| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Diagnostic Radiology, Yale University School of Medicine, 20
York St., Tomkins East 2, New Haven, CT 06520.
2 Research Department, American College of Radiology, Reston, VA.
Received December 5, 2007;
accepted after revision December 7, 2007.
Address correspondence to N. R. Nayak
(nikhil.nayak{at}yale.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We analyzed data from the American College of Radiology's 2003 Survey of Radiologists, a stratified random-sample survey that guaranteed respondents' confidentiality and achieved a 63% response rate with a total of 1,924 responses. Responses were weighted to make them representative of all radiologists and radiology practices in the United States. We used univariate analysis and multiple logistic regression.
RESULTS. In 2003, PACS, wet-reading telephone lines, film-hanging staff, and templates (standard report language) were each used in practices that encompassed approximately half of U.S. radiologists. In contrast, only 42% of radiologists were in practices that used nurse practitioners or physician assistants for tasks beyond what technologists may do, and only 18% were in practices that used speech recognition software (SRS). Twenty-one percent of radiologists were in practices reported to have neither film-hanging staff nor PACS. The percentage of practices (as opposed to radiologists) that used various technologies ranged from 13% for SRS to 49% for templates. Multiple logistic regression showed that, other factors equal, academic practices were particularly likely to use some of the technologies and solo practices and other small practices were particularly likely not to have some of the technologies.
CONCLUSION. Most operational technologies are fairly widely diffused, but a surprising number of radiologists work without some basic supports.
Keywords: digital images • practice of radiology • speech recognition software • teleradiology • wet-reading telephone lines
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article therefore examines six purportedly productivity-increasing operational technologies: PACS; wet-reading telephone lines (WRLs); film-hanging staff; nurses, physician assistants (PAs), or both to perform tasks beyond the scope of radiology technologists; speech recognition software (SRS); and templates (standard report language). WRLs are telephone lines that refer to an accessible dial-in number that a clinician can call, enter a medical record number, and hear the full dictated report before or after it has been transcribed. We call these technologies "operational" because they deal with how radiology practices and radiologists do their work rather than with what is done directly to the patient. Details about the use of a seventh operational technology, teleradiology, have been presented previously [2].
Our principal focus is on the prevalence of each of the six technologies and on which characteristics of practices seem to foster using each of the six. Also, because some of the technologies seem partial substitutes for one another—for example, if a practice acquires PACS, one would expect it to dispense with film-hanging staff, we also report the prevalence of having at least one of such a pair and analyze the extent to which, in keeping with our hypothesis, the pair actually functions as substitutes for one another.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The ACR survey asked radiologists whether their practices used each of a number of technologies, as enumerated earlier. It also asked radiologists questions about the characteristics of their practices and about themselves. Practice characteristics included in the survey and analyzed in this study are practice type, practice size, census region, care settings served, practice location, and practice ownership. Table 1 shows the analysis categories used for each characteristic. Characteristics of individual radiologists included in the survey and in this study are sex, age, if working full- or part-time, whether subspecializing, and the number of physically separate locations at which a radiologist works.
|
Statistical and Data Analysis Methods
Before this study, survey responses were weighted to make them
representative of all radiologists and of all radiology practices in the
United States, as previously described
[5,
6]. Univariate analysis was
used to measure the prevalence of each of the six technologies among all
radiology practices, among multiradiologist and solo practices, and in various
practice categories (Fig. 1 and
Table 1). Multiple logistic
regressions were run to identify the independent effects of different practice
characteristics, statistically controlling for all other factors, on the
likelihood of a radiology practice using a particular technology
(Table 2). Univariate analysis
was performed to determine the prevalence of each technology among the total
population of radiologists and for various categories of radiologists
(Fig. 1 and
Table 3).
|
|
The results of the multiple logistic regressions are, as is standard, reported as odds ratios—that is, the ratio between the probability that an event does occur and the probability that the event does not occur. Odds ratios greater than 1 imply that an event is more likely to occur for a particular group compared with the reference group, whereas odds ratios less than 1 indicate that the event is less likely to occur compared with the reference group.
We also analyzed prevalence of three technology combinations (Table 4): first, PACS and film-hanging staff; second, SRS and WRLs; and, third, PACS and teleradiology. We report the prevalence of having at least one of the technologies of the pair for all radiology practices, multiradiologist practices, solo practices, and the total population of individual radiologists (Fig. 2). We then investigate our hypothesis that the members of each pair are to some extent substitutes for each other. For technologies A and B, if P(A) is the probability of having technology A (Table 4, column 1) and P(B) the probability of technology B (Table 4, column 2), then we can infer the following about the prevalence of having both (Table 4, columns 3 and 4) on the basis of whether A and B are independent technologies, substitutes, or complements. If A and B are independent or unrelated, then P(A and B) = P(A) x P(B). If A and B are substitutes, P(A and B) < P(A) x P(B) because there is little or no reason for a facility to have both. If A and B are complements, then P(A and B) > P(A) x P(B) because the presence of one technology increases the usefulness or ease of installing the other.
|
|
Odds ratios in the logistic regressions were considered significantly
different from 1 when p 
0.10. Statistical tests at p
0.10 are the standard in the health services research literature. Readers
who prefer the p 0.05 significance level in regressions can find
it indicated in Table 2 and
also find the information needed to use an even more restrictive definition of
what is regarded as statistically significant.
