| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Research Department, American College of Radiology, 1891 Preston White Dr.,
Reston, VA 20191.
2 Department of Diagnostic Radiology, Yale University, New Haven, CT
06520.
Received October 25, 2002;
accepted after revision January 23, 2003.
Address correspondence to Y. S. Cypel.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. In May 1999, the American College of Radiology surveyed 970 practices by mail using a 65-item questionnaire. A final response rate of 66% was achieved. Weighting was used to make responses representative of all radiology practices in the United States.
RESULTS. Approximately 4400 radiology practices existed in the
country; 39% were located in nonmetropolitan or rural areas. The greatest
percentage of practices (85%) was private nonacademic radiology practices. The
greatest percentage of practices (43%) served only hospitals. However, solo
practices were particularly likely (p 
0.01) to serve only
nonhospital sites. Of all multiradiologist practices, 87% were completely
owned by their members, but approximately two thirds of academic
multiradiologist practices had outside owners. Seventy percent of
multiradiologist practices interpreted images from outside clinics or groups;
these averaged 3.2 outside clinics or groups for which they interpreted
images.
CONCLUSION. The basic characteristics of radiology practices in 1999 are described. Good agreement was seen with findings obtained from other data sources.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The ACR developed a 65-item questionnaire that was initially mailed via regular mail to these 970 practices in May 1999. In addition to the initial mailing, four follow-up mailings were conducted between June and November 1999. The ACR received a total of 708 responses. From these, approximately 90 were excluded because of incompleteness of data, because the response was from someone not intended as the target group (e.g., residents in training responded but were not considered part of the target population), or because they were duplicates. The result was a final response rate of 66%. The final number of unweighted practices used for analysis was 645.
Statistical and Data Analysis Methods
Responses were weighted so that weighted statistics would be representative
of the answers that would have been received if all practices in the United
States had been surveyed and responded. We weighted according to 28 strata
based on four census regions and seven practice-size categories (the latter
were measured by the total number of full- and part-time radiologists at the
time of the survey). These strata represented only those practices that
reported performing some diagnostic radiology services;
radiation-oncology-only practices were excluded throughout the analysis
because very few of these practices responded to the survey. Additional
information about how weights were applied to responses is presented elsewhere
[18].
Data from sample surveys are subject to sampling variability [19]. The usual measure of sampling variability is the standard error. There is a 99% probability that the true value of a population parameter (such as a population mean) lies within approximately 2.6 standard errors of the corresponding statistic for a sample drawn from a normal population. Unweighted sample sizes are presented in some of the tables containing descriptive statistics so that standard errors may be calculated. For percentages in tables, the standard error (in percentages) can generally be approximated by taking the square root of p(1 – p) / n, where p = percentage and n = unweighted sample size. Other descriptive statistics that were computed included frequencies (counts), means (averages), and percentiles (25th, 50th [median], and 75th). Descriptive statistics were not shown in a table on the number and distribution of radiology practices in the United States.
The information used in producing the results presented in the text and tables was based on the practice's interpretations of the survey questions. For example, a practice was recorded as "academic" because it self-reported "primarily academic" as its practice type.
The 1999 Survey of Practices questionnaire obtained information about a practice's self-report of whether it was one of three practice types: diagnostic-radiology-only practices, radiation-oncology-only practices, or mixed practices (those practices performing both diagnostic radiology and radiation oncology procedures). This article includes only diagnostic-radiology-only practices. Mixed practices are the subject of another planned article. We use the term "radiology practice" to refer to diagnostic-radiology-only practices (either solo or multiradiologist). Practices that contain more than one diagnostic radiologist are referred to as "multiradiologist practices." Practices containing only one member are referred to as "solo practices."
When information about practice location is discussed in the text and tables, the types of locations are center city or suburb of a large metropolitan area (area population > 1 million), center city or suburb of a smaller metropolitan area (area population, 50,000–1 million), nonmetropolitan city or town or rural area (area population < 50,000), and varied locations (no one type is principal; area population not specified). Also, "any outside ownership" is defined as "complete outside ownership" or "partial outside ownership."
We report key findings from regression analyses—namely, logistic regression and multiple regression analysis: logistic regression to determine what practice characteristics could predict a dichotomous dependent variable (such as whether the practice served a hospital or not), and multiple regression analysis to determine what practice characteristics could predict a cardinal dependent variable (such as the total number of settings served by a practice) [20].
