Socioeconomic Cynthia
H. Deitch,
PhD
Recruiting The 1991
Ronald
#{149}
G. Evens,
MD
Jonathan
H. Sunshine,
#{149}
R
ECENT studies
present
conflicting
how the supply of radiologists compares with the need. In a study that was published in the mid-1980s, the Graduate Medical Educalion National Advisory Committee (GMENAC) projected a massive surplus of more than 6,000 radiologists by 1990 (1). In contrast, one recent study from within the radiology community suggested that we entered the 1990s with a shortage of 5,000 radiologists and that the shortage would grow (2). To cast light on the surplus/shortage situation, the American College of
recruit
Radiology
smaller
ones.
Groups
filled 1,348 (71%) of these positions, while 552 positions remained unfilled, which was 2.8% of groups’ total staff. Remaining vacancies ranged from 5% of total staff in the Northeast and 3% in the Midwest to 2% in the South and 1% in the West. The percentage of positions filled was higher (76%) in nonmetropolitan areas than elsewhere. The authors conclude that the supply of radiologists is very close to balancing the number of positions available; the disparity is about 1/io of some numbers mentioned previously and is probably shrinkbig. Also, recruiting is not less suc-
cessful small
in nonmetropolitan cities
than
areas
and
estimates
gists,
Index terms: Radiology and radiologists Radiology and radiologists, departmental management #{149} Radiology and radiologists, socioeconomic issues
some
1
From
185:47-51
the Research
Department,
American
College of Radiology, 1891 Preston White Dr. Reston, VA 22091 (C.H.D.,J.H.S.); George Washington University, Washington, DC (C.H.D.); and Mallinckrodt Institute of Radiology, Washington University School of Medicine, St Louis (R.G.E.). Received May 6, 1992; accepted June 3. Address reprint requests to J.H.S. 0
RSNA,
1992
the
principal
pro-
practices
may experience
difficulty in recruiting than others. The second research question considered therefore is, for what types of positions and in what geo-
locations most
are unfilled
posi-
likely?
The interview included specific, factual questions about the number of radiologists in the group, the exact number of radiologists
the group
recruited
compare
actually
hired
with
the number
for, and
the
in the 1991 hiring season. Interviewers also elicited more subjective responses about decisions to recruit or not and reasons for not filling positions. To facilitate comparison with an earlier survey, the interview also induded the subjective question, “How does the number of radiologists . . . in your group number
that would
be ideal?” without requesting that the “ideal” figure meet any “reality test” such as actual recruiting activity. Both the quantitative and the qualitative data are discussed
in this
artide.
The ACR maintains a listing, which is updated on a continuing basis, of all radiologic practices in the United States. This data base was used to design the
sampling sample. tically
frame and to select the actual sample design ensured statisadequate representation of groups The
in each
of four
east, Midwest,
U.S.
census
South,
and
regions
West),
(North-
as well
MATERIALS and
AND Sampling
METHODS Methods
The research department of the ACR conducted a national sample telephone survey of radiology group practices in May and June of 1991. The interview focused on the 1991 hiring season-that is, hiring done of radiologists who would start their new positions in the summer of 1991. (Many of those hired completed training programs that ended June 30.) Whenever possible, the interviewer spoke directly to the president of the group. Established
telephone
ods were
employed.
survey
research
meth-
as
each of three different group sizes, based on number of radiologists in the group (two to four, five to seven, and eight or more). Within each of the 12 sampling strata (four regions with three size categories within each region), groups were selected randomly for telephone interviews. Only group practices were sampled; solo pracrices were omitted. The overall response rate was 78% . Attempts were made to contact 192 radiolrepresentation
within
each
region
of
ogy groups for telephone interviews. these, a total of 150 interviews were
Survey 1992;
(ACR),
greater
graphic
elsewhere.
