Variation in hysterectomy rates across small geographic areas of Massachusetts

Susan Haas, MD, MS; David Acker, MD,c, d Charles Donahue, MA,. and Martha Ellen Katz, MDa, b

Boston, Massachusetts

OBJECTIVE: Our objective was to determine whether small-area variation in hysterectomy rates exists within Massachusetts and, if so, whether regions with persistently high or low rates can be identified,

STUDY DESIGN: By examining data derived from standardized hospital discharge forms between 1982 and 1989, we compared hysterectomy rates among 172 small geographic areas within Massachusetts, The ratios of actual to expected number of hysterectomies for each area were also computed, Statistical significance of rate variation was established by a value of p < 0,01 (x2 test),

RESULTS: The hysterectomy rate per 1000 women varies fivefold across small-population areas of

Massachusetts, Eight areas had persistently high and five had persistently low rates compared with the statewide average.

CONCLUSION: Possible explanations for these rate variations include indigenous population factors, supply of health care resources, income, insurance coverage, methodological problems in data analysis, and variation in professional judgment with regard to the indications for hysterectomy. Only a combination of the study of local causes of high and low rates and the study of the clinical outcomes of different treatment decisions will allow physicians to develop an informed consensus on appropriate indications for surgery. (AM J OBSTET GYNECOL 1993;169:150-4.)

Key words: Small-area variation, hysterectomy, physician practice patterns, health services research

Variation in the rate at which many surgical proce­dures are performed has been documented since 1938. I Such variation has more recently been shown to occur between countries, between regions in the United States, between urban teaching hospitals, and between contiguous small geographic areas with populations as small as 45,000. 2

.10 Hypotheses to explain this variation

include differences in patient population (e.g., income, education, or rates of disease), in local health care policies (e.g., hospital admission vs outpatient treat­ment), in insurance coverage and fee reimbursement, in physician supply, and in individual physician practice style."·1S An example of the last hypothesis is the physician's choice of indications for surgery in cases in which the procedure is elective and medical manage­ment and observation without intervention are accept­able alternatives.

From the Department of Obstetrics and Gynecology, Brigham and Women's Hospital Harvard Medical School: Children's Hospital Medical Center,' Boston City Hospital,' Boston University School of Medicine/ and the Health Planning Council for Greater Boston, Inc.' Received for publication June 27, 1992; revised January I I, 1993; accepted January 27, 1993. Reprint requests: Susan Haas, MD, Department of Obstetrics and Gynecology, Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115. Copyright © 1993 by Mosby-Year Book, Inc. 0002-9378/93 $1.00 + .20 6/1/45945

150

Approximately 650,000 hysterectomies are per­formed annually in the United States.16 Despite the widely held belief that gynecologic disorders are rea­sonably uniform in their geographic distribution, rates of hysterectomy have been shown to differ among large geographic areas. For example, a woman of reproduc­tive age living in the southern United States is 2.5 times more likely to have a hysterectomy than a woman living in the Northeast. '6

More than 10,000 hysterectomies were performed annually in Massachusetts between 1982 and 1989. With the aid of a new data base that can aggregate hospital discharge records according to the patient's residence, we have documented the variation in hyster­ectomy rates among small geographic areas within Massachusetts. Further, we have defined areas that

consistently have rates significantly higher or lower than the state mean.

Material and methods

The general method of small-area analysis has been detailed extensively elsewhere. '7 This method involves

defining small geographic areas; obtaining population, age structure, and gender data from the most recent U.S. census; and counting the number of procedures of interest (i.e., hysterectomy) performed on residents of that geographic area in any year. The unique aspect of

Volume 169, Number I Am J Obstet Gynecol

this method is that all procedures performed on resi­dents of a geographic area, whether performed locally or at distant hospitals, are credited to the geographic area in which a patient resides, Thus the analysis is population-based rather than hospital-based, We used data from the Massachusetts Rate Setting Commission, which collects and analyzes information from uniform, standardized hospital discharge data abstracts. Data for each patient include age, gender, discharge diagnoses, surgical procedures, length of stay, zip code of resi­dence, hospital, and payor source. The Health Planning Council of Greater Boston aggregated the population of Massachusetts into 172 geographic subgroups with a minimum population of 15,000 people who live pre­dominantly in one hospital service area.

