Variation in Rates of Utilization of Surgical Services in the Commonwealth of Massachusetts
on
Benjamin A. Barnes, MD; Elizabeth O'Brien, MSPH; Catherine Comstock, MPH; Deborah G. D'Arpa, MBA; Charles L. Donahue, MA
This article presents an analysis of over 140,000 selected surgical procedures performed in Massachusetts in 1980, giving the per capita rates of 14 common procedures and of four less frequently performed procedures. The analysis defines 172 geographic areas for the commonly performed procedures and 45 for the less frequently performed procedures. Per capita surgical rates among the defined areas are significantly different from both a statistical and a clinical point of view. Twofold and threefold variations occur frequently across geographic areas. In certain areas, some surgical services appear to be provided at rates substantially different from the statewide rate. We discuss the importance of these data for physicians as well as the implications for the distribution and quality of clinical care and for containment of medical
care
Report
ongoing activity.1
The focus of this analysis is on surgical rather than medical services, because precisely defined and coded procedures are well suited to unam¬ biguous rate analysis and because useful comparisons may be made with other reports on surgical rates and their cost implications.
METHODS Previous reports have presented the methodology of rate analyses of medical care within small geographic areas.24 In general terms, the analysis of rates of surgical utilization includes three major
costs.
(JAMA 1985;254:371-375)
FREQUENTLY voiced concern of physicians and their professional or¬ ganizations is their lack of involve¬ ment in the development of regula¬ tions and legislation affecting the health care system. Yet it is generally accepted that physicians are uniquely qualified by their training and experiA
For editorial comment see p 407. of
to participate in the formulation regulatory mechanisms, particu¬ larly when clinical judgments are
ence
necessary,
as
in cost-containment
or
distributive
strategies for medical
From the Health Planning Council for Greater Boston, Inc (Ms O'Brien and Mr Donahue); the Institute for Health Research, Harvard School of Public Health, Boston (Dr Barnes); and the Massachusetts Health Data Consortium, Waltham (Mss
Comstock and
for Greater Boston, Inc, 294 Washington 630, Boston, MA 02108 (Mr Donahue).
D'Arpa). Reprint requests to the
Health
Planning Council St, Suite
The active involvement of physi¬ cians would help in addressing ques¬ tions about the distribution, quality, and cost of medical care, and the purpose of this report is to provide useful data toward this end. We have assembled a data base for Massachusetts that allows the scruti¬ ny of per capita rates of selected surgical procedures for specified geo¬ graphic areas and enables physicians and other analysts to describe varia¬ tions in medical practice patterns. An example of the constructive use of such data is the action taken by the Maine Medical Society in 1983 to approve and support an intensive pro¬ gram to publicize and distribute per capita surgical rate analyses of health care. Professional specialty societies in Maine and the Health Facilities Cost Review Board, which is the state agency mandated to pre¬ pare public reports on hospital dis¬ charge data, are collaborating on this
care.
