ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
SECTION EDITOR
RONALD D. MILLER
Managing Risk and Expected Financial Return from Selective
Expansion of Operating Room Capacity: Mean-Variance
Analysis of a Hospital’s Portfolio of Surgeons
Franklin Dexter, MD, PhD*, and Johannes Ledolter,
*Department of Anesthesia and †College of Business, University of Iowa, Iowa City
Downloaded from http://journals.lww.com/anesthesia-analgesia by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3i3D0OdRyi7TvSFl4Cf3VC4/OAVpDDa8KKGKV0Ymy+78= on 02/18/2022
Surgeons using the same amount of operating room
(OR) time differ in their achieved hospital contribution
margins (revenue minus variable costs) by ⬎1000%.
Thus, to improve the financial return from periopera-
tive facilities, OR strategic decisions should selectively
focus additional OR capacity and capital purchasing on
a few surgeons or subspecialties. These decisions use
estimates of each surgeon’s and/or subspecialty’s con-
tribution margin per OR hour. The estimates are subject
to uncertainty (e.g., from outliers). We account for the
uncertainties by using mean-variance portfolio analy-
sis (i.e., quadratic programming). This method charac-
terizes the problem of selectively expanding OR capac-
ity based on the expected financial return and risk of
F
inancially strong hospitals can make capital im-
provements (e.g., buy anesthesia information
management systems), grow vibrant surgical
practices (e.g., provide sustained business to the anes-
thesiology group), and provide uncompensated com-
munity benefits (e.g., teaching and indigent care). Sur-
geons using the same amount of operating room (OR)
time differ in their achieved hospital contribution
margin (revenue minus variable costs) by ⬎1000%
(1– 4). Thus, to improve financial return from periop-
erative facilities, OR strategic decisions can selectively
focus additional OR capacity and capital purchasing
on a few surgeons or subspecialties. Such strategic
decisions are based on statistical summaries of data
from OR information systems and hospital accounting
databases (1– 4). Financial summaries calculated at a
Presented, in part, at the Institute for Operations Research and
Management Science meeting, November 19, 2002, San Jose, CA,
and the Association of Anesthesia Clinical Directors meeting, March
23, 2003, Carlsbad, CA.
Accepted for publication February 25, 2003.
Address correspondence and reprint requests to Franklin Dexter,
Division of Management Consulting, Department of Anesthesia,
University of Iowa, Iowa City, IA 52242. Address e-mail to
Franklin-Dexter@UIowa.edu.
DOI: 10.1213/01.ANE.0000066263.53951.93
190 Anesth Analg 2003;97:190–5
PhD†
different portfolios of surgeons. The assessment reveals
whether the choices, of which surgeons have their OR
capacity expanded, are sensitive to the uncertainties in
the surgeons’ contribution margins per OR hour. Thus,
mean-variance analysis reduces the chance of making
strategic decisions based on spurious information. We
also assess the financial benefit of using mean-variance
portfolio analysis when the planned expansion of OR
capacity is well diversified over at least several sur-
geons or subspecialties. Our results show that, in such
circumstances, there may be little benefit from further
changing the portfolio to reduce its financial risk.
(Anesth Analg 2003;97:190 –5)
surgeon and/or subspecialty level have uncertainty
due to limited sample sizes, large variability, and out-
liers (3,4). Consequently, these strategic decisions (e.g.,
the anesthesiology group decides to hire a subspe-
cialty trained anesthesiologist) can be based on spuri-
ous information. The purpose of this paper is to de-
scribe how to consider uncertainties in financial data
in the strategic decision of which surgeons and/or
subspecialties should have their OR capacity
expanded.
The remainder of the Introduction reviews previous
work in the field, to motivate our paper.
Lack of sufficient hospital accounting data is usually
not a factor limiting these strategic analyses. Variable
costs can be estimated sufficiently accurately for pur-
poses of strategic decision-making using the patients’
OR times, hospital lengths of stay, intensive care unit
(ICU) lengths of stay, and implant costs (3). Hospitals
with a detailed hospital cost database would, of
course, use their direct estimates of variable costs.
