| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Georgetown Public Policy Institute, 3600 N St., N.W., Ste. 200, Washington, DC
20007.
2 American College of Radiology, 1981 Preston White Dr., Reston, VA
20191-4397.
Received August 27, 2001;
accepted after revision October 30, 2001.
Address correspondence to J. M. Mitchell.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We surveyed radiologists who completed training in 1997 and obtained 487 relevant responses. Multiple regression analysis was used to identify the independent effects of characteristics of the physician, his or her job and employment search, and market area characteristics of his or her practice locality.
RESULTS. Academic starting salaries were, other things equal, 6% below private practice. Residency-only graduates had incomes 7% below a typical fellowship income. Only a few fellowship fields garnered incomes that were significantly different from the typical income. More managed care in a locality was associated with lower income, and a higher percentage of elderly in the locality was associated with a higher income. We found no statistically significant (p < 0.05) effects of sex, job location constraints, local per capita income, local cost of living, or (generally) graduate quality as measured by the ranking of a graduate's residency program.
CONCLUSION. The determinants of income are multiple and varied, including physician characteristics, such as field of subspecialty training; job characteristics, such as academic versus private practice employment; and market area characteristics. However, the study yielded as many puzzling, negative findings, such as the lack of effect of physician quality or of even severe locational constraints, as positive, expected findings.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The failure to study intraspeciality determinants of income is regrettable because it leaves several important questions in radiology unanswered, such as "How much has managed care depressed incomes?" "Are women radiologists paid less than men?" "Do jobs with a negative feature—for example, an undesirable location—pay more to compensate?" and "Do graduates of better training programs command a higher salary?"
Our study attempted to address this gap in knowledge and answer such questions by ascertaining the full range of factors that affect the starting salaries of newly trained diagnostic radiologists.
New graduates are a particularly attractive group to study. In general, the greatest difficulty encountered in studies identifying the determinants of earnings is addressing the effects of quality and entrepreneurship. Both determinants presumably have very large effects but are extremely difficult to measure [5]. For new graduates, quality should be relatively well reflected by the quality of the just-completed training program, whereas entrepreneurial ability is largely unknown to potential employers, unlike the situation with physicians who have a track record in practice. Thus entrepreneurship should have little effect on the salary offered.
Predominantly using data gathered by the American College of Radiology (ACR), we evaluated the following four types of factors that are likely to be significant determinants of the graduate's income: characteristics of the graduate, such as quality and subspecialty field; the graduate's geographic constraints in his or her job search; characteristics of his or her job; and economic characteristics of the market area (locality) where the graduate is employed that might affect income, such as managed care's market share and factors affecting the demand for health insurance.
Our data (the latest available to us) are for 1997 graduates. Since then, the employment market has greatly improved (from the graduates' perspective) so that current graduates receive more employment offers and garner higher salaries. However, questions of what factors account for differences in income remain unchanged and important, and, unlike the situation with average salaries, there is no evidence that there has been a change.
Our analysis is based on multiple regression analysis. This technique measures the independent effect of each factor being considered after statistically controlling for the effects of all other factors studies. For example, the analysis shows not whether the average salary of women is less than that of men, but rather answers the question, "If the radiologists' qualifications are the same, their jobs are similar, and their localities are similar, are women paid less than men?" This is usually the more analytically appropriate and policy-relevant question to be asking.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The methodology of the 1997 survey of graduates was similar to that used in 1995 and 1996; extensive details of this methodology have been reported in previous articles [18, 21, 32]. All 1997 graduates of radiology training programs were surveyed approximately 6-12 months after the usual June graduation date. Graduates who did not respond to the initial mailing were sent as many as three additional surveys. The response rate for both residency and fellowship graduates was 65% (767/1172 and 540/834, respectively). The response rates were similar to those obtained in the two prior surveys. The item response rate (i.e., the proportion of respondents who provided usable answers to specific questions) was high—typically in excess of 90%. Item response for questions regarding income and job satisfaction was somewhat lower—generally, approximately 75-80%. The analyses presented are based on the combined sample of the 153 responding residency graduates who went directly from training to employment rather than going into a fellowship and the 540 fellowship graduates who responded to the survey.
