American Journal of Emergency Medicine 31 (2013) 1437–1442

Contents lists available at ScienceDirect

American Journal of Emergency Medicine

j ourna l homepage: www.e lsev ie r .com/ locate /a jem

Original Contribution

Variation in use of all types of computed tomography by emergency physicians☆

Matthew B. Levine, BA a, Andrew B. Moore, MS a, Chris Franck, PhD c,d, Jie Li, PhD d, Damon R. Kuehl, MD b,⁎a Virginia Tech Carilion School of Medicine, Roanoke, VAb Department of Emergency Medicine, Carilion Clinic and Virginia Tech Carilion School of Medicine, Roanoke, VAc Laboratory for Interdisciplinary Statistical Analysis, Virginia Polytechnic Institute, Blacksburg, VAd Department of Statistics, Virginia Polytechnic Institute, Blacksburg, VA

a b s t r a c ta r t i c l e i n f o

☆ A small portion of the work was presented at the AmPhysicians Scientific Assembly; Denver, CO; Octoannemergmed.com/article/S0196-0644(12)00951-1/full⁎ Corresponding author. Department of Emergency

Virginia Tech Carilion School of Medicine, Roanoke,E-mail address: damon.kuehl@gmail.com (D.R. Kueh

0735-6757/$ – see front matter © 2013 Elsevier Inc. Alhttp://dx.doi.org/10.1016/j.ajem.2013.07.003

Article history:

Received 20 December 2012Received in revised form 28 June 2013Accepted 4 July 2013

Study Objectives: Variation in computed tomography (CT) use between emergency medicine (EM) physiciansmay delineate appropriate or inappropriate use. We hypothesize that variation in all types of CT use existsbetween providers and their use in patients with common chief concerns. We determine EM physicians'variability in CT use of all types and whether high use in one area predicts use of other CT types.Methods: This was a retrospective study of EM physicians practicing at an 800-bed tertiary level 1 trauma

center over a 3.5-year period. Computed tomography rates by type and by patient chief concernweremodeledfor providers as a function of patient acuity, disposition, age, and time of day using logistic regression.Results: Of 195 801 eligible visits, 44 724 visits resulted in at least 1 CT scan. The adjusted rate of CT orderingby providers was 23.8% of patient visits, ranging from 11.5% to 32.7% The upper quartile of providers wasresponsible for 78% of the CT scans ordered above the mean. There was a large variation in use of all types ofCT and by chief concern. There was an 8-fold variation in use of CT abdomen in discharged patients. High headCT use by providers predicts high use in all other CT types.Conclusion: We demonstrate a dramatic variation in CT use among EM physicians in all types of CT andcommon chief concerns. Greater variation was present in patients who were discharged. Large deviation fromthe mean by a group of providers may suggest inappropriate use.

© 2013 Elsevier Inc. All rights reserved.

1. Introduction

National and emergency department (ED) level data show thatcomputed tomography (CT) use rates have continued to rise in EDcare [1–8]. Some of the increase in CT use can be attributed toincreased diagnostic value and subsequent changes in standard ofcare, although no unified explanation exists for this dramatic increase[9–14]. Coinciding with the rise in CT use, advanced medical imaginghas been identified as a major driver of increasing medical costs [15].Increased patient exposure to harmful radiation with CT use is alsowell documented [16–20]. Reducing head CT use is a current goal ofthe Centers for Medicare and Medicaid Services [21,22].

Evidence of variation in head CT use by EDs is also accumulating[23–26]. A recent study suggests that there is significant interphysi-cian variation of head CT ordering by emergency medicine (EM)physicians that may explain this variation [27]. Risk aversion appearsto play some role in provider's excess use of some types of CT [28–30],although a more recent study suggests that malpractice fear and risk

erican College of Emergencyber 10, 2012. http://www.text.Medicine, Carilion Clinic andVA 24014.l).

l rights reserved.

tolerance do not explain such variation [23]. Provider experience alsodoes not appear to impact head CT ordering [27].

