having both retrospective and prospective cohort components,^30,31 and one qualitative study.³²

Determinants were evaluated for their independent contribution to overuse with multivariate regression methods in seven of the studies.^{15–17,27–30} Five studies reported only bivariate analyses,^{13,14,18,25,26} and seven were entirely descriptive in their presentation of results.^19–24,31

Because the majority of identified studies focused on duplicate imaging in transferred trauma patients, we present the results as: 1) determinants of duplicate imaging in patients with trauma transferred to a trauma center, and 2) determinants of unnecessary, inappropriate or defensive imaging in the ED among diverse patients who were not transferred.

Risk of Bias

This body of literature was not of high quality. The risk of bias was determined to be moderate in 13 studies^{13–16,18,20,21,23–25,27,28,30} and high in four studies.^19,22,26,31 Only two studies^17,29 were considered to have a low risk of bias. Prominent flaws included the lack of reporting of response rates in surveys and the lack of use of validated tools for data collection. Most of the studies did not clearly describe the characteristics of the study participants at enrollment. The quality of the single qualitative study was good.³²

Determinants of Duplicated Scans in Transferred Patients with Trauma

Twelve studies described scans duplicated in patients arriving at a level 1 trauma center after transfer.^{13–16,19,22–24,26,28,29,31} Two studies described acquisition of scans at the referring facility prior to emergent transfer.^17,25

Eight out of 12 studies of duplicate scans upon arrival probed the reasons for ordering the duplicated scan; this was learned either by survey,²² by medical record review^{13,19,26,29,31} or by medical record review with adjunctive physician report.^23,24(Table 2) Two reasons were noted in all eight studies: 1) the scan was not received by the trauma center or there was inadequate transfer of data; and 2) the quality of the transferred image was inadequate for clinical care.^{13,19,22–24,26,29,31} Other reasons for duplication of imaging included physician preference for repeat imaging,^24,26,31 the belief that repeating a scan in a transferred patient was routine care,¹⁹ and a consultant’s request or convenience.²⁹

Table 2.

Determinants of Duplicate Scans in Transferred Patients with Trauma^*

Author, Year N, Overuse Events	Patient Determinants	Clinician Determinants	System and Environmental Determinants
Liepert, 201426 480 trauma patients w/144 repeated CT	NE	NE	Non-integrated health system (48% of patients) vs integrated health system (16% of patients) p = 0.0001
Bible, 201413 1427 trauma patients with spine injuries w/194 repeated CT, radiograph or MRI	Medically stable or unstable	NE	Unsent CT (50%) vs. sent (23%) CT from outside hospital p = 0.0001
Haley, 200928 410 trauma patients w/218 repeated CT	Higher ISS (6% increase odds of overuse for each point, p=0.002), age, injury mechanism, length of stay at transferring facility	NE	NE
Jones, 201229 211 trauma patients w/82 repeated CT	Higher ISS (for every unit increase in ISS, odds increase 3.6% [p = 0.0361]), injury mechanism, insurance type, sex, age	NE	NE
Farach, 201514 8658 pediatric patients w/288 repeated CT	Younger age (more overuse-repeated CT scan vs no repeated CT scan [Mean: 6.3 vs 7.2 years, p<0.01]), higher ISS (more overuse - repeated CT scan vs no repeated CT scan [Mean: 12.2 vs. 8.7, p<0.01]), GCS 13 or 14 (less overuse – repeated CT scan vs no repeated CT scan [Mean: 13.1 vs. 14.1, p<0.01), motorized trauma/penetrating trauma/assault vs fall/nonmotorized trauma/sports related injury [p<0.05], sex	NE	Greater hospital distance (more overuse[repeated CT scan vs no repeated CT scan [Mean: 69vs. 49.2 miles, p<0.01])
Emick, 201115 1375 adult patients w/820 repeated CT	ISS between 24–33 relative to ISS<15 (more overuse[OR, 1.6; 95% CI, 1.05–2.4]), triage alert criteria 1, 2, relative to 3 (more overuse[OR, 1.6; 95% CI, 1.01–2.7 for level II; OR, 2.2; 95% CI, 1.2–4.1 for level I]), age, arrival SBP, arrival GCS, mechanism of injury (motor vehicle collision or other), insurance status	NE	Helicopter versus ground transport (more overuse[OR, 1.6; 95% CI, 1.2–2.2]), referring center volume, day of transfer, year of admission
Young, 201116 2678 adult patients w/559 repeated CT	Higher ISS (for every point increase, OR, 1.02; 95% CI, 1.01–1.03; P = 0.0001]), older age (for every year increase, OR 1.006; CI, 1.003 –1.010; p = 0.0004]), race, injury mechanism, sex	NE	NE
Lee, 200825 Survey of 143 physicians	NE	ATLS certification	NE
Mohan, 201017 N =7713 4434 patients w/3303 having repeated CT	Female (more overuse [OR, 1.18; 95% CI, 1.05 –1.33]), older age (more overuse [(OR, 1.30; 95% CI, 1.07–1.58 if age 65–80 yrs; OR, 2.19; 95% CI, 1.79 –2.77 if age >80]), obesity (less overuse [(OR, 0.56; 95% CI, 0.41– 0.78]), Medicare (more overuse [OR, 1.44; 95% CI, 1.18 –1.74])) and uninsured (less overuse [OR, 0.66; 95% CI, 0.57–0.76])) relative to commercially-insured, motorcycle collision (less overuse [OR 0.81, 95% CI 0.68–0.97] or penetrating injury (less overuse [OR, 0.16; 95% CI, 0.11–0.20])) relative to motor vehicle accident, lower ISS (more overuse [OR, 1.66; 95%CI, 1.43–1.91 if ISS 16–23; OR, 1.59; 95% CI, 1.32–1.92 if ISS 24–33	ATLS certification	Greater distance from hospital (more overuse), volume of trauma patients

