Symposium on Surgical Health Care Delivery

Should Surgery Be Regionalized?

]. P. Bunker, M.D.,*
H. S. Luft, Ph. D.,t
and A. Enthoven, Ph. D.:j:

There are two reasons why regionalization of medical or surgical pro-

cedures might be desirable, first, to improve the quality of medical care and,
second, to save money. Both might reasonably be expected to be achieved
by the regionalization of procedures that involve large capital outlays and
require great professional and technical skill, such as open heart surgery.
Whatever benefits might be achieved by regionalization must be weighed
against the potential inconvenience to physician and patient, including
constraints on place and character of practice, costs of regulation, costs of
transportation, and possible losses in time from work. These considerations
have, to date, forestalled any widespread regionalization of surgical services.
It is generally assumed that, other things being equal, the quality of care
improves with the experience of those providing it; however, there are only
a few sets of data demonstrating this effect for surgery. In a 1960 article
Hotchkiss reported a comparison of mortality following surgical correction of
patent ductus in teaching and in nonteaching hospitals." By mail survey, he
identified 508 operations conducted in 68 hospitals over the previous 10-year
period. The mortality rates for ligation of the patent ductus, and for ligation
and division, in these nonteaching hospitals were compared with the mortality
rates for 3688 of these operations conducted by 49 "outstanding cardiovascular
surgeons in this country and abroad." 22 The results are presented in Table 1.
In a second set of relevant data, Chalmers and his associates at Mt. Sinai
Medical Center in New York compared case fatality rates for coronary artery
bypass graft in nationwide published reports with unpublished case fatality
rates for surgery in New York City.• The mean case fatality rates were 5.64
per cent for the unpublished series and 3.98 per cent for the published series,
as shown in Figure 1. Twelve of 14 unpublished series were from institutions

*Professor of Anesthesia and of Family, Community and Preventive Medicine, Stanford University
School of Medicine, Stanford, California
tAssociate Professor of Health Economics, Health Policy Program, University of California, San
Francisco, California
:J:M. S. Eccles Professor, Graduate School of Business, and Professor by Courtesy, Family,
Community and Preventive Medicine, School of Medicine, Stanford University, Stanford,
California

Surgical Clinics of North America-Vol. 62, No.4, August 1982

657
658 J. P. BUNKER ET AL.
Table 1. Comparison of Mortality in Surgical Treatment of Patent Ductus
Arteriosus*
TEACHING CENTERS NONTEACHING CENTERS

OPERATIONS Cases Deaths Mortality(%) Cases Deaths Mortality(%)

Ligation 1476 38 2.60 320 12 3.75

Division 2212 63 2.80 188 18 9.57
TOTALS 3688 101 2.77 508 30 5.90

*Adapted from Hotchkiss 9 and from Waterman. 22

doing fewer than 200 such operations a year, whereas only 11 of21 published
series had case loads of less than 200. 3 A third set of data comes from the
California Children's Services, reported by Cretin and Roberts, 5 and shows
an inverse relationship between volume of pediatric heart surgery and
mortality (Table 2).
A fourth set of data comes from a study conducted by us. 14 The study
examined mortality rates for 12 surgical procedur~s of varying complexity in
1498 hospitals to determine whether there is a relation between a hospital's
surgical volume and its surgical mortality. The mortality following open-
heart surgery, vascular surgery, transurethral resection of the prostate, and
coronary bypass was markedly lower in hospitals with large numbers of i
operations. Hospitals in which 200 or more of these operations were done
I.
annually had death rates, adjusted for case mix, that were 25 to 41 per cent
lower than those in hospitals with lower volumes. For other procedures the v
mortality curve flattened at lower volumes. For example, hospitals doing 50
to 100 total hip replacements attained a mortality rate for this procedure
almost as low as that of hospitals doing 200 or more. Some procedures, such
as cholecystectomy, showed no relation between volume and mortality.
Summary data from this study are presented in Table 3.