Generally, results of regression analysis provide better information about which relationships among variables are real and meaningful than do tables containing descriptive statistics, because regression analysis measures the effect of each variable after statistically controlling for the effect of all other variables included. We used explanatory (independent) variables (such as practice type, size, census region, general practice setting served, and location) that were found to be important in earlier published analyses [5, 11–13, 16] in the regression models.
We used polynomial regression techniques to better understand the relationship between practice size (one of the independent variables examined) and the specific dependent variable used in the regression models [21]. We added the square of the practice size variable [(practice size)2] to each regression model developed. We did so because the association between practice size and the dependent variable may be better approximated by a curve than by a straight line; for example, a characteristic may be most common for medium-sized practices rather than for small or large ones.
The regression analyses were conducted on all radiology practices, both solo and multiradiologist. In contrast, for those tables presenting descriptive statistics, the analyses were mainly conducted on multiradiologist practices and excluded solo practices. We limited the analysis of solo practices because the number of solo practices that responded to the 1999 survey was low, making the results of cross-tabulations with other practice characteristics statistically unreliable. Thus, in the descriptive statistics tables, estimates for solo practices are presented for practice type only.
For any analysis involving the practice characteristic "general settings served," responses in which the calculated total number of settings served by a practice did not fall within 80–125% of the reported total were excluded to minimize error resulting from invalid or missing data. Categories of practice characteristics with fewer than 20 responding practices are not presented in the descriptive tables. For example, "government" and "other" are not presented under "practice type" because their unweighted sample sizes were below 20.
Because we make many comparisons—which tends to increase the number
of results judged statistically significant although they are, in fact, due to
random chance—we used p 0.01 as the measure of statistical
significance for all statistical tests, rather than the more commonly used,
but less demanding, criterion of p 0.05. All data analyses were
conducted with statistical software (release 8.01, SAS Institute, Cary, NC).
If a finding was statistically significant, it was reported as such.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Of those practices that reported their type (weighted sample size = 4394), approximately 183 practices (4%) self-reported being primarily academic practices. Private nonacademic radiology practices were the predominant practice type, constituting approximately 3713 practices (85%). Private nonacademic multispecialty practices and government practices constituted approximately 329 (7.5%) and 43 (1%) practices, respectively. (The remaining 3% was reported as "other practices.")
Of those practices that reported their location (weighted sample size = 4398), approximately 1433 practices (33%) were practicing primarily in large metropolitan areas. Approximately 1149 practices (26%) practiced primarily in smaller metropolitan areas, and approximately 1733 (39%) were located in nonmetropolitan or rural areas. Approximately 84 practices (2%) were in varied locations with no type of location being primary.
Of those practices that reported the types of general settings they served (weighted sample size = 4326), approximately 43% reported that they served only hospitals. In contrast, 23% of radiology practices reported that they served only nonhospital settings. Approximately 34% reported that they served both types of settings.
Practice Ownership
Table 1 shows descriptive
statistics on practice ownership by various practice characteristics. Of all
multiradiologist practices (i.e., excluding solo practices), 87% were
completely owned by their members, 10% were entirely owned by outsiders, and
3% had part-outside ownership. Of all multiradiologist practices with some
outside ownership (either part or complete outside ownership), the most
frequently reported types of outside owners were universities or medical
schools (20%) and not-for-profit hospitals (30%). Smaller percentages
(5–14%) of multiradiologist practices with outside ownership reported
that physician practice management companies, government hospitals, or
for-profit hospitals were outside owners. (The total of the percentages
exceeds 100 because practices sometimes reported multiple outside owners.)
|
A high percentage (59%) of academic multiradiologist practices were entirely outsider owned, with the greatest percentages of outside ownership contributed by universities or medical schools (58% of those with outside ownership) and not-for-profit hospitals (26%).
With the effects of all other variables controlled for using logistic regression, academic practices, private nonacademic multispecialty practices, government practices, and "other" practices were significantly more likely to have any outside ownership (i.e., either complete or partial outside ownership) than private, nonacademic radiology practices. Practices in the center city of large metropolitan areas were significantly more likely to have any outside ownership than those located in the center city of smaller metropolitan areas.
Practice Location
Of all multiradiologist practices, 24% were located in the center city of
smaller metropolitan areas (Table
2); 9% were located in the suburbs of smaller metropolitan areas.