Data on the quality of diagnostic radiology residency programs were obtained by having three senior radiologists whose professional activities had given them extraordinary expertise in the quality of these programs provide independent, subjective ratings of each residency program in the United States. Each rater placed each program into one of three to five quality categories. (Numeric ranking from 1 to 200 of the approximately 200 programs would, obviously, have been impractical.)
Data on health maintenance organization (HMO) enrollment were obtained from interstudy and the area resource file. Data on other characteristics of the market area (per capita income in the county and the percentage of elderly residents in the county) were obtained from the area resource file. As a proxy for the cost of living in the market area, we used total payroll expenses per full-time equivalent hospital employee, a statistic reported in the annual survey conducted by the American Hospital Association.
Statistical Analyses
We used a multivariate regression analysis to investigate factors that may
affect the starting salaries of radiologists who recently completed either
residency or fellowship training programs. Because actual starting salary was
reported by categories of income (e.g., $150,000-159,999), we transformed the
categories into a continuous variable by assigning each category the midpoint
for its range. For example, the midpoint for $150,000-159,999 was $155,000.
Because income tends to be skewed, we used the log transformation of the
actual starting salary of diagnostic radiology residents or fellows in 1997 as
the dependent variable.
Our analysis included a number of factors that are likely to account for differences in starting salaries among newly trained radiologists. Reasons for including each of these factors and the anticipated effects of each are discussed briefly.
Candidate and Job Characteristics
Previous research has found that female physicians earn lower salaries than
their male counter-parts and that this gap in earnings can be attributed to
the fact that female physicians work significantly fewer hours than male
physicians
[37,38,39].
It is also well documented that part-time employment is associated with
considerably lower wage compensation than full-time work. Hence, we included a
variable to indicate whether the radiologist worked part-time, which would
indicate the disparity in earnings between part-time and full-time workers,
and we included a variable for gender. The latter would identify income
differences linked to gender beyond those that can be accounted for by
part-time versus full-time work as well as other variables in the model.
Academic radiologists, on average, perform a lower volume of clinical work than their nonacademic counterparts [40,41,42], presumably because of teaching, research, and other nonclinical activities. Because physicians' incomes are primarily derived from revenues from patient care, we anticipated that academic radiologists would earn lower incomes than those employed in private practice.
Jobs with undesirable attributes are likely to have some impact on compensation, but it is unclear whether they should be expected to result in higher or lower salaries. Economists argue that employers must pay higher wages to attract individuals to work in employment situations with undesirable features. For example, the wages for radiology positions in undesirable locations may need to be higher to compensate for the drawbacks of the location. Many noneconomists contend that the reverse is true; namely, that jobs with less attractive features are unattractive all around, including salary, so they typically pay less.
Our survey instrument included a series of questions regarding the presence or absence of 11 possibly undesirable job characteristics, including the following: employed by an HMO or its physician group; locum tenens is the main work; practice has too few patients; job is considered temporary by the employer; radiologist is employed by a nonradiology practice; radiologist is employed as a physician administrator; radiologist is employed by an equipment manufacturer or supplier; the position is not a radiology job; the job is primarily during the night or weekened; the job is not on the partnership track; and the job is situated in a seriously undesirable location. If the radiologist's job had one or more of these characteristics, the survey asked him or her whether he or she liked, disliked, or was neutral about the characteristics. We constructed a dummy variable equal to one if the respondent's job had one or more of these possibly unattractive job characteristics and the respondent disliked the job characteristic.
Quality of the Candidate
One would anticipate that higher quality graduates are able to obtain
higher starting salaries. We measured a graduate's quality by the rating of
the graduate's residency program. This approach to measurement reflects the
view that, on average, higher ranked programs do a better job of training
their residents, or the more highly regarded residency programs attract and
accept the most talented medical students. The ratings by our three experts
were combined to construct a mutually exhaustive set of eight rankings ranging
from "very high" to "poor." In general, we found
reasonably good concordance among the rankings of the three raters. The simple
correlation between each pair of raters was, on average, approximately 0.6.