The actual number of nonradical hysterectomies per­formed on patients in each geographic subgroup was recorded for each year from 1982 to 1989. On the basis of the most recent (1980) census figures, we calculated the crude rate and the (direct) age-adjusted hysterec­tomy rate per 1000 women. (Age adjustment controls for differences in the age distribution of women - and thus their likelihood of undergoing hysterectomy - in different communities.) The X2 test was used to assess whether the age-adjusted hysterectomy rate in each subgroup differed significantly from the state mean rate.

The expected number of hysterectomies in each sub­group is the number of surgical cases that would have occurred had the age-specific rates found in the com­munity been the same as those in the state as a whole. We calculated the expected number of hysterectomies for each subgroup by multiplying the subgroup's rate for each stratum by the statewide fraction of population in that age bracket and then deriving the sum of the results. We computed the actual-to-expected ratio for each subgroup. This ratio represents the number of hysterectomies that actually occurred in the subgroup divided by the number expected had the statewide rate been found in that population. Massachusetts sub­groups were ranked by actual-to-expected ratio for hysterectomy from highest to lowest ratio.

Geographic subgroups in which numbers of hyster­ectomy discharges were significantly different from those of other subgroups in any year were defined both statistically and clinically. Statistical significance was established by a value of p < 0.01 in a X2 test that compared the subgroup rate with the statewide rate. Clinical significance was established by an actual-to­expected ratio of > 1.2 or < 0.8. Subgroups that met the criteria for both statistical and clinical significance in three or more of the study years are reported. A test for multiple comparisons was not used because we used an initial value of p < 0.1 and because we required that a subgroup rate be significantly different from the state

Haas et al. 151

rate in a minimum of 3 separate years. The odds of a subgroup having a value of p < 0.01 in 3 years is 0.01" or 0.000001 by chance alone.

Final data are not available for 1988. Preliminary analyses indicated that results were similar to those for other years.

Results

During the years 1982 through 1987 and 1989, the mean age-adjusted Massachusetts hysterectomy rate was consistently close to 3 per 1000 women per year. The exact rates were 3.21,3.10,3.12, 3.05, 2.98, 2.96, and 2.86 for the years 1982 through 1987 and 1989, respectively. The lowest age-adjusted rate was approx­imately 1.3 per 1000 during those years. The specific annual rates for each year were 1.59, 1.69, 0.78, 1.64, 1.14, 1.48, and 1.14, respectively per 1000 women. The highest age-adjusted rates were > 5 per 1000 women per year. Specifically, for each of the study years they were 6.13, 5.14, 5.16, 5.36, 5.07, 5.64, and 4.87 per 1000, respectively,.

The highest and lowest actual-to-expected ratios in­dicate how many times higher or lower than the state­wide average the furthest outliers were. The highest ratios in the study years were 1.90, 1.65, 1.66, 1.78, 1.66, 1.91, and 1.70, respectively, and the lowest ratios were 0.50, 0.53, 0.24, 0.50, 0.36, 0.49, and 0.33, re­spectively. Thus there was generally a 70% excess of hysterectomies in the high-rate geographic areas and a 50% reduction in the low-rate geographic areas.

Eight subgroups had persistently high numbers of hysterectomies in the study years. Subgroups with rates significantly greater (p < 0.01) than the state mean and an actual-to-expected ratio > 1.2 in 3 or more of the study years are depicted in Fig. 1. The subgroups with persistently higher rates and ratios (South Dorchester, Bourne, Plymouth, Duxbury, Fitchburg, Attleboro, Hav­erhill, and Uxbridge) are located on the perimeter of a ring approximately 25 miles from Boston. Five sub­groups (Newton, Cambridge, Brighton, East Boston, and Worcester) had hysterectomy rates that were signif­icantly lower (p < 0.01) than the state mean and an actual-to-expected ratio < 0.8 for the years 1982 through 1987 and 1989. They are also depicted in Fig. 1. The subgroups with low rates are located in metro­politan areas (Boston and Worcester).