^*Statistically significant determinants of overuse are in bold, the direction of effect is in parentheses

ATLS=Advanced Trauma Life Support, CT=computed tomography scan, GCS= Glasgow Coma Score, ISS=Injury Severity Score, MRI=magnetic resonance imaging, NE=Not Examined, PGY=postgraduate year, SBP= systolic blood pressure

Patient factors

Six of the 12 studies of duplicate imaging upon arrival analyzed patient factors associated with receipt of a repeated scan.^{13–16,28,29} One of the two studies of inappropriate imaging prior to transfer also examined patient factors.¹⁷ All seven of these were cohort studies.

Six studies tested the association with age, with half finding significant associations: two reported that imaging was more likely to be performed on middle age or older patients, with the mean ages of 42.3 years¹⁶ and over 65 years.¹⁷ One reported higher rates of repeat imaging in the youngest children (mean age 6.3 years).¹⁴ Four studies tested whether repeat imaging depended on the patient’s sex.^14,16,17,29 Three did not find significant associations; the one that did reported that inappropriate imaging prior to transfer was more likely to occur in women than men.¹⁷ Race was only analyzed in a single study and was not influential.¹⁶

Six studies tested associations with Injury Severity Score (ISS) with all finding that imaging was associated with higher ISS scores.^{14–17,28,29} One cohort study looked at medically stable versus unstable transferred patients with spine injuries.¹³ Medically unstable patients were less likely to have their imaging sent and viewable to the receiving institution, and yet were not significantly more likely to have repeat imaging. The cohort study by Mohan et al. also reported that medically unstable patients were less likely to receive imaging prior to transfer than their stable counterparts.¹⁷ Mechanism of injury was a significant determinant of repeat imaging in two^14,17 of the six studies that evaluated this.^{14–17,28,29}

Insurance status was a significant factor in just one¹⁷ of three studies.^15,17,29 In multivariate analysis of 7,713 adults presenting to a non-trauma center with indications for immediate transfer to a trauma center, patients with Medicare received more duplicate CT scans while uninsured patients had fewer duplicate CT scans, when compared to patients with commercial insurance.¹⁷ Emick et al. did not find a significant association between repeated scanning in 1,375 trauma patients who had received a scan prior to transfer and their insurance status.¹⁵ Jones et al. also did not see a significant association between insurance and repeated CT scans in a multivariate analysis involving 211 patients.²⁹

Physician factors

The only physician-level factor, Advanced Trauma Life Support certification, was examined in a cohort study¹⁷ and a physician survey.²⁵ Both studies found that physicians with Advanced Trauma Life Support certification were neither more nor less likely to order imaging than physicians without.^17,25

Environmental and systemic factors

Five cohort studies analyzed the environmental or system level factors contributing to repeat imaging.^{13–15,17,26} Two studies analyzed the impact of distance from the receiving trauma center; a greater distance was associated with CT scanning upon arrival¹⁴ and with scanning prior to transfer.¹⁷ The latter study also evaluated the impact of the volume of trauma patients at the referring institution and found this was not significantly associated with scanning prior to transfer.¹⁷

