a b s t r a c t

Objective: To assess for changes in emergency department (ED) utilization of neuroimaging in Medicare fee-for- service beneficiaries from 1994 to 2015.

Methods: Using Medicare Physician Supplier Procedure Summary Master Files, annual ED volumes of head com- puted tomography (CT), magnetic resonance (MR), and carotid duplex ultrasound (CDUS) were assessed from 1994 through 2015. Annual volumes of head CT angiography (CTA), neck CTA, head MR angiography (MRA), and neck MRA studies were assessed from 2001 (first year of unique reporting codes) through 2015. Longitudinal population-based utilization rates were calculated using annual Medicare Part B enrollment, and utilization rates were normalized annually per 1000 ED visits.

Results: From 1994 through 2015, ED neuroimaging utilization rates per 1000 ED visits increased 660% overall (compound annual growth rate [CAGR] 9%); 529% for head CT (CAGR 9%); 1451% for head MRI (CAGR 14%); and by 104% for CDUS (CAGR 3%). From 2001 to 2015, rates increased 14,600% (CAGR 43%) and 17,781% (CAGR 45%) for head and neck CTA, respectively, and 525% (CAGR 14%) and 667% (CAGR 16%) for head and neck MRA, respectively. Trends were similar when volumes were normalized for annual Medicare fee-for-service enrollment. Non-contrast head CT was the most common imaging modality throughout the study period (86% of annual neuroimaging volume in 1994; 89% in 2015).

Conclusions: In Medicare beneficiaries, neuroimaging utilization in the ED grew unabated from 1994 through 2015, with growth of head and neck CTA far outpacing other modalities. Non-contrast head CT remains by far the dominant ED neuroimaging examination.

(C) 2018

Introduction

The ability of advanced neuroImaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI) to rapidly and reliably evaluate the brain and associated vasculature has made these modalities a central component of neurologic workup in the ED. While neuroimaging in the acute setting demonstrably benefits clinical deci- sion-making, recent literature addressing ED imaging use has raised concerns of increasing overutilization [1]. Prior studies have bolstered this assertion by demonstrating that rising overall advanced imaging use has not been met with a commensurate increase in actionable acute diagnoses [2]. Moreover, the unabated growth of such imaging

* Corresponding author at: Massachusetts General Hospital, Harvard Medical School, Department of Radiology, 55 Fruit Street, Boston, MA 02114, United States.

E-mail address: [email protected] (R.V. Gottumukkala).

¹ These authors contributed equally to this work and share joint first authorship.

in the ED contrasts with a recent trend of declining imaging utilization across other sites of care [1]. Implications of these findings include in- creased Healthcare costs, greater ionizing radiation exposure, and the potentially harmful diagnostic cascade resulting from identification of incidental findings.

Efforts to understand the general evolution of imaging use in the ED setting would benefit from granular analyses of population based changes specific to individual anatomic regions (such as the brain) that are frequently imaged. Prior work revealed that rates of head CT use in the ED more than doubled from 1995 to 2007 for presentations related to head trauma or neurologic complaints [3]. However, little is known about how this trend has evolved over the subsequent decade, or whether the same trend applies to all neuroimaging modalities, in- cluding MRI and neurovascular imaging. Longitudinal analysis of overall neuroimaging use, partitioned by modality, could help better contextu- alize the relative changes across these modalities, with the ultimate aim of more appropriately directing the study of corresponding evidence-

https://doi.org/10.1016/j.ajem.2017.12.057

0735-6757/(C) 2018

A.M. Prabhakar et al. / American Journal of Emergency Medicine 36 (2018) 680-683 681

based Clinical decision support tools and appropriate use criteria. There- fore, the aim of this study was to assess for overall and modality-specific changes in ED utilization of neuroimaging in the Medicare population from 1994 through 2015.

Methods

Study design, setting, and selection of participants

This study used aggregated Medicare Claims data from Centers for Medicare and Medicaid Services designated Public Use Files and was granted exempt status from our Institutional Review Board.