Patients undergoing urgent or emergent surgery are
typically excluded from these analyses for two rea-
sons. First, there is a commitment to provide timely
care to such patients once admitted to the hospital
(1– 4). Decisions regarding OR capacity expansion do
not change this. Second, cost accounting is easier
©2003 by the International Anesthesia Research Society
0003-2999/03
ANESTH ANALG ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
2003;97:190 –5
when excluding such patients, because then the vari-
able cost of each patient’s entire outpatient visit or
hospitalization can be attributed to the surgeon or
subspecialty who scheduled the patient for elective
surgery. For example, the variable costs and revenues
for a child undergoing tonsillectomy and adenoidec-
tomy can reasonably be attributed to the child’s pedi-
atric otolaryngologist. However, the costs of a 40-day
hospitalization for cerebral trauma from a motor ve-
hicle accident should not be attributed to the otolar-
yngologist who performed a tracheostomy on the 18th
hospital day.
Hospital contribution margins per OR hour vary
several-fold among surgeons (1,2), with 5th and 95th
percentiles of $310 and $3370, respectively (Fig. 1).
Thus, increasing OR capacity selectively is more ad-
vantageous financially for a hospital than increasing
capacity equally. The conceptually easiest way to in-
crease OR capacity selectively is to sort surgeons in
descending order of their contribution margins per
OR hour, increase capacity for the surgeon at the top
of the list by the maximum allowable amount, increase
capacity for the next highest surgeon by the maximum
amount, and so forth, until no more additional capac-
ity is available (1). For example, as each surgeon is
considered for an increase in his or her OR time, the
expansion of that surgeon’s OR time could be any
value between 0% and 25% (2,4). This maximum ex-
pansion of 25% represents a feasible expansion of OR
capacity without recruiting more surgeons. Alterna-
tively, each surgeon could have his or her OR time
increased by 100% (3). This represents hiring a new
surgeon modeled after a surgeon currently practicing
at the hospital (3). For example, we used a 15% total
increase in OR capacity, a 100% maximum increase in
each surgeon’s usage, and the data in Figure 1. Selec-
tive expansion of capacity increased expected contri-
bution margin by 32% versus a 15% increase achieved
by expanding capacity equally (i.e., proportional to
current workload) (2,3).
Increasing OR capacity for a subspecialty using
many ICU beds would be counter-productive if ICU
occupancy is nearly 100%. Thus, hospitals also need to
consider constraints from limited ICU and hospital
ward beds (2). Then, strategic decisions are made not
simply by ranking surgeons or subspecialties by the
contribution margin per OR hour (2). Instead, the
more sophisticated method of linear programming
should be used, because it considers constraints (2–5).
On the far right of Figure 1, we label those surgeons
who would be provided more OR capacity when the
constraints on ICU and hospital ward beds are con-
sidered along with the expected contribution margin
per OR hour. Several surgeons with the largest con-
tribution margins per OR hour do not have increases
in their OR capacity. These surgeons’ patients need
ICU beds, even though few are available (2). Several
DEXTER AND LEDOLTER 191
SELECTIVITY EXPANDING OPERATING ROOM CAPACITY
Figure 1. Contribution margin per operating room (OR) hour for 98
surgeons. This figure summarizes work published in References
2– 4. The circles show each surgeon’s ratio of mean contribution
margin among his or her patients to mean OR hours among his or
her patients. The horizontal bars extend to the upper and lower
limits of 95% confidence intervals (4,6). To simplify the plot, the
three confidence intervals extending below zero were truncated at
zero. The one upper confidence interval that was larger than $6000
was truncated at $6000. The triangles on the far right show the
surgeons who would be provided additional OR capacity based on
trying to enhance the hospital financially. Several surgeons with the
largest contribution margins per OR hour do not have increases in
OR capacity. These surgeons’ patients need intensive care unit
(ICU) beds, even though few are available (2). The analysis was
performed with linear programming (2–3), a technique that takes
into account such constraints. The linear programming was per-
formed with each surgeon’s future use of OR time increased by 0%
to 25% of his or her use during the previous fiscal year. Total OR
capacity was increased 15%, hospital ward capacity by 10%, and
ICU capacity by 5%.
surgeons with relatively small contribution margins
per OR hour receive the maximum increase in OR
capacity. These surgeons use few of the scarce hospital
ward and ICU beds relative to surgeons with larger
contribution margins per OR hour (2). In the Methods
and Results sections, the percentage increases in con-
tribution margins are less than that described in the
preceding paragraph, because of the important effects
of limited hospital ward and ICU beds.