Programs with more than minor discordance among the three ratings were placed
in the category "mixed"; only 16% of graduates in our study
completed a program identified as such. Because the "very high"
category is the reference group in our analysis, we anticipated that all other
ranking categories would have negative effects on starting salary. This
finding would indicate that fellows who graduate from the best programs
command higher salaries relative to those who complete lower ranked
programs.
Radiology Subspecialty
Certain subspecialties within radiology are widely recognized to generate
higher billings and receipts than others, and as a consequence, they may
command higher starting salaries. Two such fields are neuroradiology and
angiography or interventional radiology. Thus relative to all other fellowship
fields, we expected to find that starting salaries are higher for
neuroradiologists and interventional radiologists. Our regression analysis
included a series of dummy variables to identify the fellowship field of each
graduate (Table 1). We used
general body imaging as the reference category to which salaries of graduates
of other fellowship programs were compared, because general body imaging may
be regarded as a generalist fellowship, and it is also a common one.
|
One would expect that radiologists who complete only a residency, going into posttraining employment without a fellowship, will earn substantially less than those who are fellowship trained. We treated these residency-only radiologists, in essence, as a fellowship category for which the fellowship consisted of "none," and the regression therefore measures how their salaries compare with those of radiologists trained in general body imaging.
Job Search Constraints
Constraints on the job search can limit the potential employment
opportunities that are available to newly trained radiologists. Therefore,
such constraints are likely to have negative effects on a radiologist's
starting salary. Our model included three such job search constraints. One
constraint was that the graduate had a spouse who also had to find a job in
the same locality. A second constraint was that the radiologist chose to
restrict his or her job search to a particular area or type of place, rather
than to consider opportunities available throughout the United States. The
third constraint was that the graduate insisted on being within commuting
distance of the institution where he or she had just completed training. This
last constraint represents a particularly severe constraint and would thus be
expected to have a larger impact on starting salary.
Managed Care
Recent evidence suggests that HMOs are more likely to enter and grow in
markets that exhibit high medical care costs, excess hospital capacity, and
large numbers of physicians relative to the population
[4]. These conditions enable
HMOs to negotiate larger discounts from providers, and as a result, they can
offer a less expensive insurance package to employers. We use HMO penetration
(the percentage of the population enrolled in HMOs) in the radiologist's
metropolitan statistical area as a proxy for the strength of managed care in
general. Because managed care plans tend to negotiate large discounts with
providers, we expected that higher HMO enrollment in the market area would
result in lower starting salaries for radiologists relative to areas with
comparatively low HMO enrollment.
Other Market Area Characteristics
Our analysis included a control variable to distinguish between
metropolitan areas on the one hand and, on the other hand, rural or nonurban
regions using standard Census Bureau definitions of which counties are part of
metropolitan areas (including suburbs). We anticipated that starting salaries
for diagnostic radiologists would be lower in rural, nonurban areas relative
to larger, more densely populated cities and suburban areas.
The analysis captures variation in the demand for health care with two variables, including per capita income in the physician's county and the percentage of the population in the county 65 years old or older. The first variable recognizes the fact that better-off individuals are more likely to have health insurance and that, for any given health status, they use more health care. The second variable recognizes that the elderly use, on average, approximately four times as much care as younger persons.
Finally, the regression model included a variable to recognize variations in the cost of living associated with particular market areas. One would anticipate that starting salaries are higher in market areas where the cost of living is higher. As a proxy for cost-of-living differentials, the model included total payroll expense per full-time equivalent hospital employee for the county in which each radiologist practiced.