The cohort study by Emick et al. found that transport by helicopter was associated with greater likelihood of scanning upon arrival compared to transport by ground, after thorough adjustment for variables characterizing the patient and the facilities.¹⁵

Liepert et al. compared rates of repeated CT scanning after transfer in an integrated health system (where the referring hospital and trauma hospital are part of the same system with fully integrated medical records and radiology picture archiving) versus a more conventional health system.²⁶ The integrated health system had a significantly lower proportion of repeat imaging. In another study, transferred patients with spine injury were significantly more likely to have a CT upon arrival if the original CT scan was not viewable at the receiving institution.¹³

Determinants of Unnecessary, Inappropriate or Defensive Imaging in Non-Transferred Patients in the Emergency Department

Six studies evaluated determinants of overused imaging in the ED with overuse defined as unnecessary, inappropriate or defensive imaging.^{18,20,21,27,30,32}

Patient Factors

One cross-sectional study¹⁸, two cohort studies,^27,30 and one qualitative study³² evaluated patient factors contributing to unnecessary or inappropriate imaging in the ED. (Table 3)

Table 3.

Determinants of Unnecessary, Inappropriate or Defensive Imaging in Non-Transferred Patients in the Emergency Department^*

Author, Year Unnecessary imaging event: Count (when available)	Patient Determinants	Clinician Determinants	System and Environmental Determinants
Parma, 201418 Head CT for minor trauma in absence of head injury, loss of consciousness, amnesia, or neurologic complaint: 106 of 438 adult patients	More CT overuse among patients 41 to 64 years, or with drug use, or vehicular injury, or surgery within 24 hours (P<0.05) –only univariate analyses	NE	NE
Benarroch-Gampel, 201130 Abdominal CT for acute gallbladder disease: 234 of 562 patients	Older age, per 5-year increase (OR, 1.14; 95% CI 1.07 –1.21); elevated white blood cell count (OR, 1.67; 95% CI, 1.10–2.53); elevated amylase (OR, 2.02; 95% CI, 1.16–3.51); hypertension (OR, 2.01; 95% CI, 1.20–3.37) sex, race, previous ED visit, diagnosis (acute cholecystitis v. gallstone pancreatitis v. common bile duct stones)	Imaging in the evening (relative to daytime) (OR, 4.44; 95% CI, 2.88–6.85), weekend imaging, admitting service	Year of event from 2005–2009
Waxman, 200027 N =324 172 English speaking and 152 Non-English speaking patients with nontraumatic chest pain or abdominal pain	Non-English language relative to English speakers (more overuse of abdominal CT for pain – mean difference in proportions 10.9%, 95% CI 1.0% to 20.8%)	NE	NE
Kanzaria, 201521 CT or MRIs N=435	NE	Main perceived contributor of overuse: fear of missing a low-probability diagnosis and avoidance of malpractice litigation; other contributory factors: patient or family expectations, standard practice in medical group, standard practice in ED, the test saves time, administrative pressure to increase reimbursement, increase in personal reimbursement Most frequent “write-in comment” was request of non-ED physician to perform testing.	Greater perception of self-ordering of medically unnecessary imaging in non-California vs. California (26% vs 20% overuse), non-group based HMO vs. group-based HMO (25% vs. 18%)
Chen, 201520 1,097 CT scans were performed on 295 trauma activation patients Litigation vs. litigation free environment	NE	NE	Litigious environment (more overuse) relative to practice influenced only by clinical judgment
Melnick, 201532 focus groups of patients (four groups, 22 subjects total) and providers (three groups, 22 subjects total) Inappropriate head CT in mild head injury	Patient expectations (expects a CT), anxiety, patient engagement	Establishing trust/bedside manner (patient engagement, reassurance, listening and caring for patient as a person, identifying and addressing concerns, provider confidence/experience, ability to identify and manage patient anxiety and tolerance for uncertainty), anxiety (risk aversion), constraints (time, concussion knowledge gap), influence of other providers (PCP, consultant, resident, mid-level)

^*Statistically significant determinants of overuse are in bold the direction of effect is in parentheses

CT=computed tomography scan, ED = emergency department, ISS=Injury Severity Score, PCP = primary care physician

The cross-sectional study, of patients with mild head injury, reported only unadjusted results, finding that middle aged patients (41–64 years) were more likely to receive unnecessary head CT compared to younger (18–40 years) and older (65+ years) patients.¹⁸ Additionally, receipt of unnecessary head CT was associated with patient’s illicit drug use, vehicular injury as compared to fall or assault or other injury mechanism, and an ISS >26 compared to ISS between 15–25. Neither patients’ sex nor alcohol use were significantly associated with head CT overuse.