Emergencymedicine physician variability in use of all CT types hasyet to be evaluated. Providers' use may vary in one type of CT imagingand not others. Detailed examination of the variation of all CTordering may more directly identify physicians that overuse CTimaging in the ED and could direct efforts of reducing inappropriateCT use. We examine whether interphysician variation in CT use existsamong providers for all types of CT scans and for common chiefconcerns and determine if high use of one type of CT imaging canpredict use of other types of CT.

2. Methods

2.1. Study design

Our study was a retrospective cross-sectional study conductedbetween August 2008 and January 2012 of EM physicians and EDpatients at an 800-bed regional referral teaching hospital and level 1trauma center. The 70-bed ED is staffed only by board-certified orboard-eligible EM physicians and receives 86 000 annual visits. Thereare 2 helical CT scanners available in the ED. During the study period,only interns (postgraduate year 1 residents) from many differentspecialties rotated in the department. Attending physicians cover theEDwithout geographic localization and evaluate patients in an orderly

Table 1Demographics and distribution of the patient and physician variables in relation toordering practices of CT

Patient characteristics CT not ordered(n = 151 077)

CT ordered(n = 44 724)

Patient sex (female), n (%) 84 962 (56.2) 25 478 (57.0)Patient age (y), n (%)8-29 46 690 (30.9) 7199 (16.1)30-49 49 639 (32.9) 12 547 (28.1)50-69 34 646 (22.9) 13 383 (29.9)70-89 18 009 (11.9) 10 223 (22.9)≥90 2093 (1.4) 1372 (3.1)

ESI, n (%)5 5173 (3.4) 76 (0.2)4 41 336 (27.4) 3588 (8.0)3 74 399 (49.2) 28 707 (64.2)2 21 073 (13.9) 10 136 (22.7)1 8104 (5.4) 1917 (4.3)Chief concern, n (%)Abdominal pain (1) 11 344 (7.5) 7270 (16.3)Chest pain (2) 14 812 (9.8) 3304 (7.4)Shortness of breath (3) 7876 (5.2) 1693 (3.8)Headache (4) 3016 (2.0) 2105 (4.7)Other (5) 11 4029 (75.5) 30 352 (67.9)Physician characteristicsYears in practice, mean ± (SD) 10.7 ± 8.0 11.3 ± 8.6Visit characteristicsVisit time (crude rate/shift), n(%)6 AM–3 PM (23.9%) 64 584 (42.7) 20 279 (45.3)3 PM–12 AM (22.4%) 66 989 (44.3) 19 159 (42.8)12 AM–6 AM (21.3%) 19 504 (12.9) 5286 (11.8)

ED disposition, n(%)Discharged 112 347 (74.4) 23 972 (53.6)Admitted to hospital 36 771 (24.3) 20 442 (45.7)Othera 1959 (13.0) 310 (0.7)

a Includes AMA, death, and transfer. Observation is categorized as admitted.

Table 2CT ordering by subtypes

CT subtypes CT ordering as a percentage of visits Rangea (%) Variationb

Head 10.7 3.7-16.7 ReferenceAbdomen/Pelvis 8.1 2.3-12.6 1.4Spine 4.3 1.2-8.6 1.9Chest 3.2 0.97-7.9 2.4Extremity 0.4 0.0-0.9 1.7

a Highest/Lowest rate of ordering among physicians for each CT type showingminimum and maximum rates of ordering for that type of CT.

b Levene’s test shows a nonuniform variation between subtypes (p b 0.0001).

1438 M.B. Levine et al. / American Journal of Emergency Medicine 31 (2013) 1437–1442

process based on next-to-be-seen and acuity level. Providers areassigned day, evening, and night shifts in an equitable fashion. Sixphysicians worked predominantly nights at 1 or more times duringthe study period.

2.2. Physician characteristics

We included EM attending physicians who worked more than 200clinical hours at the study ED between August 2008 and January 2012.

Fig. 1. Overall CT orderin

Providers may work at multiple other facilities within the system butwere only included if they worked 200 hours at the study facility. Weexcluded physicians who primarily worked in the pediatric ED. Afterexclusions, 49 physicians were eligible for the study. Physiciancharacteristics were obtained from employment records. Physician-specific variables included hours worked, number of patients seen,age, and years in practice since the completion of residency.

