| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Health Policy and Practice |
1 From the Research Department, American College of Radiology, 1891 Preston White Dr, Reston, VA 20191-4397. Received April 3, 1998; revision requested June 29; final revision received November 18; accepted January 7, 1999. Supported in part by National Cancer Institute grant U01CA62462 and in part by the American College of Radiology. Address reprint requests to J.H.B.
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Three measures of costs were evaluated: (a) input resources, (b) actual payments, and (c) billed charges. A combination of methods were used for data collection from 10 sites enrolled in a large-scale, multiinstitutional, randomized controlled clinical trial.
RESULTS: Input resource cost data (42 core-needle and eight surgical biopsies) were the most difficult to obtain. Actual payments and billed charges data collection (32 core-needle and 44 surgical biopsies) was hampered by the difficulty of obtaining data from all providers involved in the procedures. Average direct input resource costs for surgical biopsy (including needle localization) were almost three times as high as those for core-needle biopsy ($698 vs $243). Actual payments ($2,398 vs $799) and billed charges ($3,764 vs $1,496) for surgical biopsy averaged two and a half to three times higher than those for core biopsy (P < .001).
CONCLUSION: There was remarkable consistency in relative costs. Input resource costs were much more difficult to obtain than were either actual payments or billed charges. However, input resource costs present a more reliable indication of the actual costs of a procedure than do the other measures. Given the difficulty in obtaining input resource costs, analyses by using actual payments may be preferred.
Index terms: Breast, biopsy, 00.125, 00.1261, 00.1267 • Cost-effectiveness
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Stereotactic core-needle biopsy has been adopted into widespread use and advocated as a lower cost, less invasive alternative to open excisional biopsy for diagnosing nonpalpable lesions detected by using mammography. This change happened without controlled scientific studies to verify its efficacy (2,3). Hence, the National Cancer Institute sought researchers to conduct a large-scale, multiinstitutional, randomized controlled clinical trial to compare core-needle breast biopsy and fine-needle aspiration with surgical breast biopsy. In its solicitation, the National Cancer Institute also required a cost analysis, which the American College of Radiology performed. This article presents the results of that cost analysis. We compared the costs of core-needle biopsy, with either stereotactic or ultrasonographic (US) guidance, with the costs of needle localization followed by excisional biopsy. (Only surgical biopsies that were preceded by needle localization were included in this analysis, because the National Cancer Institute study is directed at biopsy of nonpalpable lesions. Furthermore, fine-needle aspiration was dropped from the clinical study protocol, so it was omitted from this study also.) When the clinical phase of the research is complete and its results are reported, we hope to combine these cost results with clinical results in a cost-effectiveness analysis.
Because the funding source, the National Cancer Institute, has a high level of interest in costs, we conducted not only an analysis of costs but also applied research into the methodology for studying costs. Health care costs can be measured in three different ways—by evaluating (a) costs of input resources, (b) actual payments, and (c) billed charges.
Data reflecting input resource costs—that is, the actual dollars expended in producing a product or service—are the best cost measures. Unfortunately, in health care studies, these data are rarely used because of the difficulty and expense of obtaining them.
Actual payments—that is, the dollars expended in purchasing a product or service—are a measure of the dollars that become unavailable for purchasing other things in the economy and are widely used in health care cost studies. Actual payments are the direct costs to society for compensating health care providers for performing procedures. These payments include co-payments, deductibles, self payments, and payments by third-party payers. They are another measure that we used in this study.
The third measure of cost that is widely used in health care studies is billed charges, or the dollar amount the consumer is billed for a service or product. Outside of health care, these "list" prices are rarely used in cost analyses, because they do not reflect actual payments, which often are lower. In addition, in health care, billed charges are infrequently paid by insurers or payers, and they may have little relationship to the costs of labor, equipment, and other resource inputs in the procedure or to the actual payments received. Actual payments are often either a discount from billed charges or derived from a payer's fee schedule. In most studies, actual payments or, less satisfactorily, billed charges are used as the measure of costs. We studied all three measures of cost because they each have advantages and disadvantages.