Data analysis was performed with Intercooled Stata 9.1 (StataCorp LP) for Microsoft Windows and Microsoft Excel 2002. We used Stata's svy estimators, which use pseudolikelihood estimators and Taylor-series expansion, respectively, for logistic regression and proportions, to produce standard errors and p values suitably adjusted for weighting and survey stratification.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Thirteen percent of all radiology practices, including 13% of multiradiologist practices and 12% of solo practices, used SRS in 2003 (Fig. 1 and Table 1). Eighteen percent of radiologists were in practices with SRS (Table 3). The logistic regressions show that, holding other factors constant, academic practices had approximately four times the odds of private radiology practices of using SRS (Table 2). The regressions also indicate that, other factors equal, practices in suburbs of large metropolitan areas had approximately four times the odds of using SRS of practices in main cities of small metropolitan areas, whereas practices in the suburbs of small metropolitan areas had only 0.38 the odds of using SRS of the same reference group. In addition, other factors equal, practices entirely owned by private outsiders had three times the odds of using SRS of practice member-owned practices.
Forty percent of all radiology practices, including 47% of multiradiologist
practices and 21% of solo practices, used WRLs
(Fig. 1 and
Table 1). Fifty-five percent of
radiologists were in practices with WRLs
(Table 3). The logistic
regressions indicate that, other factors held constant, solo practices had
approximately one fourth the odds of using WRLs of large practices (practices
with 
15 radiologists) and practices with 2-4 radiologists had
approximately half the odds (Table
2). Practices serving only nonhospital settings had approximately
one fourth the odds of using WRLs of practices serving both hospital and
nonhospital settings.
Forty-four percent of practices, including 48% of multiradiologist practices and 32% of solo practices, used clerical staff to hang films (Fig. 1 and Table 1). Fifty-five percent of radiologists were in practices with such staff. The logistic regressions show that, other factors equal, practices with 2-4 radiologists had approximately half the odds of using film-hanging staff of large practices (Table 2). Practices serving only hospitals had approximately two thirds the odds of using film-hanging staff of practices that served both hospital and nonhospital settings.
The use of nurses or PAs (or both) was found in 24% of all practices, including 29% of multiradiologist practices and 10% of solo practices (Fig. 1 and Table 1). Forty-two percent of radiologists were in practices that used these types of personnel. The regressions indicate that, holding other factors constant, practices serving nonhospital settings had only approximately one fifth the odds of using nurses or PAs (or both) of practices that served both hospital and nonhospital settings and that hospital-only practices had approximately two fifths the odds (Table 2). The regressions also show that, other factors equal, nonmetropolitan practices had approximately half the odds of using such staff as did practices in the main cities of smaller metropolitan areas and that practices in the suburbs of small metropolitan areas had approximately one third the odds.
Forty-nine percent of practices, including 51% of multiradiologist practices and 42% of solo practices, used templates or standard language in reports to reduce the work of dictation (Fig. 1 and Table 1). Fifty-four percent of all radiologists were in practices that used templates. The logistic regressions show that, holding other factors constant, practices in suburbs of large metropolitan areas had 1.66 the odds of using templates of the reference group (practices in main cities of smaller metropolitan areas) and that practices that operated in varied types of locations had one fourth the odds (Table 2). Furthermore, government-owned practices had approximately one eighth the odds of physician-owned practices of using templates, and practices that were partly or entirely owned by private outsiders had approximately half the odds.
Finally, 6% of practices, including 5% of multiradiologist practices and 9% of solo practices, had none of the six technologies we studied; 3% of radiologists were in these practices (Tables 1 and 3). Logistic regression found no practice characteristics significantly associated with having none of the six technologies.
Pairs of Partly Substitutable Technologies
In 2003, 63% of practices, including 69% of multiradiologist practices and
45% of solo practices, used PACS, film-hanging staff, or both; 79% of
radiologists were in these practices (Fig.
2). Thus, 37% of all practices had neither technology for
relieving radiologists of the burden of handling films themselves and 21% of
radiologists were in these practices.
Forty-eight percent of practices, including 55% of multiradiologist practices and 30% of solo practices, had SRS, WRLs, or both; 65% of radiologists were in these practices. Thus, 52% of practices had neither of these means of making reports available to referring physicians without the conventional process of transcriptionists word processing the radiologist's dictation. Thirty-five percent of radiologists were in these practices.
Seventy-three percent of practices, including 80% of multiradiologist practices and 55% of solo practices, had teleradiology, PACS, or both; 87% of radiologists were in these practices. Thus, 27% of practices had neither of these means of digitally moving images from the locale where they were acquired. Thirteen percent of radiologists were in these practices.