16.0
~ •
LLJ
1-
•
ct
0::
•
>- 0 Figure 1. Operative fatality rates for coronary
1- artery surgery in an unpublished series collected
:J ••• from different New York hospitals (1976) and from
ct
1-
• Og all series published from nationwide hospitals in
ct 0~ 1977 are shown. The arrow indicates the Veterans
1.1.. --;·~.64
cPgo Administration cooperative study. (From Chalmers,
LLJ • oa-i·~.g· T. C., Smith, H. J., Ambroz, A., eta!.: In defense of
VI
ct
•• Oo the VA randomized control trial of coronary artery
u •• ~0 surgery. Clin. Res., 26:230, 1978, with permission.)
•
·-·•
UNPUBLISHED
Case fatality
0

PUBLISHED
Case fatality
rates reported roles notion-
by N.Y hospilols wide
(19761 (1977)
 en
:c
0
c:
t-<
0
en
c:
flt'1
Tabie 2. Actual and Expected Death Rates at CCS Centers for Open- and Closed-Heart Surgical Procedures 1976-1978*
"l:l:l><
ACTUAL ACTUAL EXPECTED NUMBER OF t'1

NUMBER OF NUMBER OF DEATH NUMBER DEATHS ABOVE

::0
ANNUAL NUMBER t'1
OF CENTERS OPERATIONS DEATHS RATE OF DEATHS (BELOW) EXPECTED 0
VOLUME
0
z
>
t-<
Open-Heart Surgery
~
0-199 27 3575 237 .066 203 34 t'1
200-399 25 6465 456 .071 368 88 0
·v
400-599 10 4490 244 .054 255 (ll)
600 and up 7 6269 246 .039 357 (ll1)
Total 69t 20,799 ll83 .057 ll83 X2 = 61.56 (P < .01)

Closed-Heart Surgery
0-99 34 2090 98 .0469 62 36
100-199 29 4250 116 .0273 126 (10)
200 and up 6 1352 14 .0104 40 (26)
Total 69t 7692 228 .0296 228 X2 = 38.73 (P < .01)

*From Cretin, S., and Roberts, N. K.: Surgical care for cardiovascular disease in California: Present status and future policy. California Policy Seminar,
Monograph No.8, Institute of Governmental Studies, University of California at Berkeley, 1981, with permission.
tEach of 23 centers observed for three years.
Source: California, Department of Health Services, Crippled Children Services, Report of Cardiac Centers, 1978 (Sacramento: 1979).

0)

""-o
 g;
0

Table 3. Actual and Expected Death Rates for Selected Procedures in High and Low Volume Hospitals
and Estimates of Excess Mortality*

AVERAGE ANNUAL EXPERIENCE 1974-1975 EXCESS AS:

(7)
(3) (4) (5) Percent of (8)
(1) (2) Actual Expected Projected (6) Deaths in Percent of
VOLUME Number Number of Death Death Death Excess Low Volume Deaths in All
OPERATION CATEGORY of Cases Hospitals Rate Rate Rate Deaths Hospitals Hospitals

Group I
Open heart surgery ,;;; 200 11,997 541 .107 .086 .066 492 38 22
> 200 15,474 46 .061 .079
Vascular surgery ,;;; 200 39,285 1,291 .111 .107 .078 1,327 30 27
> 200 5,501 18 .075 .103
TURt ,;;; 200 71,964 1,162 .011 .010 .008 396 25 22
> 200 14,750 55 .008 .010
CABG ,;;; 200 9,549 157 .057 .047 .034 224 41 28 ':"'"'
> 200 7,616 25 .034 .047 - - - - ~
1:!:1
Group II c::
z
Colectomy ,;;; 50 21,523 1,191 .074 .068 .061 291 18 12 1"1
t'1
>50 14,560 199 .061 .069 - - - ::0
t'1
Biliary tract surgery <;10 4,500 1,007 .092 .087 .081 51 12 7 "">
>10 4,457 271 .080. .086 - - - f'
Total hip replacement "'50 10,297 725 .OI9 .OI6 .Oll 85 43 32 rJl
X
>50 6,042 79 .Oll .OI6 - - - - 0
c
Resection and graft, 3,384 654 .203 .I89 .I48 209 30 24 r-
"'20 0
abdominal aortic > 20 I,240 38 .I4I .I80 - - - - rJl

aneurysm ,0
c

Vagotomy and/or ,;5 I,305 606 .OI2 .010 .009 3 I7 7 ,

1'1
~
pyloroplasty for >5 3,0ll 332 .008 .009 - - - tx:l
duodenal ulcer 1'1
tx:l
Cholecystectomy and ,;5 I,287 609 .029 .027 .025 4 I3 5 1'1
0
incision of common bile >5 2,293 285 .025 .027 - - 5z
duct >
r-
Group III
N
t'1
Vagotomy, all ,;I I6I .027 0
I6I .040 .029 2 27 I ·v
>I 8,543 947 .027 .027
Cholecystectomy ,;I 3 3 .I67 .007 .007 <I 95 0.02
>I I62,569 I,478 .010 .010