steps:
areas
sex-structure data for these areas from the US decennial census; and counting the number of procedures of interest per¬ formed on residents of these areas in a particular year. This methodology credits to the geographic area of residence of a patient the procedures performed in hospi¬ tals located in another geographic area. The Massachusetts study described herein follows these three steps. Counts of surgi¬ cal procedures in each geographic area were obtained from the Uniform Hospital Discharge Data Abstract and the Massa¬ chusetts Health Data Consortium gener¬ ated data reports. Geographic areas were defined by the aggregation of the 351 incorporated cities and towns across Mas¬ sachusetts and the 16 designated Boston neighborhoods into 172 population sub¬ groups. These 172 subgroups contain com¬ munities that, for the most part, are contiguous, have a minimum population of approximately 15,000, and usually contain within their boundaries one or more acutecare hospitals where the plurality of resi-
defining contiguous geographic primarily to the same hospital(s); obtaining population, age, and
that relate
Downloaded from www.jama.com by guest on August 1, 2009
Table 1.—Selected
Surgical Procedures and Annual Rates in Large Geographical Areas*
Massachusetts Data International Classification of Diseases, 9th Revision, Clinical Modification Codes 13.1-13.59 28.2, 28.3 36.10-36.14 37.7-37.77 45.7-45.79 47.0 49.4-49.49 51.2-51.22 Total No. Procedures Statewide In 1980 11,962
Annual Rates of Procedures per 1,000 Population Rhode
Maine Rate
2.09
Vermont
Procedure Lens extraction Tonsillectomy with
or
1980'
1.91
1975s
2.3
Island
1975*
3.6
United States
1978"
1.67
without adenoidectoi
Aortocoronary artery bypass graft
Insertion of cardiac pacemaker Partial excision of large intestine
7,213 1,824 4,688
1.26
0.32
2.97
2.96
0.44
0.82
i no
0.67
1 *V)
0.70 0.58
1.6
Appendectomy
Procedures
on
1.1
2.2
1.40
0.88 2.02 1.87
2.89
hemorrhoids
or
Cholecystectomy (partial or total) Unilateral repair of hernia (inguinal
femoral)
53.0-53.05;
53.2-53.29 60.2-60.69 68.3-68.7 with 65.565.52; 65.6-65.62 68.3-68.7 without 65.5 65.52; 65.6-65.62
2.18 2.37
0.9 2.0 2.7
2.2
2.6
2.5
Prostatectomy (all types) Hysterectomy (abdominal or vaginal) with oophorectomy Hysterectomy (abdominal or vaginal) without oophorectomy Cesarean section (all types) Reduction of fracture of femur (all types)
or destruction of intervertébral cartilage Excision of semilunar cartilage of kne Spinal fusion (all types) Total hip replacement
8,557 5,018
3.13
1.67 1.73
1.24
2.42 2.50
4.3
5.7
5.82
4.61
74.0-74.2; 74.4-74.99 79.05, 79.15, 79.25, 79.35
80.5 80.6
81.0-81.09 81.5-81.59
13,463 4,696
4.48 0.82
Excision
3,557 4,061 1,062 1,656
0.62 0.71
0.19 0.29
0.84 0.89
0.70
0.80
0.31 0.31
a
Maine for
"Comparison
sex-specific operations
of Massachusetts rates (1980) with Maine (1980), Vermont (1975), Rhode Island (1975), and United States rates are calculated on a base of the total population and are half as large as would be calculated with
(1978). The rates reported from sex-specific population base.
Table 2.—40 Subgroups With 20 Lowest and 20 Highest Mean Observed / Expected Ratios
Lowest Ratios
Highest Ratios
Hospital Service Area (Subgroup)
Amherst Back Bay
Health Systems
Agency Region*
1 4
4
Hospital Service Area (Subgroup)
Agawam
Attleboro
Health Systems
Agency Region
Brighton/Allston Cambridge
Charlestown Concord Dartmouth Harvard
4
4
Chicopee Fitchburg
Haverhill
4 5 2
4
Holyoke
Jamaica Plain
Lynn
Mansfield
Hingham Millbury
Newton
2
4 1 4 4
Palmer
Peabody
Pittsfield
Northampton
North End
Revere
Roxbury
South End Wareham
4
(Boston)
4
Raynham Southbridge Springfield Stoughton
Taunton Webster Westfield West Springfield
5 4
4
Wellesley
West End
Winthrop
Average population General fertility rate (per 1,000 women,
4
36,000
36
aged
15-44
yr)
'These regions
are
six
geographical
divisions used in state of Massachusetts health
planning
activities.