of
fessional organization of radiologists in the United States, conducted a stratified national sample of radiology group practices, studying their actual recruitment and hiring experience in 1991. The survey was designed to evaluate two concerns. The first is whether, nationally, there is an adequate supply of radiologists to meet the recruitment needs of groups. Beyond this question of an overall nalional shortage or surplus of radiolo-
lions
Radiology
PhD
Radiologists: Hiring Survey’
Recent studies present widely varying views on the radiologist staffing situation, ranging from a projected surplus of over 6,000 radiologists to an estimated shortage of 5,000. To help clarify the situation, the Amencan College of Radiology undertook a stratified random sample survey of radiology groups’ recruitment efforts and hiring results in 1991. This survey of 192 groups achieved a 78% response rate. It found that half of all groups recruited, seeking to fill a nationwide total of 1,900 positions. Larger groups were more likely to than
Issues
Of
completed, with a minimum of 10 interviews in each of the 12 sampling strata. Only five of the 192 groups actually declined an interview (some asked to have the survey mailed instead). For most of the other 37 nonresponding groups, an appropriate person to interview was not reached despite
repeated
leave
messages.
call-backs
and
efforts
to
ACR = American College of Radiology, FFE = full-time equivalent, GMENAC = Graduate Medical Education National Advisory Committee, SE = standard error. Abbreviations:
47
The responses
and group
were
weighted,
by region
practices
size, to represent all group in the United States. The per-
centage weighted
tables in this report show the population estimates and thus
show figures sponses from States.
Table
1 Success
Radiologist-level
by Group
and Vacancy
No.
Variable
Variables
(weighted)
Central),
West. Metropolitan four categories:
population
South,
over
3 million,
is, with
100,000. Group ber of radiologists any 1991 hiring.
For type
area population
size
which their
55 60 82 88
(4)
4.9(1.8)
(4)
3.3 (1.5)
Over3milhion
512
500,000-3 million 100,000-499,999 Outside a metropolitan
721 901
area Group size (no. of radiolo-
492
ology
group,
office,
were
or nonteaching
Because “private
1.8 (1.1)
(3)
0.9(0.8)
70(4) 65 (4)
3.4(1.5) 2.5(1.3)
75 (4)
2.5(1.3)
76 (3)
2.9(1.4)
539 391
79 (3) 66 (4) 64 (4)
2.4(1.2) 3.6(1.5) 2.7(1.3)
445
71 (4)
2.6 (1.3)
2,264
71 (4) 70 (4)
2.7(1.3) 3.1 (1.4)
1,251
11 Type of group Private radiology All othei4
of the following best depractice: (a) primarily acaor teaching (b) government (city or hospital, Veterans Administration, military, etc); (c) radiologists part of a private multispecialty group; (d) private radital; and (e) other. groups answered
(3)
s)
under
interviewees
hospi-
* Success rate = no. of radiologists t Vacancy rate = (no. of radiologists Numbers in parentheses are SEs. t Includes academic, government,
Table 2 Group Level
86% of the radiology
362
Recruitment
hired/no.
recruited. Numbers no. hired)/(FTE
recruited and
and Hiring
and
other”
includes
(ii) all other
ment,
and
Data
Analysis
groups,
where
the academic,
multispecialty
recruited.
The
Percentage
“all
Variable
govern-
rate is the
of unfilled positions remaining cruiting was completed as a total need. The hiring success tal number of positions filled age of the number for which
100,000-499,999 Outside metropolitan
Group
number
1). Because estimates reports
in reporting
the results
(Table
the numbers
we present
are
based on a sample from every practice,
.05) if that difference square second
48
Radiology
48* 55*
52
79*
45
82*
63 47
70 53 69
50 40
76
size (no. of radiologists)
Type of group Private radiology All other
For the
which A
in interpreting the between two significant (P
of Groups
Region Northeast
as North
of
Recruitment
Hiring
as if we had obtained reall groups in the United
Total
known
Rates for 1991
Practices
the
plus
the
analysis
of group-level
data,
is shown
in Table 2, the dependent variables are (a) for all groups, whether or not the group recruited at all, and (b) for recruiting lions were
groups, ifiled.