*The two volume categories for each procedure represent annual volumes below and above those identified by the "Hat of the curve." Actual death rates
are based on the total procedure-specific deaths in all hospitals within the volume category divided by the number of cases (column I). Expected death rates
are based on the total death rates for each age, sex, or single or multiple diagnosis cell, weighted by the case mix of patients in all hospitals within the volume
category. Projected death rates are computed by multiplying the expected death rate for low volume hospitals by the ratio of the actual to expected death rates
for high volume hospitals, for example, .066 = .0865 (.06I/.079). The number of excess deaths is derived by multiplying the number of patients in low volume
hospitals by the difference between the actual and projected death rates. Note: figures in columns 5 through 8 may not add up because of rounding.
tTUR = transurethral resection of the prostate.

g;
......
662 J.P. BUNKER ET AL.

Nearly half of the open heart operations were carried out in hospitals
averaging fewer than two such procedures per month (as can be readily
calculated from columns 1 and 2 of Table 3), with many hospitals reporting
only one or two cases a year. When frequency was observed at this low a
level, the reported mortality was as high as 18 per cent, as shown in Figure
2. A similar pattern was also observed for other procedures, including
transurethral resection of the prostate, total hip replacement, and resection
and graft of abdominal aortic aneurysm, also shown in Figure 2.
The data in this study establish, for some operations, a negative correlation
between the annual number done in a hospital and postoperative mortality,
after controlling for patients' age, sex, and single or multiple diagnoses.
These results may be explained by the effect of volume or experience on
mortality, or by the referral of a larger volume of patients to those institutions
or surgeons known to have better outcomes. The results may also reflect, in
part, differences in patient mix that are not measured by our statistical
controls; for example, institutions with better outcomes may be able to justify
operating on patients with less severe disease.
Data on the effect of experience on nonfatal outcomes of surgical care
are limited to the recent report by Farber, Kaiser, and Wenzel of a "highly
significant inverse relation ... between ... the frequency of operation and
the infection rate" for a number of common elective procedures. 7 One would
like to know, in addition, nonfatal outcomes such as the frequency of
reoperation after total hip replacement or of common duct injury following
cholecystectomy. Unfortunately, such data are not available.
What conclusions can we draw from these data? In our study of the
relation between surgical volume and mortality, we were unable to resolve
the important question of whether the observed relationship represents the
effects of volume or experience on outcome, or whether it is the result of the
attraction of more patients to centers having better than average results.*
From the patient's point of view, this may not be of very great importance.
If the patient could be guaranteed a high quality of care by choosing a
hospital with a high volume of a particular procedure, that might be all he
or she needed to know. Of course, it is not that simple. Some hospitals with
low volumes may have a low mortality. What the patient, and the referring
physician, needs to know is the quality of care itself.
If the medical profession were able to guarantee the quality of surgical,
or medical, care provided in all hospitals, whatever their size and volume of
procedures, the argument for regionalization would be reduced to one of
cost. Some physicians apparently believe that existing procedures of medical
self-regulation and peer review do provide an adequate protection of the
public against poor quality care. That this may not be the case is suggested
by recent headlines such as "Soaring Mortality Rate Halted Heart Surgery
at Med Center," 6 "Chicago Heart Surgery: Too Much and Too Lethal," 11 and
"Heart Surgery Death Rate Probed at Malden Hospital: 51% Mortality
Reported in Open Heart Operations." 10 The most shocking instances are, of

*Indeed, in subsequent analyses of the same data base, one of us (H.S.L.) has presented
evidence that for several of the operations studied-including coronary artery bypass graft,
transurethral resection of the prostate, and total hip replacement-the referral of patients to centers
with good results may be the more important cause.J3
 SHOULD SURGERY BE REGIONALIZED? 663

OPEN-HEART SURGERY CORONARY-ARTERY BYPASS

0.20
0.20

0.15

•
0.101 .........