Decisions to change or maintain resources are un-
certain for many surgeons or subspecialties when the
decisions are based on the estimated financial impact
on the hospital (4). Typically, one fiscal year of histor-
ical data is used for estimation, thereby achieving as
large a sample size as possible, while excluding an-
nual re-negotiations of managed care contracts,
changes in national reimbursement, trends in practice
patterns, and alternations in hospital accounting for
fixed costs (1– 4). However, the estimates are not al-
ways reliable. The relative standard error in estimat-
ing a surgeon’s mean contribution margin per OR
192 ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
SELECTIVITY EXPANDING OPERATING ROOM CAPACITY
hour is 23% on average (n ⫽ 98; sd ⫽ 46%) (Fig. 1) (4).
This uncertainty in each surgeon’s estimated financial
performance is sufficiently large to alter decision-
making for many surgeons (4). For example, in one
study, 45% of the surgeons not having their OR capac-
ity increased should have received more OR re-
sources, the discrepancy being due to random error
(4).
Confidence intervals for expected contribution mar-
gins are not routinely included in hospital financial
reports. However, for this application, uncertainty is
sufficiently important that it should be incorporated
into the analysis. Reports listing each surgeon’s or
subspecialty’s contribution margin per OR hour in
rank order are misleading otherwise (4). The purpose
of this new paper is to improve upon just measuring
the uncertainty, by determining how to use the meas-
ure of uncertainty to improve the quality of strategic
decisions.
Methods
The data have been described previously (2– 4). The
population consists of all patients undergoing outpa-
tient or same day admit surgery during the 2000 fiscal
year at a large, academic, multiple specialty hospital in
the United States (US). The data were extracted from
the hospital’s activity based costing system. Variable
costs, revenue, hours of OR time, hours of regular
ward time, and hours of ICU time were calculated for
each physician using year 2000 US dollars. We limited
the analysis to the 98 physicians who did at least 15
cases at the hospital during the study year (2– 4).
Linear programming (5) (Excel 2000, Microsoft,
Redmond, WA) with estimates of the surgeons’ mean
contribution margins per OR hour was used to find
the portfolio of surgeons’ OR time allocations that
maximizes expected contribution margin. As a typical
example, we included the following constraints on the
availability of resources:
• Each surgeon’s OR capacity can be expanded by
as much as 25% or 100% (see above) of the num-
ber of OR hours that he or she used during the
past fiscal year.
• OR time for a surgeon is not reduced (i.e., we
studied expansion of OR capacity).
• Total OR time used can be increased by up to 15%
by extending the workday. The additional OR
time will cost $50 more per hour more than exist-
ing OR time.
• Hospital ward use can be 10% more without hir-
ing additional staff.
• ICU use can be 5% more without hiring additional
staff.
DEXTER AND LEDOLTER ANESTH ANALG
2003;97:190 –5
Linear programming determines the single best
portfolio of surgeons to maximize expected contribu-
tion margin subject to the above constraints. The num-
ber of surgeons in the optimal portfolio, which sur-
geons are in the optimal portfolio, and the increase in
OR time provided to each surgeon are chosen auto-
matically. Generally, each increase in the maximum
expansion of OR time for each surgeon results in a
decrease in the number of surgeons in the optimal
portfolio. Likewise, each increase in the total available
OR time results in an increase in the number of sur-
geons in the optimal portfolio.
We used Fieller’s result for the statistical distribu-
tion of the ratio of two random variables to obtain the
standard error of each surgeon’s mean contribution
margin per OR hour (4,6). Variables used were the
contribution margins and OR times of all cases per-
formed by each surgeon (4). The corresponding 95%
confidence intervals are shown in Figure 1.