Approximately 15% were situated in the center city of large metropolitan
areas, 18% were in the suburbs of these cities, and 32% were in
nonmetropolitan and rural areas. Three percent reported being in varied
locations with no type of location predominant. Multiradiologist practices
with 15 or more members and academic practices were more concentrated in the
center city of large metropolitan areas than were practices of other sizes and
types.
|
After controlling for other practice characteristics using logistic regression, the following significantly increased the likelihood of practices being in the center city of large metropolitan areas: being in the West (vs being in the South) and increasing practice size. Government practices were significantly more likely to be located in the suburbs of large metropolitan areas than were private nonacademic radiology practices when the effects of other practice characteristics were controlled. Practices in the Northeast were also significantly more likely to be located in the suburbs of large metropolitan areas than were those in the South. Practices that served only nonhospital settings were significantly more likely to be in the suburbs of large metropolitan areas than were those serving both types of settings.
After controlling for all other practice characteristics using logistic regression, solo practices were significantly less likely to be in the center city of smaller metropolitan areas than were private nonacademic multiradiologist practices. Increasing practice size significantly increased the likelihood of a practice being in the suburb of a smaller metropolitan area.
After controlling for other characteristics using logistic regression, increasing practice size significantly decreased the likelihood of a practice being located in a nonmetropolitan city or town or in a rural area. This effect was less when the practice was larger. Moreover, practices serving only nonhospital settings were significantly less likely to be in rural areas than were practices serving both hospitals and nonhospital settings.
General Practice Settings Served
Fifty-six percent of multiradiologist practices served both hospitals and
nonhospital settings, 36% served only hospitals, and 9% served only
nonhospital settings (Table
3).
|
After controlling for the effects of other characteristics using logistic regression, solo practices were more likely to serve only nonhospital settings than were private nonacademic radiology practices. Radiology practices in nonmetropolitan and rural areas were more likely than practices in cities of smaller metropolitan areas to serve only hospitals. After controlling for the effects of all other characteristics, for practices with up to approximately 35 members, the likelihood that the practice served both types of sites increased significantly with increasing size. However, beyond approximately 35 members, this likelihood decreased with increasing size.
Hospitals Served by Practices
Teaching hospitals.—Overall, 16% of multiradiologist
practices served private university or teaching hospitals and 6% served
governmental (public) university or teaching hospitals
(Table 4). The mean total
number of private university or teaching hospitals served by multiradiologist
practices that did serve such hospitals was 1.2. The mean total number of
governmental (public) university or teaching hospitals served by practices
that served these hospitals was 1.0. Controlling for other factors using
logistic regression, the likelihood that a practice served teaching hospitals
significantly increased with increasing practice size for practices of up to
approximately 40–45 members, after which the likelihood decreased.
Practices in the West were significantly less likely to serve any type of
teaching hospital than were those in the South. Practices located in the
center city of large metropolitan areas were significantly more likely to
serve teaching hospitals than were those in the center city of smaller
metropolitan areas.
|
Private nonteaching not-for-profit hospitals.—Sixty percent of multiradiologist practices served private nonteaching not-for-profit hospitals (Table 4). The mean total number of hospitals of this type served by practices that served such hospitals was 1.8. After controlling for the effects of all other practice characteristics using logistic regression, rural practices were much more likely to serve such hospitals than were practices in the center city of smaller metropolitan areas. Also, the likelihood that the practice served private nonteaching, not-for-profit hospitals increased with size for practices of up to approximately 25 members; the likelihood decreased with size for practices having more than approximately 25 members.
For-profit (investor-owned) nonteaching hospitals.—Twenty-one percent of multiradiologist practices served for-profit nonteaching hospitals (Table 4). On average, multiradiologist practices that served such hospitals served a total of 1.3 hospitals of this type. After controlling for the effect of other practice characteristics using logistic regression, solo practices were less likely to serve these hospitals than were private nonacademic radiology practices. Also, practices in the Northeast, Midwest, and West were less likely to serve for-profit, nonteaching hospitals than were practices in the South.
Public nonteaching nonfederal hospitals.—Approximately 25% of multiradiologist practices served public nonteaching nonfederal hospitals (Table 4). The mean total number of such hospitals served by practices that served this kind of hospital was 1.5.
