| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Special Reviews |
1 From the Department of Radiology and Harborview Injury Prevention and Research Center, Harborview Medical Center, University of Washington, Box 359728, 325 Ninth Ave, Seattle, WA 98104. From the 2003 RSNA Annual Meeting. Received February 16, 2004; revision requested April 21; revision received May 29; accepted June 22. Supported in part by AHRQ grant K08 HS11291. Address correspondence to the author (e-mail: craige@u.washington.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
The focus will be on emergency radiology because, in many ways, emergency radiology is an ideal setting for the use of clinical prediction rules and cost-effectiveness analysis. Time constraints for critically injured patients in the emergency and trauma settings necessitate the development of protocols in advance. These protocols are not only a challenge but also an opportunity to incorporate the tenets of evidence-based practice. Emergency radiology is, in some respects, easier to study than other aspects of imaging, because the outcomes are relatively short term. In emergency radiology, we can obtain information about outcomes in days or weeks rather than the years or decades that it might take for other imaging issues (eg, cancer screening). In addition, there is a body of evidence, in the form of clinical prediction rules and results of cost-effectiveness analyses, that drives practice in emergency radiology. Finally, emergency radiology affects nearly everyone, because most medical centers have emergency departments, and the influence of the emergency department is felt in many of the different subareas within radiology. Although the focus of this review is on emergency radiology, the concepts of clinical prediction rules and cost-effectiveness analysis are generalizable to all of radiology and to all of medicine.
|
WHO SHOULD UNDERGO IMAGING? |
|---|
|
TOPABSTRACTINTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
Fortunately, in the case of the ankle injury, we have the Ottawa Ankle Rule (1), a clinical prediction rule that we can use to identify subjects at such low probability of fracture that imaging is not necessary. The Ottawa Ankle Rule is based on three simple criteria: inability to bear weight, tenderness at the posterior edge or inferior tip of the lateral malleolus, and tenderness in a similar distribution at the medial malleolus. In the example described above, the patient did have tenderness over the posterior edge of the lateral malleolus, so imaging was indicated and a fracture was diagnosed.
The Ottawa Ankle Rule is an example of a clinical prediction rule. In simple terms, a clinical prediction rule is a decision-making tool composed of multiple factors; the rule suggests a course of action or provides a probability of disease or injury. For the Ottawa Ankle Rule, the multiple factors under consideration are aspects of the history and physical examination results and include focal areas of tenderness and inability to bear weight. The course of action is whether or not to perform radiography.
To be useful, clinical prediction rules must meet several criteria (Table) (2). First, they must be clinically important. The prediction rule must address an issue that is of direct clinical relevance. Further, development and validation of clinical prediction rules are costly and time-consuming processes. Hence, the underlying clinical issue must be of sufficient importance to warrant the research necessary to derive the clinical prediction rule. In this case, radiography of the ankle is a very common procedure performed in emergency departments and is clinically relevant.
|
Clinical prediction rules must be reproducible. The performance of a clinical prediction rule by dedicated researchers in the controlled research setting where it was derived is less important than the performance of the clinical prediction rule in actual practice. Reproducibility, or agreement between observers, must be measured and demonstrated for such a tool.
Clinical prediction rules also must be easy to use by the practitioners who will make the imaging decisions. In the case of ankle radiographs, the decision to image is often made not by an experienced physician but rather by another health care professional such as a triage nurse or perhaps a medical student. Often, the decision to image is made before the patient is even examined by a physician. Therefore, to be useful in this setting, a clinical prediction rule must be easy to use by triage nurses and medical students. The Ottawa Ankle Rule is composed of relatively simple criteria that can be applied by someone with only limited medical training and so meets this criterion of ease of use.
Another critical aspect of the usefulness of a clinical prediction rule is that it must present results in a clinically relevant form—namely, the probability of a disease or injury. Clinical prediction rules are clinical tests rather than statistical tests. The medical research is replete with studies that demonstrate associations between factors such as posterior lateral malleolar tenderness and fracture. Association, or prediction, in itself is not valuable. We can determine association and perform a hypothesis test, generating a P value to show statistical significance. However, this P value is not sufficient to guide action. To act, we must understand further information such as the risk ratio or odds ratio for injury and the probability that an injury is present if the test result is positive or negative. In the case of the Ottawa Ankle Rule, rather than simply generate P values to demonstrate association, the authors of that study (1) reported the probability of fracture with a positive finding according to the prediction rule, in this case 19.6% (50 of 255), and the probability of fracture with a negative finding according to the rule, in this case 0% (zero of 198). This information can then be used to suggest a course of action. If the probability of fracture demonstrated by the prediction rule is 0%, then the course of action is obvious: Imaging is not indicated.
Clinical prediction rules are derived by researchers in controlled research settings. Accordingly, they should be regarded with skepticism until they have been tested, or validated, in a real clinical setting. This is the final stage in determining if a clinical prediction rule is useful. Validation is best performed prospectively in a sample of patients distinct from the group used to derive the prediction rule.
There are numerous clinical prediction rules in emergency radiology that meet these criteria. In addition to the Ottawa Ankle Rule, there are the Ottawa Knee Rules (3) where findings of fibular head tenderness, inferior patellar tip tenderness, limited range of motion, age, and inability to bear weight predict the need for radiography. There are also three prediction rules for appropriate use of head computed tomography (CT) in trauma cases (4–6). These revolve around Glasgow Coma Scale scores, mechanism of injury, age, presence of amnesia, and physical signs of trauma above the clavicle or in the head. Finally, there are two clinical prediction rules for cervical spine imaging (7,8). One of these, the NEXUS rule (7), has been validated and relies on the presence of midline tenderness and neurologic deficit in patients who are examinable and not intoxicated and are without another painful injury that might distract the patient’s attention from any pain or tenderness in the cervical spine.
