| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Both authors: Research Department, American College of Radiology, 1891 Preston White Dr., Reston, VA 20191.
Received April 11, 2002;
accepted after revision May 9, 2002.
Address correspondence to M. Bhargavan.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Data on procedures performed were collected from the American College of Radiology's 1999 Survey of Practices, and responses were weighted to make them representative of all radiologists in the United States. Workload measured as number of procedures per full-time equivalent diagnostic radiologist was tabulated by practice type, size, setting, and location and compared with corresponding survey results from 1995-1996. The independent effect of these factors on workload was measured using regression analysis. Changes in procedure complexity were calculated in terms of relative value units (RVUs) per procedure using Medicare files.
RESULTS. In 1998-1999, the average workload per full-time equivalent diagnostic radiologist had increased 8.5% since 1995-1996 to 12,800 (standard error = 200) procedures annually, with substantial variation by group type. For example, the average workload was 9400 procedures in academic groups and 13,600 in nonacademic private radiology practices. Even in relatively homogeneous categories of practices, radiologists at the 75th percentile of workload were typically performing at least 50% more procedures than radiologists at the 25th percentile. Average RVUs per Medicare procedure increased by 4% between 1995 and 1998, mainly because of an increase in the share of more complex techniques such as MR imaging and CT in the procedure mix.
CONCLUSION. The workload per radiologist measured in procedures and RVUs increased substantially between 1995-1996 and 1998-1999. Because much variation remains unexplained, averages or medians should not be used as norms.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Workload measured in relative value units (RVUs) is a better measure than a count of procedures because it takes into account the changes in complexity of radiologic procedures over time. However, the number of procedures is a measure much more readily supplied by practices than RVUs. To partially overcome the limitations of a simple count of procedures, our study also includes data on trends in RVUs per procedure taken from comprehensive Medicare data.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The 645 responses were weighted to represent what the responses would have been if all radiology practices in the United States had been surveyed and had responded to the survey. This weighting method has been described in detail elsewhere [1]. The practice weights were then multiplied by the number of diagnostic radiologists in each practice to make the data on workload representative of radiologists, not practices.
To study trends over time, we compared the 1998-1999 data with previously published data on workload for 1995-1996 obtained from ACR's 1996 Survey of Hiring by Groups [6].
To measure the relative complexity of procedures, we used professional component RVUs assigned to radiology procedures in the Medicare 1998 Physician/Supplier Procedure Summary (formerly Part B, Medicare Annual Database, or BMAD) file [7] and counts of procedures from that file. That file contains compete information on all physician and supplier services rendered to Medicare patients who have supplemental medical insurance (Medicare Part B). Our analysis was restricted to procedures that were performed by diagnostic radiologists, nuclear medicine specialists, and interventional radiologists, and that fell within Current Procedural Terminology [8] codes 70010-79999 for diagnostic imaging (excluding 77261-77799, which are radiation therapy codes) and codes 93303-93981 for sonography and interventional imaging procedures. For the approximately 80 supervision and interpretation codes, we developed professional component values representing a "complete procedure." Specifically, the relative value used for the supervision and interpretation code was the sum of the relative value for that code plus the relative value(s) for the other code(s) previously judged most frequently to be part of the complete procedure by ACR's in-house coding and nomenclature expert (Mabry M, personal communication, 1997).
Analysis Methods
Approximately 16% of responses to the survey reported very low numbers of
procedures performed. Specifically, for these 16%, procedures per FTE
radiologist were fewer than a third of the median number of procedures per FTE
radiologist among all responses. In addition, 2% of the practices reported
large numbers of procedures relative to number of diagnostic radiologists,
more than three times the overall median of procedures per FTE radiologist. To
avoid outliers distorting the results, we deleted all observations that were
fewer than one third or more than three times the median for procedures per
FTE radiologist. For consistency of comparison over time, we recalculated
workloads for 1995-1996 with observations restricted to those between one
third and three times the median; 3% of observations were eliminated by this
restriction.