whether
x2statistics
or not all posi-
(Table 3) of group-level data to assess which of the independent variables exammed (region, metropolitan area size, group size, and type of group) are statistically
significant
or not a recruiting
test the significance
are used to of differences between
categories
in Table
variables
of groups
We employ
logistic
regression
2.
analysis
tions cruited.
after
determinants
group
controlling
for all of the other
and the number Logistic
of whether
filled all posi-
regression
of positions
re-
is a standard
October
1992
Table
by metropolitan group size were
3
Factors
Affecting
Whether
Group
Recruiting Logistic
a
Parameter Variable Region:
Estimate
Northeast
west Area population Metropolitan million Metropolitan
and
Mid-1.29
(#{216}.54)*
and
(65%)
0.50 0.01
type: private radiology No. of radiologists recruited Intercept Note-Numbers * x2 significant
Of the
(0.91) (0.06)
groups,
50%
Another 35% (n just one position, (n = 260) recruited and about 5% (n
three
are SEs.
or more
percent
906)
=
groups
are
(whether
or not all positions
some
measured
independent
group
were
variables
as interval-level
data.
interval-level independent cluded number of positions
(The
variables recruited
inand
size.)
RESULTS Radiologist-level
Analysis
When the data across all groups, sents 17,527 (SE diologists
are aggregated the sample repre-
1,129) counting
with
and,
ra-
radiologists, an FTE of 18,007 (SE = 1,142) radiologists in a total
2,626
groups.
Overall,
these
of
groups
filled
the 1991 hiring rate (positions
as a percentage
as
rate (positions filled of positions recruited approximately 71% nation-
hiring
success
a percentage
for) was ally.
As shown
bers
varied
The
vacancy
in Table
1, these
substantially rate
was
num-
by region. highest
in the
Northeast (5%) and lowest in the West (1%). The hiring success rate was and
lowest highest
in the Northeast (55%) in the West (88%).
The differences in vacancy rates associated with other variables in Table 1-metropolitan area size, group size, and type of group-were not statistically significant. For the hiring success rate, some of the differences
Volume
185
Number
#{149}
1
silions
recruited
politan variate
area. The results analysis, conducted
two
regression,
or po-
the
that
filled
all positions
and
left with vacancies. that the percentage
actively
recruiting
Taof
in 1991
var-
ied slightly by region, size of metropolitan area, size of group practice, and type of group practice. tests indicate that only the differences in
x2
by group size were statissignificant (P < .01). In general, groups were more likely to re-
recruitment
lically larger
than
Table
smaller
groups.
2 also shows
of groups
that
ing succeeded openings.
were
percentage
actively
recruit-
in filling
At the
individual silions was South than
what
1991 hiring.
regression,
For region, observed
Northeast
and
and
the
South
Group groups
group
radiology
groups
fill all their
types ment,
were
positions
of groups
(academic,
or mullispecialty),
the
egory
were
other
govern70%
versus
Midwest
on
one
hand,
on the
other.
West
size
that
of community
are
this analysis with
499,999. The lion in opposed related cessful negative
of these categories from the equation,
convention with the most
consisted
is to omit the cases, which
data in Table the Northeast to the South to a reduced recruiting, as coefficient.
catin
of metropolitan of 100,000-
a population
3 show that locaand Midwest (as or West) was likelihood of sucindicated by the This difference
controlling ables listed.
significant (P < .05), for all of the other variOther variables were not
significant.