0.05,_, ____~

o.oal 0
I
100
I
200
_
300

TRANSURETHRAL RESECTION OF PROSTATE RESECTION AND GRAFT FOR ABDOMINAL

AORTIC ANEURYSM

::f ~~-,-.-~-~-~ . .
0 20 40 60 80 100

TOTAL HIP REPLACEMENT CHOLECYSTECTOMY

0.15~---.---.---.---.
0.08
.._.__ ACTUAl DEATH RATES
- - • EXPfCTED DEATH RATES
...... DEATH RATES BASED ON
w
I- 0.06 0.10 <;15 HOSPITALS
<{
a:
:z:
I-
<{ 0.04
w
0 0.05

NUMBER OF OPERATIONS
Figure 2. Actual and expected death rates for selected procedures by annual number of
procedures in each hospital. The volumes shown are the mean number of operations per year for
hospitals in each of the following categories: (1, 2-4, 5-10, ll-20, 21-50, 51-100, 101-200, 201 + ).
664 J. P. BUNKER ET AL.
course, the most apt to receive publicity, and it is not possible to generalize
from such sensational press releases. However, the data presented in Figure
2 do suggest that high mortality rates may occur rather frequently with
occasional surgery.
The explanation for these high mortalities is, again, not clear. Graft of
abdominal aortic aneurysm is often carried out as an emergency, and the
occasional-to-rare such procedure may well be performed only as a last resort.
For the other procedures studied, this is not an adequate explanation,
however. A coronary artery bypass graft may be carried out as an emergency
procedure immediately following myocardial infarction, particularly when
the infarct occurs incidental to coronary angiography. However, the mortality
of coronary angiography has also been shown to vary inversely with volume, 1
and it should not be carried out on an occasional basis or in hospitals not
conducting cardiac surgery on a regular basis. 18 Similar arguments can be
made for the other procedures listed.
The second argument offered in support of regionalization is to save
money. Potential savings might be achieved by the more efficient use of
capital equipment and personnel, by improving the quality of care (thus
reducing complications and death), and by improving the decision whether
or not to operate (thus, perhaps, reducing the number of procedures of small
net benefit). Recent estimates of the magnitude of potential dollar savings
vary widely, 8 • 15• 20 in part as a result of the different assumptions on which
these estimates were based.
Considerable savings are potentially to be achieved simply by concen-
trating some procedures in a relatively small number of hospitals. Regional-
ization of obstetric services, for example, has been proposed to decrease costs
as well as to improve the quality of care. Pettigrew, while with the Massa-
chusetts Department of Public Health in 1976, recommended immediately
closing all maternity units with fewer than 500 deliveries a year (except for
a few geographically isolated cases) and ultimately establishing a minimum
standard of 1500 deliveries for urban hospitals.'" In addition, she recom-
mended a statewide system of maternity units at three levels of care, as
follows: (1) uncomplicated deliveries plus minimum capabilities for stabiliz-
ing unexpected emergencies prior to transportation, (2) intermediate capa-
bilities, and (3) a full range of specialized maternal, fetal, and newborn
services. These recommendations have now been largely implemented.
Pettigrew did not report total statewide cost estimates for the existing
system compared with the recommended system. But one indication of the
possible economies available is that the costs of an efficient minimum-
capability maternity unit would have been $1245 per admission at 500
admissions per year, compared with only $653 per admission at 1200
admissions per year (1975 prices).
The number of obstetric deliveries are assumed to be fixed and would
not be altered by regionalization. Even greater relative savings might be
achieved by regionalizing procedures that do not have fixed numbers, that
is, procedures for which the indications may be, to a considerable extent,
discretionary. Thus, it may be argued that the experienced surgeon will refine
the indications for a particular operation and will be able to identify and to
exclude operations in clinical situations in which the marginal net benefit is
SHOULD SURGERY BE REGIONALIZED? 665
very small or in which risks may actually exceed benefits. The active surgeon
who performs a high volume of surgery is, in addition, sufficiently busy not
to be tempted to carry out procedures when the marginal net benefit is
perceived to be very small.
A third source of savings that is not ordinarily included as a potential
benefit of regionalization, but nevertheless is an important one and potentially
very large, is in the cost of complications and the monetary value of patients'
lives. Wound infections have been reported to occur with greater frequency
for common surgical procedures when the procedure is performed infre-
quently/ The cost of treating this complication alone is large; other compli-
cations found to occur more frequently when the operation is infrequently
performed would obviously add to these costs. There is no universally agreed
upon basis for determining the value of a human life. At a minimum, however,
this value is not less than the loss of future earnings incurred by a surgical
death. If, in fact, regionalization were to lead to a decrease in total surgical
mortality and morbidity, the savings from future earnings alone could be
huge.
What policy recommendations can be drawn from the foregoing data,
however incomplete? Some degree of regionalization is, of course, already
widely accepted for complex procedures such as organ transplantation, much
of pediatric surgery, and the surgical care of major burns. Our own data (see
Table 3) reveal that in 1974-1975 there was already a marked concentration
of cardiac procedures and considerable conce~tration of total hip replacements
and of abdominal aortic grafts. Thus, although all but 17 of the 1498 hospitals
in our study performed at least one cholecystectomy in 1974 or 1975, only
587 performed open heart surgery; furthermore, the 14 per cent of hospitals
having more than 200 annual procedures accounted for 35 per cent of all the
cholecystectomies, whereas the 8 per cent of 587 hospitals having more than
200 annual procedures accounted for 56 per cent of all open heart procedures.
Even today, regionalization of coronary artery bypass grafting is far from
complete, and large numbers of hospitals apparently fall far short of the
standards recommended by the Cardiovascular Committee of the American
College of Surgeons in 1973 and by the lntersociety Commission for Heart
Disease Resources in 1975. In agreeing that further regionalization is a worthy
goal, Longmire and Mellinkoff have urged that such further regionalization
be voluntarily undertaken, based on standards "similar to those already
established for open-heart surgery, delivery, and perinatal services. Such
standards might include case load, qualifications of the professional staff, the
essential support services, and the specialized facilities required for the
designated procedure." 12
Whether such voluntary efforts on the part of individual physicians will
succeed remains to be seen. The need to control costs may be too great to
wait for this to occur. Thus, Blue Shield of California is currently exploring
the feasibility of contracting with better qualified hospitals and physicians
for heart surgery at set prices substantially lower than "usual and customary." 19
Similar arrangements for other procedures, such as total hip replacement and
surgical procedures for cataracts, may subsequently be considered.
The regionalization of procedures that are already well established will,
no doubt, be difficult to achieve. The regionalization of new and experimental
666 J. P. BUNKER ET AL.