Federal nonteaching hospitals.—Approximately 8% of multiradiologist practices served federal nonteaching hospitals (Table 4). The mean total number of such hospitals served by practices that served this type of hospital was 1.4. After controlling for the effects of all other practice characteristics using logistic regression, government practices were significantly more likely to serve federal nonteaching hospitals than were private nonacademic radiology practices.
Nonhospital and Other Settings Served by Practices
Nonhospital multispecialty group facilities.—Of all
multiradiologist practices, 28% served nonhospital multispecialty group
facilities (Table 5). The mean
total number of such sites served by practices that served this kind of site
was 1.8. Of multiradiologist practices that are the radiology portion of
private nonacademic multispecialty practices, 87% served nonhospital
multispecialty group facilities. After controlling for the effects of all
other practice characteristics using logistic regression, private
multispecialty practices were more likely to serve nonhospital multispecialty
group facilities than were private nonacademic radiology practices.
|
Nonhospital radiology offices, radiation oncology offices, clinics, freestanding centers.—Of all multiradiologist practices, 49% served nonhospital radiology offices, radiation oncology offices, clinics, and freestanding centers (Table 5). The mean total number of such nonhospital sites served by practices that served any of them was 2.2. After controlling for the effects of all other practice characteristics using logistic regression, the likelihood that private nonacademic multispecialty practices served nonhospital offices, clinics, and freestanding centers (other than nonhospital multispecialty group facilities) was significantly less than that for private nonacademic radiology practices. The likelihood of serving these nonhospital settings was also significantly less for practices in the Midwest than for those in the South, as well as for practices located in nonmetropolitan and rural areas than for those in the center city of smaller metropolitan areas.
Mobile vans.—Of all multiradiologist practices, 26% served mobile vans (Table 5). The mean total number of mobile vans served by practices that served this type of site was 1.4. After controlling for the effects of all other practice characteristics using logistic regression, larger practices were significantly more likely to serve mobile vans than were smaller practices.
Outside clinics or groups for which a practice interprets images.—Of all multiradiologist practices, 70% served outside clinics or groups for which they interpreted images (Table 5). The mean total number of these outside clinics or groups for which interpretation was done among practices that served them was 3.2.
Total Number of Settings Served by a Practice
Table 6 presents descriptive
statistics on the total number of settings (both hospitals and nonhospital
settings) served by practices. The mean total number of settings served by
radiology practices (including solo practices) was 2.7; the means for all
multiradiologist practices and solo practices were 3.3 and 1.5,
respectively.
|
A multiple regression analysis, with the total number of settings served by a practice as the dependent variable and practice characteristics as the independent variables, indicated that, after controlling for the effects of all other practice characteristics, the total number of settings served increased with increasing practice size. For small practices, one site was added for approximately every five additional radiologists in the practice, but the increase was less for larger practices. Private nonacademic multispecialty practices served 1.25 fewer total settings than otherwise similar private nonacademic radiology practices. Practices that served only hospitals served 2.2 fewer total settings than otherwise similar practices that served both hospitals and nonhospital sites. Practices that served only nonhospital sites served 1.7 fewer total settings than otherwise similar practices that served both hospitals and nonhospital settings.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Good agreement existed in the findings obtained from the 1999 and 2000 surveys with respect to the basic distribution and size of radiology practices in the United States. The greatest percentage of practices was in the South. In both, the greatest percentage of practices constituted solo and two- to four-member practices, and percentages decreased with increasing group size.
Academic practices constituted only a small percentage of radiology practices—approximately 5% in each of the two surveys. Private nonacademic practices constituted by far the greatest percentage of radiology practices in each of the two surveys, with private multispecialty practices constituting the next largest percentage for both surveys. The percentage of government practices in the 1999 and 2000 surveys was low; the percentage of radiology practices in the 2000 survey that were government-based was approximately 6%, whereas it was approximately 1% in the 1999 survey. The much lower 1999 survey estimate probably results from how the sample was obtained. In 1999, practices were randomly selected from a listing of those that responded to a request for information on Medicare compliance program guidelines. Federal government practices may have been much less responsive to compliance issues because, in general, federal practices cannot bill Medicare. This fact may have resulted in fewer government practices being included in the original sample frame in the 1999 survey.
Like findings from the 1999 survey, findings from the ACR census conducted in 1992 indicated that the largest percentages of practices were located in the South and were nonacademic private radiology practices. Only a small percentage of practices was academic.