Thus,
was statistically
sion
in Table
sis, confirms Table
Subjective
logistic
regresanaly-
the univanate
finding
of
Questions
If a group
were the successful
indicated been
filled,
“What
main reasons in hiring?”
respondents, with unfilled swers
the
3, a multivariate
2.
asked
more
than
and
viewer
private likely to
the
the
not linear, we compared each metropolitan size category to all others.
level,
in fillsig-
we tested between
we observed
by
had
west. The differences by region ing all positions were statistically nificant (P < .05). “Ordinary,”
group
and
Because
silions
as at the
in
type is private radiology versus all others, as previously
all of their
level, success in filling p0higher in the West and in the Northeast or Mid-
are presented
logistic
areas
those groups ble 2 shows
of this multiby using
size was measured as FTE of practicing radiologists in the group prior to
differences
three more
might
3.
defined.
positions
in
for or size of metro-
one unfilled position, and 8% had more than one vacant slot. We sought to identify differences between groups that recruited cornpared with those that did no recruiting, and between the recruiting
un-
2.8% nationally.
posi-
by region
Technically, one must be omitted
cruit
the
of all positions)
was approximately The
season, remaining
differences
sitions recruited for, the more likely one or more positions would go unfilled. Among the 1,310 groups that did recruit, 66% successfully filled all their positions, 25% were left with
groups
reported recruiting to fill a total of 1,900 (SE = 197) openings in 1991. Of these 1,900 positions, 1,348 (SE = 168) were filled, leaving 552 (SE = 110) unfilled positions. After vacancy
for
one
in their to 49% of
is
in success
disappear (or become more salient) once we controlled for number of po-
differences
sought to fill Seventy-five
144)
groups
recruiting
groups
full-time part-time
=
10%
and 44% for groups seeking more radiologists. Thus, the
technique used for such an analysis, where there is a dichotomous dependent
for
for two positions, =
ample,
For
recruited
we addressed
differences
filling positions evident in the last column of Table 2 persisted after controlling for the other factors. For ex-
=
approximately
positions.
of the
(n
at all in 1991.
question
of the
Table
1,316) did no recruiting
0.18(0.75) -0.74(0.46) 2.38 (0.72)
in parentheses at P < .05.
2,626
next
which
logistic
Analysis
lion to fill were successful hiring efforts, compared
and
in areas
with a population of 500,000 to 3 mulion and highest (76%) in nonmetropolitan areas. It was higher (79%) for groups of two to four radiologists than for larger groups, for which it averaged approximately 67%.
Group-level
-0.54(0.68)
Group
filled)
suc-
0.40(0.74) area million
area size: FTE of group
variable
The
by
The
area over 3
500,000-3 Outside a metropolitan
Group
size
cess rate was lowest
Filled All Positions: Analysis
Regression
area
significant.
to this
that
not
the
inter-
do you
all po-
think
you weren’t Twenty-five
or 81 % of those groups positions, offered anopen-ended
question.
Although the small number of actual (unweighted) responses and the open-ended nature of the question make it difficult to generalize from the answers,
the
answers
nonetheless
are
differences area and
by size of metropolitan by size of group practice
of interest because they confirm some of the patterns evident from the more quantitative data and they suggest additional factors for future invesliga-
were
statistically
tion.
45%.
cal
This
difference significance (P
not
is close to statisti< .08). Observed
significant.
Radiology
#{149} 49
Most
of the answers contain a few themes. Thirty-six percent of those giving a reason cited location as a problem; 20% specified difficulty finding someone in a particular specialty; 20% mentioned that they had common
selective
very
criteria
or complained
to their
geographic
location
as
or only reason cited were located in the Northeast or Midwest regions. Of those groups in the South or West that mentioned location as a problem, one group in New Mexico cited “low academic performance” of candidates as the first reason mentioned and added location as a second reason; one group in Texas listed “poor location” as the third reason mentioned, after citing poor working conditions
and
heavy
workload.