In a multivariate analysis of patients presenting to the ED with complicated gallstone disease, wherein authors considered ultrasonography to be the appropriate diagnostic test, older age predicted inappropriate use of abdominal CT.³⁰ Other significant associations in the multivariate model were the presence of hypertension, elevated white blood count, and abnormal amylase on laboratory analyses.³⁰ Patient sex, race, previous ED visit, and other laboratory values and comorbidities were not significant predictors of receipt of abdominal CT.³⁰

A prospective cohort study reported that non-English speaking patients experienced a higher rate of abdominal CT for abdominal pain than did English language speakers. This relationship was not found for other diagnostic imaging tests, specifically abdominal radiographs (plain film) for abdominal pain, chest radiographs or echocardiography for chest pain.²⁷

A qualitative study by Melnick et al. used patient and physician focus groups, cognitive task analysis after direct observation in the ED, and interviews with four senior physicians to identify themes related to inappropriate receipt of CT for a minor head injury. Patient-level themes that were identified included patients’ expectations of receiving of CT, patients’ lack of tolerance for ambiguity (as perceived by the physician), and patients’ balance of short and long term risks.³²

Clinician Factors

One cohort study,³⁰ one cross sectional²¹ and one qualitative study³² evaluated clinician factors associated with unnecessary or inappropriate imaging in the ED. (Table 3) The cohort study, described above, of patients with complicated gallstone disease, found that the time of placement of the order for imaging by clinicians was the single largest predictor of inappropriate CT use, with inappropriate CT orders much more likely in the evening than daytime; weekend or weekday ordering of imaging did not predict inappropriate CT use.³⁰ Similarly, the admitting service to which the patient was directed did not drive CT use in this situation.

In the cross-sectional survey of ED providers’ perceptions of inappropriate imaging in their department, most respondents reported that they believed the main contributor to overuse was fear of missing a low-probability diagnosis and avoidance of malpractice issues.²¹ The less strongly endorsed responses were: perception of patient or family expectations, imaging as standard practice in their medical groups or among their closest colleagues, imaging as standard practice in emergency medicine, that the test saves time, administrative pressures to increase group reimbursement, and increased personal reimbursement. Respondents who chose to write in a reason for overuse mentioned that imaging was often done at the request of a non-ED physician.²¹

Multiple clinician-level themes emerged from the qualitative study by Melnick et al. on inappropriate head CT for minor head injury.³² The most frequently reported themes were about establishing trust as a way to minimize inappropriate head CT use: this included attending to patient engagement, providing reassurance, caring for the patient as a person, identifying and addressing patient concerns, identifying and managing patient anxiety and tolerance for uncertainty. Other factors included clinicians’ confidence and experience, their anxiety and their tendencies toward risk aversion.

Systemic and Environmental Factors

The cohort study of gallstone evaluation³⁰ and two surveys^20,21 evaluated health system or environmental factors contributing to unnecessary or inappropriate imaging in the ED. The cohort study, by Benarroch-Gampel, noted the absence of a trend in inappropriate CT use from 2005–2009 in evaluation of acute cholecystitis.³⁰ Kanzaria et al. reported from their survey of ED providers that more respondents outside of California endorsed ordering medically unnecessary CTs or magnetic resonance images (MRIs) than respondents in California and more in non-group-based HMOs than in group-based HMOs.²¹ When attending trauma surgeons were surveyed about whether they would practice similarly to their current practice in a litigation-free environment, they responded that the litigious environment in which they practiced was a major driver of their imaging orders.²⁰

DISCUSSION

In this systematic review, we synthesized the primary literature describing determinants that are positively or negatively associated with overuse of imaging in the ED. Fourteen (70%) of the included studies addressed overuse of CT scans in trauma patients presenting to a local ED with subsequent transfer to a level 1 trauma center. The authors of the primary studies considered this practice to be overuse because it generates little new information, harms patients with radiation or discomfort, delays treatment and increases the costs of care. Determinants associated with receipt of a duplicate scan include those representing clinical instability—patients at the extremes of age and/or more severely injured (Table 4). One might argue that these are not all episodes of overuse in these clinically unstable patients; but, for this review, we used the definition of overuse provided in each of the included articles.

Table 4.