Our study of costs was, given its research nature, designed to answer several questions other than how core-needle biopsy and excisional biopsy compare in cost. Specifically, we also sought to answer the following questions: (a) How feasible is it to measure costs by using each of the three measures (input resources, actual payments, and billed charges)? (b) What difficulties are encountered? (c) How can these difficulties be overcome? (d) How similar are the three measures of cost in their findings with regard to the relative costs of the two kinds of biopsy?
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
We used a combination of methods for data collection: We gathered concurrent data for resource cost estimation with time and motion studies, obtained retrospective data on actual payments and billed charges, and distributed questionnaires for estimating patient time. Because of the time and expense of implementing all of these methods, all of them are not typically used in health care research. For this project, however, they were the source of data.
Approximately 22 sites were enrolled in the clinical research project, and the seven facilities that performed the largest number of core-needle biopsies were initially designated for site visits for cost data collection. These sites performed approximately 63% (945 of 1,508) of the core-needle biopsies during the first 2 years of the study. Because surgical biopsies were relatively uncommon, we also visited three additional facilities to obtain data on larger numbers of surgical biopsies. Overall, 10 sites received site visits. Although institutional review board approval and informed consent were required at each site participating in the overall study, not all sites needed these approvals for the site visits and patient questionnaires associated with this economic analysis.
Input Resource Costs
Five components were included in our measurement of input resources: (a) labor time of physicians, nurses, technologists, and other health care workers; (b) supplies; (c) equipment; (d) facilities (room time); and (e) patient time. Data on the first four input components were collected during on-site visits, and data on patient time were obtained from answers to a separate patient questionnaire.
Data on the labor component of the resource inputs were collected during on-site time and motion studies. Information about the quantity and cost of supplies, equipment, and facilities was also obtained during this visit, when possible. Patient questionnaires with self-addressed stamped envelopes were distributed during each visit and given to approximately 10 consecutive patients who would be undergoing core-needle biopsy and 10 consecutive patients who would be undergoing surgical biopsy.
Labor.—The labor cost of the health care providers involved in the procedures was expected to be the largest portion of the total cost of each procedure, so the greatest effort was devoted to obtaining information on this input. In the time and motion studies, an observer was located in an unobtrusive location outside the procedure room to record the times the patient and each provider entered and exited the room. To maintain consistency in the collection of the time and motion data, one observer (J.H.B.) performed all site visits.
Generally, when a physician was not in the procedure room, even during an ongoing procedure, his or her time was not attributed to the procedure. This is because physicians would often interpret images from other studies or observe other patients while waiting for a patient to be positioned or film to be developed. Technologist time was treated differently and generally allocated to the procedure, even when the technologist had left the procedure room. Technologists rarely treated more than one patient at a time, but they would perform multiple tasks for the ongoing procedure, such as processing film, hanging film for the radiologist, and preparing or cleaning the procedure room.
Many of the facilities in the study were academic institutions. Consequently, residents and fellows were frequently included in the procedures for teaching purposes. Their time was treated the same as that of other physicians, and a national average income for residents was used (4).
The dollar value per minute of the provider time was calculated from the national average hourly salary or annual income for each type of provider. Published national averages (4–6) rather than site-specific salaries and incomes were used to eliminate the variation due solely to differences in salaries and incomes between sites, practice types, or regions of the country. We estimated the vacation and other fringe benefit costs at 22% of salary (American Hospital Association Trend Analysis Group, personal communication, May 1997).
Supplies.—We asked managers at each site to provide a list, with prices, of supplies used for each procedure. Supply costs for core-needle biopsy, including both stereotactic and US-guided procedures, were received from six sites and averaged, as were the supply costs from the two sites that provided information on needle localization supplies. Costs from one surgical site provided the basis for the data used to estimate the surgical biopsy supply costs.