2.3. Patient and visit characteristics

For each patient visit, patient demographics and encounterinformation were obtained from the electronic medical record(EMR; EPIC, 2010, Verona, WI). We excluded patients who wereyounger than 8 years, patients who left before being seen, andpatients recorded as trauma activations. Visit and patient character-istics were extracted from the EMR: variables included patient ageand sex, time of visit, chief concern, Emergency Severity Index (ESI),ED disposition, attending physician of record and ordering attendingphysician, all completed CT studies, and the anatomical regionscanned. The ESI is a validated measure that stratifies patients into 5levels of severity (1, most urgent; 5, least urgent) and predictsresource use and hospital admission based on presenting signs andsymptoms [31–33]. Chief concern was recorded at the time of triageby a nurse using a predetermined pick list.

Computed tomography studies are ordered through physicianorder entry in the EMR. Computed tomography ordering by internsmust be approved by an attending physician before order entry. Onlycompleted CT orders with a documented interpretation by aradiologist were included in the analysis. All CT scans ordered during

g rates by provider.

Fig. 2. Provider CT ordering rates by CT subtype.

Table 3CT ordering rates stratified by chief concern and ED disposition

Chiefconcern

Mean (95% CI), % SE(%)

Range (%) Total no. ofprovidersa

Mean providerobservations

Abdominalpain

49 397.3

All patients 38.5 (35.7-41.3) 1.5 12.1-52.5Discharged 32.8 (29.5-36.1) 1.7 7.7-49.5Admitted 53.2 (49.6-56.9) 1.9 28.6-71.1

Chest pain 44 425.0All patients 13.4 (11.4-15.3) 1.0 4.12-32.1Discharged 12.8 (10.5-15.1) 1.2 2.99-32.4Admitted 15.1 (12.8-17.4) 1.2 3.74-28.8

Shortness ofbreath

34 266.0

All patients 13.6 (11.6-15.5) 1.0 4.1-29.2Discharged 11.0 (8.9-13.0) 1.1 2.0-23.4Admitted 15.2 (12.4-18.0) 1.4 3.9-37.9

Headache 37 140.9All patients 40.9 (36.6-45.2) 2.2 16.6-76.5Discharged 38.0 (33.4-42.6) 2.4 8.1-76.3Admitted b

a Providers having more than 50 observations for each category.b Only 1 provider had more than 50 patients with a chief concern of headache that

resulted in admission.

1439M.B. Levine et al. / American Journal of Emergency Medicine 31 (2013) 1437–1442

a given patient visit were included. For a CT to be attributed to aphysician, the physician had to be the attending of record and musthave ordered the image.

2.4. Outcome measures and analysis

Ordering rates of CT and CT subtypes (head, abdomen, chest, spine,extremity) were calculated. To account for variation in visit andpatient characteristics, adjusted CT order rates using multiple logisticregression for individual providers as a function of ESI, age, time of dayof visit, and the patient's disposition were calculated. These variableswere selected a priori based on methods used in a similar studies [25]to replicate similar conditions and also known factors that affectacuity and resource use in ED patients [34].

Provider CT ordering rates in patients with common chief concerns(CT abdomen/pelvis for abdominal pain, CT chest for shortness ofbreath, CT chest for chest pain, CT head for headache) were alsodetermined, and variability between these subtypes was comparedusing the Levene’s test. Only physicians who saw a minimum of 50cases of a given chief concern were included in the analysis.

To assess if providers use CT imaging consistently across differentCTmodalities for a given chief concern, each provider was categorizedas high/low user within each complaint if the corresponding adjustedCT order rate is above/below the upper quartile for each group. Oddsratio are computed for the high versus low users for any 2 givencomplaints. Statistical analysis was performed using SAS v9.2 (SASInstitute Inc, Cary, NC). Graphics were plotted using R 2.13.2.

3. Results

Over the 3.5-year period, 44 724 (22.8%) of the 195 801 uniqueeligible ED visits resulted in at least 1 CT study, with a total CT studycount of 57 282. Of the unique visits, 8220 (18.4%) had multiple CTstudies performed during the same visit and 2499 (5.6%) of the 44 724visits resulted in at least 3 CT studies. The average practice time sincethe completion of residency was 10.9 years, ranging from 1 to 31years, with an interquartile range of 3.6 to 15.8 years. Mean number of

hours worked was 2494, and 80% of the providers worked more than1000 hours (range, 216-5514 hours). Provider experience was not apredictor of CT use (Table 1).