The data and methodology for using these claims data has been pre- viously described [4,5]. Annual Medicare Physician Supplier Procedure Summary (PSPS) Master Files were obtained from CMS for years 1994 through 2015. These files contain aggregated Part B Medicare billing data for all claims submitted by physicians and other providers nation- ally. Fields include information on procedures by Current Procedural Terminology (CPT) code, billing provider specialty, and patient location at the time of service [6].

CPT codes for a variety of neuroimaging procedures throughout this time period were identified (Table 1). We focused on both neuroana- tomical imaging (CT and MRI) and neurovascular imaging (ultrasound [US], CT angiography [CTA], and MR angiography [MRA]). CPT codes for MR angiography (MRA) and CT angiography (CTA) were first intro- duced in 2001, while other codes were available across the entire study period. We focused exclusively on services rendered in the emergency department (site code #23). Annual volumes of all targeted codes were ascertained for all noted years. Concurrent annual procedure counts for all billed emergency services evaluation and management (E&M) codes (99281 through 99285) were ascertained from PSPS Mas- ter Files and used to identify clinical ED encounters in the same popula- tion; these were used to normalize imaging frequency on an average per 1000 ED encounter basis. Annual Medicare fee-for-service (FFS) benefi- ciary enrollment for each year was separately obtained from CMS and used to calculate utilization on a per 1000 Medicare beneficiary popula- tion basis. During the study period, Medicare FFS enrollment increased from 32,305,000 in 1994 to 33,838,060 in 2015, representing approxi- mately 12.4% and 10.5% of the United States population, respectively.

Outcomes

The primary outcomes were the total and annual growth for each neuroimaging modality in the utilization rate per 1000 ED visits in the Medicare fee-for-service population. Other outcomes included: growth in overall volume for each neuroimaging modality, growth in utilization

per 1000 Medicare fee-for-service beneficiaries for each neuroimaging modality, and changes in the relative percentage of overall ED neuroim- aging volume represented by each modality.

Analysis

Annual imaging volumes for each modality were tabulated and population- and encounter-based utilization rates were calculated. Compound annual growth rates (CAGR) were calculated where appropriate. Data analysis was performed using SAS 9.4 (SAS Institute, Inc., Cary, North Carolina) and Excel 2015 (Microsoft, Redmond, Washington).

Results

Overall neuroimaging use

Between 1994 and 2015, total ED neuroimaging volume in the Medi- care fee-for-service population continuously increased from 350,721 to 3,649,072 examinations, representing 940% growth (CAGR 12%). Over- all, ED neuroimaging utilization increased 660% (CAGR 9%) per 1000 ED visits and 993% (CAGR 12%) per 1000 Medicare enrollees from 1994 to 2015. Data on neuroimaging volume for specific modalities is further detailed in Table 1. Per 1000 ED visits and per 1000 Medicare beneficiaries, growth was similar (Figs. 1 and 2, respectively).

Head CT

Total ED head CT volume continuously increased from 330,296 in 1994 to 3,273,159 in 2015, representing 891% growth (CAGR 12%). Head CT utilization increased 529% (CAGR 9%) per 1000 ED visits and 846% (CAGR 11%) per 1000 Medicare enrollees from 1994 to 2015.

Head MRI

Total ED head MRI volume continuously increased from 5228 in 1994 to 127,836 in 2015, representing 2345% growth (CAGR 16%). Head MRI utilization increased 1451% (CAGR 14%) per 1000 ED visits and 2234% (CAGR 16%) per 1000 Medicare enrollees from 1994 to 2015.

Head and neck CTA

ED CTA volume demonstrated the largest relative growth of all neu- roimaging modalities studied, increasing from 375 to 67,513 (17,903%

growth, CAGR 45%) for head CTA and from 298 to 65,258 (21,799% growth, CAGR 47%) for neck CTA from 2001 to 2015. Utilization for

Table 1

Volume of specific ED neuroimaging modalities in the Medicare fee for service population (1994-2015), with associated Current Procedural Terminology (CPT) codes.

