Availability and Quality of Computed Tomography and Magnetic Resonance Imaging Equipment in U.S. Emergency Departments
Presented at the Society for Academic Emergency Medicine Annual Meeting, Washington, DC, May 30, 2008.
Address for correspondence and reprints: Adit A. Ginde, MD, MPH; e-mail: firstname.lastname@example.org.
Objectives: The objective was to determine the availability and quality of computed tomography (CT) and magnetic resonance imaging (MRI) equipment in U.S. emergency departments (EDs). The authors hypothesized that smaller, rural EDs have less availability and lower-quality equipment.
Methods: This was a random selection of 262 (5%) U.S. EDs from the 2005 National Emergency Department Inventories (NEDI)-USA (http://www.emnet-usa.org/). The authors telephoned radiology technicians about the presence of CT and MRI equipment, availability for ED imaging, and number of slices for the available CT scanners. The analysis was stratified by site characteristics.
Results: The authors collected data from 260 institutions (99% response). In this random sample of EDs, the median annual patient visit volume was 19,872 (interquartile range = 6,788 to 35,757), 28% (95% confidence interval [CI] = 22% to 33%) were rural, and 27% (95% CI = 21% to 32%) participated in the Critical Access Hospital program. CT scanners were present in 249 (96%) institutions, and of these, 235 (94%) had 24/7 access for ED patients. CT scanner resolution varied: 28% had 1–4 slice, 33% had 5–16 slice, and 39% had a more than 16 slice. On-site MRI was available for 171 (66%) institutions, and mobile MRI for 53 (20%). Smaller, rural, and critical access hospitals had lower CT and MRI availability and less access to higher-resolution CT scanners.
Conclusions: Although access to CT imaging was high (>90%), CT resolution and access to MRI were variable. Based on observed differences, the availability and quality of imaging equipment may vary by ED size and location.
The emergency department (ED) provides acute medical care 24 hours per day for an estimated 115 million patients in the United States each year.1 Characterization of the distribution and quality of emergency services has gained greater attention as public health officials have sought to understand and reduce geographic disparities in access to high-quality emergency care. For instance, we recently created the first national inventory of U.S. EDs, a project that allowed us to describe the number, distribution, and basic characteristics of EDs.2
Increasingly, cross-sectional imaging has become an important component of the diagnostic evaluation for many ED patients. Indeed, ED utilization of computed tomography (CT) or magnetic resonance imaging (MRI) has increased from 2.4% of all ED visits in 1992 to 11.2% in 2005.1,3 National stroke and trauma guidelines recommend 24-hour availability of CT imaging and interpretation.4,5 Clinical pathways for abdominal pain and pulmonary embolism emphasize the role of CT in timely diagnosis.6,7 Additionally, multislice CT scanners have been touted for their increased quality and speed,8 but their availability for ED patients is unknown. Although not as widely utilized as CT, emergent MRI is increasingly utilized for stroke care and spinal emergencies and was performed during 0.5% of all U.S. ED visits in 2005.1
The availability and quality of diagnostic imaging equipment in U.S. EDs are important extensions of the effort to describe access to high-quality emergency care, but has not previously been characterized on a national level. In this study, we sought to characterize the availability and quality of CT and MRI equipment in U.S. EDs, with particular attention to differences based on ED characteristics. We hypothesized that smaller and rural EDs would have less availability of on-site cross-sectional imaging and less access to higher-resolution equipment.
Study Design and Population
We performed a multicenter, cross-sectional survey of radiology technologists at a random sample of U.S. hospitals with EDs. We obtained Institutional Review Board approval with waiver of informed consent.
We used the 2005 version of the National Emergency Department Inventories (NEDI)-USA (http://www.emnet-usa.org/nedi/nedi_usa.htm) to obtain a comprehensive list of all nonfederal US hospitals with EDs (n = 4,828). Methods for derivation of this database have been previously described.2 Briefly, NEDI-USA combines data from three sources: Verispan Marketing Group’s Hospital Market Profiling Solution Database, the American Hospital Association Annual Survey of Hospitals, and information collected independently by Emergency Medicine Network (Boston, MA) staff. EDs were defined as emergency care facilities that are open 24 hours per day, 7 days per week, and available for use by the general public; “urgent care” facilities known to be closed at certain hours or days were excluded.
We obtained a random sample of 262 (5%) hospitals from the 2005 NEDI-USA database, using a random number generator. Site characteristics obtained from the database included: U.S. region, urban status, annual ED visit volume, critical access hospital status (receive federal reimbursement for importance in access to care in remote areas), and academic hospital status (per Association of American Medical Colleges Council of Teaching Hospital designation). Additionally, we used data from the Joint Commission (http://www.jointcommission.org/) and the American College of Surgeons (http://www.facs.org/) to obtain site designations as primary stroke centers and trauma centers, respectively.