The adjusted mean rate of CT ordering by providers was 23.8% (95%confidence interval [CI], 23.4-24.2) of patient visits, ranging from a lowof 11.5% (95% CI, 9.2-14.4) to high of 32.7% (95% CI, 31.0-34.5) (Fig. 1).The upper quartile of providers was responsible for 78% of the CT scansordered above themean CT ordering rate. There was a large variation inuse of all types of CT (Table 2; Fig. 2). We used the Levene’s test todetermine that variation was not uniform across the CT types (p b

0.0001). Variation in CT use of chest, abdomen, and spine was greaterthan CT head. Total combined use of chest, abdomen, and spine CT was45% greater than the use of CT studies of the head (32 916 vs 22 141).

Fig. 3. Variance of CT ordering rates by chief concern.

1440 M.B. Levine et al. / American Journal of Emergency Medicine 31 (2013) 1437–1442

In Table 3, the differences in adjusted CT ordering rates for each typeof chief concern based on the patient's ED disposition are demonstrated.Themean number of observations for each provider ranged from 140 to424 for chest pain. There was a significant variation in CT orderingbetween individual physicians in the combined (Fig. 3), admitted, andthe discharged patient populations within each chief concern. Inpatients presenting with abdominal pain, there existed a 7-folddifference in interphysician-adjusted CT ordering rates among dis-charged patients (8.8%-59.8%) but less than a 3-fold difference inadmitted patients (29.6%-73.1%). There was an 8-fold difference in CTordering rates for discharged patients who presented with headache.

Providers who more frequently order a CT for a chief concern ofabdominal pain were significantly more likely to also order CT forheadaches (Table 4). Providers with high use of CT head for patientswith headache are considerably more likely to order a CT for anychief concern.

4. Discussion

We demonstrate that there is a wide variation in all types of CTordering among EM physicians. This finding persisted despiteadjusting for patient level factors that affect CT ordering. There is aneven greater variation among providers use of CT in common chief

Table 4Prediction of high CT users among paired chief concerns

Chief concern 1 | chief concern 2a Odds ratiob 95% CI P

Headache | abdominal pain 88 7.0-1108.0 .0005Total | headache 24.5 3.4-178.0 .0016Chest pain | shortness of breath 14.7 2.3-92.1 .0042Shortness of breath | abdominal pain 4.0 0.8-19.7 .0926Chest pain | abdominal pain 3.4 0.7-15.3 .107Headache | shortness of breath 5.3 0.9-31.9 .0667

a Chief concern 1 given chief concern 2.b Providers are X times more likely to be a high CT user for chief concern 2, if they are

a high user of CT for chief concern 1 compared with low users of CT for chief concern 1.

concerns, with a smaller group of providers accounting for a largeportion of EDCT use. Their behavior in onemodality of CT predicts highuse in other areas. It is notable that the bulk of this high rate appears tobe explained in use on patients who are ultimately discharged.

The variation highlights a major difference providers take in theirapproach to certain patients. Risk aversion has been shown to playsome role in providers who are high CT users compared with low CTusers [28–30], but the relative effect does not appear to account for allthe variations among providers. Within this group, overestimation ofdisease prevalence or defensive medicine practices must be involvedyet cannot readily account for the meteoric rise of imaging, or suchdramatic variation. Risk adverse providers are unlikely to havechanged their practice so dramatically over the past years.