Modality	1994	2004	2015	% change	CAGR
CT head total	330,296 (94.2)	1,566,288 (94.8)	3,273,159 (89.7)	891%	12%
CT head without contrast (CPT 70450)	302,517 (86.3)	1,541,205 (93.3)	3,256,714 (89.2)	977%	12%
CT head with contrast (CPT 70460,70470)	27,779 (7.9)	25,083 (1.5)	16,445 (0.5)	-41%	-2%
MRI head total	5228 (1.5)	29,388 (1.8)	127,836 (3.5)	2345%	16%
MRI head without contrast (CPT 70551)	2606 (0.7)	13,567 (0.8)	89,883 (2.5)	3349%	18%
MRI head with contrast (CPT 70552,70553)	2622 (0.7)	15,821 (1.0)	37,953 (1.0)	1347%	14%
CTA head (CPT 70496)		2837 (0.2)	67,513 (1.9)	17,903%	45%
CTA neck (CPT 70498)		1915 (0.1)	65,258 (1.8)	21,799%	47%
MRA head total		11,547 (0.7)	40,907 (1.1)	665%	16%
MRA head without contrast (CPT 70544)		10,979 (0.7)	38,470 (1.1)	678%	16%
MRA head with contrast (CPT 70545,70546)		568 (0.0)	2437 (0.1)	503%	14%
MRA neck total		6319 (0.4)	25,629 (0.7)	839%	17%
MRA neck without contrast (CPT 70547)		3124 (0.2)	10,989 (0.3)	419%	12%
MRA neck with contrast (CPT 70548,70549)		3195 (0.2)	14,640 (0.4)	2288%	25%
Carotid US (CPT 93880,93,882)	15,197 (4.3)	33,761 (2.0)	48,770 (1.3)	57%	6%
Total neuroimaging exams	350,721	1,652,055	3,649,072	940%	12%

Note. Values in parentheses are percentages. CAGR = compound annual growth rate.

682 A.M. Prabhakar et al. / American Journal of Emergency Medicine 36 (2018) 680-683

Fig. 1. Utilization per 1000 ED visits (1994-2015). Vertical axis scale on the left reflects CTA, MRA, and carotid duplex US. Vertical axis scale on the right reflects CT and total neuroimaging. CTA head and neck were combined for diagrammatic clarity.

head CTA increased 14,600% (CAGR 43%) per 1000 ED visits and 16,666% (CAGR 44%) per 1000 Medicare enrollees from 2001 to 2015. Utilization for neck CTA increased 17,781% (CAGR 45%) per 1000 ED visits and 20,294% (CAGR 46%) per 1000 Medicare enrollees from 2001 to 2015.

Head and neck MRA

ED MRA volume increased from 5348 to 40,907 (665% growth, CAGR 16%) for head MRA and from 2730 to 25,629 (839% growth, CAGR 17%) for neck MRA from 2001 to 2015 (Table 1). Utilization for head MRA in- creased 525% (CAGR 14%) per 1000 ED visits and 612% (CAGR 15%) per 1000 Medicare enrollees from 2001 to 2015. Utilization for neck MRA increased 667% (CAGR 16%) per 1000 ED visits and 774% (CAGR 17%)

per 1000 Medicare enrollees from 2001 to 2015.

Carotid duplex ultrasound

Total ED carotid duplex US volume continuously increased from 15,197 in 1994 to 48,770 in 2015, representing 221% growth (CAGR

6%). Carotid duplex US utilization increased 104% (CAGR 3%) per 1000 ED visits and 206% (CAGR 5%) per 1000 Medicare enrollees from 1994 to 2015.

Discussion

Using aggregated Medicare claims data from 1994 to 2015, we iden- tified continuous growth in utilization rates of all neuroimaging modal- ities in the ED. Our findings confirm the persistence of a previously identified trend of unabated growth in ED imaging use [1] and further refine our understanding of this trend by evaluating its relative magni- tude across specific neuroimaging modalities. Several important conclu- sions can be drawn from this report.