None of the groups that mentioned location as a problem were located in the largest metropolitan areas, and none were in the smallest (nonmetropolitan) areas. Thus, the respondents’ perceptions of location as a problem appear fairly consistent with the geographic differences we found in the quantitative data analysis presented earlier. Some groups reported difficulties in finding candidates with the specific qualifications group desired,
and specialties while others
the
found new graduates looking for greater specialization than the group could offer. For example, one respondent indicated that the need for a general radiologist was the problem. Another suggested that the reason for not hiring was that the group performed mostly ultrasound phy, which was
and
mammogra-
limited for new grads.” One stated that it was hard to find a clinic radiologist because “the supply is limited and the money is low.” Another respondent offered that there were “not many candidates in neuroradiology.” Yet another reported that the group had been looking “for years” for an angiography/
interventional a shortage 50
#{149} Radiology
“too
specialist in that
specialty.
but found Other
wanting more exonly boardqualified candidates, and requiring a fellowship and subspecialization as their reasons for not ifiling positions. The question about ideal staffing drew
the poor qualifications of applicants; 20% indicated salary was a problem, including one who mentioned a low conversion factor; and 16% noted a less than full-fledged search-they started late or did not advertise except by word of mouth. In addition, 8% mentioned that a radiologist had been hired but backed out at the last minute. (Some groups cited more than one reason; therefore, the responses noted above total more than 100%.) All of the groups that mentioned experiencing difficulty attracting canabout
didates the first
groups perience,
mentioned considering
response
a
2,220 radiologists than groups had cruited
radiologists
showing
a wish
for
(SE = 242) more before newly rejoined them.
The
question
shortage
expect
or Balance?
Shortage,
of whether
of radiologists
there
is a
has important
for the medical commuthe public. Within the profession, it has implications for young physicians choosing a specialty, for residency programs, and for group practices, among others. The data presented herein indicate
implications nity and
that
the
approximate
supply
of radiologists
balance
with
is in
the number
of positions available. This conclusion of approximate balance is supported by a number of considerations. First, the number of unfilled posilions (approximately 550) was less than 3% of the total number of posilions, and the number of 550 is probably an overstatement. According to groups’ own reports, some of the recruitment efforts were started late or not pursued at full force. Also, on the other side of the ledger, there are likely to be some number of radiologists who would haved liked to obtain practice positions but were unable to find them. Relative to the surplus of approximately 6,000 radiologists that GMENAC predicted (1), or the 5,000-and-growing shortage that one radiologist reported (2), 550 is a small number. It is less that one-half of the annual number of newly minted radiologists emerging from residency
programs
(diagnostic
and
each year and about equal to the annual growth in the number of radiologists in the United States (3). A useful way of recognizing how small this number of 550 is is to note that 550 radiologists working full-lime are approximately equivalent to 1 hour of work a week of the radiologists currently in groups. In other words, if current radiologists worked 1 hour a week more (or just over 10 minutes a day) that would be equivalent to having approximately 500 more radiologists at work in the United States. Another indication of approximate overall balance is the fact that the vacancy rate varies widely by region. oncology)
throughout
the
United
States,
with
regional differences being relatively small. In other words, if there really were a substantial national surplus of radiologists, on the order of that projected by GMENAC, we would not
DISCUSSION Surplus,
The Northeast vacancy rate was nearly twice the national average, while the Western vacancy rate was only one-third of the national average. If a shortage or surplus were major, one might expect it to dominate
to find the shortfall
observed
in some regions. Conversely, if there really were a substantial shortage of radiologists of the magnitude some have suggested, we would not expect to find a vacancy rate of less than 1% in one major region of the country (the West). The 1991 ACR Hiring Survey offers some limited evidence that the supply of radiologists is increasingly meeting demand. In the ACR’s 1989 Group Practice Survey (4), radiology groups indicated that they would have liked 1,100 more radiologists than they had at that time. The Group Practice Survey’s question was similar to the question on the 1991 Hiring Survey about how many radiologists would be “ideal,” in that neither question was grounded in a reality test. (A standard
finding
on
surveys
is that
ungrounded questions about what respondents would like draw more expansive answers than questions based on real actions.) Comparing answers
to the
Hiring
Survey’s
ques-
lion about “ideal” staffing with its data on recruitment suggests approximately 870 would be the 1991 number that corresponds to the number of 1,100 from 1989. (For 1991, 870 equals the 2,220 “ideally” desired radiologists minus the 1,348 actual hires.) Thus,
relative
supply
seems
to demand
between
1989 and
noted, however, and time of year the two surveys,
improvement
to have
increased
by roughly
200
1991. It should that the questions were not identical and the apparent
might
be in
be a statistical
artifact.