The patterns observed in the types of hospital settings served by multiradiologist practices based on the 1999 survey statistics paralleled statistics regarding hospitals obtained from the American Hospital Association (American Hospital Association 2000 Annual Survey, unpublished data generated at our request). For instance, American Hospital Association statistics in 2000 showed that approximately 64% of for-profit hospitals are located in the South and logistic regression analysis of 1999 survey data indicated that practices in the Northeast, Midwest, and West were less likely to serve for-profit, nonteaching hospitals than practices in the South. Moreover, estimates from the United States Census Bureau show that the greatest percentage changes in population (total resident population) for different regions of the United States, comparing 1990 with 2000, were in the South and in the West [22].
Substantive Findings from the 1999 Survey of Practices
Many findings from the 1999 Survey of Practices were expected. For
instance, one would expect that private nonacademic practices and solo
practices might tend toward less external ownership than that in other
practice types, because practices of this kind represent independent ventures;
this is what we found. Not only did academic practices have the greatest
percentage (59%) of complete outside ownership, the major contributor to this
type of outside ownership was reported to be universities or medical schools
(58%), as might be expected.
We had several less-expected findings. For example, outside ownership was particularly likely for practices located in the center city of large metropolitan areas even after the effects of practice type, such as whether the practice was academic or not, and all other factors were considered. Also, solo practices were not more or less likely to serve nonhospital offices, but were less likely to serve only hospitals, than were private nonacademic multiradiologist practices.
Strengths and Limitations of the Survey
The principal strengths of the survey were intensive follow-ups to ensure a
good response rate, telephone calls to ensure that questions designated as
"core questions" were answered and consistent, the receipt of a
relatively large number of responses, and detailed weighting to make the
responses representative of all practices in the United States. The generally
good agreement of this survey with the 2000 survey bolsters confidence in its
soundness.
However, as with all surveys, the 1999 survey was based on a sample and therefore has the limitations of sample surveys. Some of these limitations are observational in nature, such as ambiguities in the wording of survey questions. Others are related to sampling error, coverage, and nonresponse [23].
We tried to reduce observational error by minimizing question ambiguities, asking about issues that are relevant to our target population, and providing reasonable response options [24, 25]. Yet these errors may still occur because terms were largely self-defined and responses were self-reported.
Coverage error may exist because few government practices were part of the sample frame (listing from which the sample was selected); thus few government practices had an opportunity to respond to the survey. Similarly, the sample frame did not include practices that contained no ACR members. Thus, even weighted responses may not be fully representative of the United States population of radiology practices.
Response rates were comparable to other surveys of practices conducted by ACR [11]. An analysis of potential nonresponse bias was not conducted because only minimal information was obtained about the nonrespondents [18]. We tried to minimize nonresponse at the onset by conducting four follow-up mailings and then corrected for nonresponse after data collection by weighting adjustments. However, the near-total lack of response to the survey from radiation-oncology-only practices is puzzling and troubling. The omission of these practices from this article may have eliminated problems related to their nonresponse.
Acknowledgments
We thank the following for their thorough reviews, comments, and guidance
in the development of this paper: J. Bruce Hauser, ACR Board of Chancellors;
Barbara Schepps, ACR Committee on Radiologist Resources; and J. Frank Wilson,
ACR Board of Chancellors.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  J. H. Thrall Changing Relationships between Radiologists and Hospitals * Part I. Background and Major Issues Radiology, December 1, 2007; 245(3): 633 - 637. [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]
|
|
|
|
|
|
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]
|
|
|
|
|
|
D. B. Larson, Y. S. Cypel, H. P. Forman, and J. H. Sunshine A Comprehensive Portrait of Teleradiology in Radiology Practices: Results from the American College of Radiology's 1999 Survey Am. J. Roentgenol., July 1, 2005; 185(1): 24 - 35. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Sunshine and Y. S. Cypel Types of Procedures Performed by Diagnostic Radiology Practices: Past Patterns and Future Directions Am. J. Roentgenol., November 1, 2004; 183(5): 1193 - 1201. [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]
|
|
|
|
|
|
C. H. Deitch and J. H. Sunshine The Relationship of Managed Care To Business, Professional, and Organizational Aspects of Radiology Practices Am. J. Roentgenol., January 1, 2004; 182(1): 29 - 38. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. S. Cypel and J. H. Sunshine Characteristics of Mixed Diagnostic Radiology-Radiation Oncology Practices Am. J. Roentgenol., September 1, 2003; 181(3): 643 - 647. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