Comment

Our results show that for a woman living in Massa­chusetts the risk of undergoing a hysterectomy varies widely, depending on the area of the state in which she lives. During the years 1982 through 1989, Massachu­setts' small-geographic-area hysterectomy rates ranged from one to five operations per 1000 women. Eight small geographic areas had persistently high and five

152 Haas et al.

High rate areas

Ell Low rate areas

July 1993 Am J Obstet Gynecol

Fig. 1. Map of Massachusetts geographic areas. Those with hysterectomy rates significandy greater (P < 0.01) than the state mean and actual-to-expected ratios> 1.2 in at least 3 of the 8 study years are blacked out. Areas with low rates (p < 0.01) and ratios < 0.8 in ;:: 3 of 8 years are shaded.

areas had persistently low numbers of hysterectomies. Explanations for observed variation in surgical rates

fall into four categories: (1) indigenous population factors that include health status, sociodemographic differences, and demand for services; (2) supply of health care resources such as physicians, hospital beds, insurance coverage, and alternative therapies; (3) meth­odological problems inherent in the technique of data gathering and analysis; and (4) variation in physician opinion about appropriate indications for surgery.

Several authors have searched for population-based causes of differences in the rate with which surgery is performed. Roos 18 identified women from a Catholic background and those who made five or more visits to different physicians per year ("doctor shoppers") as significantly more likely to undergo hysterectomy than women without those sociodemographic characteristics. Two recent studies of cardiac angiography and coronary artery bypass grafting demonstrated higher procedure rates for white patients than for black patients when the authors controlled for all other risk factors. 19. 20 Re­searchers in Maryland evaluated this risk and did not find a race-based difference in hysterectomy rates. 21

An inadequate supply of health care resources might explain low hysterectomy rates. In Massachusetts, per-

sistently low rates were found ~mly in metropolitan areas (four areas in Boston and one in Worcester). Both cities have large medical centers with a more than adequate physician and bed supply. Thus lack of re­sources cannot explain the low surgical rates in certain areas of Massachusetts.

As the technique of small-area analysis has increased in popularity, methodological problems have been noted and addressed. Readmission for the same oper­ation, miscoding of the procedure, and use of outpa­tient surgery create problems in data analysis for other operations. Fortunately, hysterectomy is performed only once on each patient, was virtually always an inpatient procedure before 1990, and is easily under­stood by medical coders. As a result, these data-quality issues are small or nonexistent for hysterectomy. Be­cause hysterectomies are most frequently performed on women in the third to fifth decade of life, the popula­tion-age distribution in a geographic area could skew the hysterectomy rate. We controlled for this possibility by adjusting our surgical rates for the age distribution in each geographic area. Random statistical fluctuations can give rise to statistically significant differences when multiple tests are performed on the same data. We used a stringent definition of statistical significance

Volume 169, Number 1 Am J Obstet Gynecol

(P < 0.01) to decrease the incidence of positive find­ings. At this significance level, we anticipated that two subgroups would have a statistically significant differ­ence from the mean rate by chance alone in each year analyzed. By considering reportable only those geo­graphic areas with persistently high or low rates in ;::: 3 years, we avoided the impact of falsely assigning signif­icance where random variation might be the cause. The odds of a subgroup appearing in Fig. 1 as a result of chance alone is 0.01' or 0.000001. A final methodolog­ical problem is that the true denominator for hysterec­tomy rates is difficult to determine, and results may be distorted if the population base of all women is taken as the denominator. If a high proportion of women in one geographic area have already had a hysterectomy, fewer are available to undergo surgery in any year. Thus the hysterectomy rate might appear artificially low in that area. On the other hand, a rate that appears average might actually be high if it reflects many operations on the small pool of women with intact uteri who are actually available for surgery. Because we lack data on the number of women in each area who have already undergone hysterectomy, it is not possible to control for this factor in the analysis. Others have estimated that the true hysterectomy rate would be increased by 25% if

the underlying rate of previous hysterectomy could be taken into account.22. 23

A final explanation of variation in surgical rates is called "professional uncertainty"; this term refers to the understanding that physicians may legitimately differ in their professional opinion as to what constitutes an appropriate indication for surgery. For example, one physician may think that any irregular bleeding trou­blesome to the patient is an indication for hysterectomy, whereas another may believe that bleeding must be severe enough to cause anemia before a hysterectomy will be considered. Likewise one physician may think that an asymptomatic fibroid uterus larger than a cer­tain size should be removed, whereas another believes that symptoms should determine the need for hyster­ectomy. Other authors have shown that where there is professional certainty, as in the case of inguinal hernia repair or reduction of a femoral fracture, surgical rates do not vary widely across geographic areas. 7 However, where surgeons' indications vary more widely, as in the case of tonsillectomy, greater rate variation is found.