In any case, one of us (R.G.E.) has pointed out that revenues available to radiology are likely to become considerably more constrained in future years (5). This may well decrease the demand for radiologists, with radiologists a!ready in practice choosing to work an hour (or more) longer each week-or to curtail their lime off-in order to keep their incomes up. As noted, only a minor increase in working hours would be needed to change the current apparent tiny shortfall of radiologists into a corresponding excess. October
1992
Reasons Positions
Why
Groups
Do
Not
Fifi
Although we do not find evidence of a substantial national shortage from the experience of radiology groups in 1991, we do observe some problems in the matching of available radiologists with groups seeking to fill positions.
In response
to an open-
ended question about why they were unable to fill all their positions, groups most frequently cited location. The quantitative data analyses support these open-ended responses in showing significant differences by region of the United States. On severa! different measures-percentage of positions filled, remaining vacancy rate, and percentage of recruiting groups that filled all positions-we found that recruiting was more successful in the South and West than in the Northeast and Midwest. This finding is counterintuitive. Radiologist training programs are concentrated in the Northeast and Midwest, so one would expect vacancies in these areas to be easiest to fill. Also, growth of the general population is most rapid in the South and West,
Volume
185
Number
#{149}
1
should make for a more rapidly growing need, more open posilions, and more remaining vacancies. Also counterintuitive is our finding of no empirical support for the conventional belief that nonmetropolitan areas have the most difficulty in obwhich
taming
radiologists.
Our
finding
nonetheless, is quite clear; no measures show above-average difficulties in obtaining radiologists in nonmetropolitan areas. Another possible source of difficulty in filling positions that the openended answers suggest is a mismatch
between qualifications hand, and
opportunities
the specialties, the
interests,
of applicants requirements
offered
imaging
and mammography are recent examples. How training programs could “turn on” and “turn off” rapidly enough to meet such short-lived peaks of hiring demand for specific subspecialties is difficult to foresee. These are questions for which further research might be of use. #{149} likely
ACR’s research
We thank William Chan, and Michael Mabry of the department for taking on the
extensive
of carrying
Acknowledgments: Pamela Kassing,
by the recruit-
groups on the other hand. We do not have enough data to assess the scope of this problem. To the extent the problem exists, it may reflect larger trends in the radiology profession such as growing subspecializalion, or a gap between the areas em-
1.
fellowship
in radiology
residency
or
programs and the needs of radiology group practices. Also, there may be large but short-lived surges of hiring demand for certain subspecialties as particular fields undergo growth spurts. Magnetic resonance
the
survey.
WillsJ,
Garrison
LPJr,
Evens
RG.
ply of requirements
for diagnostic
apeutic
Radiology
radiologists.
The
sup-
and ther1984;
151:9-
14. 2.
Brogden
equation:
ing
phasized
out
References
and
on one of and
work
BC.
a new
The radiology manpower look. AJR 1990; 154:1111-
1115. 3.
Sunshine
accurate review radiologists.
4.
5.
JH, Evens
RG, Chan
WC.
How
was GMENAC?: a retrospective of supply projections for diagnostic Radiology
1992;
182:365-368.
Sunshine fessional,
JH, Bansal S. Operational, proand business characteristics of radiology groups in the United States. Radiology 1992; 183:535-540. Evens RG. The impact of new payment systems on the specialty of radiology. Radiology 1992; 182:613-620.
Radiology
#{149} 51