Some other undetermined provider, patient, or environmentalfactors are involved. Increases in ED visits and ED crowding are onecertainty over the same period as the increase in CT. Combined withprovider pressures to maintain throughput and productivity, crowd-ing and patient flow pressures may play a significant role. It may becertain that providers respond to this increased density or pressure onthroughput by using CT imaging in unnecessary or inappropriateways. Whether reimbursement or relative value units per patientincreases when CT (whether indicated or not) is used is unknown. Theuse of the CT scan to replace time spent with patients and a thoroughexamination is a very real possibility and alluded to in lay press buthas yet to be studied. The recent study of Kocher et al [2] suggestingthat CT adds 1 hour to the length of stay stands counter to thisreasoning. If CT does reduce admission, as suggested by some studies[10,14], then an hour increase may be trivial in comparative timesavings. Our findings of an almost 8-fold variation in provider CTordering on patients who ultimately go home will require detailedinvestigation but suggest significant indiscriminate use in thispatient population.

One of the major values and savings of ED CT imaging wouldappear to be accurate diagnosis and the allowance for a provider tomake a risk limited discharge of an ED patient [10]. This may provide acost savings over admission, but it is not clear what savings would begained from reducing the rate of ordering in this group. The fact that

1441M.B. Levine et al. / American Journal of Emergency Medicine 31 (2013) 1437–1442

providers with low rates of ordering in discharged patients existsuggests that there is opportunity. Using a conservative estimate of1000 dollars per CT scan, within our cohort, a single high user initiatesan estimated 600 000 dollars in additional charges each yearcompared with a low user. That there exists no additional increasein serious abdominal diagnosis rates despite a 2-fold increase in CTuse in EDs highlights that the increase has not improved quality orreduced admissions and thereby appears only to add to cost, notsavings [9].

Thewide variation between physician CT use not only has financialimplications but also poses significant differences in radiation risks topatients. Physician awareness and understanding of CT radiationexposure may account for some observed differences in use [35–37].Educating EM physicians may reduce iatrogenic radiation to avulnerable population that is largely unaware of its cumulativedanger yet expects its use during ED visits [38,39].

The most striking finding is our ability to predict that high provideruse in one area corresponds to heavy use in other areas. The findingthat providers with themost dramatic use of head CT are also excessiveusers of CT for abdominal pain and more likely to use CT modalities ingeneral suggests that this small group accounts for heavy CT use in allED settings. High users of chest CT for chest pain also appear to use itindiscriminately for a related complaint, shortness of breath. Our datasuggest that agencies and institutions that wish to examine provider'suse of CT could use head CT as a reliable surrogate for high use of CTimaging in general. That ED providers with high use of head CT predictshigh use in general suggests that head CT usemay be a good target. Thecurrent Centers for Medicare and Medicaid Services goal of reducinghead CT use may also have other impacts. Whether attempts atreducing CT head use have any effect on other CT use remains to bedetermined. It is important to note, though, that variation was fargreater in other types of CT use and other CT types (combined)accounted for far more CT imaging as a whole than head CT.

4.1. Limitations

Our studywas conducted at a single institution andmay not reflectother practice sites. National rates of CT ordering have been recentlyestimated at 16.7% of ED visits [3]. Our adjusted rate is slightly higherand likely reflects rates at a tertiary referral hospital with higheracuity than would be estimated from a national average. Our study'shead CT use and variation rate is very close to a study at a similartertiary ED (8.9 vs 10.1) [27], and our abdominal CT use rate is nearlyidentical to a 2011 study of abdominal CT use (8.1 vs 7.9) [4] in anacademic ED.

5. Conclusions

Quality and cost reduction initiatives surrounding CT use mustattempt to reduce variation rather than simply focusing on overallrates of use. Decision support systems appear to have some impact onreducing CT use [40]. However, failure to consider or understandpotential drivers of high use such as risk aversion, crowding, orthroughput pressure may not allow such technology to be fullyrealized. There are many providers with high CT use, and variation isheavily centered on patients who are discharged for all CT modalities.Interventions focused on certain providers and patient types mayhave dramatic impacts on CT use rates and, ultimately, cost.

References

[1] Korley FK, Pham JC, Kirsch TD. Use of advanced radiology during visits to USemergency departments for injury-related conditions, 1998-2007. JAMA2010;304:1465–71.

[2] Kocher KE, Meurer WJ, Fazel R, Scott PA, Krumholz HM, Nallamothu BK. Nationaltrends in use of computed tomography in the emergency department. Ann EmergMed 2011;58:452–62.e3.