We calculated means (averages) of workload measured in annual procedures per FTE radiologist. We also calculated the standard errors (SEs) of the means to measure the uncertainty, or "sampling variability," in the means caused by the use of data from a sample of practices rather than from all practices in the United States. Throughout the text, standard errors are placed in parentheses—for example, (SE = 400). Table 1 also shows three percentiles (25th, 50th, and 75th) to indicate the variability among practices within each practice category.
|
Variations, if any, in mean workloads by practice characteristics were assessed in two ways: using descriptive data in a table and through a regression analysis. In the table, practices were first sorted by type. Subsequent comparisons in the table were restricted to diagnostic-radiology-only nonacademic multiradiologist private practices, by far the most common practice type. By excluding academic, solo, government, and multispecialty practices, we eliminated variability due to practice type and focused on the effect of other characteristics. We compared workloads across different practice sizes, census regions in which the practice was located, hospital and nonhospital settings, and size of the geographic location of the practice. To keep the comparison homogeneous, we based all calculations on practices for which we had information on all the relevant variables. This method prevented, for example, a practice being in the analysis based on practice size but being omitted from the analysis based on census region because its region was missing.
For the practices with information on all the relevant variables, we also performed an ordinary least squares regression. Multivariate regression measures the individual impact of each variable studied while statistically controlling for the effects of all the other variables included in the analysis. Thus, multivariate regression is the best technique we have for studying which variables really affect workload.
Independent variables in the regression were the natural logarithm of FTE radiologists in the practice, census region, practice type, practice setting, and size of geographic area served. The dependent variable was the natural logarithm of the annual number of procedures performed by the practice in 1998-1999. We performed this regression on responses from the 1995-1996 and the 1998-1999 surveys to compare the effects of practice characteristics in the two survey years.
The coefficients from this regression measure the multiplicative factor by which each independent variable affects the annual number of procedures performed by a practice. The coefficient of the logarithm of the number of FTE diagnostic radiologists measures the percentage of change in procedures performed for each 1% of change in the number of radiologists in the practice. All other variables, such as region, are categoric variables. For them, we created dummy variables for each value except a reference category. For example, the South was chosen as the reference region, and dummy variables were created for the Northeast, the Midwest, and the West. The category with the highest frequency of occurrence was chosen as the reference category. The effect of each of the other categories was measured relative to this reference category.
We used a p value of less than 0.05 (two-tailed) as the test of statistical significance of the difference in mean workload across years. For the regression analysis, we used the more conservative p value of less than 0.01 as the test of significance. We did this because we used several categoric variables in the analysis, and any one category of a categoric variable (e.g., some one region) might have been significant at a p value of less than 0.05, but the dummies for all the categories may not have been collectively significant at the same level. Using the more stringent p value of less than 0.01 partly corrected for this issue with categoric variables. To further confirm our results, we checked for joint significance of all the categories that made up each categoric variable (e.g., all the regions combined) with an F test; we reported results in the text and significance in the table only if the entire variable was significant at least at a p value of less than 0.05.
The survey yielded measures of the numbers of procedures performed per FTE radiologist. The Medicare Physician/Supplier Procedure Summary data [7] were used to supplement this information with measures of the complexity of each procedure. Complexity was measured in terms of professional component RVUs per procedure, which we calculated for all of radiology and separately by technique. In addition, we calculated the percentage share of procedures by technique for the procedures in the 1998 Medicare files.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In diagnostic-radiology-only nonacademic multiradiologist private practices, the most common type of practice, diagnostic radiologists performed an average of 13,600 (SE = 300) procedures per year. Those working in the radiology departments of nonacademic multispecialty groups averaged 12,400 (SE = 1000) procedures annually; those in primarily academic settings averaged far fewer procedures, 9400 (SE = 600) annually. Diagnostic radiologists in solo practices averaged 14,400 (SE = 900) procedures annually. Diagnostic radiologists in practices that also have radiation oncologists averaged 12,100 (SE = 700) procedures annually per FTE radiologist. Except for academic groups, within each of these categories of practices, the number of procedures at the 75th percentile of workload exceeded the number at the 25th percentile by approximately 50% and, more typically, exceeded it by approximately 70%. For academic groups, the 75th percentile averaged 25% more procedures. Fewer than 10 practices reported working for "government" or in "other" settings, and we do not report data for those practices.