The standard deviation of the expected return quan-
tifies the risk of a portfolio, whether of stocks and
bonds or of surgeons (5). For a specified level of ex-
pected return, a rational investor chooses the portfolio
having the least risk. A portfolio is “efficient” if there
is no portfolio having the same return with a lesser
standard deviation, and vice versa. The “efficient fron-
tier” is the collection of all efficient portfolios.
We calculated efficient frontiers by minimizing the
variance of resulting hospital contribution margins,
subject to the constraint that the expected return was
at least equal to a specified value (5). This mean-
variance portfolio analysis was accomplished using
Excel’s Solver function’s quadratic programming (5).
The analysis automatically chooses the number of sur-
geons in the portfolio, which surgeons are included in
the portfolio, and the increase in OR time provided to
each surgeon in the portfolio.
We also examined the sensitivity of our results to
the number of surgeons in the portfolio of surgeons
with expanded OR capacity. We varied the numbers
of surgeons by progressively increasing the maximum
allowable increase in each surgeon’s capacity in incre-
ments of 0.1%. We evaluated sensitivity of our results
to constraints from limited ICU and hospital ward
capacity by repeating the sensitivity analysis while
excluding these two constraints.
Results
Each point along an efficient frontier represents a port-
folio of surgeons. Each portfolio differs in the number
of surgeons provided additional OR capacity and/or
in the increase in OR capacity provided to each such
surgeon.
The portfolio that corresponds to the peak of the
efficiency frontier guarantees maximum expected con-
tribution margin, but usually at the expense of sizable
ANESTH ANALG ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
2003;97:190 –5
risk. The peak is at the point at the far right of Figure
2. That portfolio is the same as that obtained by linear
programming (see Introduction).
Moving to the left on the efficiency frontier, the
portfolios have smaller risk, but also smaller expected
financial return (i.e., contribution margin). The objec-
tive of the mean-variance portfolio analysis is to draw
the efficient frontier and then choose a portfolio along
the frontier that has an acceptable expected return, but
appreciably less risk. The desired portfolio is identi-
fied by choosing a point along the curve where, from
right to left, the initial steepness of the curve has
leveled off. There is certainly no good reason for a
portfolio to be selected that corresponds to points that
are to the left of this value.
For example, with a 100% maximum increase in
each surgeon’s OR time, the desired point where the
steepness levels off has a risk of 1.7% and expected
return of 13.7%. The smallest attainable risk is 0.2%
less, but at a relatively large expense in expected re-
turn. In contrast, the risk at the peak of the efficiency
frontier is 0.2% more, but with little (only 0.4% in-
crease) effect on the expected return.
The portfolios, of surgeons provided expanded OR
capacity, differ between the peak of the efficient fron-
tier and the point at which the slope in the efficient
frontier has started to level off. Differences in the
portfolios reveal which surgeons’ uncertainties in con-
tribution margins per OR hour influence strategic de-
cisions. For example, with a 100% maximum increase
in each surgeon’s OR time, the portfolio at the peak of
the efficient frontier increases the OR capacity of 16
surgeons. The portfolio at the point where the slope of
the curve starts to level off (moving from right to left)
increases the allocation of 14 surgeons.
The two excluded surgeons had the largest uncer-
tainties in the mean contribution margin per OR hour
among all surgeons. The large uncertainties were not
the result of overly small sample sizes. The two sur-
geons had samples of 72 and 74 cases per year, respec-
tively. Instead, the primary reason for their large stan-
dard errors was the presence of outliers with large
positive contribution margins relative to the used OR
time. Specifically, both surgeons had two patients
with diagnosis related groups 483 (tracheostomy ex-
cept for face, mouth, and neck diagnosis). This diag-
nosis related group has the second largest Diagnosis
Related Groups’ Medicare reimbursement. Yet, the
four patients had lengths of stays of only 3 to 10 days.
Our exploration of Figure 2 illustrates how the
mean-variance portfolio analysis is used in practice.
The remaining results describe sensitivity analyses
that we performed to test hypotheses about possible
reasons for our findings.