Equipment.—Information on the prices of equipment used for needle localizations and core biopsies was obtained from three radiology department managers and equipment manufacturer representatives. Data on the prices of equipment used in surgery suites and recovery rooms and on actual patient load statistics were collected from one surgery department manager. This formed the basis for the surgical biopsy equipment cost component. By using (a) the price information, (b) an expected equipment useful life of 5 years for both core-needle and excisional biopsy equipment, and (c) an estimated annual cost for capital and maintenance of 5% each, we calculated the cost of equipment per procedure while operating the equipment at 100% of capacity (ie, full capacity). Full capacity was defined as the capacity of equipment used 8 hours per day, 5 days per week, and 52 weeks per year. On the basis of estimates from the time and motion studies, eight core-needle biopsies or five surgical biopsies per 8-hour day would be performed in a facility (room) functioning at full capacity.
Facility.—Facility (procedure room) costs were calculated by multiplying the estimated cost per minute by the average time per procedure. The average time expended per procedure was collected during the time and motion studies by recording the time that a room was devoted to the actual procedure. This included room set-up time, procedure time, and room cleanup time after the procedure. Per-minute costs were calculated by estimating the average room size, construction costs (Paul Durance, Huron Systems, Ann Arbor, Mich, personal communication, May 1997), capital costs, and maintenance costs for radiology rooms, surgical suites, and preoperative/recovery rooms. From these estimates, the cost per minute for operating the rooms at full capacity was used to compute the facility cost per procedure for the radiology, surgery, and preoperative/recovery rooms.
Patient time.—The cost of the patient's time in undergoing and recovering from the biopsy procedure was the fifth component of the input resource costs. Questionnaires, on which patients were asked how many days (or hours if less than 1 day) it took before she was able to resume usual daily activities after the biopsy, were left with personnel in the radiology department during each site visit. These were to be distributed to 10 consecutive core-needle biopsy patients and 10 consecutive needle localization patients who would be undergoing surgical biopsy. The cost in patient time was calculated by using the 1994 median earnings published by the U.S. Bureau of the Census for year-round, full-time female workers (7).
Indirect and overhead costs.—In hospital accounting systems, adding in indirect and overhead costs is an entire discipline in itself, is unavoidably arbitrary in some of its allocations, and has been heavily criticized. Therefore, we ignored all indirect and overhead costs when we calculated input resource costs. For example, we counted only the purchase price of supplies, ignoring the internal warehousing costs and the labor costs of moving the supplies from bulk cartons to the procedure/operating rooms. For labor costs, we assumed that physicians and other health care personnel have no "down time" between procedures and spend no time before or after procedures communicating with colleagues or patients. Similar omissions for other categories of cost were also made. Because of the omission of indirect and overhead costs, the absolute amounts in our input resource cost analysis are low, but the ratios should be relatively sound.
Sensitivity Analyses for Input Resource Costs
Not all core-needle and surgical biopsies are performed under the same conditions. Variation occurs in equipment (US vs stereotactic guidance), facilities (freestanding breast clinic vs hospital, minor surgery facility vs surgical suite), supplies (vacuum-assisted biopsy probes vs long-throw, core-biopsy needles), and facility operating workloads. Therefore, we performed sensitivity analyses by changing various assumptions to see what difference they would make with regard to input resource costs. For one, we calculated the effect of using vacuum-assisted biopsy probes rather than long-throw biopsy needles. Two of the six sites from which we obtained supply cost data used vacuum-assisted biopsy probes for stereotactic breast biopsies rather than the less expensive long-throw, core-biopsy needles.
Second, we explored the costs per procedure of a facility functioning at the actual operating level of one of the sites in the study (67% and 64% of capacity for excisional biopsy and core-needle biopsy, respectively) rather than at the assumed 100% of capacity. This is equivalent to performing 3.4 surgical biopsies daily instead of five at full capacity and to performing 5.2 core-needle biopsies out of a possible eight per day.
Third, because many of the patients involved in this study were either retired or full-time homemakers, we computed the effect of eliminating the cost of patient time. Some authorities take the view that the cost assigned to patient time should be the patients' actual wages. If so, a cost of $0 would be assigned to the time of retirees and full-time homemakers.