Overall, in practices with more than one radiologist, the number of procedures per FTE radiologist increased by a statistically significant 8.5% since 1995-1996. Moreover, an increase occurred for every category and subcategory of practice, as Table 1 shows, although many of the increases were not statistically significant.
Tabular Analysis of Variation in Workload
Table 1 also addresses the
relationship between workload and practice characteristics such as size and
location. One difficulty in such an analysis is correlation of characteristics
that might affect workload. For example, academic groups are typically large
and tend to be located in the Northeast and in urban areas. So a low average
workload in the Northeast might be solely a result of the concentration of
academic groups there rather than a true regional characteristic. To avoid
such correlations leading to confusion about which factors have an effect on
workload, we limited most of Table
1 to diagnostic-radiology-only nonacademic multiradiologist
private practices. This category, often labeled simply "private
radiology groups," is by far the most common category.
Among these practices, no obvious pattern of workload differences by practice size was seen. Diagnostic radiologists in the states in the Northeast averaged the fewest procedures per FTE radiologist, 12,500 (SE = 400). The greatest number of procedures per FTE diagnostic radiologist was performed in the Midwest, 15,400 procedures (SE = 700). In 1995, the regional pattern was broadly similar. Diagnostic radiologists in practices that function in both hospital and nonhospital settings averaged 13,900 procedures per year (SE = 400). Those in hospital-only practices averaged 13,400 (SE = 400) procedures per year, which is not significantly different. But those in practices that practiced only in imaging centers and offices averaged only 11,100 (SE = 1300). The pattern in 1995-1996 was similar. We did not find significant variation in workload by type or size of geographic area that a practice served.
The subcategories considered in the previous paragraph are relatively homogeneous. They are all composed of private nonacademic nonsolo diagnostic-radiology-only groups, which are then divided into subcategories that are homogeneous by region or practice setting. Nonetheless, considerable variation exists in workload in each homogeneous subcategory. (Each of these subcategories corresponds to a row in Table 1, and considerable variation can be seen among the practices that constitute each row.) Specifically, the number of procedures at the 75th percentile of workload is typically at least about 40% greater than that at the 25th percentile.
Results from Multivariate Regression
The results of the multivariate regression are shown in
Table 2. Controlling for all
other factors, diagnostic radiologists in academic practices performed about
one-fourth fewer procedures than radiologists in nonacademic practices in both
1998-1999 and 1995-1996.
|
Understandably, the greater the number of FTE radiologists in a practice, the greater the number of procedures the practice performs. A 1% increase in the number of FTE radiologists in a practice resulted in a 0.98% increase in the number of procedures performed in 1998-1999, which was not significantly different from a one-for-one linear increase in size. In contrast, in 1995-1996, for each 1% increase in size, a 0.88% increase occurred in the number of procedures performed, indicating that, with other variables being equal, larger practices performed somewhat fewer procedures per FTE radiologist.
After controlling for other factors, in 1998-1999, practices in the Midwest performed significantly fewer procedures than did those in the South (the reference region). In 1995-1996, on the other hand, practices in the Northeast and the West performed significantly fewer procedures than those in the South.
In 1998-1999, no significant difference was seen in workload based on practice setting, but in 1995-1996, practices that operated exclusively in nonhospital settings performed about one-fourth fewer procedures than those that operated in both hospital and nonhospital settings.
Finally, none of the variables describing the type of geographic location of a practice had a significant impact on workload in 1995-1996 or 1998-1999.
Change in RVUs per Procedure
In 1998, the overall average number of professional component RVUs per
radiology procedure (measured using the 1998 Medicare files) for procedures
performed by diagnostic radiologists on Medicare patients was 0.78
(Table 3), which represented an
increase of 4% since 1995, when the value was 0.75. No difference was noted in
the average RVUs per procedure between procedures performed at hospitals and
those performed at offices or imaging centers.
|
Overall average RVUs per procedure can increase as a result of either or both of two kinds of shifts: a shift in the mix of procedures performed so that a larger share of total procedures uses techniques with relatively high RVUs per procedure, or a shift within each technique to procedures having higher RVUs. The increase in RVUs per procedure between 1995 and 1998 was overwhelmingly due to the first factor—that is, a shift to higher RVU techniques. The average number of RVUs per procedure is at most one third as much for general radiology as for any other technique (Table 3), and general radiology procedures declined from 74.0% of all procedures to 70.6%. In contrast, CT and MR imaging procedures, which average a relatively high number of RVUs per procedure, increased at a particularly rapid rate, growing from 12.9% of all procedures to 15.3%. Angiography and interventional procedures, which have the highest average RVUs per procedure, had the second fastest rate of increase. On the other hand, within techniques, the average number of RVUs per procedure declined for every technique, with CT and MR imaging as the sole exceptions.