First, non-contrast head CT continues to represent the dominant neuroimaging modality in the ED among Medicare beneficiaries, ac- counting for 89% of all ED neuroimaging exams in 2015 (up from 86% in 1994, despite growth in other modalities). While this indicates that the role of head CT as a workhorse for rapid triage of trauma and neuro- logic symptoms has not been diminished by MRI growth, it also under- scores the need to evaluate reasons for the unchecked rise in head CT use. Contributing factors could range from theorized causes of imaging overutilization (e.g. pressure to increase ED throughput or medicolegal concerns) to evolving Clinical demographics that might justify growing use (e.g. increase in ED visits for traumatic brain injury). However, our data did not permit analysis of underlying causes. Non-uniform adop- tion of established clinical decision tools and appropriate use criteria among providers might represent an area for further investigation, as

Fig. 2. Utilization per 1000 Medicare beneficiaries (1994-2015). Vertical axis scale on the left reflects CTA, MRA, and carotid duplex US. Vertical axis scale on the right reflects CT and total neuroimaging. CTA head and neck were combined for diagrammatic clarity.

A.M. Prabhakar et al. / American Journal of Emergency Medicine 36 (2018) 680-683 683

suggested by work demonstrating multifold variation in head CT order- ing rates among ED physicians, even within the same institution [7]. Along the same lines, a study by Raja et al. noted a decrease in ED imag- ing utilization at a single large institution from 2007 to 2012 [8],a time period during which ED imaging utilization continued to increase na- tionally [1], indicating substantial variation in utilization trends across institutions. The authors of that study postulated that the downturn in ED imaging use at their institution, based on temporal proximity, might have been partially attributable to the implementation of clinical decision support and institutional initiatives to reduce duplicative imag- ing [8].

Second, head MRI utilization in the ED for Medicare beneficiaries rose dramatically (by over 20-fold) from 2001 to 2015. Although we could not study the factors underlying this trend, potential sources in- clude the rapidly increasing role of MRI in acute stroke workup [9] as well as increasing accessibility to MRI services over time. Although ris- ing utilization of an expensive modality should ostensibly drive up cost, prior work has suggested that increased ED head MRI use might ac- tually reduce inpatient admission rates and inpatient length of stay for stroke rule-out indications [10], implying the potential for an overall de- crease in resource utilization owing to expedited workup. Further anal- ysis quantifying the downstream impact of increased ED MRI utilization, employing methods such as time-driven activity based costing [11] at the single institutional level or patient-specific claims data at the na- tional level, could further explore the interplay between ED and inpa- tient imaging utilization for specific conditions.

Third, the immense growth of noninvasive neurovascular imaging in

the ED for Medicare beneficiaries from 2001 to 2015 - particularly for CTA - has far outpaced that of the other studied neuroimaging modali- ties. This finding confirms that the previously identified upward trajec- tory in overall head and neck CTA and MRA use from 2002 to 2007 [12] has showed no signs of deceleration in the ED over the subsequent de- cade. Moreover, a recent analysis of overall CTA use in the Medicare population indicated that growth was swiftest in the ED as compared to other sites of care (e.g. inpatient, outpatient, or office), a finding that also applies to head and neck CTA [5]. The disproportionate growth of head and neck CTA (CAGR 45-47%) compared to head and neck MRA (CAGR 16-17%) has several potential explanations, including the wider availability of CT, quicker scan time, and higher spatial resolution with fewer false positive results compared to MRA. The brisk growth in both CTA and MRA has also overshadowed the comparatively slower rise in carotid duplex US use (CAGR 6%), likely owing to the ability of the former modalities to survey a vastly larger anatomic extent and de- finitively answer a broader range of clinical questions.

Our analysis was subject to several limitations. First, this compre- hensive 100% Medicare fee-for-service dataset did not permit analysis of the nuances of clinical care delivery (e.g. indications for studies). Sec- ond, the dataset was limited to the Medicare fee-for-service population and did not include Medicare managed care, private payer, or Uninsured patients. Third, while aggregated claims data are useful for longitudinal analysis of national trends in Utilization of services, their lack of patient- specific and provider-specific information precludes outcomes analysis (e.g. impact on clinical care, downstream effects of imaging, rate of re- peat imaging). Finally, the lengthy timeframe encompassed by our study period (1994-2015) had the potential to introduce framing bias, whereby logistical factors (e.g. lack of widespread availability of CT and MRI) that may have limited the use of certain imaging modalities early in the study period could have amplified the overall Percentage change in utilization rates. However, this longitudinal analysis also

provides a more comprehensive and historical illustration of temporal changes than would have been possible with a more narrowly focused timeframe.