Eleven percent of hysterectomies are performed for nonelective indications (i.e., cancer or obstetric catas­trophe), whereas 89% are performed for elective indi­cations such as fibroids, uterine prolapse, abnormal bleeding, pelvic pain, and endometriosis. Given the high percentage of elective indications for hysterec­tomy, our data, as well as data from Maryland,21 show a relatively low degree of variation compared with that for other surgical procedures.

Haas et al. 153

Wennberg and Gittelsohn12 have described the "sur­gical signature" left on some geographic areas by the particular combination of specialties and preferences of surgeons practicing there. We wondered whether geo­graphic areas of Massachusetts with high or low hyster­ectomy rates would also have persistently high or low cesarean section rates. Only one area identified as a high cesarean section area in our previous works

-

Attleboro - also has a persistently high hysterectomy rate. Of the other 13 areas with a high or low cesarean section rate and the 12 areas with a high or low hysterectomy rate, no area was found to have high or low rates for both procedures.

How shall obstetrician-gynecologists understand and use small-area variation data whether the data are for hysterectomy, cesarean section, or any other procedure? We believe a four-step process is necessary. The first step is to understand the computation and significance of the rates. The mean rate is simply the statewide average. There is no reason to assume that it is the "correct" or "best" rate. The lowest or the highest rate in any state may be correct, and the best rate may vary, depending on the population under study.

The second step is to compare areas with high and low rates and to attempt to determine what unmea­sured variables are responsible for the differences. Lo­cal causes of excess or deficiency can then be addressed at the local level. When information about high hyster­ectomy rates in one Maine city was communicated back to local gynecologists, the rate dropped within 2 years from approximately five to three operations per 1000 women per year. 24

Third, more studies of clinical outcome are critically needed to aid in individual decision making and in consensus formation by the specialty as a whole. The literature contains no large comparative studies that answer such questions as "Is the risk of delayed diag­nosis of ovarian cancer increased by not removing a 14-week fibroid uterus?" or "In what percentage of women with pelvic pain does hysterectomy as opposed to simple observation improve the quality of life?"

Finally, obstetrician-gynecologists must continue to examine the processes of decision making and delivery of care, and they must begin to undertake extensive studies that describe long-term surgical outcomes. In this way, "appropriate" indications for surgery will be derived from outcome data and from specialty-wide consensus. As more practice-variation data become available, both the Health Care Financing Agency and insurers are likely to make increasing use of such data for the purpose of practice evaluation and regulation. Only a combination of the study of local rate variation, individual medical decision making, and clinical out­comes of different treatment decisions will allow gyne­cologists to continue to evolve an informed consensus

154 Haas et al.

of "appropriate" indications for surgery. With such a consensus, physicians will be able to respond to at­tempts by insurers and regulators to manage physician decision making.

We thank Elizabeth O'Brien for valuable assistance in data preparation.

REFERENCES 1. Glover JA. The incidence of tonsillectomy in school chil­

dren. Proc R Soc Med 1938;31:95-112. 2. Pearson~, Smedby B, Berfenstam R. Hospital care loads

in Liverpool, New England, and Uppsala: an international comparison. Lancet 1968;2:559-66.

3. McPherson K, Wennberg JE, Hovind OB, Clifford P. Small-area variations in the use of common surgical procedures: an international comparison of New England, England, and NOIway. N Engl J Med 1982;307:1310-4.

4. Vayda E. A comparison of surgical rates in Canada and in England and Wales. N Engl J Med 1973;289:1224-9.

5. Perrin JM, Homer CJ, BelWick DM, Woolf AD, Freeman JL, Wennberg JE. Variations in rates of hospitalization of children in three urban communities. N Engl J Med 1989;320: 1183-7.