Summary of Determinants Associated with Duplicate Scans in Transferred Patients with Trauma

Factors (Number of Studies)	Determinants Examined	Studies Reporting Significance/Studies Evaluating the Association	Key Findings
Patient Factor (6 Studies)	Age	3/6	Younger age was associated in 1 study and older age in 2 studies.
	Sex	1/4	Female sex was associated in 1 study.
	ISS score	6/6	Higher ISS scores was associated in 6 studies.
	Insurance status	1/3	Insurance status was associated in 1 study.
	Mechanism of injury	2/6	Mechanism of injury was associated in 2 studies.
System and Environmental Determinants (5 Studies)	Health System	1/1	Non-integrated health system was associated in 1 study compared integrated health system.
	Scan sent outside	1/1	Whether the outside hospital had sent the scan to the receiving hospital, (Unsent CT) was associated in 1 study.
	Distance	2/2	A greater distance was associated in 2 studies.
	Transport mode	1/1	Transport by helicopter was associated in 1 study compared to transport by ground.

Other findings suggest that there may be systematic biases regarding who gets more imaging, such as determined by insurance status. Others identified determinants included environmental factors like distance between hospitals and whether the referring and receiving hospitals were part of an integrated system where there is more seamless transfer of records and imaging files.

The remaining six studies were more varied in the types of imaging that were studied. Indeed, while three studies used guideline-determined definitions of non-indicated CT scans for mild head injuries and acute gallbladder disease, the remaining three used vaguer definitions like physicians’ self-perception about unnecessary imaging or testing. Like in the repeated testing studies, overuse was greater in patients who were older or who had more comorbid disease.

Other determinants, like time of day, suggests that overuse is related to staffing deficiencies (such as lack of ultrasonography technicians at night) or crowding of the ED, but this was not explicitly examined. This is supported by the qualitative study, which suggested that overuse of head CT would be reduced by engaging patients and managing patient expectations and anxiety—practices that necessitate greater time at the bedside. Defensive medicine as a driver of overuse, including imaging, has been previously described and was noted by physicians as influencing their behavior. Finally, as with the repeated CT scan studies, less health system-level integration also predisposed to imaging overuse in the ED (Table 5).

Table 5.

Summary of Determinants Associated with Unnecessary, Inappropriate or Defensive Imaging in Non-Transferred Patients in the Emergency Department

Categories (Number of Studies)	Determinants Examined	Studies Reporting Significance/Studies Evaluating the Association	Key Findings
Patient Factors (4 Studies)	Age	2/2	Middle age was associated in 1 study and older age was in other study.
	Drug use	1/1	Drug use was associated in 1 study.
	Mechanism of injury	1/1	Mechanism of injury was associated in 1 study.
	laboratory values and comorbidities	1/1	Presence of hypertension, elevated white blood count, and abnormal amylase on laboratory analyses were associated in 1 study.
	Language	1/1	Language barrier was associated in 1 study.
	ISS score	1/1	Higher ISS scores was associated in 1 study.
	Patient-level themes	1/1	Patient expectations (expects a CT), anxiety, patient engagement were associated in 1 study.
Clinician Factors 3 studies	Time of imaging	1/1	Imaging in the evening was associated in 1 study.
	Factors believed to contribute most to unnecessary imaging	1/1	Main contributor was fear of missing a low-probability diagnosis and avoidance of potential malpractice suits. Frequent “write-in comment” was request of non-ED physician to perform testing was associated in 1 study.
	Clinician-level themes	1/1	Establishing trust/bedside manner was associated in 1 study.
System and Environmental (3 studies)	Practice location, practice type, reimbursement method	1/1	Self-ordering of medically unnecessary CT/MRIs in non-California states than in California and in non-group based HMOs than in group-based HMO were associated in 1 study.
	Litigious environment	1/1	Current litigious environment was perceived to be a major driver of imaging orders in 1 study.

CT=computed tomography scan, ED = emergency department, ISS=Injury Severity Score, PCP = primary care physician

To our knowledge, there are no other systematic reviews on this topic. There are many studies that describe variation in use of imaging without probing the determinants of this variation and without exploring whether the variation represents overuse or underuse. Levine and colleagues studied ED physicians at a single level 1 trauma center and found rates of ordering of imaging varied across physicians three-fold.³³ Marin and colleagues, using the Nationwide Emergency Department Sample, found 40 percent variation in CT use for children with head trauma across EDs, with differences between academic and non-academic medical centers.³⁴ Patient race and insurance type influence the diagnostic workup received by patients in the ED, but it is unclear, again, whether this represents underuse or overuse of services.^35,36 Other studies have also demonstrated more intensive health service utilization, besides imaging, in the ED for patients who speak languages other than English, possibly in an effort to overcome diagnostic challenges posed by communication difficulties.^37,38