[3] Kirsch TD, Hsieh YH, Horana L, Holtzclaw SG, Silverman M, Chanmugam A.Computed tomography scan utilization in emergency departments: a multi-stateanalysis. J Emerg Med 2010.

[4] Raja AS, Mortele KJ, Hanson R, et al. Abdominal imaging utilization in theemergency department: trends over two decades. Int J Emerg Med 2011;4:19.

[5] Lee J, Kirschner J, Pawa S,Wiener DE, Newman DH, Shah K. Computed tomographyuse in the adult emergency department of an academic urban hospital from 2001to 2007. Ann Emerg Med 2010;56:591–6.

[6] Rao VM, Levin DC, Parker L, Frangos AJ, Sunshine JH. Trends in utilization rates ofthe various imaging modalities in emergency departments: nationwide Medicaredata from 2000 to 2008. J Am Coll Radiol 2011;8:706–9.

[7] Broder J, Warshauer DM. Increasing utilization of computed tomography in theadult emergency department, 2000-2005. Emerg Radiol 2006;13:25–30.

[8] Larson DB, Johnson LW, Schnell BM, Salisbury SR, Forman HP. National trendsin CT use in the emergency department: 1995-2007. Radiology 2011;258:164–73.

[9] Pines JM. Trends in the rates of radiography use and important diagnoses inemergency department patients with abdominal pain. Med Care 2009 Jul;47(7):782–6.

[10] Abujudeh HH, Kaewlai R, McMahon PM, Binder W, Novelline RA, Gazelle GS, et al.Abdominopelvic CT increases diagnostic certainty and guides managementdecisions: a prospective investigation of 584 patients in a large academic medicalcenter. AJR Am J Roentgenol 2011;196(2):238–43.

[11] Stein PD, Fowler SE, Goodman LR, et al. Multidetector computed tomography foracute pulmonary embolism. N Engl J Med 2006;354:2317–27.

[12] Raja AS, Wright C, Sodickson AD, et al. Negative appendectomy rate in the era ofCT: an 18-year perspective. Radiology 2010;256:460–5.

[13] Rosen MP, Sands DZ, Longmaid 3rd HE, Reynolds KF, Wagner M, Raptopoulos V.Impact of abdominal CT on the management of patients presenting to theemergency department with acute abdominal pain. AJR Am J Roentgenol2000;174(5):1391–6.

[14] Iglehart JK. Health insurers and medical-imaging policy—a work in progress.N Engl J Med 2009;360:1030–7.

[15] Smith-Bindman R, Miglioretti DL, Johnson E, et al. Use of diagnostic imagingstudies and associated radiation exposure for patients enrolled in large integratedhealth care systems, 1996-2010 diagnostic imaging and radiation exposure. JAMA2012;307:2400–9.

[16] Amis ES, Butler PF, Applegate KE, et al. American College of Radiology white paperon radiation dose in medicine. J Am Coll Radiol 2007;4:272–84.

[17] Einstein AJ, Henzlova MJ, Rajagopalan S. Estimating risk of cancer associated withradiation exposure from 64-slice computed tomography coronary angiography.JAMA 2007;298:317–23.

[18] Sodickson A, Baeyens PF, Andriole KP, et al. Recurrent CT, cumulative radiationexposure, and associated radiation-induced cancer risks from CT of adults.Radiology 2009;251:175–84.

[19] Brenner DJ, Hall EJ. Computed tomography—an increasing source of radiationexposure. N Engl J Med 2007;357:2277–84.

[20] OP-15 Technical Specifications. Accessed from Qualitynet.org on May 21,2012. http://www.qualitynet.org/dcs/ContentServer?c=Page&pagename=QnetPublic%2FPage%2FQnetTier2&cid=1228695266120.

[21] Harris B, Hwang U, Lee WS, Richardson LD. Disparities in use of computedtomography for patients presenting with headache. Am J Emerg Med 2009;27:333–6.

[22] http://www.fda.gov/Radiation-EmittingProducts/RadiationSafety/RadiationDoseReduction/ucm199994.htm. Accessed on 6/22/2013.

[23] Andruchow JE, Raja AS, Prevedello LM, Zane RD, Khorasani R. Variation in headcomputed tomography use for emergency department trauma patients andphysician risk tolerance. Arch Intern Med 2012;172:660–1.