In conclusion, neuroimaging utilization in the ED has continued to grow dramatically through 2015, with CTA of the head and neck dem- onstrating the largest relative growth and non-contrast head CT re- maining the dominant ED neuroimaging modality. Our findings could serve as the basis for more detailed analyses employing research identi- fiable claims data, focusing on specific conditions and geographic varia- tion. Doing so could help further elucidate forces underlying the identified trends in ED neuroimaging use and inform more targeted methods to curb imaging overutilization.

Funding source(s) statement

Dr. Duszak received research support from the Harvey L. Neiman Health Policy Institute.

Conflict(s) of interest statements

Authors do not have any conflicts of interest.

Meetings

A preliminary version of this work was presented as an abstract at the 2016 American College of Radiology Annual Conference, May 14 to 19th, Washington, D.C.

References

1. Levin DC, Rao VM, Parker L, Frangos AJ. Continued growth in emergency department imaging is bucking the overall trends. J Am Coll Radiol 2014;11(11):1044-7. https:// doi.org/10.1016/j.jacr.2014.07.008.
2. Korley FK, Pham JC, Kirsch TD. Use of advanced radiology during visits to US emer- gency departments for injury-related conditions, 1998-2007. JAMA 2010;304(13): 1465. https://doi.org/10.1001/jama.2010.1408.
3. Larson DB, Johnson LW, Schnell BM, Salisbury SR, Forman HP. National trends in CT use in the emergency department: 1995-2007. Radiology 2011;258(1):164-73. https://doi.org/10.1148/radiol.10100640.
4. Rosenkrantz AB, Bilal NH, Hughes DR, Duszak R. National specialty trends in billable diagnostic ultrasound in the ED: analysis of Medicare claims data. Am J Emerg Med 2014;32(12):1470-5. https://doi.org/10.1016/j.ajem.2014.09.002.
5. Prabhakar AM, Misono AS, Hemingway J, Hughes DR, Duszak R. Medicare utilization of CT angiography from 2001 through 2014: continued growth by radiologists. J Vasc Interv Radiol 2016;27(10):1554-60. https://doi.org/10.1016/j.jvir.2016.05.031.
6. Rosenkrantz AB, Hughes DR, Duszak R. Medicare claims data resources: a primer for policy-focused radiology health services researchers. J Am Coll Radiol May 2017. https://doi.org/10.1016/j.jacr.2017.04.005.
7. Prevedello LM, Raja AS, Zane RD, et al. Variation in use of head computed tomogra- phy by emergency physicians. Am J Med 2012;125(4):356-64. https://doi.org/10. 1016/j.amjmed.2011.06.023.
8. Raja AS, Ip IK, Sodickson AD, et al. Radiology utilization in the emergency depart- ment: trends of the past 2 decades. Am J Roentgenol 2014;203(2):355-60. https:// doi.org/10.2214/AJR.13.11892.
9. Wintermark M, Sanelli PC, Albers GW, et al. Imaging recommendations for acute stroke and transient ischemic attack patients. J Am Coll Radiol 2013;10(11): 828-32. https://doi.org/10.1016/j.jacr.2013.06.019.
10. Redd V, Levin S, Toerper M, Creel A, Peterson S. Effects of fully accessible magnetic resonance imaging in the emergency department. In: Griffey RT, editor. Acad Emerg Med, 22; 2015. p. 741-9 (6) https://doi.org/10.1111/acem.12686.
11. Yun BJ, Prabhakar AM, Warsh J, et al. Time-driven activity-based costing in emergen- cy medicine. Ann Emerg Med 2016;67(6):765-72. https://doi.org/10.1016/j. annemergmed.2015.08.004.
12. Friedman DP, Levin DC, Rao VM. Trends in the utilization of CT angiography and MR angiography of the head and neck in the Medicare population. J Am Coll Radiol 2010;7(11):854-8. https://doi.org/10.1016/j.jacr.2010.05.007.