Limitations and strengths

Our review has several limitations. First, a limitation is the varying definitions of overuse across the studies. Second, some may argue that repeating scans in transferred patients is frequently appropriate (and not overuse) due to a clinical status change and this may not have been easily captured in the studies that used administrative data. Third, the studies examining utilization in the ED (non-transferred patients) used a variety of definitions and yet many of the determinants were consistent across studies. Fourth, our search strategy focused on terms related to “overuse” and may have missed articles focused on guideline-discordant care, although additional hand searching did not identify other relevant articles. Fifth, we did not search for gray literature so publication bias is a possibility; that is, analyses of wasteful services may be conducted within health systems and not published, such as when there are null findings or findings that reflect poorly on the health system. However, we expect that our exclusion of unpublished literature would not importantly change our conclusions given that our synthesis was entirely qualitative and missing a single study would not change point estimates and confidence intervals as it might in a quantitative synthesis of study results.

We note that the quality of the included studies was only fair; only two of the included studies were deemed to have a low risk of bias. Additionally, the response rates in the surveys were often not reported and might be presumed to have been low. It is hard to know in which direction this biased the results. We recognize that many of the studies described factors associated with overuse that may not in fact be causal determinants; the methods used in the primary studies methods do not permit definitive comments about causality.

Study strengths include our use of a second reviewer, which minimizes the likelihood of incorrect extraction of data about determinants. Additionally, our inclusion of surveys in the review allowed for the association between determinants and perceptions of determinants of overuse to be included.

Future research

As high resolution radiologic imaging becomes more widespread in hospitals, there will be an even greater need for understanding its value to patients. The ABIM, Academy Health, the Robert Wood Johnson Foundation and the Agency for Healthcare Research and Quality have all recognized the harmful impact of overuse on patient health and health systems. A research agenda set forth by Morgan, et al, emphasized gaps in our understanding of determinants, including their relative importance, interaction and the potential value of changing any one determinant in isolation.³⁹ These authors wonder in particular about the role of communication, clinical uncertainty and cognitive biases as contributors to overuse. Additional research is needed that explores a wider breadth of determinants and their modifiers, preferably using a consistent definition of overuse. Studies that can quantify the impact of determinants on overuse will further enable evidence-based policies to be implemented and evaluated.

Several of the common determinants of repeated imaging—scans not sent with the patient, incomplete transfer of data, poor quality of scans, lack of health system integration—can be improved with standardization of protocols and improved technological interoperability between referring and receiving sites. As health information exchanges (HIE) gain functionality,^40,41 there should be more opportunity to avoid duplication of scans ordered because of incomplete transfer of data.

Targeting physician ordering behaviors has been a core strategy in multiple interventions designed to reduce inappropriate imaging in hospitals. Decision support systems often use existing infrastructure of the electronic medical record; ordering physicians may receive a “low-value” message if they order a test that may be considered inappropriate based on the patient’s characteristics.⁴² Multifaceted approaches are likely to be more impactful than reliance on messaging alone; the literature strongly supports “alert-fatigue” that lessens the impact of many messaging systems.^43,44

Promisingly, overuse of imaging in the ED has been addressed by several societies contributing recommendations to the Choosing Wisely Initiative. This national campaign, initiated by the ABIM Foundation and supported by the American College of Physicians, promotes conversations between clinicians and patients to help them choose care that is “supported by evidence, not duplicative of other tests or procedures already received, free from harm and truly necessary”.⁴⁵ The ACEP cautions to avoid CT scans of the head in patients with minor head injury not meeting criteria for imaging by validated decision rules, and to avoid CT of the abdomen and pelvis in young patients in the ED with known kidney stones and symptoms of renal colic. The College recommends avoidance of lumbar spine imaging in the ED for adults with non-traumatic back pain and no red-flag symptoms, avoidance of CT pulmonary angiography in patients with a low likelihood of pulmonary embolism, and avoidance of CT of the head in asymptomatic adult patients with syncope or minor trauma and a normal neurological evaluation.

CONCLUSION

Overuse of ED imaging is common yet there is little consistency in the determinants described across studies of this topic. The wide range of determinants identified in these studies represent avenues for future research and then intervention. This review highlights the need for more precise definitions of overuse of imaging tests in the ED, and more thorough investigation of factors driving overuse in this setting.

Appendix Figure 1. Summary of the literature search.

*Reviewers did not need to agree on reason for exclusion