Descriptive results throughout this article are reported as estimated percentages. Results from the multivariate regression are expressed as the percentage of increase or decrease relative to the reference group for dummy variables and as elasticities for continuous variables, such as HMO market penetration. The elasticity indicates the percentage increase (decrease) in starting salary that is associated with a 1% increase (decrease) in an independent variable (e.g., HMO penetration). When reporting results of the multivariate regression, we reported findings only for those independent variables that were statistically significant at a p value of less than 0.05.
Because item response rates in the survey were less than 100%, data on some variables used in the analysis were missing for some radiologists, and these graduates had to be omitted from the analysis. The actual number of graduates included in the multivariate analysis was 487 (359 fellows and 128 residents).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A relatively large percentage reported job search constraints. Thirty-five percent of graduates reported that they had a spouse who also had to find a job in the same area, and 66% indicated that they had a non—spouse-related geographic constraint on locations they would consider. Further, 17% would look for a job only within commuting distance of the location of their just-completed training program.
On average, 22% of the population in the metropolitan statistical area where each graduate practiced was enrolled in HMOs; the standard deviation was 16%. Approximately 16% of graduates were located in nonurban areas. The mean per capita income in the radiologist's county was just under $27,000.
In a direct, nonregression comparison, we found that the mean starting salary for radiology fellowship graduates who had full-time positions in private practice was approximately 9% higher than their counterparts who took an academic position. In private practice, full-time fellowship graduates earned on average approximately 10% more than diagnostic radiologists who had completed only residency training. These comparisons involve only simple direct comparisons and do not take account of the effects of other variables.
Multivariate Regression Estimates
The F statistic for the regression shows that the overall model fit was
highly significant at a p value of less than 0.01, which implies that
the independent variables are jointly significant predictors of starting
salaries for newly trained radiologists. The adjusted R2
statistic from the regression indicates that the explanatory variables
accounted for 28% of the variation in starting salaries of diagnostic
radiologists; this finding is regarded as a substantial amount of variation
considering that the regressions are based on individual-level data.
Table 2 reports the percentage
change in starting salary associated with each characteristic that was
significant at the 5% level after controlling for the effect of other
variables included in the analysis.
|
As anticipated, working part-time has a substantial negative effect on income; other things being equal, radiologists who work part-time earned on average 27% lower starting salaries relative to those employed in full-time positions. On the other hand, gender and working in a job with at least one or more undesirable attributes that the radiologist actively disliked did not have statistically significant independent effects on starting salaries. Also, contrary to expectations, employment constraints did not have significant effects on starting salaries.
Consistent with expectations, we found that radiologists who accepted academic positions earned, other things being equal, 6% lower starting salaries, on average, than those who chose employment in private practice. With one exception, we found that the ranking of the graduate's residency program had no significant effect on starting salary. The exception was that radiologists who were the product of a residency program with a poor ranking earned starting salaries that were 8% higher than those who graduated from the most highly regarded residency programs, other things being equal.
The field of subspecialization within diagnostic radiology resulted in a few significant starting salary differentials. Fellowship graduates trained in either angiography or interventional radiology or in neuroradiology had, other things being equal, 6% and 8% higher starting salaries, respectively, than those who were fellowship-trained in general body imaging (the reference category). In contrast, those fellowship-trained in pediatric radiology had starting salaries 9% lower than those fellowship-trained in general body imaging. None of the other fields of fellowship specialization impacted salaries significantly. Radiologists who completed only a residency had, other things being equal, a starting salary averaging 7% lower than those who were fellowship-trained in general body imaging.