The peaks of the efficient frontiers in Figure 2 were
obtained by increasing OR capacity of 34 and 16 of the
98 surgeons, when the maximum increases were 25%
DEXTER AND LEDOLTER 193
SELECTIVITY EXPANDING OPERATING ROOM CAPACITY
Figure 2. Efficient frontiers for selectively expanding operating
room (OR) capacity. Risk of a portfolio is quantified as the standard
deviation of the maximum increase in hospital contribution margin.
Efficient frontiers were calculated by specifying the expected return
shown on the horizontal axis and then performing quadratic pro-
gramming to minimize the variance of resulting hospital contribu-
tion margins (5). Increasing each surgeon’s OR capacity by up to
25% represents a feasible expansion of OR capacity without recruit-
ing more surgeons (2). Increasing OR time by 100% represents
hiring a new surgeon modeled after a surgeon currently practicing
at the hospital (3).
and 100% of baseline OR capacity, respectively. There
was some reduction in the risk by accepting a smaller
expected contribution margin. However, the risk re-
duction in Figure 2 was not particularly impressive.
We know that if OR capacity was expanded based
solely on the data from the one surgeon with the
largest mean contribution margin per OR hour, then
the risk would be very high. Thus, we hypothesized
that the number of surgeons in the peak portfolio of
Figure 2 was sufficiently large to have provided a
portfolio that was diversified enough to account for
the limited incremental reductions in risk achieved by
the quadratic programming.
To test this hypothesis, the maximum possible in-
crease in each surgeon’s OR capacity was varied to
achieve a wide range in the numbers of surgeons with
expanded OR capacity (Fig. 3). With the decision
based on data from progressively fewer surgeons, the
risk increased by ⬍1% (lower pane, lower line), even
though the expected return increased from 5% to 17%
(upper pane, lower line). Each reduction in the num-
ber of surgeons’ data used for the strategic decision
caused the ratio of the risk (lower pane) to the ex-
pected increase in contribution margin (upper pane)
to be reduced until the strategic decision was based on
data from fewer than five surgeons.
The small risk reductions for portfolios correspond-
ing to points on the left of the efficient frontier in
Figure 2 may also have been caused, in part, by ICU
194 ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
SELECTIVITY EXPANDING OPERATING ROOM CAPACITY
Figure 3. Sensitivity analyses to explain results of Figure 2. The
maximum allowable increase in each surgeon’s capacity was varied
over a wide range. At each value, linear programming (5) was run
to find the optimal portfolio of surgeons to maximized expected
hospital contribution margin. One resulting dependent variable, the
number of surgeons with expanded operating room (OR) capacity,
was plotted on the horizontal axis. The other two resulting depen-
dent variables (increase in contribution margin and risk in contri-
bution margin) were plotted on the vertical axes. With constraints
on intensive care unit (ICU) and hospital ward beds, reductions in
the number of surgeons’ data used in the strategic decision resulted
in an increase in risk of ⬍1% (lower pane, lower line), even though
the expected return increased from 5% to 17% (upper pane, lower
line). When constraints were eliminated, the large increase in risk
arose for fewer than eight surgeons.
and hospital ward constraints leading to already di-
versified portfolios with a sufficient number of sur-
geons’ data. To test this hypothesis, the sensitivity
analysis was repeated without the constraints on ICU
and hospital ward use. The risk did not increase by
more than 1% until fewer than eight surgeons’ data
were used for the strategic decision (lower pane, up-
per line). When the risk was expressed as a percentage
of the expected increase, the risk was progressively
reduced until the strategic decision was based on data
from fewer than eight surgeons.
Discussion
The vibrancy of an anesthesiology group is inextrica-
bly linked to the financial strength of the hospital(s)
DEXTER AND LEDOLTER ANESTH ANALG
2003;97:190 –5
where its anesthesia providers practice. In this paper,
we continued previous work in improving methods of
linking hospital financial data and OR information
systems data for making good strategic decisions
(1– 4).