dents receive their health care. Further aggregation of these 172 population sub¬ groups produced 45 population groups, which were used in computing the rates of four procedures with a statewide rate of less than 0.5 per 1,000 population. Table 1 lists the surgical procedures for which rates were determined. In 1980, an incon¬ sequential number of these procedures were performed in ambulatory care facili¬ ties (in contrast to the number of tooth extractions or dilations and curettages, which are not considered in this report). We selected these procedures because they were sufficiently frequent to permit calcu¬ lation of stable rates for comparison with other studies or were of interest as innova¬ tions in therapy (eg, total hip replace¬ ment). Rates reported in other studies of the New England states and of the United States are also shown. We note, for exam¬ ple, that Massachusetts has a lower tonsillectomy rate and higher prostatectomy rate than reported in the other four stud¬ ies. We calculated an index to rank the 172 population subgroups according to their overall surgical activity as judged by the 14 procedures with a statewide rate great¬ er than 0.5 per 1,000 population. The purpose of the index is to weigh the specific types of procedures equally so that the high or low rate of a frequently performed procedure, such as cesarean section or prostatectomy, would not over¬ whelm the influence of a less frequently
Downloaded from www.jama.com by guest on August 1, 2009
—
=
Highest Mean O/E Ratios (n=20)
Lowest Mean O/E Ratios (n=20) Standard Error
=
--
I
o Q-
O O o
ra
er
1L
"-fr"
'I
/
cf
J_
_L
/
^
J
i il è /
O
Sà
i#
/
<*
/
s
T
j-
/
OQ &
£1 ?1.2 and
Procedure
Lens extraction
Statewide (A) 55,085
XJ>6.63) (B)
2,080 1,661 8,727 1,994
892
%
(BX100/A)
3.8
11.8 12.7 1.8 2.0
Tonsillectomy and adenoidectomy
Pacemaker insertion Partial colectomy
Appendectomy Cholecystectomy Inguinal hernia repair Prostatectomy
Hysterectomy Hysterectomy
and
oophorectomy
Cesarean section Fracture of femur reduction Intervertébral disk excision
Knee cartilage excision
Total
14,105 68,711 110,904 43,671 138,037 77,544 145,351 49,209 45,422 94,878 150,761 49,578 26,572 1,069,828
4,727
649
3.4
0.8 1.9
2,735 1,149 2,272 7,260 6,324 3,240 1,198 44,908
2.3
5.0 7.7
4.2
6.5
4.5
4.2
such as pacemaker insertion or intervertebral disk excision, which may have an unusual rate as well. Thus, for each of the population sub¬ groups, 14 observed/expected (O/E) ratios were averaged; the 172 population sub¬ groups were then ranked by these mean O/E ratios.
performed procedure,
RESULTS
observing the symmetrical bell-shaped distribution of the mean O/E ratios of the 172 population subgroups, we identified the 40 areas with the 20 highest and 20 lowest mean values. These are listed alpha-
After
except for the prostatectomy rates. The difference in the rates of the remaining procedures have a statisti¬ cal significance, with P<.05 in every comparison and P<.001 for nine of the procedures. For tonsillectomy, insertion of pacemaker, and interver¬ tebral disk excision, the ratio of the average rates of the high-use popula¬ tion subgroups to the average rates of the low-use population subgroups ex¬ ceeds two. Hysterectomy without oophorectomy and knee cartilage ex¬ cision were performed over 90% more frequently in the high-rate group than in the low-rate group. The least relative differences in rates are seen in prostatectomy (not statistically significant), lens extraction, hysterec¬ tomy with oophorectomy, reduction of
betically in Table 2. In Fig 1 the mean rates of the lowest and highest sub¬ groups have been plotted for the 14 commonly performed procedures. The two curves are a profile of the aver¬ age surgical activity in the two cate¬ gories of 20 subgroups. The 14 paired points on the two curves are statisti¬ cally distinct by the t test (á/=39),
Downloaded from www.jama.com by guest on August 1, 2009
fracture of femur, and partial colectomy. We are comparing here the aver¬ age of 20 rates for a specific proce¬ dure in the 20 population subgroups with the highest mean O/E ratios to the average in the 20 population sub¬ groups with the lowest mean O/E ratios, and this calculation should not be confused with the outdated "ex¬ tremal quotient" concept discussed by
made to the state rate, which possibly gives the impression that this overall rate has some authoritative, norma¬
Willemain.7
The cost implications of variations in the rates of surgical procedures throughout the state are estimated conservatively in Table 3. We have tabulated the additional bed days that have accrued for the 14 proce¬ dures when their rate of performance in a population subgroup was at least 20% greater than the 1980 Massachu¬ setts rate and when this higher rate had a statistical significance with a P value of <.01. For example, lens extractions were performed during
tive value. This is not the correct interpretation in our opinion, since further clinical investigations may modify medical perceptions as to the indications for any procedure. These modifications could result in a de¬ crease or increase in utilization in successive years. In recent history, tonsillectomy and