Finally, we turn to the impact of market-area characteristics on starting salaries of newly trained diagnostic radiologists. Increased HMO market penetration had a significant negative effect on starting salaries. The regression results indicate that a radiologist who accepted a position in a market area where HMO enrollment was 45% (double the sample mean), other things being equal, received a salary 4% lower than his or her colleague who worked in a market area where HMO penetration was 23%, the actual sample mean. Differences in the percentage of elderly residents in the radiologist's county also were associated with significant salary differences. To illustrate, if the percentage of elderly residents in the county were to double from the actual sample mean of 11.7-23.4%, the expected corresponding difference in starting salary, holding all else constant, would be approximately an 8% increase. In contrast, neither practicing in non-urban areas, nor variations in county per capita income, nor variations in the county cost of living had a statistically significant effect on starting salaries.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Changes in the radiology job market since 1997 (the graduation year covered by our data) clearly have greatly increased average starting salaries. However, the question we investigated—namely, "What factors give rise to differences in starting salaries?"—remains as important as ever, and there is no particular reason to think that findings, such as the apparent lack of financial recognition of quality, have changed. The near ubiquity with which practices now complain of the difficulty of attracting the radiologists they need suggests that patterns of differences in pay have not changed.
Physician, Job Search, and Job Characteristics
The 27% lower income of part-timers relative to full-time radiology
graduates closely matches a previous finding that, for radiologists of all
levels of experience, part-timers average approximately 30% fewer hours than
full-timers [14]. However, we
do not have information on work hours of those part-timers who have just
completed training, so the similarity may be coincidental.
From the standpoint of concern about possible discrimination, it is reassuring that we found that gender had no impact on the starting salary of just-graduated radiologists. This finding corroborates results of a recent cross-specialty study [1]. Hadley and Mitchell [1] found that after controlling for other confounding factors, female physicians earned about $23,000 less per year than their male counterparts, but essentially all of the salary differential could be attributed to fewer hours worked, because the difference in hourly earnings between male and female physicians was small and not statistically significant. Previous analysis of the survey of 1997 graduates found that relative to men, women radiology graduates were more likely to work part-time [32].
As expected, we found that radiologists employed in academic positions command lower salaries than those employed in private practice. However, the 6% salary difference found is much smaller than expected. Academic radiologists perform, on average, about two thirds as much clinical work as their private practice counterparts [40,41,42], and we would have expected a similarly large difference in salaries. Perhaps the explanation is that, in private practice, income of new members of the practice increases by a large amount after 2-3 years as they became partners, whereas in academia, incomes increase more slowly and by less.
Also surprising was the absence of a relationship between graduates' ability and salary. Even if residency program quality is only an imperfect proxy for graduates' ability, as is almost certainly true, one would expect a general positive relationship between the two. Instead, we found a general lack of a relationship. Moreover, the one statistically significant relationship that we found is very puzzling, that graduates of the poorest programs have, other things being equal, higher incomes than graduates of the best rated programs. This finding is not explicable by the hypothesis that graduates of the poorest rated programs enter private practice, where salaries are higher than in academia. Because we used multiple regression analysis, effects of all other variables considered in our studies, including private practice versus an academic job, were controlled for. In other words, the puzzling finding constitutes the difference in salary after controlling for what type of job the graduate has taken (and for all other factors we studied).
Equally difficult to understand is the absence of a negative relationship between any of the three job search limitations we studied and starting salary. Perhaps the absence of a relationship is explicable for the general non—spouse-related geographic limitation. Among graduates, this limitation ranged from refusing to consider only a small part of the United States to being willing to locate in only a few areas, and such a varied constraint might not have a uniform effect. However, the limitation of considering jobs only within commuting distance of one's just-completed training program inevitably rules out all but a small percentage of jobs and thus must impose stringent constraints on job choice.
Being employed in a position with a presumably undesirable job characteristic that the graduate dislikes did not affect starting salaries in either a positive or a negative way. As noted earlier, either relationship might be plausible.