Actual Payments and Billed Charges
Although actual payments and billed charges are two different ways of measuring costs, in this study, the method of obtaining information about them was the same, so they are reported together. One of the difficulties in obtaining accurate payment and charge information from available data sets such as Part B Medicare Annual Data files or commercial claims sources is the Physicians' Current Procedural Terminology coding system. Prior to 1994, the costs of the two biopsy procedures—that is, fine-needle aspiration and core-needle breast biopsy—could be submitted to third-party payers for payment by using the same Current Procedural Terminology code. Furthermore, there were multiple ways to code core-needle biopsy, so obtaining payment and charge data from Part B of the Medicare annual data files or commercial sources was unreliable.
Because preexisting claims databases could not be used owing to coding inconsistencies, patient encounter logs were reviewed during site visits to obtain the names of approximately 20 patients who had actually undergone core-needle breast biopsy and 20 who were documented to have undergone excisional breast biopsy. We then requested that hospital departments and physicians' billing offices provide us with the total billed charges and total actual payments from all payers for the approximately 20 patients in each biopsy group who we identified from the patient encounter logs. Each patient had undergone the procedure at least 6 months previously, so providers would have had sufficient time to receive essentially all the payments.
Billed charges and actual payments can differ systematically among sites. For example, they may be systematically higher in high-cost cities. Avoiding this source of variation requires analysis in addition to that of the overall averages we present. To eliminate this variation, we compared actual payments and billed charges separately at each of the two institutions that provided data on patients who had undergone multiple core-needle biopsies and on those who had undergone multiple excisional biopsies.
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Approximately 160 questionnaires were left at the sites to be distributed to patients. Sixty-two surveys were returned, and 44 were usable for analysis. The others were unusable, primarily because the type of procedure was not indicated.
Three of the 10 sites visited were able to provide complete billed charges and actual payments data both on patients who had undergone core-needle biopsy and on those who had undergone surgical biopsy. One of these three sites reported data on only one patient who had undergone surgical biopsy and on seventeen patients who had undergone core-needle biopsy. A fourth site provided data only on patients who had undergone surgical biopsy. In total, we obtained complete actual payments and billed charges data on 32 core-needle biopsy and 44 surgical biopsy cases.
Difficulties in Data Collection
Collecting cost data in all three ways was both time-consuming and expensive. Despite the expenditure of approximately 40 observer days in the field and additional time at the office, data were obtained on a relatively small number of cases. Many pitfalls and problems were encountered as data—on input resource costs, actual payments, and billed charges—were being collected for this project. Some were related to the willingness of contacts to provide data, whereas others stemmed from the availability of the data we needed.
Due to the time necessary to perform time and motion studies for each procedure, considerably fewer surgical biopsies than core-needle biopsies were observed during site visits. If appointments were fully scheduled in a day (which they rarely were), four to six core biopsies, which lasted about an hour, could be studied. However, only one surgical biopsy could be observed in a day, because it usually required 5-6 hours of elapsed time (for needle localization plus surgical biopsy). At one site, because of unusual circumstances, no surgical or core biopsies were performed during the entire site visit, precluding the obtaining of any data for the time and motion study.
Because input resource measurements included those in both radiologic procedures and surgical procedures, approval from radiology departments and surgery departments was necessary to observe the providers during the time and motion studies. Even with surgery department approval, some surgeons were still unwilling to allow the American College of Radiology staff member to be outside the operating room to record provider time.
It was difficult to obtain detailed cost information about supplies, particularly those for surgery, because our contacts with surgical personnel were more indirect. Normally, surgery departments would refer us to the medical supply department for data on prices paid for supplies, and the supply departments did not spend the time to collect the data that we needed. As a result, we were unable to perform statistical comparisons of the supply data.
Billing systems presented problems in data collection. Frequently, there were separate billing offices with separate computer systems for hospital and physician charges, and there often was no interaction between these systems. One site had separate billing systems for the radiology department, surgery department, radiology professional services, and surgery professional services. This made it necessary for us to navigate between numerous administrative units, frequently unsuccessfully, to collect data from all sources of billed charges from and actual payments to providers (including radiologists, surgeons, anesthesiologists, hospitals, and pathologists) for patient procedures.