aortocoronary artery bypass graft
decreasing and increasing rates, respectively, and the overall state rate represents a moving aver¬ age over time, hardly a normative
have had
standard. We use the state rate as a point of reference to facilitate com¬ parisons and to help us define the distributions of rates, which neces¬ sarily are grouped around the state rate. If medical opinions agree on an
and low-rate areas through studies such as this one may serve to draw attention to medical practices that could be modified for the benefit of the patient and as a contribution to cost containment. When these tar¬ geted, special reviews uncover sub¬ stantial differences in professional opinions as to the optimal rate for a procedure, expert opinion or informa¬ tion derived from the medical litera¬ ture could determine a preference for a particular rate, high or low. In some situations controlled, clinical trials would be justified to provide solid evidence for these determinations. The differences in procedure rates between the 20 high- and 20 low-use
high-
population subgroups (Fig 1) repre¬ sent statistically significant depar¬
tures from the statewide rates. These disparate rates are of clinical concern
hospitalizations totaling 55,085 days. Of that number, 2,080 days were the result of hospitalizations in sub¬ groups where the procedure was being performed at a rate at least
rate, quite possibly different from the state rate, the
appropriate
new
20% greater than the state rate of 2.09 per 1,000 population, and where a shorter average length of stay did not compensate for the effect of this increased surgical rate on total bed occupancy. The total of these addi¬ tional bed days is 44,900, or about 4% of the total number of bed days utilized across the state for these 14 procedures. The estimated costs for these 44,900 days at an average cost of $358 per diem in 1980 for nongov¬
cost estimate of $16 million based on Table 3 conveys only a general
care
analysis could use this new rate reference point. Consequently,
as a
the
zation 20% greater than the state average induce and emphasizes the importance of reviewing high rates as part of a cost-containment strategy.
impression of the size of the health expenditures that rates of utili¬
A professional consensus defining a lower rate as optimal would result in a larger cost estimate, and vice
versa.
ernmental, short-term, general hospi¬
tal services in Massachusetts8 are in excess of $16 million. A more refined estimate would use procedure-specific costs, but these are not available for the statewide experience in 1980. Estimates of costs for a specific ser¬ vice cannot be directly equated with dollar savings if the service were terminated, because hospitals have substantial administrative problems in reducing overhead costs in propor¬ tion to service cuts. Thus, the savings realized from decreasing a service would seldom equal its original esti¬ mated costs.
COMMENT
We have two concerns regarding the interpretation of the study's results. First, reference is repeatedly
Second, many factors enter into the determination of the utilization of medical care. It would be unwise to conclude a simple explanation is gen¬ erally available. The balance between demand from patients and supply by surgeons is a subtle one, with the expectations of the patient depending on health status, education, socioeconomic factors, and cultural values, among many other variables. The interest of physicians is focused on the role played by professional judg¬ ment in the determination of the rates of procedures. In situations where judgments are largely account¬ able for hospital admission, eg, for prostatectomy for nonmalignant con¬ ditions, physicians may wish to be particularly certain that the indica¬ tions for hospital admission are con¬ sistent and rational across communi¬ ties. The identification of outlying
and importance since variation by a factor of two or more in the utiliza¬ tion of medical care across these communities requires some explana¬ tion. For example, an analysis was made of the differing cesarean section rates noted in the two categories of subgroups in Table 2. This analysis disclosed that the mean general fer¬ tility rate (an average of the number of live births per 1,000 females, aged 15 to 44 years) of the 20 subgroups with high cesarean section rates was 64, whereas the mean general fertility rate of the 20 subgroups with low cesarean section rates was 36. This almost twofold difference in fertility rates explains the difference between the cesarean section rates for these two categories of subgroups. When cesarean section rates are expressed per live births, there is no significant difference between the high and low rates. With the exception of this
procedure,
our
have found no differences in the pro¬ cedure rates depicted in Fig 1 that are explained by other simple circum¬
stances.