The increase of 4% in average RVUs for each procedure between 1995 and 1998 was, on average, 1.3% per year. (This number is calculated as the compounded average per year that results in the 4% change for the 3-year period 1995-1998.) This change was less than the change between 1991 and 1995, when the increase was 2.1% per year [5].
Overall Increase in Workload
With radiologists performing an average of 12,800 procedures annually, and
procedures averaging 0.78 professional component RVUs, the average
radiologist's annual workload was approximately 10,000 professional component
RVUs in 1998-1999.
The combination of an 8.5% increase in procedures over 3 years and a 4%
increase in RVUs per procedure meant the average radiologist's workload
increased by approximately 13%, measured in RVUs, over 3 years (108.5% x
104% 
113%). We estimate the total FTE of radiologists was increasing by
approximately 1.6% annually
[2], which would make the total
increase in the work performed by all radiologists in the United States over
the 3-year period approximately 13.8% measured in procedures and 18.3%
measured in professional component RVUs, which translates into 4.4% annual
growth in procedures and 5.8% annual growth in RVUs. Growth and aging of the
United States population together generate an increase of approximately 1.2%
annually [9,
10]. Accounting for this 1.2%
related to population growth and aging, the procedures per age-standardized
American increased at approximately 3.2% annually, and RVUs per
age-standardized American increased at approximately 4.5% annually.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Interpretation of Findings
Our data show that for every category of practice group (large or small,
academic or nonacademic), a substantial variation in workload exists, with
diagnostic radiologists at the 75th percentile of workload typically
performing at least 1.5 times as many procedures as those at the 25th
percentile (Table 1). Because
of these wide variations, the mean or the median should not be used as a
standard or norm that groups believe they must match.
We estimate that procedures per age-standardized American increased at 3.2% annually and RVUs increased 4.5% annually. This measure of increase is sometimes called "technologic progress" because it is due to advances in medicine rather than to growth or aging of the population. Elsewhere, we estimated [2] that in the period 1995-1998, technologic progress for the Medicare population was 2.5% annually in procedures and 4.1% in RVUs. The estimates coincide with each other fairly well.
Our findings of increasing workload per radiologist reinforce evidence of a shortage. Note, however, that an increase in workload does not necessarily mean an equivalent increase in the number of hours worked. It is possible that radiologist productivity (measured as number of procedures or RVUs performed per hour) also increased.
As was found in the 1995-1996 survey, diagnostic radiologists in academic groups averaged substantially fewer procedures annually than other diagnostic radiologists (Tables 1 and 2), presumably because academics spend more of their time in activities other than clinical work, particularly in research and teaching. Although the number of procedures performed in academic practices increased between 1995-1996 and 1998-1999, the absolute number of procedures performed in those practices in each of those years was smaller than the number performed in nonacademic practices.
Study Strengths and Weaknesses
Like all survey data, the data from ACR's 1999 Survey of Practices have
strengths and weaknesses. The principal strengths are that the sample was
large (n = 970) and carefully drawn, the survey had a high response
rate (70%), and responses were weighted accurately to reflect the total
population of radiologists. Also, when findings from other studies can be
compared, the results generally reinforce each other.
However, weaknesses remain. Most notably, more than 15% of those responding reported implausibly small workloads per year. We could not discern any systematic pattern in these. We used a standard data cleaning technique to eliminate outliers and thereby dismissed these responses as errors made by respondents. (We rechecked data entry from the survey forms into our data files and did not find that to be the source of the outliers.) It is possible that we are mistaken and that these extremely low responses, indicating fewer than 4000 procedures per FTE radiologist annually, correctly report practices' operations. If so, our estimates of workload are too high.