Survey Content and Administration
The survey asked whether the hospital had CT and MRI equipment available for imaging ED patients. Affirmative responses were recorded for “mobile” MRI units, if they were available for ED patients when on site. For hospitals with access to equipment, we asked about hours of availability for ED patients, including hours that technologists were at the hospital or on-call from home. Finally, we asked about the resolutions (in slices) of CT scanners.
Three study investigators (AF, DMR, MV) attempted to contact CT and MRI technologists by telephone during usual business hours. These investigators were unaware of the primary hypothesis at the time of data collection. If referred, responses from radiology department supervisors or physician radiologists were accepted. After verbal consent was obtained, we administered the <5-minute survey (see Data Supplement S1, available online at http://www.blackwell-synergy.com/doi/suppl/10.1111/j.1553-2712.2008.00192.x/suppl_file/acem_192_sm_DataSupplementS1.pdf) by telephone and collected responses using a standardized data abstraction form. Additionally, a second study investigator, blinded to the responses from the initial survey, repeated the survey for a random 10% subsample to evaluate interrater agreement.
Our primary hypotheses were based on descriptive outcomes (i.e., availability and quality of imaging equipment). Thus, we based our estimated sample size requirements on the stability of 95% confidence intervals (CIs) around the prevalence of characteristics at increasing sample sizes. We calculated that an overall sample size of approximately 250 would yield two-tailed 95% CIs that span ≤15%, which we deemed adequate precision for this study.
We performed statistical analyses using Stata 9.0 (StataCorp, College Station, TX) and summarized data using descriptive statistics. We calculated the kappa statistic to evaluate interrater agreement for the subgroup of surveys administered by two investigators. We evaluated the magnitude and statistical significance of associations between site characteristics and imaging availability and quality by calculating 95% CIs for the difference in proportions.
We collected responses from 260 (99%) of the 262 hospitals, which represent a 5% random sample of all U.S. EDs. Interobserver agreement for survey responses was high (kappa = 0.77 to 1.00).
Site characteristics, compared to the overall NEDI-USA database, are presented in Table 1. Sample characteristics were similar to the overall population. In our sample, 56% of critical access hospitals had volumes under 5,000 visits per year, all had more than 20,000 visits per year, and 76% were rural. Additionally, all teaching hospitals, trauma centers, and 93% of stroke centers had visit volumes ≥20,000 per year; none of these designations were present in rural areas.
Table 1. Site Characteristics of Representative Sample of U.S. Hospitals with EDs
|ED annual visit volume|
| <5,000||48||18% (14, 23)||883 (18%)|
| 5,000–9,999||46||15% (11, 20)||657 (14%)|
| 10,000–19,999||42||16% (12, 21)||1,052 (21%)|
| 20,000–29,999||41||16% (11, 20)||762 (16%)|
| 30,000–39,999||41||16% (11, 20)||579 (12%)|
| ≥40,000||48||18% (14, 23)||895 (19%)|
| Northeast||50||19% (15, 24)||667 (14%)|
| Midwest||81||31% (25, 37)||1,410 (29%)|
| South||95||36% (30, 42)||1,844 (38%)|
| West||36||14% (10, 19)||907 (19%)|
| Urban||140||53% (48, 60)||2,785 (58%)|
| Suburban||48||18% (14, 23)||846 (18%)|
| Rural||72||28% (22, 33)||1,197 (25%)|
|Critical access hospital||70||27% (21, 32)||1,267 (26%)|
|Teaching hospital||17||7% (4, 10)||289 (6%)|
|Primary stroke center||29||11% (8, 16)||NA|
|Trauma center||14||5% (3, 9)||NA|
Survey responses on availability and quality of imaging equipment are presented in Table 2. Lack of CT capability was more common in EDs with fewer than 10,000 annual patient visits (difference = 13%; 95% CI = 7 to 21), rural hospitals (difference = 8%; 95% CI = 2 to 17), and critical access hospitals (difference = 8%; 95% CI = 2 to 17). Low-resolution CT (≤4 slices) was more common in EDs with <10,000 annual patient visits (difference = 46%; 95% CI = 33 to 57), rural hospitals (difference = 40%; 95% CI = 26 to 52), critical access hospitals (difference = 41%; 95% CI = 27 to 53), nonteaching hospitals (difference = 23%; 95% CI = 2 to 31), and nonstroke centers (difference = 23%; 95% CI = 7 to 31).