preliminary inquiries
The importance of these findings for the definition and practice of high-quality medical care is the possi¬ ble unrecognized existence of overutilization or underutilization. If ei¬ ther phenomenon is present, the patient is not receiving optimal care. The medical profession should, there¬ fore, be concerned about the causes of the variations in the rates of proce¬ dures and also about the definition of
Downloaded from www.jama.com by guest on August 1, 2009
optimal rate or range of rates for a specific procedure. We believe physicians' decisions are of critical importance here, but this report is not primarily concerned with the evidence in favor of profes¬ sional judgment playing a command¬ ing role in the determination of pro¬ cedure rates. In brief, this evidence has four parts. Clinical judgments of physicians have traditionally played a
an
dominant role in the selection of therapy, a role entirely appropriate and expected considering their pro¬ fessional interests in caring for patients. The hierarchy of procedures ranked according to the variation of their per capita rates parallels in a convincing fashion the discretionary nature of the surgical intervention, a correlation noted by Wennberg and Gittelsohn' and McPherson et al.10 Modification of the rates of proce¬ dures has followed deliberate educa¬ tional programs to alter physicians' perceptions of the indications for sur¬
dures within population groups, the fivefold difference between the rates of the two operations and the absence of a significant correlation strongly support the part played by opposite professional judgments in determin¬ ing the utilization of these two proce¬ dures. Possible variations in the inci¬ dence of ovarian disease or in other indications for oophorectomy, al¬ though not ruled out by this study, seem to be distinctly less likely
causes.
originally reported by Wennberg.'3 Limiting the count of additional days in Table 3 to only those population subgroups that have a surgical rate for a specific procedure 20% or more
The health care costs induced by high rates of utilization are substan¬ tial and raise issues of equity and cross-subsidization of health care, as
gical interventions, as reported by Wennberg" and Dyck et al.12 The
correlation of rates of certain proce¬ dures performed by the same special¬ ists within their communities is low. For example, in this study the corre¬ lation coefficient of the rate of hys¬ terectomy without oophorectomy and the rate of hysterectomy with oopho¬
45 population groups (P=.3). The ratio of the simple hysterectomy rate to the rate of hysterectomy with oophorectomy ranged from 0.5 to 2.8. Since the surgical specialists, hospital facilities,
rectomy
was
across
-.16
and
access
to medical
care are essen¬
tially identical for these
two proce-
above the state rate with P<.01 underestimates the number of addi¬ tional bed days. The estimated in¬ duced cost of $16 million clearly sug¬ gests that monitoring high rates might be a rewarding part of a costcontainment strategy. Of course, whether such costs are warranted depends on the judgment as to the "best" or "proper" rate. The importance of these findings for cost-containment strategies seems clear. The difference in health care expenditures by public and private funds resulting from different rates of procedures and consequent hospi¬ talizations far exceeds the relatively minor cost effects of adjusting lengths of hospital stay associated with a given procedure."1 Thus, a hospital endeavoring to comply with a
References
mandated cap on its budget, such as required by the legislation of Chapter 372 in Massachusetts, may conduct staff reviews of the indications for admission for patients recommended for procedures known to be performed at a rate substantially in excess of the statewide rate in communities the hospital serves. This administrative strategy will not be available to all hospitals equally, since some are located in communities where the responsibility for providing a particu¬ lar procedure is shared among several institutions, as demonstrated in the Massachusetts Patient Origin Re¬ port." On the other hand, where one or two hospitals dominate a popula¬ tion subgroup, as is generally the case outside of major urban areas, such reviews may be helpful. Active collab¬ oration by the professional staffs or administrations of several responsi¬ ble hospitals involved could pinpoint discordant practices and develop cost-
saving measures.