The findings that only a few subspecialty fields had starting salaries that differed significantly from salaries in a generalist reference category (general body imaging), and that these differences were modest, are consistent with two important facts. First, although billings and revenues differ by a large amount—as much as 2:1—across subspecialties, the near-universal norm in radiology practices is that each partner receives an equal share of income (Muroff L, personal communication). The underlying rationale is that to obtain a hospital contract, a practice needs to be able to provide a full range of services; hospital-based services constitute about three fourths of radiologists' clinical work. Second, the ACR's annual surveys of the hiring conducted by radiology groups have measured the strength or weakness of the employment market for each subspecialty by comparing the percentage of available positions that groups were able to fill in each subspecialty with the overall percentage filled [9, 12, 20, 26]. Similarly, the annual surveys of graduates have examined the percentage of graduates trained in each subspecialty who had any serious postgraduation employment difficulties [18, 21, 32]. These surveys consistently find few, if any, fields in which the percentages differ from the overall average by a statistically significant amount. Moreover, the fields in which a statistically significant difference occurs change from year to year, indicating that the phenomenon is transient, rather than a long-term strength or weakness of the employment market in a particular subspecialty. In this context, the survey of 1997 hiring [26] found the market for pediatric radiologists considerably weaker than average, consonant with our finding of relatively low average starting salaries that year.
Market Area Characteristics
Our finding that higher HMO penetration is associated with lower starting
salaries for radiologists provides support for the hypothesis that managed
care in general, or at least HMOs in particular, has reduced radiologists'
income. Other studies have found similar effects. Hadley and Mitchell
[1] analyzed data for a
national sample of physicians in 1990 and found that a doubling of HMO
penetration in the physician's market area was associated with a 7-11%
reduction in physicians' annual earnings. Simon et al.
[4] analyzed the change in
median physician income across states between 1985 and 1993 as a function of
the change in HMO penetration and reported that this relationship was highly
significant and negative for the sub-sample consisting of radiologists,
anesthesiologists, and pathologists.
A higher percentage of elderly residents, which increases the demand for care, resulted in higher starting salaries for radiologists. Although this finding was understandable, it is puzzling why higher per capita income, which also increases the demand for care, was not associated with higher starting salaries. It was also surprising to find that starting salaries were not higher in areas where the cost of living is relatively high.
Study Limitations
The response rate to the survey of 1997 graduates is higher than usually
attained in surveys of physicians, and studies that have analyzed data from
the ACR's previous surveys of graduates indicated that they were free of
nonresponse bias [18,
21]. Nonetheless, our findings
may be subject to sample selection bias; such bias would be problematic if the
relationships among the variables are different among the one third of
graduates who did not respond to the survey than among the two thirds who did
respond.
More important, our study covers only one specialty, only physicians just beginning practice, and only during 1 year. The determinants of physician incomes for other groups of specialties, at other points in time, and for different levels of experience may be different. Some managed care plans tend to encourage the use of primary care services and impose restrictions on the use of specialists' services. Thus, for primary care physicians greater managed care penetration may result in higher starting salaries, the reverse of the pattern found for radiologists. We have already noted possible differences within radiology for more experienced physicians. Finally, the employment market for diagnostic radiologists clearly has moved to a situation of much greater shortage now than when our data were gathered, and changes of this type, in addition to increasing the income of all graduates, may have changed what factors affect beginning salaries and to what extent.
Some of the variables used in the analysis are imperfect measures of what we seek to study. For example, we would like to know the effect of managed care overall, but small area data are available only for HMOs. Because data on enrollment in preferred provider arrangements and point-of-service plans are not available, we may have captured the effects of HMOs only. Furthermore, we would be more confident in our indicators of residency program quality if the correlations among the three experts' ratings were higher. In general, better measures might yield somewhat different results.
The number of graduates each year is too small to permit investigation of the distinct effect of each of the 11 possibly undesirable job characteristics or of having more than one. Finally, our findings are based on cross-sectional data. Thus our results may not accurately portray what would happen in response to changes over time in market-level variables, such as HMO penetration.
In conclusion, our overall finding that the determinants of physicians' income are many and varied implies that future research needs to consider at least as broad a range of factors as we have analyzed here. The failure to address a wide range of factors is a major limitation of prior research. It is our hope that future studies will lead to a more thorough understanding of the factors that determine physicians' income and will help to explain why our study has produced as many negative and puzzling findings as anticipated positive findings.
Acknowledgments
We thank Harvey L. Neiman and William T. Thorwarth, Jr., for thoughtful
advice.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