Second, the weighting process corrects for differential response by practice size and region, and it is not obvious that nonresponders to the survey were anything but a random subsample of the practices that we contacted. Nonetheless, it is possible that, differentially, the busier practices did not respond. Sunshine and Burkhardt [11] described this problem in practices' response regarding RVUs. If this were the case, our estimates would be too low.
Third, despite the large size of the survey, the standard errors—which measure how much uncertainty exists in our estimates because we used a sample—are nontrivial, and this limits our ability to detect relatively small differences.
Fourth, the Medicare Physician/Supplier Procedure Summary data [7] provide accurate, detailed, and complete information about Medicare services but may not be representative of the non-Medicare patients who constitute about two thirds of radiologists' workload. The procedure mix across techniques in Medicare may not be the same as the procedure mix across the population of non-Medicare patients. Trends over time, such as the 4% increase in average RVUs per procedure between 1995 and 1998, might broadly generalize to the non-Medicare population. However, some reports on limitedscale commercial insurance data suggest that the growth rate of MR imaging utilization might have been much greater than for the Medicare population (Poenitszke A, personal communication, 2001). If so, the growth rate of average RVUs per procedure would have been larger among the non-Medicare population.
Finally, data that measured radiologists' workload in RVUs, not by the number of procedures, might be more informative and accurate than the data we obtained, but our survey did not collect this information. Practices' ability to report RVUs is limited, so seeking this information results in poor response rates and, possibly, a bias with respect to which practices respond [11, 12].
Acknowledgments
We thank the radiology practices that responded to our survey; their
responses make information available to the entire profession. We thank Chris
Hogan for the meticulous initial testing, editing, and documentation of the
survey data; and we thank William T. Thorwarth and J. Bruce Hauser for reading
our manuscript and offering their comments.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Meghea and J. H. Sunshine Retirement patterns and plans of radiologists. Am. J. Roentgenol., December 1, 2006; 187(6): 1405 - 1411. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Sunshine, and C. Meghea, How could the radiologist shortage have eased? Am. J. Roentgenol., November 1, 2006; 187(5): 1160 - 1165. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. S. Lewis, J. H. Sunshine,, and M. Bhargavan A portrait of breast imaging specialists and of the interpretation of mammography in the United States. Am. J. Roentgenol., November 1, 2006; 187(5): W456 - W468. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Matin, D. W. Bates, A. Sussman, P. Ros, R. Hanson, and R. Khorasani Inpatient Radiology Utilization: Trends over the Past Decade Am. J. Roentgenol., January 1, 2006; 186(1): 7 - 11. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Merewitz and J. H. Sunshine A Portrait of Pediatric Radiologists in the United States Am. J. Roentgenol., January 1, 2006; 186(1): 12 - 22. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Sunshine, R. S. Lewis, and M. Bhargavan A Portrait of Interventional Radiologists in the United States Am. J. Roentgenol., November 1, 2005; 185(5): 1103 - 1112. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Bhargavan and J. H. Sunshine Workload of Radiologists in the United States in 2002-2003 and Trends Since 1991-1992 Radiology, September 1, 2005; 236(3): 920 - 931. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. T. Bramson and G. A. Taylor SOS: Can We Save Pediatric Radiology? Radiology, June 1, 2005; 235(3): 719 - 722. [Full Text] [PDF]
|
|
|
|
|
|
M. Bhargavan and J. H. Sunshine Utilization of Radiology Services in the United States: Levels and Trends in Modalities, Regions, and Populations Radiology, March 1, 2005; 234(3): 824 - 832. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Vertinsky and B. Forster Prevalence of Eye Strain Among Radiologists: Influence of Viewing Variables on Symptoms Am. J. Roentgenol., February 1, 2005; 184(2): 681 - 686. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Hogan and J. H. Sunshine Financial Ratios in Diagnostic Radiology Practices: Variability and Trends Radiology, March 1, 2004; 230(3): 774 - 782. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. D. Saketkhoo, J. H. Sunshine, A. M. Covey, and H. P. Forman Findings in 2002 from a Help Wanted Index of Job Advertisements: Is the Job-Market Shortage of Diagnostic Radiologists Easing? Am. J. Roentgenol., August 1, 2003; 181(2): 351 - 357. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