Table 2. Access to Imaging Equipment among a Representative Sample of U.S. Hospitals with EDs
|CT available (n = 260)||249||96% (93, 98)|
|CT hours (n = 249)|
| 24/7 (on-site technologist)||235||94% (91, 97)|
| 24/7 (on-call technologist)||12||5% (3, 8)|
| 7 days/week (<24 hours/day)||2||1% (0, 3)|
|CT resolution, slices (n = 246)|
| 1||35||14% (10, 19)|
| 2–4||33||13% (9, 18)|
| 5–16||81||33% (27, 39)|
| >16||97||39% (33, 46)|
|MRI available (n = 260)|
| On-site||171||66% (60, 72)|
| Mobile||52||20% (15, 25)|
|MRI hours (n = 223)|
| 24/7 (on-site technologist)||29||13% (9, 17)|
| 24/7 (on-call technologist)||59||26% (21, 32)|
| 6–7 days/week (<24 hours/day)||30||13% (9, 18)|
| 5 days/week||50||22% (17, 28)|
| <5 days/week||55||25% (19, 30)|
Lack of on-site MRI capability was more common in EDs with fewer than 10,000 annual patient visits (difference = 72%; 95% CI = 61 to 80), rural hospitals (difference = 60%; 95% CI = 48 to 70), critical access hospitals (difference = 70%; 95% CI = 59 to 78), nonteaching hospitals (difference = 37%; 95% CI = 17 to 43), nonstroke centers (difference = 38%; 95% CI = 25 to 45), and nontrauma centers (difference = 36%; 95% CI = 14 to 42).
We found that while ED access to CT imaging was nearly universal for patients in U.S. EDs, the resolution of this CT imaging and access to MRI were quite variable. Basic characteristics of the EDs (e.g., visit volume, location) were associated with important differences in the availability and quality of imaging.
Given the increased utilization of MRI in U.S. EDs, particularly for stroke care,1 access to MRI equipment has become increasingly important. Moreover, the resolution of CT imaging can affect the speed and quality of imaging and disparities in distribution of this technology may indicate a modifiable barrier to the provision of higher-quality emergency care. These differences must, however, be balanced in the context of competing concerns, such as increased utilization and cumulative radiation exposure that may result from greater availability of imaging equipment.
While the theoretical risks of increased exposure to diagnostic radiation have garnered recent attention,9 multislice CT scanners offer the opportunity to more accurately and rapidly diagnose a variety of serious medical conditions, such as pulmonary embolism, appendicitis, and traumatic injuries.6–8 New applications, such as evaluation of possible acute coronary syndrome, may enhance their importance in emergency care.10 The added value of routine ED use of higher-resolution equipment is not known and merits further investigation.
In the initial report of NEDI-USA,2 we found significant variation in distribution and use of U.S. EDs, which suggested potential variation in the quality and availability of services, particularly in smaller, rural EDs. Indeed, these EDs had less access to CT, especially higher-slice scanners, and MRI. Increasing availability of higher technology imaging equipment in these EDs must be balanced with the cost–benefit of this decision versus regionalized referral care. However, disparities in imaging availability and quality for critical access hospitals are of particular concern. These institutions are, by definition, more than 35 miles away from the nearest hospital and, therefore, transfer to another facility, let alone a referral center, is more challenging. While EDs of critical access hospitals are of low volume, they typically serve as the only source of acute care for their large catchment area and are designated to provide higher levels of service than would be expected based on their size.
The NEDI-USA database is limited by the quality of information available from data sources, as previously described.2 Selection of a 5% sample may have created bias, but random sampling and the similar characteristics of our sample compared to the overall database support the generalizability of our estimates. Data on availability and quality of imaging equipment are limited by reliance on self-report. However, the selected respondents were knowledgeable on imaging equipment at their institutions, and high interobserver agreement indicates reproducibility of data collected by independent observers, on different days, and usually from different respondents. We assumed that on-site and higher-resolution imaging equipment is a quality marker, but the accuracy of this assumption and the ideal distribution of imaging resources are unknown. Associations between site characteristics and imaging availability and quality of imaging equipment are not causal and probably are confounded or modified by other factors. However, disparities in availability of imaging suggest important variability in potential quality of care, based on site characteristics.
Although access to CT imaging was high, CT resolution and access to MRI were quite variable in this nationwide study. Smaller, rural EDs have more limited availability and quality of imaging equipment. Further evaluation of the impact of these differences on the efficiency and quality of emergency care for common conditions, such as stroke, is warranted.