Actual payments and billed charges for all providers were requested during data collection. Two of the sites provided actual payments and billed charges data with Current Procedural Terminology code detail, so it was obvious that all providers, including pathologists, were represented. However, the small sample size in our study, especially for surgical biopsies, introduced a possible response bias, because some sites that were supposed to provide data did not. Our comparison of core-needle and surgical biopsy costs at individual institutions that provided data on both types of biopsy addresses this response bias issue. The findings indicate that this was not a problem.
Input Resource Costs
Data on each component of input resource costs in a baseline case are shown in Table 1. The total average input resource cost for surgical biopsy was almost three times as high as that for core-needle biopsy ($698 vs $243). Moreover, for every component, surgical biopsy appeared to have a higher average cost than did core-needle biopsy. The difference was real and statistically significant (P < .001) for the components for which a statistical comparison was possible—provider time, facilities, and patient time. (Because only one site reported the cost of surgical supplies and equipment, we could not compute an SD and thus could not perform a statistical comparison of the average supply or equipment costs of the two types of breast biopsy.)
|
|
|
|
|
|
Institutions differed in the supplies used for core-needle biopsy. As Table 7 shows, the estimated average cost for core-needle biopsy with a long-throw needle was $215, and that for core-needle biopsy with a vacuum-assisted probe was $341. The cost of excisional biopsy was unaffected.
|
Actual Payments
The comparison of actual payments for surgical biopsy and core-needle biopsy is shown in the Figure. For all institutions combined, the average actual payment for surgical biopsy preceded by needle localization ($2,398) was three times as high as that for core-needle biopsy ($799) (P < .001). At the two institutions that provided information on more than one excisional biopsy patient and on more than one core-needle biopsy patient, the actual payments for the former group were at least two and a half times as high as those for the latter patients (P < .01).
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Interpretation of Findings
Our principal finding about relative costs—that the billed charges, actual payments, and resource inputs of surgical breast biopsy were greater than those for core-needle biopsy—was no surprise. However, the magnitude of the difference has not been well documented previously in the literature. The difference between core-needle biopsy and surgical biopsy in our baseline case input resource cost analysis was almost three-fold. In all the sensitivity analyses, the costs for surgical breast biopsy were also between two and three and a half times as high as those for core-needle biopsy. The difference in billed charges and payments for the two types of biopsy also was almost threefold.
In all three measures, surgical breast biopsy is approximately two and a half to three times as costly as core-needle biopsy. This suggests that actual payments or billed charges may be reasonable measures for estimating relative costs, even though they both inconsistently reflect true costs and billed charges are an overestimation of costs. It may not be necessary to spend the time and money collecting input cost data, as we did, if the goal is to compare the relative savings of one procedure over another in percentages. However, measuring actual input resource costs, as we did, seems necessary to confirm findings about billed charges or actual payments, because the charges and payments may be arbitrary.
The purpose of this article is not to calculate the effect of altering the choice of breast biopsy on total health spending in the United States. However, one must consider that the difference in cost (as measured by payments) for the approximately one million breast biopsies performed in the United States annually (8) would be about $1.6 billion.
In reality, savings of $1.6 billion should not be expected. Core-needle biopsy may not be appropriate for all patients. It is, however, less invasive and less costly than surgical biopsy. If its diagnostic accuracy is adequate to replace surgical biopsy in most cases, then, as occurred when laparoscopic cholecystectomy replaced surgical cholecystectomy (9,10), core-needle biopsy will be performed at a lower threshold of suspicion than will surgical biopsy. This means that some of the health spending reduction that would be achieved by a simple one-for-one replacement will not occur. Instead, some of the potential reduction will be spent to achieve better care as women who previously did not undergo biopsy because it was too expensive or too invasive gain from undergoing this diagnostic procedure.
Data for use in cost studies are available in several different forms. For our study, we used three different ways of measuring costs and found remarkable consistency in relative costs. Given the difficulty of obtaining resource input cost data—including the necessity of time and motion studies because the needed information is not otherwise available—analyses with actual payments may be preferred; however, this must be recognized as a less reliable method.