These clinical prediction rules are all designed for a single purpose, and that is to determine those subjects in whom imaging is not indicated. To meet this goal, the sensitivity of these imaging exclusion criteria must approach 100%. The specificity then provides an estimate of the usefulness of the prediction rule. Prediction rules of high specificity will have fewer false-positive results, meaning that fewer subjects who are positive according to the prediction rule will have negative radiographs. Thus, specificity provides an estimate of the reduction in unnecessary radiography.
|
HOW SHOULD IMAGING BE PERFORMED? |
|---|
|
TOPABSTRACTINTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
For example, approximately 10 000 cervical spinal cord injuries occur per year in the United States, at a cost of $3.4 billion (9). Cervical spine fractures may be occult, in that patients are neurologically intact at the time of presentation. Therefore, imaging is performed to exclude unstable cervical spine fracture in trauma patients and to prevent the occurrence of spinal cord injury. Decisions regarding who should undergo imaging to exclude cervical spine fracture can be made by using a clinical prediction rule such as that of the NEXUS rule, as discussed previously.
Once the decision has been made that imaging is appropriate, however, there are two approaches that can be used. Traditionally, imaging of the cervical spine is performed with radiography. Since the mid-1990s, however, CT screening in high-risk subjects has achieved prominence. CT is a more sensitive and more specific imaging modality than is radiography and may be faster, particularly in subjects who are already on the CT gantry positioned for head CT. However, the direct costs of CT are higher. Therefore, to utilize CT appropriately, it is necessary to determine which subjects are at high risk of fracture. To achieve this goal, a clinical prediction rule, the Harborview Cervical Spine CT Criteria, was developed and validated at our institution (10–12). The validation study demonstrated that the subjects identified by using the prediction rule had a high probability of fracture, in this case 12.8% (77 of 601), whereas subjects that did not meet the criteria had a probability of fracture of 0.2% (seven of 3684). Thus, the Harborview CT Screening Criteria do facilitate identification of a high-risk group (12).
To be useful, a clinical prediction rule should suggest a course of action. The Harborview CT Screening Criteria suggest that high-energy trauma with head injury yields a high probability of fracture of 12.8%. However, this probability alone does not indicate the appropriate imaging approach. To choose between competing imaging strategies, it is necessary to integrate the results of the clinical prediction rule with those of the cost-effectiveness analysis. Probability of injury becomes a key component of the cost-effectiveness analysis, and the cost-effectiveness analysis provides information about which imaging choice is appropriate. For CT screening of the cervical spine, when all short- and long-term costs and outcomes are considered, CT is cost-effective if the risk of fracture is greater than 4% (13). The cost-effectiveness of CT is due to several factors, including the higher frequency of inadequate radiographs in high-risk subjects (14), the higher cost of performance of radiography in high-risk subjects (15), and the extreme cost in dollars and outcome should a missed fracture progress to paralysis. Thus, by integrating the clinical prediction rule with the cost-effectiveness analysis results, the appropriate imaging approach in a patient with high-energy trauma and a high probability of injury becomes clear: CT is indicated.
In summary, clinical prediction rules in the form of imaging exclusion criteria can be used in the emergency radiology setting to answer the question of who should undergo imaging. The Ottawa Ankle Rule is an example of a developed and well-validated study for this objective. In addition, integration of clinical prediction rules with cost-effectiveness analysis can enable determination of how imaging should be performed. For example, the Harborview CT Screening Criteria help identify a high-risk group in whom CT has been shown to be cost-effective.
It should be emphasized that implementation of clinical prediction rules for imaging requires a multidisciplinary approach that includes both radiologists and referring physicians. Fortunately, both clinical prediction rules and evidence-based medical practice are common in other medical disciplines, particularly internal medicine and emergency medicine (2,16–21). However, radiologists should be thoroughly familiar with the concepts of clinical prediction rules so that they can be in a position to inform discussions of the who and how in imaging.
The responsible use of imaging is critical to the future of radiology. Radiation risk, overutilization, and cost containment are issues that will continue to confront radiologists indefinitely. To ensure the continued leadership of radiology in medical imaging, radiologists must embrace the evidence-based approach to selection of imaging strategies. Radiologists must focus not simply on the possibilities for imaging but rather on the realities of clinical prediction rules, cost-effectiveness analysis, and the other health services research tools that can help optimize imaging for our patients.
|
ESSENTIALS |
|---|
|
TOPABSTRACTINTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
To be useful, clinical prediction rules should be clinically important, have face validity, be reproducible and easy to use, be clinically relevant, and suggest a course of action.
To ensure generalizability, clinical prediction rules should be validated in subjects distinct from those used to develop the rule.
Radiologists must embrace the evidence-based imaging approach to ensure the continued leadership of radiology in medical imaging.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONWHO SHOULD UNDERGO IMAGING?HOW SHOULD IMAGING BE...ESSENTIALSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Brink, J. Deunk, H. M. Dekker, D. R. Kool, M. J. R. Edwards, A. B. van Vugt, and J. G. Blickman Added Value of Routine Chest MDCT After Blunt Trauma: Evaluation of Additional Findings and Impact on Patient Management Am. J. Roentgenol., June 1, 2008; 190(6): 1591 - 1598. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