6. Lewis CEo Variations in the incidence of surgery. N EnglJ Med 1969;281:880-4.

7. Barnes BA, O'Brien E, Comstock C, D'Arpa DG, Donahue C. Report on variation in rates of utilization of surgical services in the Commonwealth of Massachusetts. JAMA 1985;254:371-5.

8. Acker DB, Haas S, O'Brien E, Donahue CLJr, Porell MM, Sachs BP. Cesarean birth rate: small-geographic-area anal­ysis. AM J OBSTET GYNECOL 1988;159:386-8.

9. Nattinger AB, Gottlieb MS, Veum J, Yahnke D, Goodwin JS. Geographic variation in the use of breast-conserving treatment for breast cancer. N Engl J Med 1992;326: 1102-7.

10. Farrow DC, Hunt WC, Samet JM. Geographic variation in the treatment of localized breast cancer. N Engl J Med 1992;326: 1 097 -101.

11. Roos NP, Roos LL. Surgical rate variations: do they reflect

Bound volumes available to subscribers

July 1993 Am J Obstet Gynecol

the health or socioeconomic characteristics of the popula­tion? Med Care 1982;20:945-58.

12. Wennberg J, Gittelsohn A. Variations in medical care among small areas. Sci Am 1982;246:120-9.

13. Roos NP, Roos LL. High and low surgical rates: risk factors for area residents. Am J Public Health 1981; 71: 591-600.

14. Pasley B, Vernon P, Gibson G, McCauley M, Andoh J. Geographic variations in elderly hospital and surgical discharge rates, New York State. Am J Public Health 1987;77:679-84.

15. Wennberg JE, Fowler FL. A test of consumer contribution to small area variations in health care delivery. J Maine Med Assoc 1977;68:275-9.

16. Pokras R, Hufnagel VG. Hysterectomies in the United States, 1965-84. Vital Health Stat [13] 1987;92:3.

17. Wennberg J, Gittelsohn AM. Small area variations in health care delivery. Science 1973; 182: 11 02-8.

18. Roos NP. Hysterectomy: variations in rates across small areas and across physicians' practices. Am J Public Health 1984;74:327-35.

19. Wenneker MB, Epstein AM. Racial inequalities in the use of procedures for patients with ischemic heart disease in Massachusetts. JAMA 1989;261 :253-7.

20. Goldberg KC, Hartz AJ, Jacobsen SJ, et al. Racial and community factors influencing coronary artery bypass graft surgery rates for all 1986 Medicare patients. JAMA 1992;267:1473-7.

21. Maryland Department of Health and Mental Hygiene. Variations in the use of medical and surgical services: the Maryland population. Report of the Division of Medical Practice Patterns Analysis. Office of Policy Analysis and Program Evaluation. Baltimore, Maryland: Maryland De­partment of Health and Mental Hygiene, 1986.

22. Pokras R, Hufnagel VG. Hysterectomy in the United States: 1964-84. Am J Public Health 1988;78:852-3.

23. Lyon JL, Gardner JW. The rising frequency of hysterectomy: its effect on uterine cancer rates. Am J Epidemiol 1977; 1 05:439-43.

24. Keller RB, Soule DN. Annual progress report 1991. Au­gusta, Maine: Maine Medical Assessment Foundation, 1991 Dec.

Bound volumes of the AMERICAN JOURNAL OF OBSTETRICS AND GYNECOLOGY are avail­able to subscribers (only) for the 1993 issues from the Publisher, at a cost of $74.50 for domestic, $102.72 for Canada, and $97.50 for international for Vol. 168 (January-June) and Vol. 169 (July-December). Shipping charges are included. Each bound volume contains a subject and author index and all advertising is removed. Copies are shipped within 60 days after publication of the last issue in the volume. The binding is durable buckram with the JOURNAL name, volume number, and year stamped in gold on the spine. Payment must accompany all orders. Contact Mosby, Subscription Services, 11830 Westline Industrial Drive, St. Louis, MO 63146-3318, USA; phone (800) 453-4351 or (314) 453-4351.

Subscriptions must be in force to qualify. Bound volumes are not available in place of a regular JOURNAL subscription.