rates of
The determination of per capita procedures and of medical diagnoses annually will provide the medical profession with information about health care consumption on a community basis and about the varia¬ tions in these rates, which would in turn require special inquiries. With¬ out such a data base, rational plan¬ ning by physicians and others will be insensitive to specific community needs.
This study was supported in part by the Health Planning Council for Greater Boston Inc. We are indebted to Gertrude Garelles for her graphics and to Pearl K. Russo and David Gute, PhD, for their careful readings of the manu¬
script.
1. Incidence Patterns of Common Surgical Procedures in Maine 1978 & 1980. Augusta, Health Facilities Cost Review Board, State of
Maine, 1981. 2. Wennberg J, Gittelsohn AM: Small
area
variations in health care delivery. Science 1973; 182:1102-1108. 3. Barnes BA: Population-based small-unit analysis of health care, in Rothberg DL (ed): Regional Variations in Hospital Use: Geographic and Temporal Patterns of Care in the United States. Lexington, Mass, DC Heath & Co, 1982. 4. Louis T, DerSimonian R: Health statistics based on discrete population groups, in Rothberg DL (ed): Regional Variations in Hospital Use: Geographic and Temporal Patterns of Care in the United States. Lexington, Mass, DC Heath & Co, 1982. 5. McPherson K, Wennberg JE, Hovind OB, et al: Small area variations in the use of common surgical procedures: An international comparison of New England, England and Norway. N
Engl J Med 1982;307:1310-1314. 6. Haupt BJ: Detailed Diagnoses and Surgical Procedures for Patients Discharged From Short\x=req-\ Study Hospitals, 1978, publication (PHS) 80-1274. US Dept of Health and Human Services, 1978. 7. Willemain TR: On the comparison of high-
est and lowest surgery rates in small-area
tions in Hospital Use: Geographic and Temporal Patterns of Care in the United States. Lexington, Mass, DC Heath & Co, 1982. 8. Average cost to community hospitals per patient, Massachusetts, 1980, in Statistical Abstract of the United States: 1982-83, ed 103. US Bureau of the Census, 1982, Table 175. 9. Wennberg J, Gittelsohn AM: Health care delivery in Maine: I. Patterns of use of common surgical procedures. J Maine Med Assoc 1975;
studies, in Rothberg DL (ed): Regional Varia-
66:123-130, 149.
10. McPherson K, Strong PM, Epstein A, et al: Regional variations in the use of common surgical procedures: Within and between England and
Effect of surveillance on the number of hysterectomies in the province of Saskatchewan. N Engl J Med 1977;296:1326-1328. 13. Wennberg JE: Should the cost of insurance reflect the cost of use in local hospital markets? N Engl J Med 1982;307:1374-1381. 14. Wennberg JE, Gittelsohn AM: A Small Area Approach to the Analysis of Health System Performance, 1980, publication (HRA) 80-14012. US Dept of Health and Human Services, 1980. 15. Massachusetts Patient Origin Report Fiscal Year 1980: Community Based Small Area Analyses. Reports 2, 3, and 4: vols 6-8. Waltham, Massachusetts Health Data Consortium Inc, 1980.
Wales, Canada and the United States of AmeriSoc Sci Med 1981;15A:273-288. 11. Wennberg JE: Physician uncertainty, specialty ideology, and a second opinion prior to tonsillectomy. Pediatrics 1977;59:952. 12. Dyck FJ, Murphy FA, Murphy JK, et al:
ca.
Downloaded from www.jama.com by guest on August 1, 2009