[24] Klassen TP, Reed MH, Stiell IG, et al. Variation in utilization of computedtomography scanning for the investigation of minor head trauma in children: aCanadian experience. Acad Emerg Med 2000;7:739–44.

[25] Stiell IG, Wells GA, Vandemheen K, et al. Variation in ED use of computedtomography for patients with minor head injury. Ann Emerg Med 1997;30:14–22.

[26] RyuWHA, Feinstein A, Colantonio A, Streiner DL, Dawson D. Regional variability inthe use of CT for patients with suspected mild traumatic brain injury. Can J NeurolSci 2009;36:42–6.

[27] Prevedello LM, Raja AS, Zane RD, et al. Variation in use of head computedtomography by emergency physicians. Am J Med 2012;125:356–64.

[28] Pines JM, Hollander JE, Isserman JA, Chen EH, Dean AJ, Shofer FS, et al. Theassociation between physician risk tolerance and imaging use in abdominal pain.Am J Emerg Med 2009;27(5):552–7.

[29] Rohacek M, Buatsi J, Szucs-Farkas Z, Kleim B, Zimmermann H, Exadaktylos A, et al.Ordering CT pulmonary angiography to exclude pulmonary embolism: defenseversus evidence in the, emergency room. Intensive Care Med 2012 Aug;38(8):1345–51.

[30] Wong AC, Kowalenko T, Roahen-Harrison S, Smith B, Maio RF, Stanley RM. Asurvey of emergency physicians' fear of malpractice and its association with thedecision to order computed tomography scans for children with minor headtrauma. Pediatr Emerg Care 2011;27(3):182–5.

[31] Van der Wulp I, Rullmann HAA, Leenen LPH, van Stel HF. Associations of theEmergency Severity Index triage categories with patients' vital signs at triage: aprospective observational study. Emerg Med J 2010.

[32] Van der Wulp I, Schrijvers AJP, van Stel HF. Predicting admission and mortalitywith the Emergency Severity Index and the Manchester Triage System: aretrospective observational study. Emerg Med J 2009;26:506–9.

1442 M.B. Levine et al. / American Journal of Emergency Medicine 31 (2013) 1437–1442

[33] Tanabe P, Gimbel R, Yarnold PR, Kyriacou DN, Adams JG. Reliability and validity ofscores on the Emergency Severity Index version 3. Acad Emerg Med 2004;11:59–65.

[34] Tandberg D, Qualls C. Time series forecasts of emergency department patientvolume, length of stay, and acuity. Ann Emerg Med 1994;23:299–306.

[35] McBride JF, Wardrop III RM, Paxton BE, Mandrekar J, Fletcher JG. Effect onexamination ordering by physician attitude, common knowledge, and practicebehavior regarding CT radiation exposure. Clin Imaging 2012;36(5):455–61.e1.

[36] Puri S, Hu R, Quazi RR, Voci S, Veazie P, Block R. Physicians' andmidlevel providers'awareness of lifetime radiation-attributable cancer risk associated with common-ly performed CT studies: relationship to practice behavior. AJR Am J Roentgenol2012;199(6):1328–36.

[37] Gimbel RW, Fontelo P, Stephens MB, Olsen CH, Bunt C, Ledford CJ, et al. Radiationexposure and cost influence physician medical image decision making: arandomized controlled trial. Med Care 2013;51(7):628–32.

[38] Baumann BM, Chen EH, Mills AM, Glaspey L, Thompson NM, Jones MK, et al.Patient perceptions of computed tomographic imaging and their understanding ofradiation risk and exposure. Ann Emerg Med 2011;58(1):1–7.e2.

[39] Horowitz JM, Yaghmai V, Miller FH, Russell EJ. Will CT ordering practices change ifwe educate residents about the potential effects of radiation exposure?Experience at a large academic medical center. Acad Radiol 2011;18(11):1447–52.

[40] Raja AS, Lp IK, Prevedello LM, et al. Effect of computerized clinical decision supporton the use and yield of CT pulmonary angiography in the emergency department.Radiology 2012;262(2):468–74.