Strengths and Limitations of the Study
As with any study, our study has both strengths and limitations. The principal strength of our study is that cost was measured in a variety of ways. When we calculated costs in terms of input resource costs, actual payments, and billed charges, we found remarkable consistency in relative costs among the measures: Surgical biopsy, including needle localization, was uniformly two and a half to three times more expensive than core-needle biopsy. Having three different measures of cost adds strength to the results: If the findings are in agreement with one another, as ours were, they lend credence to the observed relationships. Furthermore, Hillner et al (2), by using cost-to-charge ratios specific to their institution, found a similar consistency in ratios of surgical biopsy and core-needle biopsy costs in their study.
Liberman et al (11) and Lee et al (12) used a different approach in computing costs for their economic models. They based their estimates on the Current Procedural Terminology codes that are commonly used for submitting charges to Medicare for stereotactic core-needle breast biopsy and surgical breast biopsy. Although this is a readily obtainable source for developing estimates, the inconsistency in coding among providers, which prompted the American College of Radiology to issue coding guidelines for core-needle breast biopsy, make this an inaccurate representation of the actual charges for both procedures. Furthermore, actual Medicare charges or payments are not accurate representations of the costs of procedures. Compared with the measures used and reported in this study, these easily accessible cost data methods have the disadvantage of being less reliable.
Although actual payments and billed charges are more easily accessible than are input resource costs, limiting our research to these two measures would have eroded the validity of our findings. The calculated costs obtained from the time and motion studies give strength to the observed relationships and ratios.
As demonstrated in each of the three referenced cost studies (2,11,12), modeling is a popular method used in cost studies. If correct assumptions are made, modeling can be a feasible alternative to the more time-consuming and expensive direct study that we performed. However, as we have described, assumptions often are seriously problematic.
At some facilities, core-needle breast biopsies were so widely accepted by all physicians that surgical breast biopsies were rarely performed. Therefore, it was impossible to observe excisional biopsies for the time and motion studies or to collect actual payments and billed charges data on surgical biopsies. This is an aspect of the "moving target" problem that is frequent in fields such as radiology where technology changes rapidly.
In spite of making additional site visits to increase the number of surgical biopsies observed in the time and motion studies, we faced a possible response bias due to the small number of observations. Our comparison of core-needle and surgical biopsy costs at individual institutions, which provided data on both procedures, addressed this response bias issue. The findings of this comparison indicate that it was not a problem.
Most of the institutions involved in the study were academic facilities and had residents and/or fellows either participating in or performing the procedures. Consequently, the times for both core-needle biopsy and surgical biopsy may be higher than are those normally found in the nonacademic community.
Because the study was conducted only at sites that participated in the clinical research study, there was further potential for bias in the results. Specifically, a common problem in observing patients involved in a study is that procedures take longer when a protocol is followed. In our research, only a portion of the core-needle biopsy patients observed for this project were "protocol patients," and none of the surgical biopsy patients were protocol patients. We did not limit our selection of patients to observe for the time and motion studies to only those enrolled in the clinical study. Rather, we chose all patients who were undergoing core-needle breast biopsy during the week of the site visit. This may have introduced some protocol-based bias in the length of time to perform the procedure, because one-third of the core-needle patients observed were enrolled in the clinical study.
On the other hand, our method may have reduced selection bias, because all patients were observed. In any case, the bias would be against core-needle biopsy, because following a protocol (presumably) lengthened the time to perform some core-needle biopsies, whereas the time for none of the surgical biopsies was lengthened by following a protocol. Thus, the differences in input resource costs between surgical biopsy and core-needle biopsy may be even larger than those we observed.
We calculated costs in three different ways—in terms of input resource costs, actual payments, and billed charges–and found remarkable consistency in relative costs among the measures, with surgical biopsy, including needle localization, uniformly costing two and a half to three times as much as core-needle biopsy. Input resource cost data were much more labor-intensive and difficult to obtain than were either actual payments or billed charges data. However, input resource costs present a more reliable indication of the actual cost of a procedure than do the other measures. Given the difficulty in obtaining resource input costs, analyses with actual payments may be preferred.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