The premise of mean-variance portfolio manage-
ment is that the economically rational investor accepts
a smaller reduction in expected return to achieve a
large reduction in risk (5). However, with respect to
expanding OR capacity selectively, basing decision-
making on trying to reduce that uncertainty is a low-
yield proposition when at least several surgeons’ or
subspecialties’ data are in the optimal portfolio. This is
evident from the relatively unimportant absolute re-
duction in risk along the vertical axis of Figure 2.
Previously, linear programming was used to iden-
tify the portfolio of surgeons providing the maximum
expected return (that is, the peaks in Fig. 2) (2,3). Such
linear programming provides the mean part of mean-
variance portfolio analysis. Next, we learned how to
create graphs such as Figure 1, displaying uncertainty
in each surgeon’s contribution margin per OR hour
(4). That measured the variance part of mean-variance
portfolio analysis. In this study, we showed that
mean-variance portfolio analysis (i.e., quadratic pro-
gramming) can be used to combine both the mean and
the variance into decision-making. Specifically, the
analysis can be used to assess the sensitivity of the
portfolio with the maximum expected return to uncer-
tainty in each surgeon’s contribution margin per OR
hour. The assessment is made by comparing the port-
folios of surgeons at the peak and at the point where
the slope of the curve, which initially is steep, starts to
level off.
Mean-variance portfolio analysis can also include
the correlations in uncertainty among surgeons (5).
Suppose that among 10 surgeons planned for expan-
sion in OR capacity, 4 were cardiac surgeons in the
same surgical practice. Then, a decision for one sur-
geon could affect the decision for others (i.e., risk
could be correlated). A limitation of our work is that
currently we do not know how to estimate such cor-
relations using data from OR information systems and
hospital accounting databases.
To simplify the presentation, we considered the
amount by which each surgeon could expand his or
her use of OR time to be the same for all surgeons. In
practice, these values should be adjusted based on
marketing data and knowledge about each of the
surgeons.
In conclusion, there is marked uncertainty in each
surgeon’s mean contribution margin per OR hour. We
showed that this is, to a large extent, due to the large
inherent variability in the contribution margins of a
surgeon’s procedures and also due to the influence of
outliers. Outliers are probably an inevitable part of
health care economics. Consequently, the decision to
ANESTH ANALG ECONOMICS, EDUCATION, AND HEALTH SYSTEMS RESEARCH
2003;97:190 –5
maintain or expand capacity will always be uncertain
for each particular surgeon or subspecialty. The pri-
mary advance of this paper is its use of mean-variance
portfolio analysis as a method to reduce the likelihood
of making strategic decisions on the basis of spurious
information. However, we also learned that if selective
expansion of OR capacity is well-diversified over at
least several surgeons or subspecialties, any additional
efforts to change decisions to reduce risk are unlikely
to yield financially important benefits.
References
1. Macario A, Dexter F, Traub RD. Hospital profitability per hour of
operating room time can vary among surgeons. Anesth Analg
2001;93:669 –75.
DEXTER AND LEDOLTER 195
SELECTIVITY EXPANDING OPERATING ROOM CAPACITY
2. Dexter F, Blake JT, Penning DH, Lubarsky DA. Calculating a
potential increase in hospital margin for elective surgery by
changing operating room time allocations or increasing nursing
staffing to permit completion of more cases: a case study. Anesth
Analg 2002;94:138 – 42.
3. Dexter F, Blake JT, Penning DH, et al. Use of linear programming
to estimate impact of changes in a hospital’s operating room time
allocation on perioperative variable costs. Anesthesiology 2002;
96:718 –24.
4. Dexter F, Lubarsky DA, Blake JT. Sampling error can signifi-
cantly affect measured hospital financial performance of sur-
geons and resulting operating room time allocations. Anesth
Analg 2002;95:184 – 8.
5. Ragsdale CT. Spreadsheet modeling and decision analysis, a
practical introduction to management science. 2nd ed. Cincin-
nati, OH: South-Western College Publishing, 1998:45– 64, 128 – 41,
334 – 41.
6. Briggs AH, Mooney CZ, Wonderling DE. Constructing confi-
dence intervals for cost-effectiveness ratios: an evaluation of
parametric and non-parametric techniques using Monte-Carlo
simulation. Stat Med 1999;18:3245– 62.