The actual prevalence of having both members of each of the pairs of technologies is close to the prevalence that would be expected if having one technology is unrelated to having the other and not substantially lower than would be expected if the technologies were substitutes (Table 4). Indeed, for the combination of teleradiology and PACS, the actual prevalence indicates some degree of complementarity. Chi-square tests of association between the presence of technologies confirmed the statistical significance of these impressions (Table 4).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In contrast, approximately four fifths of radiologists were in practices that had teleradiology in 2003 [2]. A detailed analysis of teleradiology use showed it was most commonly used to transmit images to radiologists' homes, thereby easing the burden of call [2]. Thus, easing the burden of call seems to have had a higher priority among radiologists than any of the efficiencies to which the technologies included in this study are supposed to produce.
However, when film-hanging staff and PACS are considered together as different technologies that achieve the same goal—namely, relieving radiologists of the task of hanging and taking down films, we also find approximately four fifths of radiologists benefiting. Indeed, we were surprised that one fifth of radiologists apparently were handling films in 2003. Providing either PACS or film-hanging staff to radiologists hanging films seems an obvious opportunity for improved productivity.
Somewhat similarly, the approximately one third of radiologists who work without WRLs or SRS also represent an important opportunity for improvement, but here the opportunity is for improvement in the speed with which findings are made available to referring physicians. It is unclear whether any improvement in radiologists' productivity per se—that is, in the amount of work a radiologist completes per hour—would result. However, in terms of the health system as a whole, a change that improves service to referring physicians without causing deteriorations in radiologists' productivity is definitely desirable.
Because the logistic regressions show the effect of each factor included in the regression while statistically controlling for the effects of all other factors included, they are helpful in disclosing which relationships may well be causal rather than purely coincidental. For a number of technologies, we found that solo practices and, sometimes, other small practices were relatively unlikely to have the technology. This finding may be related to the considerable capital costs of some technologies, most notably PACS. We also found a tendency for academic practices to be more likely to have technologies. Perhaps the academic environment fosters an interest in technology. Finally, one specific relationship we found—low prevalence of WRLs among practices that function only in nonhospital settings—seems to have an obvious explanation. Such practices do not have urgent or emergent patients, so speed in getting information to referring physicians is of lower importance than in other practices.
At first it seems odd that, contrary to our hypothesis, the two technologies in the pairs we studied do not seem to function as substitutes. Rather, we mostly found that whether a practice has one technology of a pair seems unrelated to whether it has the other. For example, a practice that has PACS is no less likely to have film-hanging staff than a practice that does not have PACS. This puzzling situation probably occurs because most practices function at multiple sites and may have different arrangements at different sites. For example, a practice may have PACS at some of its hospitals but may have film-hanging staff at other locations.
The complementarity we find between PACS and teleradiology is that practices with PACS are more likely to have teleradiology than practices without PACS. This finding is understandable. With PACS, images are already in digital form, an integral piece of a teleradiology system. In the absence of digital images, additional steps are necessary to make a teleradiology service operational.
Strengths and Limitations of the Study
The primary strengths and limitations of the 2003 Survey, including its
improvements over previous surveys, have been discussed elsewhere
[1,
3]. The strengths of this study
include a high response rate, large sample size, sophisticated statistical
weighting process that accurately represents a national sample, and meticulous
cleaning of data to improve quality.
One limitation of the survey (and of all surveys) is a certain level of sampling variability as measured by standard errors. Standard errors tend to be large for categories with relatively few respondents. There may also be nonresponse bias with respect to characteristics not considered in developing the weights. Despite quality measures and data cleaning, there may be some errors in the data.
The technologies studied are widely believed to increase productivity; for example, SRS and WRLs presumably increase productivity because they decrease the number of reading room interruptions due to telephone calls. However, this study does not examine whether the technologies actually increase productivity, nor does it examine their effect on quality or patient outcomes.
In conclusion, our results indicate that each of the practice characteristics studied was significantly associated in logistic regression with the use of at least one of the six technologies we investigated. In particular, practice type and practice size seem to be the most important variables affecting whether practices have technologies; other factors equal, academic practices are more likely to have some of the six technologies that we studied than are other practices, whereas small practices, especially solo practices, are less likely. Also, practices in the least urban settings—the suburbs of small metropolitan areas and nonmetropolitan areas—repeatedly were, other factors equal, less likely to have a technology than practices located elsewhere. We plan further research to try to understand the reasons for these patterns.
Because the 2003 Survey is the first to accurately quantify the prevalence of operational technologies in a carefully representative national sample and to analyze the practice characteristics and radiologist characteristics associated with variations in prevalence, it can serve as a useful benchmark. Future observations can be compared with it as these technologies evolve or become antiquated and as new technologies become available.
Acknowledgments
We thank all those who responded to the ACR's 2003 Survey of Radiologists.
By contributing the time needed to complete the questionnaire, they have
helped make important information available to the entire profession.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