procedures may be a more realistic goal. When a new treatment is in an

experimental phase it should, almost by definition, be limited to those
institutions qualified to carry out a reliable evaluation. This is, of course,
precisely how new drugs are introduced. Although the problems encountered
in the introduction of an operation are very different from those encountered
with new drugs, nevertheless, there is a great and currently unmet need for
a standard process by which to evaluate new operations. With the recent
demise of the National Center for Health Care Technology and the current
acute shortage of other funds to support the evaluation of new procedures,
we have proposed the creation of a private "Institute for Health Care
Evaluation.'' 2 Such an institute, if adequately funded, might be able to
identify the resources and skills appropriate to given procedures and thus to
help in the safe, rational, and cost-effective diffusion of new techniques.
We strongly urge that new surgical procedures be carried out initially in
selected institutions, to allow for their efficient evaluation. In making this
recommendation, it is acknowledged that longer term regionalization, or at
least partial regionalization, is a likely consequence. This has already oc-
curred in the case of total hip replacement as a result of the selective
awarding of investigational new drug licenses (INDs) for the use of methyl-
methacrylate by the Food and Drug Administration between 1969 and 1971; 2
it can be expected to occur in the future for operations that involve the use
of devices, and that therefore also come under the jurisdiction of the Food
and Drug Administration.
We also encourage regionalization of complex procedures, such as
coronary artery bypass grafting and transurethral resection of the prostate,
for which it has been or can be demonstrated that mortality is inversely
related to the volume of experience. In making this recommendation, we do
not suggest that all such complex procedures be concentrated in large
hospitals. Indeed, our data suggest that after controlling for the volume of
the procedure in question, large hospitals offer no advantage in terms of
mortality. 14 Regionalization might occur in relatively small hospitals, with
certain procedures being carried out in one group of hospitals, others in
another.
Whatever the merits of specialization, and whether or not regionalization
is undertaken by choice or by imposition, the medical profession as a whole,
and surgeons in particular, have a responsibility to the public that they can
and should undertake at once: to make public their current volume of
experience. A physician or surgeon has the responsibility to inform his or
her patient when a new procedure is to be undertaken for the first time, or
during the early clinical trials, and it is assumed that they fulfill this
responsibility. By the same token, a physician, or a hospital, should undertake
the responsibility to inform prospective patients of their experience with
particular procedures. This may seem an unfair obligation with which to
saddle the hospital that has good results despite a small volume. It can be
assumed, however, that such a hospital can establish a reputation as good as
its results, certainly with referring physicians, who can be expected to
continue to play a dominant role in the referral process, and indirectly with
their patients.
Finally, we should call attention to the potential conflict of the goal of
SHOULD SURGERY BE REGIONALIZED? 667
surgical regionalization with the goal of "improved geographic distribution
of truly trained surgeons." 17 The numbers of surgeons trained per year
continues to increase and in 1979 was double that necessary to meet national
needs, as estimated by the American Surgical Association and the American
College of Surgeons in their Study on Surgical Services for the United States
(SOSSUS). 21 Regionalization will certainly place some constraints on the
practice opportunities for new surgeons. Resolution of this potential conflict
can be accomplished only to the extent that the problems of organization and
manpower of medicine as a whole are addressed.
Nevertheless, most of the practice of surgery will remain unaffected by
a regionalization that is limited to a small number of complex procedures.
For a hospital carrying out a half dozen heart procedures a year, for example-
and there are apparently many of them-regionalization will not seriously
disrupt the services offered by that hospital, or of the surgeons involved, for
whom so few cases must represent a very small part of their practice. Thus,
regionalization can have small overall impact on surgical practice, but at the
same time it can have a large impact on the adverse consequences of high
risk operations that are performed only occasionally.

REFERENCES
1. Adams, D. F., Fraser, D. B., and Abrams, H. L.: The complications of coronary arteriography.
Circulation, 48:609, 1973.
2. Bunker, J. P., and Fowles, J.: Assessment of medical technology strategies: Coverage and
reimbursement options for experimental medical and surgical procedures. Washington, D.C.,
Office of Technology Assessment, U.S. Congress, September 1981.
3. Chalmers, T. C.: Personal communication.
4. Chalmers, T. C., Smith, H., Jr., Ambroz, A. et al.: In defense of the VA randomized control
trial of coronary artery surgery. Clin. Res., 26:230, 1978.
5. Cretin, S., and Roberts, N. K.: Surgical care for cardiovascular disease in California: Present
status and future policy. California Policy Seminar, Monograph No. 8, Institute of Govern-
mental Studies, University of California at Berkeley, 1981.
6. Divoky, D.: The Sacramento Bee, June 10, 1981.
7. Farber, B. F., Kaiser, D. L., and Wenzel, R. P.: Relation between surgical volume and incidence
of postoperative wound infection. N. Engl. J. Med., 305:200, 1981.
8. Finkler, S. A.: Cost-effectiveness of regionalization: The heart surgery example. Inquiry,
16:264, 1979.
9. Hotchkiss, W. S.: Patent ductus arteriosus and the occasional cardiac surgeon. }.A.M.A., 173:244,
1960.
10. Knox, R.: Boston Globe, February 23, 1976.
11. Kotulak, R.: Chicago Tribune, May 4, 1980.
12. Longmire, W. P., and Mellinkoff, S. M.: Regionalization of operations. N. Engl. J. Med.,
301:1393, 1979
13. Luft, H. S.: The relation between surgical volume and mortality: An exploration of causal
factors and alternative models. Med. Care, 18:940, 1980.
14. Luft, H. S., Bunker, J. P., and Enthoven, A. C.: Should operations be regionalized? N. Engl.
J. Med., 301:1364, 1979.
15. McGregor, M., and Pelletier, G.: Planning of specialized health facilities. N. Engl. J. Med.,
299: 179, 1978.
16. Pettigrew, A. H.: Final report of the Massachusetts maternity and newborn regionalization
project. Boston, Massachusetts Department of Public Health, August 1, 1974-September 30,
1976.
17. Polk, H. C.: The Sky is not Falling. Presidential Address, Society of University Surgeons.
Surgery, 88: l, 1980.
18. Public Health Service: National Guidelines for Health Planning. Washington, D.C., Department
of Health, Education and Welfare, Federal Register, Vol. 43, No. 60, March 28, 1978.
668 J.P. BUNKER ET AL.

19. Schaffarzick, R.: Personal communication.

20. Schwarz, W. B., and Joskow, P. L.: Duplicated hospital facilities. N. Engl. J. Med., 303:1449,
1980.
21. Surgery in the United States: A summary report of the study on surgical services for the United
States. American College of Surgeons, Chicago, 1975.
22. Waterman, D. H., Samson, P. C., and Bailey, C. P.: The surgery of patent ductus arteriosus.
Presented in the Symposium on Cardiovascular Surgery, 21st Annual Meeting of the American
College of Chest Physicians, Atlantic City, New Jersey, June 4, 1955.

Department of Family, Community and Preventive Medicine

Stanford Medical Center
Stanford, California 94305